DIFF_Transformer.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "PXoKw0D4C8kB"
      },
      "source": [
        "## **Differential Transformer**\n",
        "This is a lightweight (Differential nanoGPT) implementation of the newly proposed \"Differential Tranformer\". You can find offical implementation here: https://github.com/microsoft/unilm/tree/master/Diff-Transformer\n",
        "\n",
        "\n",
        "<img src=\"assets/dtm.png\"/>\n",
        "<img src=\"assets/intro.png\"/>\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "SfGcL_qXTVEC",
        "outputId": "564cd78b-4ccc-45e0-970a-277757526a5e"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "\u001b[1;30;43mStreaming output truncated to the last 5000 lines.\u001b[0m\n",
            "iter 45051: loss 1.4397, time 9.76ms, mfu 0.49%\n",
            "iter 45052: loss 1.4657, time 9.47ms, mfu 0.51%\n",
            "iter 45053: loss 1.5523, time 9.30ms, mfu 0.52%\n",
            "iter 45054: loss 1.5854, time 9.46ms, mfu 0.51%\n",
            "iter 45055: loss 1.4762, time 9.33ms, mfu 0.52%\n",
            "iter 45056: loss 1.4950, time 9.29ms, mfu 0.52%\n",
            "iter 45057: loss 1.5420, time 9.42ms, mfu 0.51%\n",
            "iter 45058: loss 1.4430, time 10.23ms, mfu 0.47%\n",
            "iter 45059: loss 1.4006, time 9.75ms, mfu 0.49%\n",
            "iter 45060: loss 1.5294, time 9.42ms, mfu 0.51%\n",
            "iter 45061: loss 1.4692, time 9.44ms, mfu 0.51%\n",
            "iter 45062: loss 1.4937, time 9.67ms, mfu 0.50%\n",
            "iter 45063: loss 1.4826, time 9.21ms, mfu 0.52%\n",
            "iter 45064: loss 1.3480, time 9.20ms, mfu 0.52%\n",
            "iter 45065: loss 1.3764, time 9.34ms, mfu 0.52%\n",
            "iter 45066: loss 1.5065, time 15.16ms, mfu 0.32%\n",
            "iter 45067: loss 1.4512, time 9.37ms, mfu 0.51%\n",
            "iter 45068: loss 1.4891, time 9.32ms, mfu 0.52%\n",
            "iter 45069: loss 1.5007, time 9.51ms, mfu 0.51%\n",
            "iter 45070: loss 1.5254, time 12.73ms, mfu 0.38%\n",
            "iter 45071: loss 1.4674, time 12.01ms, mfu 0.40%\n",
            "iter 45072: loss 1.5017, time 10.60ms, mfu 0.45%\n",
            "iter 45073: loss 1.4758, time 9.28ms, mfu 0.52%\n",
            "iter 45074: loss 1.5616, time 9.30ms, mfu 0.52%\n",
            "iter 45075: loss 1.4940, time 9.35ms, mfu 0.52%\n",
            "iter 45076: loss 1.3655, time 9.38ms, mfu 0.51%\n",
            "iter 45077: loss 1.4854, time 9.42ms, mfu 0.51%\n",
            "iter 45078: loss 1.4398, time 14.61ms, mfu 0.33%\n",
            "iter 45079: loss 1.4823, time 13.51ms, mfu 0.36%\n",
            "iter 45080: loss 1.4162, time 9.25ms, mfu 0.52%\n",
            "iter 45081: loss 1.5406, time 9.16ms, mfu 0.53%\n",
            "iter 45082: loss 1.5811, time 9.16ms, mfu 0.53%\n",
            "iter 45083: loss 1.5537, time 9.31ms, mfu 0.52%\n",
            "iter 45084: loss 1.6165, time 10.33ms, mfu 0.47%\n",
            "iter 45085: loss 1.3815, time 9.29ms, mfu 0.52%\n",
            "iter 45086: loss 1.4962, time 13.24ms, mfu 0.36%\n",
            "iter 45087: loss 1.4173, time 9.56ms, mfu 0.50%\n",
            "iter 45088: loss 1.4978, time 9.54ms, mfu 0.51%\n",
            "iter 45089: loss 1.4408, time 9.22ms, mfu 0.52%\n",
            "iter 45090: loss 1.4617, time 9.23ms, mfu 0.52%\n",
            "iter 45091: loss 1.4660, time 9.27ms, mfu 0.52%\n",
            "iter 45092: loss 1.4114, time 9.33ms, mfu 0.52%\n",
            "iter 45093: loss 1.5485, time 9.41ms, mfu 0.51%\n",
            "iter 45094: loss 1.5520, time 9.32ms, mfu 0.52%\n",
            "iter 45095: loss 1.4951, time 9.19ms, mfu 0.53%\n",
            "iter 45096: loss 1.5048, time 9.30ms, mfu 0.52%\n",
            "iter 45097: loss 1.4564, time 9.11ms, mfu 0.53%\n",
            "iter 45098: loss 1.5091, time 9.21ms, mfu 0.52%\n",
            "iter 45099: loss 1.6052, time 9.09ms, mfu 0.53%\n",
            "iter 45100: loss 1.5506, time 9.14ms, mfu 0.53%\n",
            "iter 45101: loss 1.5058, time 10.72ms, mfu 0.45%\n",
            "iter 45102: loss 1.5991, time 9.82ms, mfu 0.49%\n",
            "iter 45103: loss 1.5151, time 9.56ms, mfu 0.50%\n",
            "iter 45104: loss 1.5568, time 9.47ms, mfu 0.51%\n",
            "iter 45105: loss 1.5627, time 9.66ms, mfu 0.50%\n",
            "iter 45106: loss 1.3952, time 9.56ms, mfu 0.50%\n",
            "iter 45107: loss 1.4168, time 9.60ms, mfu 0.50%\n",
            "iter 45108: loss 1.5016, time 14.64ms, mfu 0.33%\n",
            "iter 45109: loss 1.4564, time 11.25ms, mfu 0.43%\n",
            "iter 45110: loss 1.4697, time 10.65ms, mfu 0.45%\n",
            "iter 45111: loss 1.5585, time 9.37ms, mfu 0.51%\n",
            "iter 45112: loss 1.5544, time 9.47ms, mfu 0.51%\n",
            "iter 45113: loss 1.5032, time 9.32ms, mfu 0.52%\n",
            "iter 45114: loss 1.4948, time 9.44ms, mfu 0.51%\n",
            "iter 45115: loss 1.4993, time 9.61ms, mfu 0.50%\n",
            "iter 45116: loss 1.5127, time 9.52ms, mfu 0.51%\n",
            "iter 45117: loss 1.5509, time 9.34ms, mfu 0.52%\n",
            "iter 45118: loss 1.5366, time 9.43ms, mfu 0.51%\n",
            "iter 45119: loss 1.4418, time 11.30ms, mfu 0.43%\n",
            "iter 45120: loss 1.5157, time 9.44ms, mfu 0.51%\n",
            "iter 45121: loss 1.4720, time 9.23ms, mfu 0.52%\n",
            "iter 45122: loss 1.5045, time 9.46ms, mfu 0.51%\n",
            "iter 45123: loss 1.4276, time 10.77ms, mfu 0.45%\n",
            "iter 45124: loss 1.5629, time 10.99ms, mfu 0.44%\n",
            "iter 45125: loss 1.4007, time 9.57ms, mfu 0.50%\n",
            "iter 45126: loss 1.4368, time 9.24ms, mfu 0.52%\n",
            "iter 45127: loss 1.4843, time 9.42ms, mfu 0.51%\n",
            "iter 45128: loss 1.5548, time 9.40ms, mfu 0.51%\n",
            "iter 45129: loss 1.6107, time 13.11ms, mfu 0.37%\n",
            "iter 45130: loss 1.6025, time 11.10ms, mfu 0.43%\n",
            "iter 45131: loss 1.4015, time 9.76ms, mfu 0.49%\n",
            "iter 45132: loss 1.5249, time 9.36ms, mfu 0.52%\n",
            "iter 45133: loss 1.4811, time 9.65ms, mfu 0.50%\n",
            "iter 45134: loss 1.3647, time 9.30ms, mfu 0.52%\n",
            "iter 45135: loss 1.3896, time 9.46ms, mfu 0.51%\n",
            "iter 45136: loss 1.4797, time 9.44ms, mfu 0.51%\n",
            "iter 45137: loss 1.5270, time 9.32ms, mfu 0.52%\n",
            "iter 45138: loss 1.4307, time 9.54ms, mfu 0.51%\n",
            "iter 45139: loss 1.5855, time 9.32ms, mfu 0.52%\n",
            "iter 45140: loss 1.5065, time 9.28ms, mfu 0.52%\n",
            "iter 45141: loss 1.3338, time 9.37ms, mfu 0.51%\n",
            "iter 45142: loss 1.5177, time 9.33ms, mfu 0.52%\n",
            "iter 45143: loss 1.4020, time 9.23ms, mfu 0.52%\n",
            "iter 45144: loss 1.4803, time 9.51ms, mfu 0.51%\n",
            "iter 45145: loss 1.4032, time 9.62ms, mfu 0.50%\n",
            "iter 45146: loss 1.5004, time 9.91ms, mfu 0.49%\n",
            "iter 45147: loss 1.3138, time 9.36ms, mfu 0.52%\n",
            "iter 45148: loss 1.5277, time 9.35ms, mfu 0.52%\n",
            "iter 45149: loss 1.5480, time 9.22ms, mfu 0.52%\n",
            "iter 45150: loss 1.5063, time 9.72ms, mfu 0.50%\n",
            "iter 45151: loss 1.4638, time 14.06ms, mfu 0.34%\n",
            "iter 45152: loss 1.5606, time 9.87ms, mfu 0.49%\n",
            "iter 45153: loss 1.4891, time 9.37ms, mfu 0.51%\n",
            "iter 45154: loss 1.4536, time 9.34ms, mfu 0.52%\n",
            "iter 45155: loss 1.4500, time 9.41ms, mfu 0.51%\n",
            "iter 45156: loss 1.5755, time 9.35ms, mfu 0.52%\n",
            "iter 45157: loss 1.4079, time 9.34ms, mfu 0.52%\n",
            "iter 45158: loss 1.4852, time 9.60ms, mfu 0.50%\n",
            "iter 45159: loss 1.4355, time 9.37ms, mfu 0.51%\n",
            "iter 45160: loss 1.5161, time 10.06ms, mfu 0.48%\n",
            "iter 45161: loss 1.4575, time 10.02ms, mfu 0.48%\n",
            "iter 45162: loss 1.4654, time 9.56ms, mfu 0.50%\n",
            "iter 45163: loss 1.4927, time 9.41ms, mfu 0.51%\n",
            "iter 45164: loss 1.5729, time 9.20ms, mfu 0.52%\n",
            "iter 45165: loss 1.5152, time 9.33ms, mfu 0.52%\n",
            "iter 45166: loss 1.4339, time 10.15ms, mfu 0.48%\n",
            "iter 45167: loss 1.5362, time 11.60ms, mfu 0.42%\n",
            "iter 45168: loss 1.2830, time 9.59ms, mfu 0.50%\n",
            "iter 45169: loss 1.4372, time 9.50ms, mfu 0.51%\n",
            "iter 45170: loss 1.4197, time 9.58ms, mfu 0.50%\n",
            "iter 45171: loss 1.5181, time 10.29ms, mfu 0.47%\n",
            "iter 45172: loss 1.4703, time 16.03ms, mfu 0.30%\n",
            "iter 45173: loss 1.5089, time 12.78ms, mfu 0.38%\n",
            "iter 45174: loss 1.4863, time 9.84ms, mfu 0.49%\n",
            "iter 45175: loss 1.5904, time 9.26ms, mfu 0.52%\n",
            "iter 45176: loss 1.5010, time 9.23ms, mfu 0.52%\n",
            "iter 45177: loss 1.4700, time 9.41ms, mfu 0.51%\n",
            "iter 45178: loss 1.5554, time 12.16ms, mfu 0.40%\n",
            "iter 45179: loss 1.4770, time 9.53ms, mfu 0.51%\n",
            "iter 45180: loss 1.4945, time 9.57ms, mfu 0.50%\n",
            "iter 45181: loss 1.4413, time 14.97ms, mfu 0.32%\n",
            "iter 45182: loss 1.5515, time 13.50ms, mfu 0.36%\n",
            "iter 45183: loss 1.5254, time 9.49ms, mfu 0.51%\n",
            "iter 45184: loss 1.4988, time 10.14ms, mfu 0.48%\n",
            "iter 45185: loss 1.5053, time 9.66ms, mfu 0.50%\n",
            "iter 45186: loss 1.3791, time 9.49ms, mfu 0.51%\n",
            "iter 45187: loss 1.5016, time 9.27ms, mfu 0.52%\n",
            "iter 45188: loss 1.5078, time 9.31ms, mfu 0.52%\n",
            "iter 45189: loss 1.4256, time 9.28ms, mfu 0.52%\n",
            "iter 45190: loss 1.5161, time 9.34ms, mfu 0.52%\n",
            "iter 45191: loss 1.5697, time 9.60ms, mfu 0.50%\n",
            "iter 45192: loss 1.4926, time 14.45ms, mfu 0.33%\n",
            "iter 45193: loss 1.4638, time 9.89ms, mfu 0.49%\n",
            "iter 45194: loss 1.4558, time 9.54ms, mfu 0.51%\n",
            "iter 45195: loss 1.5836, time 9.53ms, mfu 0.51%\n",
            "iter 45196: loss 1.4805, time 9.26ms, mfu 0.52%\n",
            "iter 45197: loss 1.5667, time 9.31ms, mfu 0.52%\n",
            "iter 45198: loss 1.5747, time 9.28ms, mfu 0.52%\n",
            "iter 45199: loss 1.4343, time 9.24ms, mfu 0.52%\n",
            "iter 45200: loss 1.4229, time 9.30ms, mfu 0.52%\n",
            "iter 45201: loss 1.4511, time 9.60ms, mfu 0.50%\n",
            "iter 45202: loss 1.4660, time 10.45ms, mfu 0.46%\n",
            "iter 45203: loss 1.5049, time 9.61ms, mfu 0.50%\n",
            "iter 45204: loss 1.4222, time 9.68ms, mfu 0.50%\n",
            "iter 45205: loss 1.4193, time 9.83ms, mfu 0.49%\n",
            "iter 45206: loss 1.3756, time 9.45ms, mfu 0.51%\n",
            "iter 45207: loss 1.5909, time 13.66ms, mfu 0.35%\n",
            "iter 45208: loss 1.4670, time 11.94ms, mfu 0.40%\n",
            "iter 45209: loss 1.5798, time 9.65ms, mfu 0.50%\n",
            "iter 45210: loss 1.4290, time 9.97ms, mfu 0.48%\n",
            "iter 45211: loss 1.5863, time 9.78ms, mfu 0.49%\n",
            "iter 45212: loss 1.4515, time 9.66ms, mfu 0.50%\n",
            "iter 45213: loss 1.5235, time 14.65ms, mfu 0.33%\n",
            "iter 45214: loss 1.4023, time 13.12ms, mfu 0.37%\n",
            "iter 45215: loss 1.4699, time 9.56ms, mfu 0.50%\n",
            "iter 45216: loss 1.4964, time 9.63ms, mfu 0.50%\n",
            "iter 45217: loss 1.4612, time 9.73ms, mfu 0.50%\n",
            "iter 45218: loss 1.4030, time 9.29ms, mfu 0.52%\n",
            "iter 45219: loss 1.4444, time 9.41ms, mfu 0.51%\n",
            "iter 45220: loss 1.4876, time 9.38ms, mfu 0.51%\n",
            "iter 45221: loss 1.4535, time 9.81ms, mfu 0.49%\n",
            "iter 45222: loss 1.4350, time 9.40ms, mfu 0.51%\n",
            "iter 45223: loss 1.4021, time 9.29ms, mfu 0.52%\n",
            "iter 45224: loss 1.5537, time 11.07ms, mfu 0.44%\n",
            "iter 45225: loss 1.5682, time 9.47ms, mfu 0.51%\n",
            "iter 45226: loss 1.4302, time 9.26ms, mfu 0.52%\n",
            "iter 45227: loss 1.4971, time 9.43ms, mfu 0.51%\n",
            "iter 45228: loss 1.3306, time 9.42ms, mfu 0.51%\n",
            "iter 45229: loss 1.3931, time 9.25ms, mfu 0.52%\n",
            "iter 45230: loss 1.4983, time 10.55ms, mfu 0.46%\n",
            "iter 45231: loss 1.4970, time 9.47ms, mfu 0.51%\n",
            "iter 45232: loss 1.5963, time 9.42ms, mfu 0.51%\n",
            "iter 45233: loss 1.5234, time 9.37ms, mfu 0.51%\n",
            "iter 45234: loss 1.5532, time 12.00ms, mfu 0.40%\n",
            "iter 45235: loss 1.4744, time 10.30ms, mfu 0.47%\n",
            "iter 45236: loss 1.3971, time 9.60ms, mfu 0.50%\n",
            "iter 45237: loss 1.4118, time 9.48ms, mfu 0.51%\n",
            "iter 45238: loss 1.4750, time 10.58ms, mfu 0.46%\n",
            "iter 45239: loss 1.4704, time 9.30ms, mfu 0.52%\n",
            "iter 45240: loss 1.4280, time 9.22ms, mfu 0.52%\n",
            "iter 45241: loss 1.4216, time 9.30ms, mfu 0.52%\n",
            "iter 45242: loss 1.3580, time 9.44ms, mfu 0.51%\n",
            "iter 45243: loss 1.5596, time 10.72ms, mfu 0.45%\n",
            "iter 45244: loss 1.4419, time 10.82ms, mfu 0.45%\n",
            "iter 45245: loss 1.3826, time 9.53ms, mfu 0.51%\n",
            "iter 45246: loss 1.4856, time 9.30ms, mfu 0.52%\n",
            "iter 45247: loss 1.4494, time 9.44ms, mfu 0.51%\n",
            "iter 45248: loss 1.4559, time 9.25ms, mfu 0.52%\n",
            "iter 45249: loss 1.5049, time 9.45ms, mfu 0.51%\n",
            "iter 45250: loss 1.4996, time 9.41ms, mfu 0.51%\n",
            "iter 45251: loss 1.4850, time 9.80ms, mfu 0.49%\n",
            "iter 45252: loss 1.4172, time 10.90ms, mfu 0.44%\n",
            "iter 45253: loss 1.4658, time 9.47ms, mfu 0.51%\n",
            "iter 45254: loss 1.6222, time 9.64ms, mfu 0.50%\n",
            "iter 45255: loss 1.4989, time 11.90ms, mfu 0.41%\n",
            "iter 45256: loss 1.5879, time 11.22ms, mfu 0.43%\n",
            "iter 45257: loss 1.5293, time 9.29ms, mfu 0.52%\n",
            "iter 45258: loss 1.4755, time 9.57ms, mfu 0.50%\n",
            "iter 45259: loss 1.4475, time 9.80ms, mfu 0.49%\n",
            "iter 45260: loss 1.6077, time 9.57ms, mfu 0.50%\n",
            "iter 45261: loss 1.5159, time 11.19ms, mfu 0.43%\n",
            "iter 45262: loss 1.3554, time 9.53ms, mfu 0.51%\n",
            "iter 45263: loss 1.5776, time 9.70ms, mfu 0.50%\n",
            "iter 45264: loss 1.4542, time 9.53ms, mfu 0.51%\n",
            "iter 45265: loss 1.4970, time 10.90ms, mfu 0.44%\n",
            "iter 45266: loss 1.5452, time 9.26ms, mfu 0.52%\n",
            "iter 45267: loss 1.6508, time 10.18ms, mfu 0.47%\n",
            "iter 45268: loss 1.4648, time 9.41ms, mfu 0.51%\n",
            "iter 45269: loss 1.5128, time 9.73ms, mfu 0.50%\n",
            "iter 45270: loss 1.4753, time 11.30ms, mfu 0.43%\n",
            "iter 45271: loss 1.4697, time 12.00ms, mfu 0.40%\n",
            "iter 45272: loss 1.4443, time 10.19ms, mfu 0.47%\n",
            "iter 45273: loss 1.4200, time 9.29ms, mfu 0.52%\n",
            "iter 45274: loss 1.5097, time 9.76ms, mfu 0.49%\n",
            "iter 45275: loss 1.3480, time 9.63ms, mfu 0.50%\n",
            "iter 45276: loss 1.4826, time 14.01ms, mfu 0.34%\n",
            "iter 45277: loss 1.5108, time 9.65ms, mfu 0.50%\n",
            "iter 45278: loss 1.3544, time 9.54ms, mfu 0.51%\n",
            "iter 45279: loss 1.4604, time 9.59ms, mfu 0.50%\n",
            "iter 45280: loss 1.5781, time 9.49ms, mfu 0.51%\n",
            "iter 45281: loss 1.4669, time 9.63ms, mfu 0.50%\n",
            "iter 45282: loss 1.5115, time 11.92ms, mfu 0.40%\n",
            "iter 45283: loss 1.4621, time 16.59ms, mfu 0.29%\n",
            "iter 45284: loss 1.4435, time 11.08ms, mfu 0.44%\n",
            "iter 45285: loss 1.4790, time 9.63ms, mfu 0.50%\n",
            "iter 45286: loss 1.4916, time 9.30ms, mfu 0.52%\n",
            "iter 45287: loss 1.5402, time 9.46ms, mfu 0.51%\n",
            "iter 45288: loss 1.5165, time 9.24ms, mfu 0.52%\n",
            "iter 45289: loss 1.5806, time 9.86ms, mfu 0.49%\n",
            "iter 45290: loss 1.4591, time 9.53ms, mfu 0.51%\n",
            "iter 45291: loss 1.5704, time 10.23ms, mfu 0.47%\n",
            "iter 45292: loss 1.4957, time 9.52ms, mfu 0.51%\n",
            "iter 45293: loss 1.4959, time 9.40ms, mfu 0.51%\n",
            "iter 45294: loss 1.4868, time 9.51ms, mfu 0.51%\n",
            "iter 45295: loss 1.5131, time 9.37ms, mfu 0.51%\n",
            "iter 45296: loss 1.3993, time 10.64ms, mfu 0.45%\n",
            "iter 45297: loss 1.4047, time 13.92ms, mfu 0.35%\n",
            "iter 45298: loss 1.5776, time 9.42ms, mfu 0.51%\n",
            "iter 45299: loss 1.4229, time 9.36ms, mfu 0.52%\n",
            "iter 45300: loss 1.4972, time 9.40ms, mfu 0.51%\n",
            "iter 45301: loss 1.4399, time 15.07ms, mfu 0.32%\n",
            "iter 45302: loss 1.4448, time 14.26ms, mfu 0.34%\n",
            "iter 45303: loss 1.5080, time 20.55ms, mfu 0.23%\n",
            "iter 45304: loss 1.5195, time 16.87ms, mfu 0.29%\n",
            "iter 45305: loss 1.4791, time 14.33ms, mfu 0.34%\n",
            "iter 45306: loss 1.4257, time 22.79ms, mfu 0.21%\n",
            "iter 45307: loss 1.4336, time 21.90ms, mfu 0.22%\n",
            "iter 45308: loss 1.4182, time 20.35ms, mfu 0.24%\n",
            "iter 45309: loss 1.3899, time 18.01ms, mfu 0.27%\n",
            "iter 45310: loss 1.4402, time 19.37ms, mfu 0.25%\n",
            "iter 45311: loss 1.5017, time 17.10ms, mfu 0.28%\n",
            "iter 45312: loss 1.4249, time 30.91ms, mfu 0.16%\n",
            "iter 45313: loss 1.5181, time 21.88ms, mfu 0.22%\n",
            "iter 45314: loss 1.5477, time 16.97ms, mfu 0.28%\n",
            "iter 45315: loss 1.5409, time 16.72ms, mfu 0.29%\n",
            "iter 45316: loss 1.4609, time 29.94ms, mfu 0.16%\n",
            "iter 45317: loss 1.4192, time 21.44ms, mfu 0.22%\n",
            "iter 45318: loss 1.4748, time 22.39ms, mfu 0.22%\n",
            "iter 45319: loss 1.4344, time 25.28ms, mfu 0.19%\n",
            "iter 45320: loss 1.4727, time 22.37ms, mfu 0.22%\n",
            "iter 45321: loss 1.4229, time 15.65ms, mfu 0.31%\n",
            "iter 45322: loss 1.3283, time 16.41ms, mfu 0.29%\n",
            "iter 45323: loss 1.4438, time 17.08ms, mfu 0.28%\n",
            "iter 45324: loss 1.4894, time 15.13ms, mfu 0.32%\n",
            "iter 45325: loss 1.3354, time 15.08ms, mfu 0.32%\n",
            "iter 45326: loss 1.4595, time 15.23ms, mfu 0.32%\n",
            "iter 45327: loss 1.3641, time 15.59ms, mfu 0.31%\n",
            "iter 45328: loss 1.4376, time 17.00ms, mfu 0.28%\n",
            "iter 45329: loss 1.5572, time 15.88ms, mfu 0.30%\n",
            "iter 45330: loss 1.4612, time 15.40ms, mfu 0.31%\n",
            "iter 45331: loss 1.5058, time 16.96ms, mfu 0.28%\n",
            "iter 45332: loss 1.5070, time 21.42ms, mfu 0.23%\n",
            "iter 45333: loss 1.3481, time 16.41ms, mfu 0.29%\n",
            "iter 45334: loss 1.4558, time 16.12ms, mfu 0.30%\n",
            "iter 45335: loss 1.6102, time 23.86ms, mfu 0.20%\n",
            "iter 45336: loss 1.4914, time 15.49ms, mfu 0.31%\n",
            "iter 45337: loss 1.4185, time 25.62ms, mfu 0.19%\n",
            "iter 45338: loss 1.3652, time 15.15ms, mfu 0.32%\n",
            "iter 45339: loss 1.4089, time 16.74ms, mfu 0.29%\n",
            "iter 45340: loss 1.3378, time 15.79ms, mfu 0.31%\n",
            "iter 45341: loss 1.4639, time 15.74ms, mfu 0.31%\n",
            "iter 45342: loss 1.6262, time 14.88ms, mfu 0.32%\n",
            "iter 45343: loss 1.5334, time 15.83ms, mfu 0.30%\n",
            "iter 45344: loss 1.4672, time 17.41ms, mfu 0.28%\n",
            "iter 45345: loss 1.5637, time 17.72ms, mfu 0.27%\n",
            "iter 45346: loss 1.4729, time 28.58ms, mfu 0.17%\n",
            "iter 45347: loss 1.4852, time 16.50ms, mfu 0.29%\n",
            "iter 45348: loss 1.3982, time 16.47ms, mfu 0.29%\n",
            "iter 45349: loss 1.5384, time 16.78ms, mfu 0.29%\n",
            "iter 45350: loss 1.5186, time 26.82ms, mfu 0.18%\n",
            "iter 45351: loss 1.5389, time 18.63ms, mfu 0.26%\n",
            "iter 45352: loss 1.3881, time 19.44ms, mfu 0.25%\n",
            "iter 45353: loss 1.4291, time 20.32ms, mfu 0.24%\n",
            "iter 45354: loss 1.5025, time 40.93ms, mfu 0.12%\n",
            "iter 45355: loss 1.4782, time 31.63ms, mfu 0.15%\n",
            "iter 45356: loss 1.3691, time 17.05ms, mfu 0.28%\n",
            "iter 45357: loss 1.4445, time 20.69ms, mfu 0.23%\n",
            "iter 45358: loss 1.5962, time 18.22ms, mfu 0.26%\n",
            "iter 45359: loss 1.5404, time 15.65ms, mfu 0.31%\n",
            "iter 45360: loss 1.4504, time 15.23ms, mfu 0.32%\n",
            "iter 45361: loss 1.5214, time 12.90ms, mfu 0.37%\n",
            "iter 45362: loss 1.4817, time 13.87ms, mfu 0.35%\n",
            "iter 45363: loss 1.5050, time 15.14ms, mfu 0.32%\n",
            "iter 45364: loss 1.4601, time 13.15ms, mfu 0.37%\n",
            "iter 45365: loss 1.5340, time 14.08ms, mfu 0.34%\n",
            "iter 45366: loss 1.4582, time 13.26ms, mfu 0.36%\n",
            "iter 45367: loss 1.4501, time 15.03ms, mfu 0.32%\n",
            "iter 45368: loss 1.3424, time 15.65ms, mfu 0.31%\n",
            "iter 45369: loss 1.4559, time 16.39ms, mfu 0.29%\n",
            "iter 45370: loss 1.3826, time 16.07ms, mfu 0.30%\n",
            "iter 45371: loss 1.5013, time 14.68ms, mfu 0.33%\n",
            "iter 45372: loss 1.6992, time 13.31ms, mfu 0.36%\n",
            "iter 45373: loss 1.4964, time 12.99ms, mfu 0.37%\n",
            "iter 45374: loss 1.5173, time 14.34ms, mfu 0.34%\n",
            "iter 45375: loss 1.5143, time 12.26ms, mfu 0.39%\n",
            "iter 45376: loss 1.3771, time 12.14ms, mfu 0.40%\n",
            "iter 45377: loss 1.5503, time 12.03ms, mfu 0.40%\n",
            "iter 45378: loss 1.5246, time 12.03ms, mfu 0.40%\n",
            "iter 45379: loss 1.5204, time 12.03ms, mfu 0.40%\n",
            "iter 45380: loss 1.5214, time 12.07ms, mfu 0.40%\n",
            "iter 45381: loss 1.4438, time 13.27ms, mfu 0.36%\n",
            "iter 45382: loss 1.4251, time 14.12ms, mfu 0.34%\n",
            "iter 45383: loss 1.4628, time 12.46ms, mfu 0.39%\n",
            "iter 45384: loss 1.5465, time 12.79ms, mfu 0.38%\n",
            "iter 45385: loss 1.5059, time 12.54ms, mfu 0.38%\n",
            "iter 45386: loss 1.3872, time 12.34ms, mfu 0.39%\n",
            "iter 45387: loss 1.4423, time 12.17ms, mfu 0.40%\n",
            "iter 45388: loss 1.4828, time 12.07ms, mfu 0.40%\n",
            "iter 45389: loss 1.4700, time 12.28ms, mfu 0.39%\n",
            "iter 45390: loss 1.5602, time 12.13ms, mfu 0.40%\n",
            "iter 45391: loss 1.5254, time 12.33ms, mfu 0.39%\n",
            "iter 45392: loss 1.4895, time 17.26ms, mfu 0.28%\n",
            "iter 45393: loss 1.3697, time 15.05ms, mfu 0.32%\n",
            "iter 45394: loss 1.4536, time 12.16ms, mfu 0.40%\n",
            "iter 45395: loss 1.3974, time 12.43ms, mfu 0.39%\n",
            "iter 45396: loss 1.5678, time 12.22ms, mfu 0.39%\n",
            "iter 45397: loss 1.4136, time 22.10ms, mfu 0.22%\n",
            "iter 45398: loss 1.3056, time 12.82ms, mfu 0.38%\n",
            "iter 45399: loss 1.3884, time 16.98ms, mfu 0.28%\n",
            "iter 45400: loss 1.4487, time 14.43ms, mfu 0.33%\n",
            "iter 45401: loss 1.6070, time 13.42ms, mfu 0.36%\n",
            "iter 45402: loss 1.5059, time 11.63ms, mfu 0.41%\n",
            "iter 45403: loss 1.5603, time 11.69ms, mfu 0.41%\n",
            "iter 45404: loss 1.5108, time 13.40ms, mfu 0.36%\n",
            "iter 45405: loss 1.4131, time 12.36ms, mfu 0.39%\n",
            "iter 45406: loss 1.5066, time 11.84ms, mfu 0.41%\n",
            "iter 45407: loss 1.3827, time 11.63ms, mfu 0.41%\n",
            "iter 45408: loss 1.4198, time 11.60ms, mfu 0.42%\n",
            "iter 45409: loss 1.6182, time 11.78ms, mfu 0.41%\n",
            "iter 45410: loss 1.3923, time 11.57ms, mfu 0.42%\n",
            "iter 45411: loss 1.4370, time 12.41ms, mfu 0.39%\n",
            "iter 45412: loss 1.2972, time 15.15ms, mfu 0.32%\n",
            "iter 45413: loss 1.5352, time 19.35ms, mfu 0.25%\n",
            "iter 45414: loss 1.5280, time 18.01ms, mfu 0.27%\n",
            "iter 45415: loss 1.4458, time 12.87ms, mfu 0.37%\n",
            "iter 45416: loss 1.4816, time 13.97ms, mfu 0.35%\n",
            "iter 45417: loss 1.4642, time 12.25ms, mfu 0.39%\n",
            "iter 45418: loss 1.3858, time 11.82ms, mfu 0.41%\n",
            "iter 45419: loss 1.5735, time 11.60ms, mfu 0.42%\n",
            "iter 45420: loss 1.4206, time 11.60ms, mfu 0.42%\n",
            "iter 45421: loss 1.5351, time 11.58ms, mfu 0.42%\n",
            "iter 45422: loss 1.4698, time 11.54ms, mfu 0.42%\n",
            "iter 45423: loss 1.4243, time 13.04ms, mfu 0.37%\n",
            "iter 45424: loss 1.4673, time 13.39ms, mfu 0.36%\n",
            "iter 45425: loss 1.3452, time 12.45ms, mfu 0.39%\n",
            "iter 45426: loss 1.3967, time 12.88ms, mfu 0.37%\n",
            "iter 45427: loss 1.4088, time 11.89ms, mfu 0.41%\n",
            "iter 45428: loss 1.5338, time 12.25ms, mfu 0.39%\n",
            "iter 45429: loss 1.4708, time 19.37ms, mfu 0.25%\n",
            "iter 45430: loss 1.4855, time 17.32ms, mfu 0.28%\n",
            "iter 45431: loss 1.3880, time 17.74ms, mfu 0.27%\n",
            "iter 45432: loss 1.5189, time 15.42ms, mfu 0.31%\n",
            "iter 45433: loss 1.4034, time 15.14ms, mfu 0.32%\n",
            "iter 45434: loss 1.4607, time 17.80ms, mfu 0.27%\n",
            "iter 45435: loss 1.4369, time 14.75ms, mfu 0.33%\n",
            "iter 45436: loss 1.4606, time 12.24ms, mfu 0.39%\n",
            "iter 45437: loss 1.4630, time 15.03ms, mfu 0.32%\n",
            "iter 45438: loss 1.4447, time 12.48ms, mfu 0.39%\n",
            "iter 45439: loss 1.4610, time 11.81ms, mfu 0.41%\n",
            "iter 45440: loss 1.4909, time 14.33ms, mfu 0.34%\n",
            "iter 45441: loss 1.5445, time 19.20ms, mfu 0.25%\n",
            "iter 45442: loss 1.3871, time 12.30ms, mfu 0.39%\n",
            "iter 45443: loss 1.5267, time 17.92ms, mfu 0.27%\n",
            "iter 45444: loss 1.5620, time 18.91ms, mfu 0.25%\n",
            "iter 45445: loss 1.4783, time 15.68ms, mfu 0.31%\n",
            "iter 45446: loss 1.5352, time 12.15ms, mfu 0.40%\n",
            "iter 45447: loss 1.3953, time 12.21ms, mfu 0.39%\n",
            "iter 45448: loss 1.5608, time 12.30ms, mfu 0.39%\n",
            "iter 45449: loss 1.4464, time 13.96ms, mfu 0.35%\n",
            "iter 45450: loss 1.4873, time 16.23ms, mfu 0.30%\n",
            "iter 45451: loss 1.5368, time 13.16ms, mfu 0.37%\n",
            "iter 45452: loss 1.4549, time 15.33ms, mfu 0.31%\n",
            "iter 45453: loss 1.4375, time 12.05ms, mfu 0.40%\n",
            "iter 45454: loss 1.5441, time 11.82ms, mfu 0.41%\n",
            "iter 45455: loss 1.4740, time 12.14ms, mfu 0.40%\n",
            "iter 45456: loss 1.5048, time 14.16ms, mfu 0.34%\n",
            "iter 45457: loss 1.6143, time 13.69ms, mfu 0.35%\n",
            "iter 45458: loss 1.3950, time 19.00ms, mfu 0.25%\n",
            "iter 45459: loss 1.4824, time 16.46ms, mfu 0.29%\n",
            "iter 45460: loss 1.5465, time 12.20ms, mfu 0.40%\n",
            "iter 45461: loss 1.4263, time 12.56ms, mfu 0.38%\n",
            "iter 45462: loss 1.5081, time 12.16ms, mfu 0.40%\n",
            "iter 45463: loss 1.4811, time 12.37ms, mfu 0.39%\n",
            "iter 45464: loss 1.2927, time 12.22ms, mfu 0.39%\n",
            "iter 45465: loss 1.6491, time 12.19ms, mfu 0.40%\n",
            "iter 45466: loss 1.5392, time 12.52ms, mfu 0.39%\n",
            "iter 45467: loss 1.4730, time 13.07ms, mfu 0.37%\n",
            "iter 45468: loss 1.3810, time 12.58ms, mfu 0.38%\n",
            "iter 45469: loss 1.6061, time 14.41ms, mfu 0.33%\n",
            "iter 45470: loss 1.4776, time 12.16ms, mfu 0.40%\n",
            "iter 45471: loss 1.4251, time 13.70ms, mfu 0.35%\n",
            "iter 45472: loss 1.4980, time 14.36ms, mfu 0.34%\n",
            "iter 45473: loss 1.4981, time 13.52ms, mfu 0.36%\n",
            "iter 45474: loss 1.4769, time 16.95ms, mfu 0.28%\n",
            "iter 45475: loss 1.5092, time 17.96ms, mfu 0.27%\n",
            "iter 45476: loss 1.5057, time 12.26ms, mfu 0.39%\n",
            "iter 45477: loss 1.3371, time 12.10ms, mfu 0.40%\n",
            "iter 45478: loss 1.3403, time 12.35ms, mfu 0.39%\n",
            "iter 45479: loss 1.3645, time 12.09ms, mfu 0.40%\n",
            "iter 45480: loss 1.4701, time 12.50ms, mfu 0.39%\n",
            "iter 45481: loss 1.5256, time 12.29ms, mfu 0.39%\n",
            "iter 45482: loss 1.5432, time 12.15ms, mfu 0.40%\n",
            "iter 45483: loss 1.4660, time 14.81ms, mfu 0.33%\n",
            "iter 45484: loss 1.4071, time 12.75ms, mfu 0.38%\n",
            "iter 45485: loss 1.4567, time 12.80ms, mfu 0.38%\n",
            "iter 45486: loss 1.5622, time 11.67ms, mfu 0.41%\n",
            "iter 45487: loss 1.5180, time 11.80ms, mfu 0.41%\n",
            "iter 45488: loss 1.5279, time 11.83ms, mfu 0.41%\n",
            "iter 45489: loss 1.6311, time 11.86ms, mfu 0.41%\n",
            "iter 45490: loss 1.4948, time 15.95ms, mfu 0.30%\n",
            "iter 45491: loss 1.4702, time 15.66ms, mfu 0.31%\n",
            "iter 45492: loss 1.4928, time 12.85ms, mfu 0.38%\n",
            "iter 45493: loss 1.5196, time 14.89ms, mfu 0.32%\n",
            "iter 45494: loss 1.5452, time 12.00ms, mfu 0.40%\n",
            "iter 45495: loss 1.4948, time 11.64ms, mfu 0.41%\n",
            "iter 45496: loss 1.4509, time 11.86ms, mfu 0.41%\n",
            "iter 45497: loss 1.4172, time 11.61ms, mfu 0.42%\n",
            "iter 45498: loss 1.4311, time 11.64ms, mfu 0.41%\n",
            "iter 45499: loss 1.5645, time 12.32ms, mfu 0.39%\n",
            "step 45500: train loss 1.3512, val loss 1.5860\n",
            "iter 45500: loss 1.5858, time 21.02ms, mfu 0.23%\n",
            "iter 45501: loss 1.4990, time 11.81ms, mfu 0.41%\n",
            "iter 45502: loss 1.4923, time 11.68ms, mfu 0.41%\n",
            "iter 45503: loss 1.4715, time 11.67ms, mfu 0.41%\n",
            "iter 45504: loss 1.3792, time 12.01ms, mfu 0.40%\n",
            "iter 45505: loss 1.4002, time 11.78ms, mfu 0.41%\n",
            "iter 45506: loss 1.4594, time 14.67ms, mfu 0.33%\n",
            "iter 45507: loss 1.5245, time 12.42ms, mfu 0.39%\n",
            "iter 45508: loss 1.4517, time 12.42ms, mfu 0.39%\n",
            "iter 45509: loss 1.4994, time 11.98ms, mfu 0.40%\n",
            "iter 45510: loss 1.4529, time 11.83ms, mfu 0.41%\n",
            "iter 45511: loss 1.3361, time 11.88ms, mfu 0.41%\n",
            "iter 45512: loss 1.6062, time 11.96ms, mfu 0.40%\n",
            "iter 45513: loss 1.5550, time 11.95ms, mfu 0.40%\n",
            "iter 45514: loss 1.5696, time 11.92ms, mfu 0.40%\n",
            "iter 45515: loss 1.3907, time 11.89ms, mfu 0.41%\n",
            "iter 45516: loss 1.5138, time 12.08ms, mfu 0.40%\n",
            "iter 45517: loss 1.3698, time 14.27ms, mfu 0.34%\n",
            "iter 45518: loss 1.5786, time 16.34ms, mfu 0.30%\n",
            "iter 45519: loss 1.3789, time 16.32ms, mfu 0.30%\n",
            "iter 45520: loss 1.4079, time 12.36ms, mfu 0.39%\n",
            "iter 45521: loss 1.4821, time 15.75ms, mfu 0.31%\n",
            "iter 45522: loss 1.5263, time 22.23ms, mfu 0.22%\n",
            "iter 45523: loss 1.5026, time 16.25ms, mfu 0.30%\n",
            "iter 45524: loss 1.4168, time 15.17ms, mfu 0.32%\n",
            "iter 45525: loss 1.4098, time 15.19ms, mfu 0.32%\n",
            "iter 45526: loss 1.5081, time 13.75ms, mfu 0.35%\n",
            "iter 45527: loss 1.3722, time 12.33ms, mfu 0.39%\n",
            "iter 45528: loss 1.4425, time 12.50ms, mfu 0.39%\n",
            "iter 45529: loss 1.4184, time 13.29ms, mfu 0.36%\n",
            "iter 45530: loss 1.5226, time 13.02ms, mfu 0.37%\n",
            "iter 45531: loss 1.6135, time 12.94ms, mfu 0.37%\n",
            "iter 45532: loss 1.5031, time 11.94ms, mfu 0.40%\n",
            "iter 45533: loss 1.5456, time 11.83ms, mfu 0.41%\n",
            "iter 45534: loss 1.4502, time 11.68ms, mfu 0.41%\n",
            "iter 45535: loss 1.5305, time 15.54ms, mfu 0.31%\n",
            "iter 45536: loss 1.4061, time 14.79ms, mfu 0.33%\n",
            "iter 45537: loss 1.5666, time 22.90ms, mfu 0.21%\n",
            "iter 45538: loss 1.5345, time 20.61ms, mfu 0.23%\n",
            "iter 45539: loss 1.4266, time 15.46ms, mfu 0.31%\n",
            "iter 45540: loss 1.4604, time 14.79ms, mfu 0.33%\n",
            "iter 45541: loss 1.6109, time 12.92ms, mfu 0.37%\n",
            "iter 45542: loss 1.6225, time 16.96ms, mfu 0.28%\n",
            "iter 45543: loss 1.3880, time 13.45ms, mfu 0.36%\n",
            "iter 45544: loss 1.4690, time 12.03ms, mfu 0.40%\n",
            "iter 45545: loss 1.5261, time 16.96ms, mfu 0.28%\n",
            "iter 45546: loss 1.4296, time 24.22ms, mfu 0.20%\n",
            "iter 45547: loss 1.3723, time 15.45ms, mfu 0.31%\n",
            "iter 45548: loss 1.5130, time 15.57ms, mfu 0.31%\n",
            "iter 45549: loss 1.5273, time 17.31ms, mfu 0.28%\n",
            "iter 45550: loss 1.4990, time 22.17ms, mfu 0.22%\n",
            "iter 45551: loss 1.4584, time 21.94ms, mfu 0.22%\n",
            "iter 45552: loss 1.4423, time 13.58ms, mfu 0.36%\n",
            "iter 45553: loss 1.5152, time 14.63ms, mfu 0.33%\n",
            "iter 45554: loss 1.4521, time 13.72ms, mfu 0.35%\n",
            "iter 45555: loss 1.4924, time 13.25ms, mfu 0.36%\n",
            "iter 45556: loss 1.3001, time 14.55ms, mfu 0.33%\n",
            "iter 45557: loss 1.4566, time 17.51ms, mfu 0.28%\n",
            "iter 45558: loss 1.4686, time 14.32ms, mfu 0.34%\n",
            "iter 45559: loss 1.4942, time 13.59ms, mfu 0.35%\n",
            "iter 45560: loss 1.4778, time 14.03ms, mfu 0.34%\n",
            "iter 45561: loss 1.3996, time 13.96ms, mfu 0.35%\n",
            "iter 45562: loss 1.4651, time 13.58ms, mfu 0.36%\n",
            "iter 45563: loss 1.4763, time 17.13ms, mfu 0.28%\n",
            "iter 45564: loss 1.4924, time 18.51ms, mfu 0.26%\n",
            "iter 45565: loss 1.5079, time 15.73ms, mfu 0.31%\n",
            "iter 45566: loss 1.5257, time 16.73ms, mfu 0.29%\n",
            "iter 45567: loss 1.4031, time 14.56ms, mfu 0.33%\n",
            "iter 45568: loss 1.5326, time 14.68ms, mfu 0.33%\n",
            "iter 45569: loss 1.4453, time 14.77ms, mfu 0.33%\n",
            "iter 45570: loss 1.3756, time 14.37ms, mfu 0.34%\n",
            "iter 45571: loss 1.3514, time 15.00ms, mfu 0.32%\n",
            "iter 45572: loss 1.3548, time 14.36ms, mfu 0.34%\n",
            "iter 45573: loss 1.4897, time 14.79ms, mfu 0.33%\n",
            "iter 45574: loss 1.4662, time 14.57ms, mfu 0.33%\n",
            "iter 45575: loss 1.4123, time 14.48ms, mfu 0.33%\n",
            "iter 45576: loss 1.4713, time 14.47ms, mfu 0.33%\n",
            "iter 45577: loss 1.4833, time 16.81ms, mfu 0.29%\n",
            "iter 45578: loss 1.5370, time 21.46ms, mfu 0.22%\n",
            "iter 45579: loss 1.4394, time 16.21ms, mfu 0.30%\n",
            "iter 45580: loss 1.5294, time 10.49ms, mfu 0.46%\n",
            "iter 45581: loss 1.4852, time 11.35ms, mfu 0.42%\n",
            "iter 45582: loss 1.4946, time 10.17ms, mfu 0.47%\n",
            "iter 45583: loss 1.5101, time 10.90ms, mfu 0.44%\n",
            "iter 45584: loss 1.5306, time 9.51ms, mfu 0.51%\n",
            "iter 45585: loss 1.3617, time 9.45ms, mfu 0.51%\n",
            "iter 45586: loss 1.4795, time 10.13ms, mfu 0.48%\n",
            "iter 45587: loss 1.4579, time 9.78ms, mfu 0.49%\n",
            "iter 45588: loss 1.5873, time 9.30ms, mfu 0.52%\n",
            "iter 45589: loss 1.5303, time 9.55ms, mfu 0.50%\n",
            "iter 45590: loss 1.4002, time 9.54ms, mfu 0.51%\n",
            "iter 45591: loss 1.4985, time 9.32ms, mfu 0.52%\n",
            "iter 45592: loss 1.4615, time 9.54ms, mfu 0.51%\n",
            "iter 45593: loss 1.4745, time 9.29ms, mfu 0.52%\n",
            "iter 45594: loss 1.5071, time 9.45ms, mfu 0.51%\n",
            "iter 45595: loss 1.5211, time 9.24ms, mfu 0.52%\n",
            "iter 45596: loss 1.4343, time 9.39ms, mfu 0.51%\n",
            "iter 45597: loss 1.5102, time 9.56ms, mfu 0.50%\n",
            "iter 45598: loss 1.3669, time 10.59ms, mfu 0.46%\n",
            "iter 45599: loss 1.4627, time 12.08ms, mfu 0.40%\n",
            "iter 45600: loss 1.5384, time 9.47ms, mfu 0.51%\n",
            "iter 45601: loss 1.3936, time 9.30ms, mfu 0.52%\n",
            "iter 45602: loss 1.5700, time 9.46ms, mfu 0.51%\n",
            "iter 45603: loss 1.4519, time 9.21ms, mfu 0.52%\n",
            "iter 45604: loss 1.4580, time 9.43ms, mfu 0.51%\n",
            "iter 45605: loss 1.5227, time 9.81ms, mfu 0.49%\n",
            "iter 45606: loss 1.4648, time 9.36ms, mfu 0.52%\n",
            "iter 45607: loss 1.4980, time 9.59ms, mfu 0.50%\n",
            "iter 45608: loss 1.4478, time 9.17ms, mfu 0.53%\n",
            "iter 45609: loss 1.5283, time 9.19ms, mfu 0.52%\n",
            "iter 45610: loss 1.4857, time 9.28ms, mfu 0.52%\n",
            "iter 45611: loss 1.5288, time 9.25ms, mfu 0.52%\n",
            "iter 45612: loss 1.4981, time 9.42ms, mfu 0.51%\n",
            "iter 45613: loss 1.5601, time 9.77ms, mfu 0.49%\n",
            "iter 45614: loss 1.5147, time 9.65ms, mfu 0.50%\n",
            "iter 45615: loss 1.4829, time 9.97ms, mfu 0.48%\n",
            "iter 45616: loss 1.4785, time 9.70ms, mfu 0.50%\n",
            "iter 45617: loss 1.2907, time 9.51ms, mfu 0.51%\n",
            "iter 45618: loss 1.4680, time 9.88ms, mfu 0.49%\n",
            "iter 45619: loss 1.4838, time 10.33ms, mfu 0.47%\n",
            "iter 45620: loss 1.4153, time 13.28ms, mfu 0.36%\n",
            "iter 45621: loss 1.3296, time 11.60ms, mfu 0.42%\n",
            "iter 45622: loss 1.4575, time 9.56ms, mfu 0.50%\n",
            "iter 45623: loss 1.5929, time 9.28ms, mfu 0.52%\n",
            "iter 45624: loss 1.5113, time 9.30ms, mfu 0.52%\n",
            "iter 45625: loss 1.5305, time 9.89ms, mfu 0.49%\n",
            "iter 45626: loss 1.4276, time 12.12ms, mfu 0.40%\n",
            "iter 45627: loss 1.4181, time 11.96ms, mfu 0.40%\n",
            "iter 45628: loss 1.4596, time 10.71ms, mfu 0.45%\n",
            "iter 45629: loss 1.2938, time 12.39ms, mfu 0.39%\n",
            "iter 45630: loss 1.4854, time 9.55ms, mfu 0.51%\n",
            "iter 45631: loss 1.4806, time 9.30ms, mfu 0.52%\n",
            "iter 45632: loss 1.5328, time 9.22ms, mfu 0.52%\n",
            "iter 45633: loss 1.4568, time 9.27ms, mfu 0.52%\n",
            "iter 45634: loss 1.5417, time 9.26ms, mfu 0.52%\n",
            "iter 45635: loss 1.3960, time 9.46ms, mfu 0.51%\n",
            "iter 45636: loss 1.4807, time 9.39ms, mfu 0.51%\n",
            "iter 45637: loss 1.4475, time 9.51ms, mfu 0.51%\n",
            "iter 45638: loss 1.5916, time 9.46ms, mfu 0.51%\n",
            "iter 45639: loss 1.5286, time 9.22ms, mfu 0.52%\n",
            "iter 45640: loss 1.3532, time 14.07ms, mfu 0.34%\n",
            "iter 45641: loss 1.6204, time 9.65ms, mfu 0.50%\n",
            "iter 45642: loss 1.4555, time 9.22ms, mfu 0.52%\n",
            "iter 45643: loss 1.4397, time 9.16ms, mfu 0.53%\n",
            "iter 45644: loss 1.5091, time 9.23ms, mfu 0.52%\n",
            "iter 45645: loss 1.5115, time 9.35ms, mfu 0.52%\n",
            "iter 45646: loss 1.4871, time 9.35ms, mfu 0.52%\n",
            "iter 45647: loss 1.6176, time 9.47ms, mfu 0.51%\n",
            "iter 45648: loss 1.4978, time 9.37ms, mfu 0.51%\n",
            "iter 45649: loss 1.4105, time 9.29ms, mfu 0.52%\n",
            "iter 45650: loss 1.4303, time 9.36ms, mfu 0.52%\n",
            "iter 45651: loss 1.3906, time 9.36ms, mfu 0.52%\n",
            "iter 45652: loss 1.5379, time 9.31ms, mfu 0.52%\n",
            "iter 45653: loss 1.4794, time 9.16ms, mfu 0.53%\n",
            "iter 45654: loss 1.3742, time 9.13ms, mfu 0.53%\n",
            "iter 45655: loss 1.3993, time 9.26ms, mfu 0.52%\n",
            "iter 45656: loss 1.3798, time 9.26ms, mfu 0.52%\n",
            "iter 45657: loss 1.4565, time 9.30ms, mfu 0.52%\n",
            "iter 45658: loss 1.4165, time 9.41ms, mfu 0.51%\n",
            "iter 45659: loss 1.4340, time 9.35ms, mfu 0.52%\n",
            "iter 45660: loss 1.5861, time 10.32ms, mfu 0.47%\n",
            "iter 45661: loss 1.4280, time 17.76ms, mfu 0.27%\n",
            "iter 45662: loss 1.4609, time 14.66ms, mfu 0.33%\n",
            "iter 45663: loss 1.5226, time 9.95ms, mfu 0.48%\n",
            "iter 45664: loss 1.4825, time 9.50ms, mfu 0.51%\n",
            "iter 45665: loss 1.5138, time 9.82ms, mfu 0.49%\n",
            "iter 45666: loss 1.4876, time 9.27ms, mfu 0.52%\n",
            "iter 45667: loss 1.4363, time 9.37ms, mfu 0.51%\n",
            "iter 45668: loss 1.3492, time 9.16ms, mfu 0.53%\n",
            "iter 45669: loss 1.5012, time 9.21ms, mfu 0.52%\n",
            "iter 45670: loss 1.5821, time 9.22ms, mfu 0.52%\n",
            "iter 45671: loss 1.5315, time 9.43ms, mfu 0.51%\n",
            "iter 45672: loss 1.4090, time 9.49ms, mfu 0.51%\n",
            "iter 45673: loss 1.4867, time 9.33ms, mfu 0.52%\n",
            "iter 45674: loss 1.5763, time 9.28ms, mfu 0.52%\n",
            "iter 45675: loss 1.4413, time 9.26ms, mfu 0.52%\n",
            "iter 45676: loss 1.4432, time 9.24ms, mfu 0.52%\n",
            "iter 45677: loss 1.5078, time 9.21ms, mfu 0.52%\n",
            "iter 45678: loss 1.5274, time 9.41ms, mfu 0.51%\n",
            "iter 45679: loss 1.5108, time 9.26ms, mfu 0.52%\n",
            "iter 45680: loss 1.4689, time 9.32ms, mfu 0.52%\n",
            "iter 45681: loss 1.4247, time 9.46ms, mfu 0.51%\n",
            "iter 45682: loss 1.4641, time 13.11ms, mfu 0.37%\n",
            "iter 45683: loss 1.5218, time 13.52ms, mfu 0.36%\n",
            "iter 45684: loss 1.3928, time 10.84ms, mfu 0.44%\n",
            "iter 45685: loss 1.4501, time 9.76ms, mfu 0.49%\n",
            "iter 45686: loss 1.4919, time 9.39ms, mfu 0.51%\n",
            "iter 45687: loss 1.4747, time 9.48ms, mfu 0.51%\n",
            "iter 45688: loss 1.4427, time 10.63ms, mfu 0.45%\n",
            "iter 45689: loss 1.4221, time 9.61ms, mfu 0.50%\n",
            "iter 45690: loss 1.4423, time 9.38ms, mfu 0.51%\n",
            "iter 45691: loss 1.4971, time 9.49ms, mfu 0.51%\n",
            "iter 45692: loss 1.5821, time 9.41ms, mfu 0.51%\n",
            "iter 45693: loss 1.4795, time 9.46ms, mfu 0.51%\n",
            "iter 45694: loss 1.4351, time 9.28ms, mfu 0.52%\n",
            "iter 45695: loss 1.5917, time 9.21ms, mfu 0.52%\n",
            "iter 45696: loss 1.4746, time 10.48ms, mfu 0.46%\n",
            "iter 45697: loss 1.4949, time 9.45ms, mfu 0.51%\n",
            "iter 45698: loss 1.4225, time 9.49ms, mfu 0.51%\n",
            "iter 45699: loss 1.4424, time 9.95ms, mfu 0.48%\n",
            "iter 45700: loss 1.4406, time 10.76ms, mfu 0.45%\n",
            "iter 45701: loss 1.4815, time 9.78ms, mfu 0.49%\n",
            "iter 45702: loss 1.4124, time 9.45ms, mfu 0.51%\n",
            "iter 45703: loss 1.5038, time 14.20ms, mfu 0.34%\n",
            "iter 45704: loss 1.5747, time 9.81ms, mfu 0.49%\n",
            "iter 45705: loss 1.4592, time 9.79ms, mfu 0.49%\n",
            "iter 45706: loss 1.4187, time 9.75ms, mfu 0.49%\n",
            "iter 45707: loss 1.3888, time 9.58ms, mfu 0.50%\n",
            "iter 45708: loss 1.3861, time 9.35ms, mfu 0.52%\n",
            "iter 45709: loss 1.5271, time 9.15ms, mfu 0.53%\n",
            "iter 45710: loss 1.5057, time 9.24ms, mfu 0.52%\n",
            "iter 45711: loss 1.4178, time 9.70ms, mfu 0.50%\n",
            "iter 45712: loss 1.6415, time 9.82ms, mfu 0.49%\n",
            "iter 45713: loss 1.4773, time 9.32ms, mfu 0.52%\n",
            "iter 45714: loss 1.4459, time 9.44ms, mfu 0.51%\n",
            "iter 45715: loss 1.4928, time 9.43ms, mfu 0.51%\n",
            "iter 45716: loss 1.5642, time 9.32ms, mfu 0.52%\n",
            "iter 45717: loss 1.4553, time 9.38ms, mfu 0.51%\n",
            "iter 45718: loss 1.4809, time 10.03ms, mfu 0.48%\n",
            "iter 45719: loss 1.5823, time 9.33ms, mfu 0.52%\n",
            "iter 45720: loss 1.4605, time 9.55ms, mfu 0.50%\n",
            "iter 45721: loss 1.3429, time 9.83ms, mfu 0.49%\n",
            "iter 45722: loss 1.5048, time 12.04ms, mfu 0.40%\n",
            "iter 45723: loss 1.4680, time 10.19ms, mfu 0.47%\n",
            "iter 45724: loss 1.4923, time 13.14ms, mfu 0.37%\n",
            "iter 45725: loss 1.4932, time 10.89ms, mfu 0.44%\n",
            "iter 45726: loss 1.3843, time 13.21ms, mfu 0.36%\n",
            "iter 45727: loss 1.4138, time 12.02ms, mfu 0.40%\n",
            "iter 45728: loss 1.5332, time 9.51ms, mfu 0.51%\n",
            "iter 45729: loss 1.5711, time 9.35ms, mfu 0.52%\n",
            "iter 45730: loss 1.5076, time 9.41ms, mfu 0.51%\n",
            "iter 45731: loss 1.5160, time 9.53ms, mfu 0.51%\n",
            "iter 45732: loss 1.4363, time 9.52ms, mfu 0.51%\n",
            "iter 45733: loss 1.5492, time 9.31ms, mfu 0.52%\n",
            "iter 45734: loss 1.4942, time 9.29ms, mfu 0.52%\n",
            "iter 45735: loss 1.4633, time 9.37ms, mfu 0.51%\n",
            "iter 45736: loss 1.4815, time 12.05ms, mfu 0.40%\n",
            "iter 45737: loss 1.5181, time 9.50ms, mfu 0.51%\n",
            "iter 45738: loss 1.5728, time 9.81ms, mfu 0.49%\n",
            "iter 45739: loss 1.4527, time 9.66ms, mfu 0.50%\n",
            "iter 45740: loss 1.5302, time 10.53ms, mfu 0.46%\n",
            "iter 45741: loss 1.4789, time 9.69ms, mfu 0.50%\n",
            "iter 45742: loss 1.4095, time 9.85ms, mfu 0.49%\n",
            "iter 45743: loss 1.4961, time 9.47ms, mfu 0.51%\n",
            "iter 45744: loss 1.4873, time 9.35ms, mfu 0.52%\n",
            "iter 45745: loss 1.4502, time 14.12ms, mfu 0.34%\n",
            "iter 45746: loss 1.4556, time 9.92ms, mfu 0.49%\n",
            "iter 45747: loss 1.5093, time 9.81ms, mfu 0.49%\n",
            "iter 45748: loss 1.4911, time 9.47ms, mfu 0.51%\n",
            "iter 45749: loss 1.4049, time 9.47ms, mfu 0.51%\n",
            "iter 45750: loss 1.5405, time 9.60ms, mfu 0.50%\n",
            "iter 45751: loss 1.4191, time 9.29ms, mfu 0.52%\n",
            "iter 45752: loss 1.3682, time 9.48ms, mfu 0.51%\n",
            "iter 45753: loss 1.3713, time 9.43ms, mfu 0.51%\n",
            "iter 45754: loss 1.5152, time 9.25ms, mfu 0.52%\n",
            "iter 45755: loss 1.5212, time 9.29ms, mfu 0.52%\n",
            "iter 45756: loss 1.4311, time 9.66ms, mfu 0.50%\n",
            "iter 45757: loss 1.5692, time 9.79ms, mfu 0.49%\n",
            "iter 45758: loss 1.4072, time 9.30ms, mfu 0.52%\n",
            "iter 45759: loss 1.4471, time 9.87ms, mfu 0.49%\n",
            "iter 45760: loss 1.4683, time 9.50ms, mfu 0.51%\n",
            "iter 45761: loss 1.4383, time 9.83ms, mfu 0.49%\n",
            "iter 45762: loss 1.4990, time 9.67ms, mfu 0.50%\n",
            "iter 45763: loss 1.4436, time 9.62ms, mfu 0.50%\n",
            "iter 45764: loss 1.4938, time 17.79ms, mfu 0.27%\n",
            "iter 45765: loss 1.4133, time 14.53ms, mfu 0.33%\n",
            "iter 45766: loss 1.5430, time 12.00ms, mfu 0.40%\n",
            "iter 45767: loss 1.5768, time 9.91ms, mfu 0.49%\n",
            "iter 45768: loss 1.4237, time 9.92ms, mfu 0.49%\n",
            "iter 45769: loss 1.5713, time 9.54ms, mfu 0.51%\n",
            "iter 45770: loss 1.5352, time 9.59ms, mfu 0.50%\n",
            "iter 45771: loss 1.5239, time 9.58ms, mfu 0.50%\n",
            "iter 45772: loss 1.5592, time 9.68ms, mfu 0.50%\n",
            "iter 45773: loss 1.4467, time 9.89ms, mfu 0.49%\n",
            "iter 45774: loss 1.4191, time 9.46ms, mfu 0.51%\n",
            "iter 45775: loss 1.5711, time 9.47ms, mfu 0.51%\n",
            "iter 45776: loss 1.4529, time 9.34ms, mfu 0.52%\n",
            "iter 45777: loss 1.4987, time 9.48ms, mfu 0.51%\n",
            "iter 45778: loss 1.4264, time 9.42ms, mfu 0.51%\n",
            "iter 45779: loss 1.3485, time 9.37ms, mfu 0.51%\n",
            "iter 45780: loss 1.4963, time 9.32ms, mfu 0.52%\n",
            "iter 45781: loss 1.6013, time 9.63ms, mfu 0.50%\n",
            "iter 45782: loss 1.4309, time 9.38ms, mfu 0.51%\n",
            "iter 45783: loss 1.5159, time 9.51ms, mfu 0.51%\n",
            "iter 45784: loss 1.5631, time 9.47ms, mfu 0.51%\n",
            "iter 45785: loss 1.4348, time 11.00ms, mfu 0.44%\n",
            "iter 45786: loss 1.5048, time 14.93ms, mfu 0.32%\n",
            "iter 45787: loss 1.3927, time 15.61ms, mfu 0.31%\n",
            "iter 45788: loss 1.5230, time 10.71ms, mfu 0.45%\n",
            "iter 45789: loss 1.4879, time 10.78ms, mfu 0.45%\n",
            "iter 45790: loss 1.5299, time 9.43ms, mfu 0.51%\n",
            "iter 45791: loss 1.4234, time 9.60ms, mfu 0.50%\n",
            "iter 45792: loss 1.4847, time 9.61ms, mfu 0.50%\n",
            "iter 45793: loss 1.5430, time 9.80ms, mfu 0.49%\n",
            "iter 45794: loss 1.4698, time 9.33ms, mfu 0.52%\n",
            "iter 45795: loss 1.4598, time 9.73ms, mfu 0.50%\n",
            "iter 45796: loss 1.4389, time 9.34ms, mfu 0.52%\n",
            "iter 45797: loss 1.5946, time 9.88ms, mfu 0.49%\n",
            "iter 45798: loss 1.3052, time 9.55ms, mfu 0.50%\n",
            "iter 45799: loss 1.5139, time 9.39ms, mfu 0.51%\n",
            "iter 45800: loss 1.5159, time 9.68ms, mfu 0.50%\n",
            "iter 45801: loss 1.5417, time 9.41ms, mfu 0.51%\n",
            "iter 45802: loss 1.5595, time 9.32ms, mfu 0.52%\n",
            "iter 45803: loss 1.3012, time 9.30ms, mfu 0.52%\n",
            "iter 45804: loss 1.5916, time 9.33ms, mfu 0.52%\n",
            "iter 45805: loss 1.4484, time 9.31ms, mfu 0.52%\n",
            "iter 45806: loss 1.5168, time 9.52ms, mfu 0.51%\n",
            "iter 45807: loss 1.5221, time 13.20ms, mfu 0.37%\n",
            "iter 45808: loss 1.5643, time 9.55ms, mfu 0.50%\n",
            "iter 45809: loss 1.4363, time 9.24ms, mfu 0.52%\n",
            "iter 45810: loss 1.5527, time 9.37ms, mfu 0.51%\n",
            "iter 45811: loss 1.5384, time 9.53ms, mfu 0.51%\n",
            "iter 45812: loss 1.4237, time 9.28ms, mfu 0.52%\n",
            "iter 45813: loss 1.5133, time 10.13ms, mfu 0.48%\n",
            "iter 45814: loss 1.6071, time 9.77ms, mfu 0.49%\n",
            "iter 45815: loss 1.4859, time 9.51ms, mfu 0.51%\n",
            "iter 45816: loss 1.3830, time 9.54ms, mfu 0.51%\n",
            "iter 45817: loss 1.5443, time 9.64ms, mfu 0.50%\n",
            "iter 45818: loss 1.4610, time 9.57ms, mfu 0.50%\n",
            "iter 45819: loss 1.3760, time 9.32ms, mfu 0.52%\n",
            "iter 45820: loss 1.4562, time 9.32ms, mfu 0.52%\n",
            "iter 45821: loss 1.5245, time 11.07ms, mfu 0.44%\n",
            "iter 45822: loss 1.4537, time 9.46ms, mfu 0.51%\n",
            "iter 45823: loss 1.5872, time 9.30ms, mfu 0.52%\n",
            "iter 45824: loss 1.3863, time 9.42ms, mfu 0.51%\n",
            "iter 45825: loss 1.4324, time 10.41ms, mfu 0.46%\n",
            "iter 45826: loss 1.4432, time 12.57ms, mfu 0.38%\n",
            "iter 45827: loss 1.4209, time 10.07ms, mfu 0.48%\n",
            "iter 45828: loss 1.4527, time 13.91ms, mfu 0.35%\n",
            "iter 45829: loss 1.5043, time 10.76ms, mfu 0.45%\n",
            "iter 45830: loss 1.4634, time 9.58ms, mfu 0.50%\n",
            "iter 45831: loss 1.5532, time 9.46ms, mfu 0.51%\n",
            "iter 45832: loss 1.5314, time 9.36ms, mfu 0.52%\n",
            "iter 45833: loss 1.5036, time 9.28ms, mfu 0.52%\n",
            "iter 45834: loss 1.3974, time 9.38ms, mfu 0.51%\n",
            "iter 45835: loss 1.5325, time 9.36ms, mfu 0.52%\n",
            "iter 45836: loss 1.3640, time 9.53ms, mfu 0.51%\n",
            "iter 45837: loss 1.4843, time 9.17ms, mfu 0.53%\n",
            "iter 45838: loss 1.4671, time 9.25ms, mfu 0.52%\n",
            "iter 45839: loss 1.5190, time 9.53ms, mfu 0.51%\n",
            "iter 45840: loss 1.4191, time 10.16ms, mfu 0.47%\n",
            "iter 45841: loss 1.4162, time 9.48ms, mfu 0.51%\n",
            "iter 45842: loss 1.5195, time 9.57ms, mfu 0.50%\n",
            "iter 45843: loss 1.3809, time 9.29ms, mfu 0.52%\n",
            "iter 45844: loss 1.5182, time 9.25ms, mfu 0.52%\n",
            "iter 45845: loss 1.4791, time 9.47ms, mfu 0.51%\n",
            "iter 45846: loss 1.4527, time 9.76ms, mfu 0.49%\n",
            "iter 45847: loss 1.5159, time 9.36ms, mfu 0.52%\n",
            "iter 45848: loss 1.4577, time 9.55ms, mfu 0.51%\n",
            "iter 45849: loss 1.3968, time 9.43ms, mfu 0.51%\n",
            "iter 45850: loss 1.4747, time 13.77ms, mfu 0.35%\n",
            "iter 45851: loss 1.6354, time 9.55ms, mfu 0.51%\n",
            "iter 45852: loss 1.5647, time 9.56ms, mfu 0.50%\n",
            "iter 45853: loss 1.4194, time 9.35ms, mfu 0.52%\n",
            "iter 45854: loss 1.4715, time 9.33ms, mfu 0.52%\n",
            "iter 45855: loss 1.3868, time 9.41ms, mfu 0.51%\n",
            "iter 45856: loss 1.4453, time 9.41ms, mfu 0.51%\n",
            "iter 45857: loss 1.3701, time 9.57ms, mfu 0.50%\n",
            "iter 45858: loss 1.4081, time 9.56ms, mfu 0.50%\n",
            "iter 45859: loss 1.5325, time 9.53ms, mfu 0.51%\n",
            "iter 45860: loss 1.4289, time 9.38ms, mfu 0.51%\n",
            "iter 45861: loss 1.4250, time 9.62ms, mfu 0.50%\n",
            "iter 45862: loss 1.5457, time 9.42ms, mfu 0.51%\n",
            "iter 45863: loss 1.5782, time 9.44ms, mfu 0.51%\n",
            "iter 45864: loss 1.4236, time 10.03ms, mfu 0.48%\n",
            "iter 45865: loss 1.4038, time 9.62ms, mfu 0.50%\n",
            "iter 45866: loss 1.5213, time 12.10ms, mfu 0.40%\n",
            "iter 45867: loss 1.4956, time 16.00ms, mfu 0.30%\n",
            "iter 45868: loss 1.4745, time 11.21ms, mfu 0.43%\n",
            "iter 45869: loss 1.3979, time 9.50ms, mfu 0.51%\n",
            "iter 45870: loss 1.4602, time 9.79ms, mfu 0.49%\n",
            "iter 45871: loss 1.5102, time 16.12ms, mfu 0.30%\n",
            "iter 45872: loss 1.4736, time 10.51ms, mfu 0.46%\n",
            "iter 45873: loss 1.4574, time 9.51ms, mfu 0.51%\n",
            "iter 45874: loss 1.5130, time 9.55ms, mfu 0.50%\n",
            "iter 45875: loss 1.5589, time 10.75ms, mfu 0.45%\n",
            "iter 45876: loss 1.4068, time 9.78ms, mfu 0.49%\n",
            "iter 45877: loss 1.4305, time 9.60ms, mfu 0.50%\n",
            "iter 45878: loss 1.4055, time 9.78ms, mfu 0.49%\n",
            "iter 45879: loss 1.5182, time 9.94ms, mfu 0.48%\n",
            "iter 45880: loss 1.4184, time 10.05ms, mfu 0.48%\n",
            "iter 45881: loss 1.4462, time 9.81ms, mfu 0.49%\n",
            "iter 45882: loss 1.3865, time 9.55ms, mfu 0.50%\n",
            "iter 45883: loss 1.4560, time 9.52ms, mfu 0.51%\n",
            "iter 45884: loss 1.3867, time 10.94ms, mfu 0.44%\n",
            "iter 45885: loss 1.4896, time 10.06ms, mfu 0.48%\n",
            "iter 45886: loss 1.4084, time 9.58ms, mfu 0.50%\n",
            "iter 45887: loss 1.4107, time 10.70ms, mfu 0.45%\n",
            "iter 45888: loss 1.4857, time 9.65ms, mfu 0.50%\n",
            "iter 45889: loss 1.4772, time 9.86ms, mfu 0.49%\n",
            "iter 45890: loss 1.3961, time 9.76ms, mfu 0.49%\n",
            "iter 45891: loss 1.5264, time 9.62ms, mfu 0.50%\n",
            "iter 45892: loss 1.4620, time 14.17ms, mfu 0.34%\n",
            "iter 45893: loss 1.4473, time 9.83ms, mfu 0.49%\n",
            "iter 45894: loss 1.4376, time 10.08ms, mfu 0.48%\n",
            "iter 45895: loss 1.5151, time 9.58ms, mfu 0.50%\n",
            "iter 45896: loss 1.4275, time 9.52ms, mfu 0.51%\n",
            "iter 45897: loss 1.5706, time 9.82ms, mfu 0.49%\n",
            "iter 45898: loss 1.6069, time 9.55ms, mfu 0.50%\n",
            "iter 45899: loss 1.4336, time 9.45ms, mfu 0.51%\n",
            "iter 45900: loss 1.5004, time 9.44ms, mfu 0.51%\n",
            "iter 45901: loss 1.4907, time 9.56ms, mfu 0.50%\n",
            "iter 45902: loss 1.4847, time 9.44ms, mfu 0.51%\n",
            "iter 45903: loss 1.3780, time 9.32ms, mfu 0.52%\n",
            "iter 45904: loss 1.4419, time 9.58ms, mfu 0.50%\n",
            "iter 45905: loss 1.5112, time 9.67ms, mfu 0.50%\n",
            "iter 45906: loss 1.3742, time 9.68ms, mfu 0.50%\n",
            "iter 45907: loss 1.3871, time 9.52ms, mfu 0.51%\n",
            "iter 45908: loss 1.4541, time 9.36ms, mfu 0.52%\n",
            "iter 45909: loss 1.4656, time 9.36ms, mfu 0.52%\n",
            "iter 45910: loss 1.3618, time 12.06ms, mfu 0.40%\n",
            "iter 45911: loss 1.4256, time 9.48ms, mfu 0.51%\n",
            "iter 45912: loss 1.4842, time 9.52ms, mfu 0.51%\n",
            "iter 45913: loss 1.4362, time 12.83ms, mfu 0.38%\n",
            "iter 45914: loss 1.5099, time 10.80ms, mfu 0.45%\n",
            "iter 45915: loss 1.5890, time 9.95ms, mfu 0.48%\n",
            "iter 45916: loss 1.4092, time 9.23ms, mfu 0.52%\n",
            "iter 45917: loss 1.4388, time 9.87ms, mfu 0.49%\n",
            "iter 45918: loss 1.5286, time 9.30ms, mfu 0.52%\n",
            "iter 45919: loss 1.5252, time 9.31ms, mfu 0.52%\n",
            "iter 45920: loss 1.5554, time 10.91ms, mfu 0.44%\n",
            "iter 45921: loss 1.4422, time 9.27ms, mfu 0.52%\n",
            "iter 45922: loss 1.4817, time 9.32ms, mfu 0.52%\n",
            "iter 45923: loss 1.4726, time 9.32ms, mfu 0.52%\n",
            "iter 45924: loss 1.3980, time 9.29ms, mfu 0.52%\n",
            "iter 45925: loss 1.5224, time 11.33ms, mfu 0.43%\n",
            "iter 45926: loss 1.4500, time 11.85ms, mfu 0.41%\n",
            "iter 45927: loss 1.4924, time 9.96ms, mfu 0.48%\n",
            "iter 45928: loss 1.5142, time 9.28ms, mfu 0.52%\n",
            "iter 45929: loss 1.5234, time 9.14ms, mfu 0.53%\n",
            "iter 45930: loss 1.4358, time 9.27ms, mfu 0.52%\n",
            "iter 45931: loss 1.4395, time 9.21ms, mfu 0.52%\n",
            "iter 45932: loss 1.5401, time 9.35ms, mfu 0.52%\n",
            "iter 45933: loss 1.4609, time 9.23ms, mfu 0.52%\n",
            "iter 45934: loss 1.5152, time 10.15ms, mfu 0.48%\n",
            "iter 45935: loss 1.4290, time 12.95ms, mfu 0.37%\n",
            "iter 45936: loss 1.5108, time 9.57ms, mfu 0.50%\n",
            "iter 45937: loss 1.3870, time 9.99ms, mfu 0.48%\n",
            "iter 45938: loss 1.4871, time 9.20ms, mfu 0.52%\n",
            "iter 45939: loss 1.5578, time 9.60ms, mfu 0.50%\n",
            "iter 45940: loss 1.4028, time 9.85ms, mfu 0.49%\n",
            "iter 45941: loss 1.5210, time 9.34ms, mfu 0.52%\n",
            "iter 45942: loss 1.5102, time 9.28ms, mfu 0.52%\n",
            "iter 45943: loss 1.5495, time 9.20ms, mfu 0.52%\n",
            "iter 45944: loss 1.4196, time 9.23ms, mfu 0.52%\n",
            "iter 45945: loss 1.4848, time 9.16ms, mfu 0.53%\n",
            "iter 45946: loss 1.3701, time 9.27ms, mfu 0.52%\n",
            "iter 45947: loss 1.4081, time 9.42ms, mfu 0.51%\n",
            "iter 45948: loss 1.4688, time 12.02ms, mfu 0.40%\n",
            "iter 45949: loss 1.4739, time 9.54ms, mfu 0.51%\n",
            "iter 45950: loss 1.5663, time 9.31ms, mfu 0.52%\n",
            "iter 45951: loss 1.5357, time 9.42ms, mfu 0.51%\n",
            "iter 45952: loss 1.4233, time 9.25ms, mfu 0.52%\n",
            "iter 45953: loss 1.4166, time 9.17ms, mfu 0.53%\n",
            "iter 45954: loss 1.3564, time 9.14ms, mfu 0.53%\n",
            "iter 45955: loss 1.6140, time 11.15ms, mfu 0.43%\n",
            "iter 45956: loss 1.5423, time 12.46ms, mfu 0.39%\n",
            "iter 45957: loss 1.4717, time 9.76ms, mfu 0.49%\n",
            "iter 45958: loss 1.4776, time 9.41ms, mfu 0.51%\n",
            "iter 45959: loss 1.4703, time 9.23ms, mfu 0.52%\n",
            "iter 45960: loss 1.5253, time 9.38ms, mfu 0.51%\n",
            "iter 45961: loss 1.4762, time 9.64ms, mfu 0.50%\n",
            "iter 45962: loss 1.4864, time 9.26ms, mfu 0.52%\n",
            "iter 45963: loss 1.4433, time 9.31ms, mfu 0.52%\n",
            "iter 45964: loss 1.4248, time 9.59ms, mfu 0.50%\n",
            "iter 45965: loss 1.5059, time 9.31ms, mfu 0.52%\n",
            "iter 45966: loss 1.5145, time 9.52ms, mfu 0.51%\n",
            "iter 45967: loss 1.4359, time 9.59ms, mfu 0.50%\n",
            "iter 45968: loss 1.4386, time 9.63ms, mfu 0.50%\n",
            "iter 45969: loss 1.3819, time 16.12ms, mfu 0.30%\n",
            "iter 45970: loss 1.5061, time 12.91ms, mfu 0.37%\n",
            "iter 45971: loss 1.5686, time 9.44ms, mfu 0.51%\n",
            "iter 45972: loss 1.3774, time 9.36ms, mfu 0.51%\n",
            "iter 45973: loss 1.5437, time 9.19ms, mfu 0.52%\n",
            "iter 45974: loss 1.5423, time 9.33ms, mfu 0.52%\n",
            "iter 45975: loss 1.3718, time 10.65ms, mfu 0.45%\n",
            "iter 45976: loss 1.5295, time 11.96ms, mfu 0.40%\n",
            "iter 45977: loss 1.5089, time 9.54ms, mfu 0.51%\n",
            "iter 45978: loss 1.4807, time 9.22ms, mfu 0.52%\n",
            "iter 45979: loss 1.5507, time 9.18ms, mfu 0.53%\n",
            "iter 45980: loss 1.5561, time 9.20ms, mfu 0.52%\n",
            "iter 45981: loss 1.4824, time 9.53ms, mfu 0.51%\n",
            "iter 45982: loss 1.4842, time 9.25ms, mfu 0.52%\n",
            "iter 45983: loss 1.5118, time 9.28ms, mfu 0.52%\n",
            "iter 45984: loss 1.4555, time 9.20ms, mfu 0.52%\n",
            "iter 45985: loss 1.5829, time 9.26ms, mfu 0.52%\n",
            "iter 45986: loss 1.4686, time 11.00ms, mfu 0.44%\n",
            "iter 45987: loss 1.5319, time 9.48ms, mfu 0.51%\n",
            "iter 45988: loss 1.5452, time 9.55ms, mfu 0.50%\n",
            "iter 45989: loss 1.4474, time 10.21ms, mfu 0.47%\n",
            "iter 45990: loss 1.3990, time 9.35ms, mfu 0.52%\n",
            "iter 45991: loss 1.6069, time 9.72ms, mfu 0.50%\n",
            "iter 45992: loss 1.5330, time 9.38ms, mfu 0.51%\n",
            "iter 45993: loss 1.5296, time 9.61ms, mfu 0.50%\n",
            "iter 45994: loss 1.4360, time 9.19ms, mfu 0.52%\n",
            "iter 45995: loss 1.5045, time 9.52ms, mfu 0.51%\n",
            "iter 45996: loss 1.5028, time 10.62ms, mfu 0.45%\n",
            "iter 45997: loss 1.4040, time 15.58ms, mfu 0.31%\n",
            "iter 45998: loss 1.4458, time 9.91ms, mfu 0.49%\n",
            "iter 45999: loss 1.5388, time 9.59ms, mfu 0.50%\n",
            "step 46000: train loss 1.4431, val loss 1.5469\n",
            "iter 46000: loss 1.4151, time 17.84ms, mfu 0.27%\n",
            "iter 46001: loss 1.4580, time 9.39ms, mfu 0.51%\n",
            "iter 46002: loss 1.4807, time 9.64ms, mfu 0.50%\n",
            "iter 46003: loss 1.4936, time 9.25ms, mfu 0.52%\n",
            "iter 46004: loss 1.5429, time 9.41ms, mfu 0.51%\n",
            "iter 46005: loss 1.5881, time 9.31ms, mfu 0.52%\n",
            "iter 46006: loss 1.5022, time 9.36ms, mfu 0.51%\n",
            "iter 46007: loss 1.4578, time 9.67ms, mfu 0.50%\n",
            "iter 46008: loss 1.5340, time 9.28ms, mfu 0.52%\n",
            "iter 46009: loss 1.4529, time 9.19ms, mfu 0.52%\n",
            "iter 46010: loss 1.5533, time 9.28ms, mfu 0.52%\n",
            "iter 46011: loss 1.4507, time 9.20ms, mfu 0.52%\n",
            "iter 46012: loss 1.5280, time 9.45ms, mfu 0.51%\n",
            "iter 46013: loss 1.4961, time 9.24ms, mfu 0.52%\n",
            "iter 46014: loss 1.5209, time 9.44ms, mfu 0.51%\n",
            "iter 46015: loss 1.3294, time 9.30ms, mfu 0.52%\n",
            "iter 46016: loss 1.4150, time 11.64ms, mfu 0.41%\n",
            "iter 46017: loss 1.4292, time 11.32ms, mfu 0.43%\n",
            "iter 46018: loss 1.4756, time 9.68ms, mfu 0.50%\n",
            "iter 46019: loss 1.4271, time 9.35ms, mfu 0.52%\n",
            "iter 46020: loss 1.4795, time 9.27ms, mfu 0.52%\n",
            "iter 46021: loss 1.5378, time 11.04ms, mfu 0.44%\n",
            "iter 46022: loss 1.5348, time 9.34ms, mfu 0.52%\n",
            "iter 46023: loss 1.5061, time 9.45ms, mfu 0.51%\n",
            "iter 46024: loss 1.5231, time 9.43ms, mfu 0.51%\n",
            "iter 46025: loss 1.5078, time 9.95ms, mfu 0.48%\n",
            "iter 46026: loss 1.5098, time 11.97ms, mfu 0.40%\n",
            "iter 46027: loss 1.5520, time 12.01ms, mfu 0.40%\n",
            "iter 46028: loss 1.4204, time 9.38ms, mfu 0.51%\n",
            "iter 46029: loss 1.5284, time 9.34ms, mfu 0.52%\n",
            "iter 46030: loss 1.5009, time 9.30ms, mfu 0.52%\n",
            "iter 46031: loss 1.3821, time 9.46ms, mfu 0.51%\n",
            "iter 46032: loss 1.4587, time 9.30ms, mfu 0.52%\n",
            "iter 46033: loss 1.4616, time 9.55ms, mfu 0.50%\n",
            "iter 46034: loss 1.5841, time 9.33ms, mfu 0.52%\n",
            "iter 46035: loss 1.5909, time 9.36ms, mfu 0.52%\n",
            "iter 46036: loss 1.4535, time 9.28ms, mfu 0.52%\n",
            "iter 46037: loss 1.5197, time 12.61ms, mfu 0.38%\n",
            "iter 46038: loss 1.4985, time 11.21ms, mfu 0.43%\n",
            "iter 46039: loss 1.4811, time 9.60ms, mfu 0.50%\n",
            "iter 46040: loss 1.5691, time 10.48ms, mfu 0.46%\n",
            "iter 46041: loss 1.4509, time 9.75ms, mfu 0.49%\n",
            "iter 46042: loss 1.5713, time 9.38ms, mfu 0.51%\n",
            "iter 46043: loss 1.5046, time 9.37ms, mfu 0.51%\n",
            "iter 46044: loss 1.5377, time 9.28ms, mfu 0.52%\n",
            "iter 46045: loss 1.5242, time 9.30ms, mfu 0.52%\n",
            "iter 46046: loss 1.5174, time 9.45ms, mfu 0.51%\n",
            "iter 46047: loss 1.5248, time 9.41ms, mfu 0.51%\n",
            "iter 46048: loss 1.4082, time 9.53ms, mfu 0.51%\n",
            "iter 46049: loss 1.4683, time 12.56ms, mfu 0.38%\n",
            "iter 46050: loss 1.5509, time 11.00ms, mfu 0.44%\n",
            "iter 46051: loss 1.4621, time 9.67ms, mfu 0.50%\n",
            "iter 46052: loss 1.4309, time 9.45ms, mfu 0.51%\n",
            "iter 46053: loss 1.3711, time 9.76ms, mfu 0.49%\n",
            "iter 46054: loss 1.4913, time 9.91ms, mfu 0.49%\n",
            "iter 46055: loss 1.5582, time 9.47ms, mfu 0.51%\n",
            "iter 46056: loss 1.4793, time 9.29ms, mfu 0.52%\n",
            "iter 46057: loss 1.4233, time 10.05ms, mfu 0.48%\n",
            "iter 46058: loss 1.4191, time 12.73ms, mfu 0.38%\n",
            "iter 46059: loss 1.5484, time 10.38ms, mfu 0.46%\n",
            "iter 46060: loss 1.4819, time 9.47ms, mfu 0.51%\n",
            "iter 46061: loss 1.4570, time 9.69ms, mfu 0.50%\n",
            "iter 46062: loss 1.5906, time 9.34ms, mfu 0.52%\n",
            "iter 46063: loss 1.3226, time 9.67ms, mfu 0.50%\n",
            "iter 46064: loss 1.3733, time 9.74ms, mfu 0.49%\n",
            "iter 46065: loss 1.4513, time 9.31ms, mfu 0.52%\n",
            "iter 46066: loss 1.6139, time 9.85ms, mfu 0.49%\n",
            "iter 46067: loss 1.3527, time 9.64ms, mfu 0.50%\n",
            "iter 46068: loss 1.4144, time 9.53ms, mfu 0.51%\n",
            "iter 46069: loss 1.5262, time 9.42ms, mfu 0.51%\n",
            "iter 46070: loss 1.5155, time 9.33ms, mfu 0.52%\n",
            "iter 46071: loss 1.4515, time 14.76ms, mfu 0.33%\n",
            "iter 46072: loss 1.4846, time 14.53ms, mfu 0.33%\n",
            "iter 46073: loss 1.4470, time 9.57ms, mfu 0.50%\n",
            "iter 46074: loss 1.4554, time 9.46ms, mfu 0.51%\n",
            "iter 46075: loss 1.4344, time 9.59ms, mfu 0.50%\n",
            "iter 46076: loss 1.4100, time 9.34ms, mfu 0.52%\n",
            "iter 46077: loss 1.5700, time 9.48ms, mfu 0.51%\n",
            "iter 46078: loss 1.4618, time 10.35ms, mfu 0.47%\n",
            "iter 46079: loss 1.5101, time 13.15ms, mfu 0.37%\n",
            "iter 46080: loss 1.5701, time 9.65ms, mfu 0.50%\n",
            "iter 46081: loss 1.3924, time 9.33ms, mfu 0.52%\n",
            "iter 46082: loss 1.5976, time 9.38ms, mfu 0.51%\n",
            "iter 46083: loss 1.4708, time 9.41ms, mfu 0.51%\n",
            "iter 46084: loss 1.4421, time 9.36ms, mfu 0.52%\n",
            "iter 46085: loss 1.4961, time 9.33ms, mfu 0.52%\n",
            "iter 46086: loss 1.4628, time 12.26ms, mfu 0.39%\n",
            "iter 46087: loss 1.4499, time 9.41ms, mfu 0.51%\n",
            "iter 46088: loss 1.3861, time 10.50ms, mfu 0.46%\n",
            "iter 46089: loss 1.4717, time 9.49ms, mfu 0.51%\n",
            "iter 46090: loss 1.4501, time 10.20ms, mfu 0.47%\n",
            "iter 46091: loss 1.6215, time 9.94ms, mfu 0.48%\n",
            "iter 46092: loss 1.4811, time 9.55ms, mfu 0.50%\n",
            "iter 46093: loss 1.5200, time 9.52ms, mfu 0.51%\n",
            "iter 46094: loss 1.6483, time 9.38ms, mfu 0.51%\n",
            "iter 46095: loss 1.4629, time 9.38ms, mfu 0.51%\n",
            "iter 46096: loss 1.5497, time 9.37ms, mfu 0.51%\n",
            "iter 46097: loss 1.4300, time 9.49ms, mfu 0.51%\n",
            "iter 46098: loss 1.4835, time 9.64ms, mfu 0.50%\n",
            "iter 46099: loss 1.5365, time 13.22ms, mfu 0.36%\n",
            "iter 46100: loss 1.4391, time 11.10ms, mfu 0.43%\n",
            "iter 46101: loss 1.4052, time 9.55ms, mfu 0.50%\n",
            "iter 46102: loss 1.4671, time 9.57ms, mfu 0.50%\n",
            "iter 46103: loss 1.4375, time 9.32ms, mfu 0.52%\n",
            "iter 46104: loss 1.4661, time 9.39ms, mfu 0.51%\n",
            "iter 46105: loss 1.4274, time 9.66ms, mfu 0.50%\n",
            "iter 46106: loss 1.5246, time 9.77ms, mfu 0.49%\n",
            "iter 46107: loss 1.3566, time 9.52ms, mfu 0.51%\n",
            "iter 46108: loss 1.5199, time 9.77ms, mfu 0.49%\n",
            "iter 46109: loss 1.5561, time 9.29ms, mfu 0.52%\n",
            "iter 46110: loss 1.4574, time 9.61ms, mfu 0.50%\n",
            "iter 46111: loss 1.5652, time 9.36ms, mfu 0.52%\n",
            "iter 46112: loss 1.4668, time 9.34ms, mfu 0.52%\n",
            "iter 46113: loss 1.4625, time 9.63ms, mfu 0.50%\n",
            "iter 46114: loss 1.4530, time 9.40ms, mfu 0.51%\n",
            "iter 46115: loss 1.4816, time 9.24ms, mfu 0.52%\n",
            "iter 46116: loss 1.6060, time 9.26ms, mfu 0.52%\n",
            "iter 46117: loss 1.3517, time 9.64ms, mfu 0.50%\n",
            "iter 46118: loss 1.3616, time 9.51ms, mfu 0.51%\n",
            "iter 46119: loss 1.4801, time 9.18ms, mfu 0.53%\n",
            "iter 46120: loss 1.4581, time 9.26ms, mfu 0.52%\n",
            "iter 46121: loss 1.5096, time 15.81ms, mfu 0.31%\n",
            "iter 46122: loss 1.4672, time 9.41ms, mfu 0.51%\n",
            "iter 46123: loss 1.4989, time 9.29ms, mfu 0.52%\n",
            "iter 46124: loss 1.4035, time 9.24ms, mfu 0.52%\n",
            "iter 46125: loss 1.6180, time 9.34ms, mfu 0.52%\n",
            "iter 46126: loss 1.5795, time 11.46ms, mfu 0.42%\n",
            "iter 46127: loss 1.4822, time 11.71ms, mfu 0.41%\n",
            "iter 46128: loss 1.4840, time 12.35ms, mfu 0.39%\n",
            "iter 46129: loss 1.4032, time 9.28ms, mfu 0.52%\n",
            "iter 46130: loss 1.5045, time 9.24ms, mfu 0.52%\n",
            "iter 46131: loss 1.3872, time 9.28ms, mfu 0.52%\n",
            "iter 46132: loss 1.4519, time 9.19ms, mfu 0.52%\n",
            "iter 46133: loss 1.4111, time 9.22ms, mfu 0.52%\n",
            "iter 46134: loss 1.5120, time 9.59ms, mfu 0.50%\n",
            "iter 46135: loss 1.5415, time 9.23ms, mfu 0.52%\n",
            "iter 46136: loss 1.5338, time 9.57ms, mfu 0.50%\n",
            "iter 46137: loss 1.4915, time 9.28ms, mfu 0.52%\n",
            "iter 46138: loss 1.5179, time 9.22ms, mfu 0.52%\n",
            "iter 46139: loss 1.4490, time 9.17ms, mfu 0.53%\n",
            "iter 46140: loss 1.5024, time 9.41ms, mfu 0.51%\n",
            "iter 46141: loss 1.5904, time 9.97ms, mfu 0.48%\n",
            "iter 46142: loss 1.4385, time 12.82ms, mfu 0.38%\n",
            "iter 46143: loss 1.5522, time 11.49ms, mfu 0.42%\n",
            "iter 46144: loss 1.3935, time 9.15ms, mfu 0.53%\n",
            "iter 46145: loss 1.4418, time 9.20ms, mfu 0.52%\n",
            "iter 46146: loss 1.4909, time 9.23ms, mfu 0.52%\n",
            "iter 46147: loss 1.4131, time 9.50ms, mfu 0.51%\n",
            "iter 46148: loss 1.4357, time 9.56ms, mfu 0.50%\n",
            "iter 46149: loss 1.5087, time 9.43ms, mfu 0.51%\n",
            "iter 46150: loss 1.4673, time 9.20ms, mfu 0.52%\n",
            "iter 46151: loss 1.5041, time 9.38ms, mfu 0.51%\n",
            "iter 46152: loss 1.5159, time 9.30ms, mfu 0.52%\n",
            "iter 46153: loss 1.3373, time 9.21ms, mfu 0.52%\n",
            "iter 46154: loss 1.5565, time 9.11ms, mfu 0.53%\n",
            "iter 46155: loss 1.4681, time 9.15ms, mfu 0.53%\n",
            "iter 46156: loss 1.5121, time 9.15ms, mfu 0.53%\n",
            "iter 46157: loss 1.4178, time 9.25ms, mfu 0.52%\n",
            "iter 46158: loss 1.5620, time 9.47ms, mfu 0.51%\n",
            "iter 46159: loss 1.4921, time 9.30ms, mfu 0.52%\n",
            "iter 46160: loss 1.5782, time 9.43ms, mfu 0.51%\n",
            "iter 46161: loss 1.5477, time 9.32ms, mfu 0.52%\n",
            "iter 46162: loss 1.5493, time 9.62ms, mfu 0.50%\n",
            "iter 46163: loss 1.5131, time 9.24ms, mfu 0.52%\n",
            "iter 46164: loss 1.5279, time 14.43ms, mfu 0.33%\n",
            "iter 46165: loss 1.4784, time 9.80ms, mfu 0.49%\n",
            "iter 46166: loss 1.4541, time 9.42ms, mfu 0.51%\n",
            "iter 46167: loss 1.4527, time 9.19ms, mfu 0.52%\n",
            "iter 46168: loss 1.4282, time 12.74ms, mfu 0.38%\n",
            "iter 46169: loss 1.5353, time 9.41ms, mfu 0.51%\n",
            "iter 46170: loss 1.4385, time 9.24ms, mfu 0.52%\n",
            "iter 46171: loss 1.4804, time 9.44ms, mfu 0.51%\n",
            "iter 46172: loss 1.3802, time 9.32ms, mfu 0.52%\n",
            "iter 46173: loss 1.4261, time 9.23ms, mfu 0.52%\n",
            "iter 46174: loss 1.4355, time 11.25ms, mfu 0.43%\n",
            "iter 46175: loss 1.3324, time 14.46ms, mfu 0.33%\n",
            "iter 46176: loss 1.5409, time 11.87ms, mfu 0.41%\n",
            "iter 46177: loss 1.4730, time 10.80ms, mfu 0.45%\n",
            "iter 46178: loss 1.4367, time 9.44ms, mfu 0.51%\n",
            "iter 46179: loss 1.5723, time 9.41ms, mfu 0.51%\n",
            "iter 46180: loss 1.4671, time 9.12ms, mfu 0.53%\n",
            "iter 46181: loss 1.4614, time 9.19ms, mfu 0.52%\n",
            "iter 46182: loss 1.3447, time 9.89ms, mfu 0.49%\n",
            "iter 46183: loss 1.3594, time 9.17ms, mfu 0.53%\n",
            "iter 46184: loss 1.4389, time 10.27ms, mfu 0.47%\n",
            "iter 46185: loss 1.4987, time 19.10ms, mfu 0.25%\n",
            "iter 46186: loss 1.5713, time 12.15ms, mfu 0.40%\n",
            "iter 46187: loss 1.3705, time 9.62ms, mfu 0.50%\n",
            "iter 46188: loss 1.4475, time 9.78ms, mfu 0.49%\n",
            "iter 46189: loss 1.4626, time 10.28ms, mfu 0.47%\n",
            "iter 46190: loss 1.4719, time 9.39ms, mfu 0.51%\n",
            "iter 46191: loss 1.5070, time 9.29ms, mfu 0.52%\n",
            "iter 46192: loss 1.5221, time 9.34ms, mfu 0.52%\n",
            "iter 46193: loss 1.5456, time 9.64ms, mfu 0.50%\n",
            "iter 46194: loss 1.4228, time 9.27ms, mfu 0.52%\n",
            "iter 46195: loss 1.3738, time 9.14ms, mfu 0.53%\n",
            "iter 46196: loss 1.3428, time 9.22ms, mfu 0.52%\n",
            "iter 46197: loss 1.3837, time 9.45ms, mfu 0.51%\n",
            "iter 46198: loss 1.4982, time 9.57ms, mfu 0.50%\n",
            "iter 46199: loss 1.3401, time 9.57ms, mfu 0.50%\n",
            "iter 46200: loss 1.5057, time 10.01ms, mfu 0.48%\n",
            "iter 46201: loss 1.4550, time 9.30ms, mfu 0.52%\n",
            "iter 46202: loss 1.4071, time 9.22ms, mfu 0.52%\n",
            "iter 46203: loss 1.3687, time 9.30ms, mfu 0.52%\n",
            "iter 46204: loss 1.5625, time 11.19ms, mfu 0.43%\n",
            "iter 46205: loss 1.4616, time 11.91ms, mfu 0.41%\n",
            "iter 46206: loss 1.4587, time 9.32ms, mfu 0.52%\n",
            "iter 46207: loss 1.4556, time 12.29ms, mfu 0.39%\n",
            "iter 46208: loss 1.3910, time 9.59ms, mfu 0.50%\n",
            "iter 46209: loss 1.4164, time 9.16ms, mfu 0.53%\n",
            "iter 46210: loss 1.3934, time 9.39ms, mfu 0.51%\n",
            "iter 46211: loss 1.4559, time 9.40ms, mfu 0.51%\n",
            "iter 46212: loss 1.4479, time 9.20ms, mfu 0.52%\n",
            "iter 46213: loss 1.5532, time 9.39ms, mfu 0.51%\n",
            "iter 46214: loss 1.4850, time 9.61ms, mfu 0.50%\n",
            "iter 46215: loss 1.5054, time 9.39ms, mfu 0.51%\n",
            "iter 46216: loss 1.4876, time 9.22ms, mfu 0.52%\n",
            "iter 46217: loss 1.4596, time 9.60ms, mfu 0.50%\n",
            "iter 46218: loss 1.5061, time 9.58ms, mfu 0.50%\n",
            "iter 46219: loss 1.5005, time 9.40ms, mfu 0.51%\n",
            "iter 46220: loss 1.3773, time 9.25ms, mfu 0.52%\n",
            "iter 46221: loss 1.4237, time 9.29ms, mfu 0.52%\n",
            "iter 46222: loss 1.5132, time 11.07ms, mfu 0.44%\n",
            "iter 46223: loss 1.4692, time 9.22ms, mfu 0.52%\n",
            "iter 46224: loss 1.4850, time 9.31ms, mfu 0.52%\n",
            "iter 46225: loss 1.5720, time 9.22ms, mfu 0.52%\n",
            "iter 46226: loss 1.5106, time 16.39ms, mfu 0.29%\n",
            "iter 46227: loss 1.5644, time 12.51ms, mfu 0.39%\n",
            "iter 46228: loss 1.4352, time 9.44ms, mfu 0.51%\n",
            "iter 46229: loss 1.3873, time 9.16ms, mfu 0.53%\n",
            "iter 46230: loss 1.4850, time 9.25ms, mfu 0.52%\n",
            "iter 46231: loss 1.4716, time 9.22ms, mfu 0.52%\n",
            "iter 46232: loss 1.4647, time 9.18ms, mfu 0.53%\n",
            "iter 46233: loss 1.5503, time 9.28ms, mfu 0.52%\n",
            "iter 46234: loss 1.5409, time 9.10ms, mfu 0.53%\n",
            "iter 46235: loss 1.5034, time 9.28ms, mfu 0.52%\n",
            "iter 46236: loss 1.4638, time 9.39ms, mfu 0.51%\n",
            "iter 46237: loss 1.5881, time 9.20ms, mfu 0.52%\n",
            "iter 46238: loss 1.4102, time 9.49ms, mfu 0.51%\n",
            "iter 46239: loss 1.4609, time 9.36ms, mfu 0.52%\n",
            "iter 46240: loss 1.5472, time 9.17ms, mfu 0.53%\n",
            "iter 46241: loss 1.4618, time 10.05ms, mfu 0.48%\n",
            "iter 46242: loss 1.4486, time 9.49ms, mfu 0.51%\n",
            "iter 46243: loss 1.4449, time 9.26ms, mfu 0.52%\n",
            "iter 46244: loss 1.4830, time 9.45ms, mfu 0.51%\n",
            "iter 46245: loss 1.3450, time 9.29ms, mfu 0.52%\n",
            "iter 46246: loss 1.6595, time 9.21ms, mfu 0.52%\n",
            "iter 46247: loss 1.5529, time 9.22ms, mfu 0.52%\n",
            "iter 46248: loss 1.3797, time 14.82ms, mfu 0.33%\n",
            "iter 46249: loss 1.4358, time 9.54ms, mfu 0.51%\n",
            "iter 46250: loss 1.4206, time 9.32ms, mfu 0.52%\n",
            "iter 46251: loss 1.4293, time 9.30ms, mfu 0.52%\n",
            "iter 46252: loss 1.5540, time 9.43ms, mfu 0.51%\n",
            "iter 46253: loss 1.5397, time 9.45ms, mfu 0.51%\n",
            "iter 46254: loss 1.3656, time 9.28ms, mfu 0.52%\n",
            "iter 46255: loss 1.5038, time 9.17ms, mfu 0.53%\n",
            "iter 46256: loss 1.4702, time 9.15ms, mfu 0.53%\n",
            "iter 46257: loss 1.3791, time 9.48ms, mfu 0.51%\n",
            "iter 46258: loss 1.4062, time 9.32ms, mfu 0.52%\n",
            "iter 46259: loss 1.3874, time 9.52ms, mfu 0.51%\n",
            "iter 46260: loss 1.4733, time 11.21ms, mfu 0.43%\n",
            "iter 46261: loss 1.4830, time 10.87ms, mfu 0.44%\n",
            "iter 46262: loss 1.4248, time 9.65ms, mfu 0.50%\n",
            "iter 46263: loss 1.4645, time 9.18ms, mfu 0.53%\n",
            "iter 46264: loss 1.5263, time 9.22ms, mfu 0.52%\n",
            "iter 46265: loss 1.6120, time 9.35ms, mfu 0.52%\n",
            "iter 46266: loss 1.4726, time 9.63ms, mfu 0.50%\n",
            "iter 46267: loss 1.5511, time 9.21ms, mfu 0.52%\n",
            "iter 46268: loss 1.3681, time 9.93ms, mfu 0.49%\n",
            "iter 46269: loss 1.4888, time 10.09ms, mfu 0.48%\n",
            "iter 46270: loss 1.4544, time 13.97ms, mfu 0.35%\n",
            "iter 46271: loss 1.5864, time 10.62ms, mfu 0.45%\n",
            "iter 46272: loss 1.5516, time 9.58ms, mfu 0.50%\n",
            "iter 46273: loss 1.5234, time 9.25ms, mfu 0.52%\n",
            "iter 46274: loss 1.4905, time 9.34ms, mfu 0.52%\n",
            "iter 46275: loss 1.5159, time 9.17ms, mfu 0.53%\n",
            "iter 46276: loss 1.5575, time 9.27ms, mfu 0.52%\n",
            "iter 46277: loss 1.5170, time 10.70ms, mfu 0.45%\n",
            "iter 46278: loss 1.3883, time 14.56ms, mfu 0.33%\n",
            "iter 46279: loss 1.4861, time 12.40ms, mfu 0.39%\n",
            "iter 46280: loss 1.5014, time 9.36ms, mfu 0.52%\n",
            "iter 46281: loss 1.5732, time 9.13ms, mfu 0.53%\n",
            "iter 46282: loss 1.3750, time 9.31ms, mfu 0.52%\n",
            "iter 46283: loss 1.4290, time 9.26ms, mfu 0.52%\n",
            "iter 46284: loss 1.4009, time 9.22ms, mfu 0.52%\n",
            "iter 46285: loss 1.4582, time 9.12ms, mfu 0.53%\n",
            "iter 46286: loss 1.4766, time 9.18ms, mfu 0.53%\n",
            "iter 46287: loss 1.4913, time 9.39ms, mfu 0.51%\n",
            "iter 46288: loss 1.5582, time 10.93ms, mfu 0.44%\n",
            "iter 46289: loss 1.5631, time 9.67ms, mfu 0.50%\n",
            "iter 46290: loss 1.4301, time 9.57ms, mfu 0.50%\n",
            "iter 46291: loss 1.4610, time 14.99ms, mfu 0.32%\n",
            "iter 46292: loss 1.3865, time 10.41ms, mfu 0.46%\n",
            "iter 46293: loss 1.5477, time 9.34ms, mfu 0.52%\n",
            "iter 46294: loss 1.4079, time 9.75ms, mfu 0.49%\n",
            "iter 46295: loss 1.4440, time 9.40ms, mfu 0.51%\n",
            "iter 46296: loss 1.3707, time 9.30ms, mfu 0.52%\n",
            "iter 46297: loss 1.5761, time 9.35ms, mfu 0.52%\n",
            "iter 46298: loss 1.4091, time 10.72ms, mfu 0.45%\n",
            "iter 46299: loss 1.4324, time 9.29ms, mfu 0.52%\n",
            "iter 46300: loss 1.4262, time 9.26ms, mfu 0.52%\n",
            "iter 46301: loss 1.4747, time 9.18ms, mfu 0.53%\n",
            "iter 46302: loss 1.4802, time 9.40ms, mfu 0.51%\n",
            "iter 46303: loss 1.4214, time 10.04ms, mfu 0.48%\n",
            "iter 46304: loss 1.4120, time 9.51ms, mfu 0.51%\n",
            "iter 46305: loss 1.5343, time 9.33ms, mfu 0.52%\n",
            "iter 46306: loss 1.5082, time 9.19ms, mfu 0.52%\n",
            "iter 46307: loss 1.5221, time 9.43ms, mfu 0.51%\n",
            "iter 46308: loss 1.5394, time 9.39ms, mfu 0.51%\n",
            "iter 46309: loss 1.4882, time 9.48ms, mfu 0.51%\n",
            "iter 46310: loss 1.4553, time 9.32ms, mfu 0.52%\n",
            "iter 46311: loss 1.4110, time 12.41ms, mfu 0.39%\n",
            "iter 46312: loss 1.5707, time 10.42ms, mfu 0.46%\n",
            "iter 46313: loss 1.6037, time 11.87ms, mfu 0.41%\n",
            "iter 46314: loss 1.4905, time 9.18ms, mfu 0.53%\n",
            "iter 46315: loss 1.5195, time 9.42ms, mfu 0.51%\n",
            "iter 46316: loss 1.4481, time 9.29ms, mfu 0.52%\n",
            "iter 46317: loss 1.3596, time 9.44ms, mfu 0.51%\n",
            "iter 46318: loss 1.4710, time 9.37ms, mfu 0.51%\n",
            "iter 46319: loss 1.5351, time 9.36ms, mfu 0.52%\n",
            "iter 46320: loss 1.5673, time 9.23ms, mfu 0.52%\n",
            "iter 46321: loss 1.4102, time 9.71ms, mfu 0.50%\n",
            "iter 46322: loss 1.5225, time 9.38ms, mfu 0.51%\n",
            "iter 46323: loss 1.4968, time 11.20ms, mfu 0.43%\n",
            "iter 46324: loss 1.5432, time 9.36ms, mfu 0.52%\n",
            "iter 46325: loss 1.4265, time 9.72ms, mfu 0.50%\n",
            "iter 46326: loss 1.4875, time 9.67ms, mfu 0.50%\n",
            "iter 46327: loss 1.4986, time 9.82ms, mfu 0.49%\n",
            "iter 46328: loss 1.4589, time 12.21ms, mfu 0.39%\n",
            "iter 46329: loss 1.4703, time 12.57ms, mfu 0.38%\n",
            "iter 46330: loss 1.4379, time 9.65ms, mfu 0.50%\n",
            "iter 46331: loss 1.3544, time 9.33ms, mfu 0.52%\n",
            "iter 46332: loss 1.3993, time 9.34ms, mfu 0.52%\n",
            "iter 46333: loss 1.5237, time 11.51ms, mfu 0.42%\n",
            "iter 46334: loss 1.5228, time 11.51ms, mfu 0.42%\n",
            "iter 46335: loss 1.4721, time 9.39ms, mfu 0.51%\n",
            "iter 46336: loss 1.4128, time 9.27ms, mfu 0.52%\n",
            "iter 46337: loss 1.4902, time 9.67ms, mfu 0.50%\n",
            "iter 46338: loss 1.5025, time 9.38ms, mfu 0.51%\n",
            "iter 46339: loss 1.4866, time 9.31ms, mfu 0.52%\n",
            "iter 46340: loss 1.5036, time 9.57ms, mfu 0.50%\n",
            "iter 46341: loss 1.4600, time 9.25ms, mfu 0.52%\n",
            "iter 46342: loss 1.3777, time 10.03ms, mfu 0.48%\n",
            "iter 46343: loss 1.5012, time 10.11ms, mfu 0.48%\n",
            "iter 46344: loss 1.4951, time 9.56ms, mfu 0.50%\n",
            "iter 46345: loss 1.5250, time 9.57ms, mfu 0.50%\n",
            "iter 46346: loss 1.5109, time 9.58ms, mfu 0.50%\n",
            "iter 46347: loss 1.4648, time 9.85ms, mfu 0.49%\n",
            "iter 46348: loss 1.5904, time 9.32ms, mfu 0.52%\n",
            "iter 46349: loss 1.5405, time 9.69ms, mfu 0.50%\n",
            "iter 46350: loss 1.5274, time 9.31ms, mfu 0.52%\n",
            "iter 46351: loss 1.4504, time 9.40ms, mfu 0.51%\n",
            "iter 46352: loss 1.3912, time 9.38ms, mfu 0.51%\n",
            "iter 46353: loss 1.6033, time 9.36ms, mfu 0.52%\n",
            "iter 46354: loss 1.4643, time 10.39ms, mfu 0.46%\n",
            "iter 46355: loss 1.4206, time 13.23ms, mfu 0.36%\n",
            "iter 46356: loss 1.5304, time 9.75ms, mfu 0.49%\n",
            "iter 46357: loss 1.4402, time 9.39ms, mfu 0.51%\n",
            "iter 46358: loss 1.4332, time 9.23ms, mfu 0.52%\n",
            "iter 46359: loss 1.4898, time 9.84ms, mfu 0.49%\n",
            "iter 46360: loss 1.3962, time 9.38ms, mfu 0.51%\n",
            "iter 46361: loss 1.3828, time 9.28ms, mfu 0.52%\n",
            "iter 46362: loss 1.4763, time 9.41ms, mfu 0.51%\n",
            "iter 46363: loss 1.4556, time 9.90ms, mfu 0.49%\n",
            "iter 46364: loss 1.4794, time 9.55ms, mfu 0.51%\n",
            "iter 46365: loss 1.4513, time 9.33ms, mfu 0.52%\n",
            "iter 46366: loss 1.4206, time 9.94ms, mfu 0.48%\n",
            "iter 46367: loss 1.5179, time 9.57ms, mfu 0.50%\n",
            "iter 46368: loss 1.5126, time 9.43ms, mfu 0.51%\n",
            "iter 46369: loss 1.4318, time 9.53ms, mfu 0.51%\n",
            "iter 46370: loss 1.4766, time 9.43ms, mfu 0.51%\n",
            "iter 46371: loss 1.4652, time 9.83ms, mfu 0.49%\n",
            "iter 46372: loss 1.5556, time 9.68ms, mfu 0.50%\n",
            "iter 46373: loss 1.5054, time 9.49ms, mfu 0.51%\n",
            "iter 46374: loss 1.4244, time 9.64ms, mfu 0.50%\n",
            "iter 46375: loss 1.3951, time 12.58ms, mfu 0.38%\n",
            "iter 46376: loss 1.4683, time 12.91ms, mfu 0.37%\n",
            "iter 46377: loss 1.4193, time 9.61ms, mfu 0.50%\n",
            "iter 46378: loss 1.6223, time 9.44ms, mfu 0.51%\n",
            "iter 46379: loss 1.3836, time 9.57ms, mfu 0.50%\n",
            "iter 46380: loss 1.4963, time 9.84ms, mfu 0.49%\n",
            "iter 46381: loss 1.4189, time 16.77ms, mfu 0.29%\n",
            "iter 46382: loss 1.4376, time 12.18ms, mfu 0.40%\n",
            "iter 46383: loss 1.4450, time 9.43ms, mfu 0.51%\n",
            "iter 46384: loss 1.6842, time 9.30ms, mfu 0.52%\n",
            "iter 46385: loss 1.5002, time 9.34ms, mfu 0.52%\n",
            "iter 46386: loss 1.5358, time 9.25ms, mfu 0.52%\n",
            "iter 46387: loss 1.4127, time 9.41ms, mfu 0.51%\n",
            "iter 46388: loss 1.5607, time 9.63ms, mfu 0.50%\n",
            "iter 46389: loss 1.5214, time 12.04ms, mfu 0.40%\n",
            "iter 46390: loss 1.5101, time 9.43ms, mfu 0.51%\n",
            "iter 46391: loss 1.5304, time 10.65ms, mfu 0.45%\n",
            "iter 46392: loss 1.3976, time 9.59ms, mfu 0.50%\n",
            "iter 46393: loss 1.4509, time 13.23ms, mfu 0.36%\n",
            "iter 46394: loss 1.3674, time 9.41ms, mfu 0.51%\n",
            "iter 46395: loss 1.4734, time 13.73ms, mfu 0.35%\n",
            "iter 46396: loss 1.5805, time 9.82ms, mfu 0.49%\n",
            "iter 46397: loss 1.4425, time 9.46ms, mfu 0.51%\n",
            "iter 46398: loss 1.5217, time 9.83ms, mfu 0.49%\n",
            "iter 46399: loss 1.4246, time 9.85ms, mfu 0.49%\n",
            "iter 46400: loss 1.4833, time 9.91ms, mfu 0.49%\n",
            "iter 46401: loss 1.4448, time 9.60ms, mfu 0.50%\n",
            "iter 46402: loss 1.5115, time 9.64ms, mfu 0.50%\n",
            "iter 46403: loss 1.3563, time 9.60ms, mfu 0.50%\n",
            "iter 46404: loss 1.4299, time 9.62ms, mfu 0.50%\n",
            "iter 46405: loss 1.5277, time 9.83ms, mfu 0.49%\n",
            "iter 46406: loss 1.4607, time 9.70ms, mfu 0.50%\n",
            "iter 46407: loss 1.4229, time 9.61ms, mfu 0.50%\n",
            "iter 46408: loss 1.4708, time 9.66ms, mfu 0.50%\n",
            "iter 46409: loss 1.5154, time 10.01ms, mfu 0.48%\n",
            "iter 46410: loss 1.4286, time 9.76ms, mfu 0.49%\n",
            "iter 46411: loss 1.3571, time 9.39ms, mfu 0.51%\n",
            "iter 46412: loss 1.5263, time 9.56ms, mfu 0.50%\n",
            "iter 46413: loss 1.5174, time 9.60ms, mfu 0.50%\n",
            "iter 46414: loss 1.3931, time 9.51ms, mfu 0.51%\n",
            "iter 46415: loss 1.4526, time 9.93ms, mfu 0.49%\n",
            "iter 46416: loss 1.3951, time 13.16ms, mfu 0.37%\n",
            "iter 46417: loss 1.3853, time 11.19ms, mfu 0.43%\n",
            "iter 46418: loss 1.6101, time 9.35ms, mfu 0.52%\n",
            "iter 46419: loss 1.4880, time 10.31ms, mfu 0.47%\n",
            "iter 46420: loss 1.4223, time 9.35ms, mfu 0.52%\n",
            "iter 46421: loss 1.3929, time 9.53ms, mfu 0.51%\n",
            "iter 46422: loss 1.4366, time 11.09ms, mfu 0.43%\n",
            "iter 46423: loss 1.4390, time 9.30ms, mfu 0.52%\n",
            "iter 46424: loss 1.5807, time 9.34ms, mfu 0.52%\n",
            "iter 46425: loss 1.3882, time 9.54ms, mfu 0.51%\n",
            "iter 46426: loss 1.4878, time 9.49ms, mfu 0.51%\n",
            "iter 46427: loss 1.4793, time 11.55ms, mfu 0.42%\n",
            "iter 46428: loss 1.4405, time 12.24ms, mfu 0.39%\n",
            "iter 46429: loss 1.4919, time 10.32ms, mfu 0.47%\n",
            "iter 46430: loss 1.5256, time 9.37ms, mfu 0.51%\n",
            "iter 46431: loss 1.4230, time 9.46ms, mfu 0.51%\n",
            "iter 46432: loss 1.4742, time 9.71ms, mfu 0.50%\n",
            "iter 46433: loss 1.4135, time 9.25ms, mfu 0.52%\n",
            "iter 46434: loss 1.5366, time 9.36ms, mfu 0.51%\n",
            "iter 46435: loss 1.3866, time 9.31ms, mfu 0.52%\n",
            "iter 46436: loss 1.5268, time 9.33ms, mfu 0.52%\n",
            "iter 46437: loss 1.5068, time 13.32ms, mfu 0.36%\n",
            "iter 46438: loss 1.5054, time 10.00ms, mfu 0.48%\n",
            "iter 46439: loss 1.3984, time 9.38ms, mfu 0.51%\n",
            "iter 46440: loss 1.4059, time 9.27ms, mfu 0.52%\n",
            "iter 46441: loss 1.5894, time 10.04ms, mfu 0.48%\n",
            "iter 46442: loss 1.6104, time 9.68ms, mfu 0.50%\n",
            "iter 46443: loss 1.4195, time 9.43ms, mfu 0.51%\n",
            "iter 46444: loss 1.4928, time 9.32ms, mfu 0.52%\n",
            "iter 46445: loss 1.4126, time 9.40ms, mfu 0.51%\n",
            "iter 46446: loss 1.3745, time 9.40ms, mfu 0.51%\n",
            "iter 46447: loss 1.4215, time 9.54ms, mfu 0.51%\n",
            "iter 46448: loss 1.5608, time 9.47ms, mfu 0.51%\n",
            "iter 46449: loss 1.3661, time 9.69ms, mfu 0.50%\n",
            "iter 46450: loss 1.4396, time 9.34ms, mfu 0.52%\n",
            "iter 46451: loss 1.4587, time 9.54ms, mfu 0.51%\n",
            "iter 46452: loss 1.5248, time 9.40ms, mfu 0.51%\n",
            "iter 46453: loss 1.3872, time 9.41ms, mfu 0.51%\n",
            "iter 46454: loss 1.5626, time 9.30ms, mfu 0.52%\n",
            "iter 46455: loss 1.4133, time 9.38ms, mfu 0.51%\n",
            "iter 46456: loss 1.5670, time 9.49ms, mfu 0.51%\n",
            "iter 46457: loss 1.4612, time 9.42ms, mfu 0.51%\n",
            "iter 46458: loss 1.3927, time 9.28ms, mfu 0.52%\n",
            "iter 46459: loss 1.4762, time 14.89ms, mfu 0.32%\n",
            "iter 46460: loss 1.4333, time 10.21ms, mfu 0.47%\n",
            "iter 46461: loss 1.3680, time 9.65ms, mfu 0.50%\n",
            "iter 46462: loss 1.4416, time 9.62ms, mfu 0.50%\n",
            "iter 46463: loss 1.4694, time 9.58ms, mfu 0.50%\n",
            "iter 46464: loss 1.4452, time 9.53ms, mfu 0.51%\n",
            "iter 46465: loss 1.4907, time 9.44ms, mfu 0.51%\n",
            "iter 46466: loss 1.4341, time 9.86ms, mfu 0.49%\n",
            "iter 46467: loss 1.4480, time 9.36ms, mfu 0.52%\n",
            "iter 46468: loss 1.4350, time 9.86ms, mfu 0.49%\n",
            "iter 46469: loss 1.4559, time 9.34ms, mfu 0.52%\n",
            "iter 46470: loss 1.4284, time 9.32ms, mfu 0.52%\n",
            "iter 46471: loss 1.4555, time 9.41ms, mfu 0.51%\n",
            "iter 46472: loss 1.4035, time 11.98ms, mfu 0.40%\n",
            "iter 46473: loss 1.4878, time 9.60ms, mfu 0.50%\n",
            "iter 46474: loss 1.5662, time 9.28ms, mfu 0.52%\n",
            "iter 46475: loss 1.4908, time 9.38ms, mfu 0.51%\n",
            "iter 46476: loss 1.4929, time 9.42ms, mfu 0.51%\n",
            "iter 46477: loss 1.4134, time 9.46ms, mfu 0.51%\n",
            "iter 46478: loss 1.4108, time 9.33ms, mfu 0.52%\n",
            "iter 46479: loss 1.4221, time 9.68ms, mfu 0.50%\n",
            "iter 46480: loss 1.5029, time 14.10ms, mfu 0.34%\n",
            "iter 46481: loss 1.5487, time 11.87ms, mfu 0.41%\n",
            "iter 46482: loss 1.5279, time 10.32ms, mfu 0.47%\n",
            "iter 46483: loss 1.5201, time 15.65ms, mfu 0.31%\n",
            "iter 46484: loss 1.4856, time 13.35ms, mfu 0.36%\n",
            "iter 46485: loss 1.4003, time 9.86ms, mfu 0.49%\n",
            "iter 46486: loss 1.5101, time 9.49ms, mfu 0.51%\n",
            "iter 46487: loss 1.4576, time 9.39ms, mfu 0.51%\n",
            "iter 46488: loss 1.4768, time 9.88ms, mfu 0.49%\n",
            "iter 46489: loss 1.4521, time 11.21ms, mfu 0.43%\n",
            "iter 46490: loss 1.4731, time 9.54ms, mfu 0.51%\n",
            "iter 46491: loss 1.5309, time 10.53ms, mfu 0.46%\n",
            "iter 46492: loss 1.4996, time 9.60ms, mfu 0.50%\n",
            "iter 46493: loss 1.5205, time 9.66ms, mfu 0.50%\n",
            "iter 46494: loss 1.5133, time 9.54ms, mfu 0.51%\n",
            "iter 46495: loss 1.3966, time 9.49ms, mfu 0.51%\n",
            "iter 46496: loss 1.4328, time 9.31ms, mfu 0.52%\n",
            "iter 46497: loss 1.4112, time 9.39ms, mfu 0.51%\n",
            "iter 46498: loss 1.4789, time 11.68ms, mfu 0.41%\n",
            "iter 46499: loss 1.5356, time 15.42ms, mfu 0.31%\n",
            "step 46500: train loss 1.2686, val loss 1.6509\n",
            "iter 46500: loss 1.4232, time 27.78ms, mfu 0.17%\n",
            "iter 46501: loss 1.4203, time 16.78ms, mfu 0.29%\n",
            "iter 46502: loss 1.4850, time 12.82ms, mfu 0.38%\n",
            "iter 46503: loss 1.4298, time 9.83ms, mfu 0.49%\n",
            "iter 46504: loss 1.3994, time 9.53ms, mfu 0.51%\n",
            "iter 46505: loss 1.4312, time 10.16ms, mfu 0.47%\n",
            "iter 46506: loss 1.4896, time 9.44ms, mfu 0.51%\n",
            "iter 46507: loss 1.4974, time 9.44ms, mfu 0.51%\n",
            "iter 46508: loss 1.6304, time 9.43ms, mfu 0.51%\n",
            "iter 46509: loss 1.3446, time 9.66ms, mfu 0.50%\n",
            "iter 46510: loss 1.5478, time 9.70ms, mfu 0.50%\n",
            "iter 46511: loss 1.4740, time 9.62ms, mfu 0.50%\n",
            "iter 46512: loss 1.3379, time 10.17ms, mfu 0.47%\n",
            "iter 46513: loss 1.4724, time 9.54ms, mfu 0.51%\n",
            "iter 46514: loss 1.4394, time 9.68ms, mfu 0.50%\n",
            "iter 46515: loss 1.5112, time 9.54ms, mfu 0.51%\n",
            "iter 46516: loss 1.6485, time 10.16ms, mfu 0.47%\n",
            "iter 46517: loss 1.3296, time 12.64ms, mfu 0.38%\n",
            "iter 46518: loss 1.4769, time 11.05ms, mfu 0.44%\n",
            "iter 46519: loss 1.6680, time 14.73ms, mfu 0.33%\n",
            "iter 46520: loss 1.4715, time 9.29ms, mfu 0.52%\n",
            "iter 46521: loss 1.4192, time 9.32ms, mfu 0.52%\n",
            "iter 46522: loss 1.4827, time 9.36ms, mfu 0.52%\n",
            "iter 46523: loss 1.3331, time 10.91ms, mfu 0.44%\n",
            "iter 46524: loss 1.4822, time 12.38ms, mfu 0.39%\n",
            "iter 46525: loss 1.3975, time 12.39ms, mfu 0.39%\n",
            "iter 46526: loss 1.4052, time 9.65ms, mfu 0.50%\n",
            "iter 46527: loss 1.4612, time 9.47ms, mfu 0.51%\n",
            "iter 46528: loss 1.5679, time 10.10ms, mfu 0.48%\n",
            "iter 46529: loss 1.4587, time 12.47ms, mfu 0.39%\n",
            "iter 46530: loss 1.4169, time 9.76ms, mfu 0.49%\n",
            "iter 46531: loss 1.4757, time 9.37ms, mfu 0.51%\n",
            "iter 46532: loss 1.6810, time 9.31ms, mfu 0.52%\n",
            "iter 46533: loss 1.3968, time 9.32ms, mfu 0.52%\n",
            "iter 46534: loss 1.4661, time 9.50ms, mfu 0.51%\n",
            "iter 46535: loss 1.4729, time 10.08ms, mfu 0.48%\n",
            "iter 46536: loss 1.5753, time 9.42ms, mfu 0.51%\n",
            "iter 46537: loss 1.4580, time 12.56ms, mfu 0.38%\n",
            "iter 46538: loss 1.4820, time 11.19ms, mfu 0.43%\n",
            "iter 46539: loss 1.5490, time 9.41ms, mfu 0.51%\n",
            "iter 46540: loss 1.4604, time 9.38ms, mfu 0.51%\n",
            "iter 46541: loss 1.5504, time 9.41ms, mfu 0.51%\n",
            "iter 46542: loss 1.3898, time 9.29ms, mfu 0.52%\n",
            "iter 46543: loss 1.5151, time 9.79ms, mfu 0.49%\n",
            "iter 46544: loss 1.4366, time 9.46ms, mfu 0.51%\n",
            "iter 46545: loss 1.4279, time 9.54ms, mfu 0.51%\n",
            "iter 46546: loss 1.3832, time 9.53ms, mfu 0.51%\n",
            "iter 46547: loss 1.4153, time 9.50ms, mfu 0.51%\n",
            "iter 46548: loss 1.5518, time 9.54ms, mfu 0.51%\n",
            "iter 46549: loss 1.3983, time 9.59ms, mfu 0.50%\n",
            "iter 46550: loss 1.4929, time 9.40ms, mfu 0.51%\n",
            "iter 46551: loss 1.4369, time 10.45ms, mfu 0.46%\n",
            "iter 46552: loss 1.3697, time 9.43ms, mfu 0.51%\n",
            "iter 46553: loss 1.4853, time 9.46ms, mfu 0.51%\n",
            "iter 46554: loss 1.4958, time 9.52ms, mfu 0.51%\n",
            "iter 46555: loss 1.5195, time 9.71ms, mfu 0.50%\n",
            "iter 46556: loss 1.4544, time 9.37ms, mfu 0.51%\n",
            "iter 46557: loss 1.4962, time 9.47ms, mfu 0.51%\n",
            "iter 46558: loss 1.5133, time 9.80ms, mfu 0.49%\n",
            "iter 46559: loss 1.5009, time 14.87ms, mfu 0.32%\n",
            "iter 46560: loss 1.5743, time 9.97ms, mfu 0.48%\n",
            "iter 46561: loss 1.4241, time 10.19ms, mfu 0.47%\n",
            "iter 46562: loss 1.5300, time 9.50ms, mfu 0.51%\n",
            "iter 46563: loss 1.6066, time 9.38ms, mfu 0.51%\n",
            "iter 46564: loss 1.5095, time 9.46ms, mfu 0.51%\n",
            "iter 46565: loss 1.4856, time 9.63ms, mfu 0.50%\n",
            "iter 46566: loss 1.5222, time 9.41ms, mfu 0.51%\n",
            "iter 46567: loss 1.5617, time 9.36ms, mfu 0.52%\n",
            "iter 46568: loss 1.4784, time 9.35ms, mfu 0.52%\n",
            "iter 46569: loss 1.3944, time 9.85ms, mfu 0.49%\n",
            "iter 46570: loss 1.5139, time 9.73ms, mfu 0.50%\n",
            "iter 46571: loss 1.4954, time 15.57ms, mfu 0.31%\n",
            "iter 46572: loss 1.5163, time 14.63ms, mfu 0.33%\n",
            "iter 46573: loss 1.5723, time 13.63ms, mfu 0.35%\n",
            "iter 46574: loss 1.3958, time 9.74ms, mfu 0.50%\n",
            "iter 46575: loss 1.4955, time 9.54ms, mfu 0.51%\n",
            "iter 46576: loss 1.4329, time 9.65ms, mfu 0.50%\n",
            "iter 46577: loss 1.4232, time 12.33ms, mfu 0.39%\n",
            "iter 46578: loss 1.4920, time 22.07ms, mfu 0.22%\n",
            "iter 46579: loss 1.5502, time 21.66ms, mfu 0.22%\n",
            "iter 46580: loss 1.5191, time 17.16ms, mfu 0.28%\n",
            "iter 46581: loss 1.4590, time 19.85ms, mfu 0.24%\n",
            "iter 46582: loss 1.4627, time 17.63ms, mfu 0.27%\n",
            "iter 46583: loss 1.4643, time 15.03ms, mfu 0.32%\n",
            "iter 46584: loss 1.5072, time 14.89ms, mfu 0.32%\n",
            "iter 46585: loss 1.4834, time 15.23ms, mfu 0.32%\n",
            "iter 46586: loss 1.4683, time 14.93ms, mfu 0.32%\n",
            "iter 46587: loss 1.4842, time 14.18ms, mfu 0.34%\n",
            "iter 46588: loss 1.4201, time 12.35ms, mfu 0.39%\n",
            "iter 46589: loss 1.4822, time 15.14ms, mfu 0.32%\n",
            "iter 46590: loss 1.5772, time 13.89ms, mfu 0.35%\n",
            "iter 46591: loss 1.4055, time 15.03ms, mfu 0.32%\n",
            "iter 46592: loss 1.5659, time 13.88ms, mfu 0.35%\n",
            "iter 46593: loss 1.5579, time 11.84ms, mfu 0.41%\n",
            "iter 46594: loss 1.4987, time 14.61ms, mfu 0.33%\n",
            "iter 46595: loss 1.3897, time 15.84ms, mfu 0.30%\n",
            "iter 46596: loss 1.5419, time 13.52ms, mfu 0.36%\n",
            "iter 46597: loss 1.5163, time 14.24ms, mfu 0.34%\n",
            "iter 46598: loss 1.4781, time 12.84ms, mfu 0.38%\n",
            "iter 46599: loss 1.5635, time 13.27ms, mfu 0.36%\n",
            "iter 46600: loss 1.4694, time 14.74ms, mfu 0.33%\n",
            "iter 46601: loss 1.4880, time 12.36ms, mfu 0.39%\n",
            "iter 46602: loss 1.4531, time 12.25ms, mfu 0.39%\n",
            "iter 46603: loss 1.3574, time 12.39ms, mfu 0.39%\n",
            "iter 46604: loss 1.4925, time 14.51ms, mfu 0.33%\n",
            "iter 46605: loss 1.5882, time 12.39ms, mfu 0.39%\n",
            "iter 46606: loss 1.3918, time 12.28ms, mfu 0.39%\n",
            "iter 46607: loss 1.4234, time 25.99ms, mfu 0.19%\n",
            "iter 46608: loss 1.4406, time 17.45ms, mfu 0.28%\n",
            "iter 46609: loss 1.5230, time 14.37ms, mfu 0.34%\n",
            "iter 46610: loss 1.4287, time 12.39ms, mfu 0.39%\n",
            "iter 46611: loss 1.5201, time 12.24ms, mfu 0.39%\n",
            "iter 46612: loss 1.5412, time 12.26ms, mfu 0.39%\n",
            "iter 46613: loss 1.4187, time 12.76ms, mfu 0.38%\n",
            "iter 46614: loss 1.5062, time 12.08ms, mfu 0.40%\n",
            "iter 46615: loss 1.5866, time 12.30ms, mfu 0.39%\n",
            "iter 46616: loss 1.5244, time 12.59ms, mfu 0.38%\n",
            "iter 46617: loss 1.4960, time 13.07ms, mfu 0.37%\n",
            "iter 46618: loss 1.4767, time 12.69ms, mfu 0.38%\n",
            "iter 46619: loss 1.4629, time 12.41ms, mfu 0.39%\n",
            "iter 46620: loss 1.4173, time 12.40ms, mfu 0.39%\n",
            "iter 46621: loss 1.3498, time 14.86ms, mfu 0.32%\n",
            "iter 46622: loss 1.5333, time 13.81ms, mfu 0.35%\n",
            "iter 46623: loss 1.4183, time 12.61ms, mfu 0.38%\n",
            "iter 46624: loss 1.4187, time 12.96ms, mfu 0.37%\n",
            "iter 46625: loss 1.4551, time 13.10ms, mfu 0.37%\n",
            "iter 46626: loss 1.5084, time 13.26ms, mfu 0.36%\n",
            "iter 46627: loss 1.5436, time 11.67ms, mfu 0.41%\n",
            "iter 46628: loss 1.4480, time 11.69ms, mfu 0.41%\n",
            "iter 46629: loss 1.5588, time 11.65ms, mfu 0.41%\n",
            "iter 46630: loss 1.3515, time 11.79ms, mfu 0.41%\n",
            "iter 46631: loss 1.4821, time 12.17ms, mfu 0.40%\n",
            "iter 46632: loss 1.4981, time 11.96ms, mfu 0.40%\n",
            "iter 46633: loss 1.4209, time 11.84ms, mfu 0.41%\n",
            "iter 46634: loss 1.5658, time 11.83ms, mfu 0.41%\n",
            "iter 46635: loss 1.5010, time 11.74ms, mfu 0.41%\n",
            "iter 46636: loss 1.5416, time 11.68ms, mfu 0.41%\n",
            "iter 46637: loss 1.5529, time 11.86ms, mfu 0.41%\n",
            "iter 46638: loss 1.4687, time 15.19ms, mfu 0.32%\n",
            "iter 46639: loss 1.3712, time 13.52ms, mfu 0.36%\n",
            "iter 46640: loss 1.4395, time 11.78ms, mfu 0.41%\n",
            "iter 46641: loss 1.3922, time 11.72ms, mfu 0.41%\n",
            "iter 46642: loss 1.5471, time 11.68ms, mfu 0.41%\n",
            "iter 46643: loss 1.4287, time 11.74ms, mfu 0.41%\n",
            "iter 46644: loss 1.3948, time 11.77ms, mfu 0.41%\n",
            "iter 46645: loss 1.4151, time 11.64ms, mfu 0.41%\n",
            "iter 46646: loss 1.4375, time 11.80ms, mfu 0.41%\n",
            "iter 46647: loss 1.5018, time 11.67ms, mfu 0.41%\n",
            "iter 46648: loss 1.4565, time 16.57ms, mfu 0.29%\n",
            "iter 46649: loss 1.4421, time 11.72ms, mfu 0.41%\n",
            "iter 46650: loss 1.5005, time 11.80ms, mfu 0.41%\n",
            "iter 46651: loss 1.4835, time 12.34ms, mfu 0.39%\n",
            "iter 46652: loss 1.5485, time 11.80ms, mfu 0.41%\n",
            "iter 46653: loss 1.5442, time 15.05ms, mfu 0.32%\n",
            "iter 46654: loss 1.4816, time 17.99ms, mfu 0.27%\n",
            "iter 46655: loss 1.4369, time 15.66ms, mfu 0.31%\n",
            "iter 46656: loss 1.3537, time 15.66ms, mfu 0.31%\n",
            "iter 46657: loss 1.5818, time 23.04ms, mfu 0.21%\n",
            "iter 46658: loss 1.5313, time 24.67ms, mfu 0.20%\n",
            "iter 46659: loss 1.4279, time 17.54ms, mfu 0.27%\n",
            "iter 46660: loss 1.4799, time 17.49ms, mfu 0.28%\n",
            "iter 46661: loss 1.5106, time 17.09ms, mfu 0.28%\n",
            "iter 46662: loss 1.5202, time 12.79ms, mfu 0.38%\n",
            "iter 46663: loss 1.5211, time 12.95ms, mfu 0.37%\n",
            "iter 46664: loss 1.6221, time 12.31ms, mfu 0.39%\n",
            "iter 46665: loss 1.4304, time 15.18ms, mfu 0.32%\n",
            "iter 46666: loss 1.4962, time 14.92ms, mfu 0.32%\n",
            "iter 46667: loss 1.5202, time 20.11ms, mfu 0.24%\n",
            "iter 46668: loss 1.5683, time 15.12ms, mfu 0.32%\n",
            "iter 46669: loss 1.4348, time 14.72ms, mfu 0.33%\n",
            "iter 46670: loss 1.5072, time 13.46ms, mfu 0.36%\n",
            "iter 46671: loss 1.5078, time 12.46ms, mfu 0.39%\n",
            "iter 46672: loss 1.3906, time 12.21ms, mfu 0.39%\n",
            "iter 46673: loss 1.4737, time 12.35ms, mfu 0.39%\n",
            "iter 46674: loss 1.6233, time 13.35ms, mfu 0.36%\n",
            "iter 46675: loss 1.6590, time 13.10ms, mfu 0.37%\n",
            "iter 46676: loss 1.6532, time 12.96ms, mfu 0.37%\n",
            "iter 46677: loss 1.4857, time 13.27ms, mfu 0.36%\n",
            "iter 46678: loss 1.5357, time 11.79ms, mfu 0.41%\n",
            "iter 46679: loss 1.5338, time 11.74ms, mfu 0.41%\n",
            "iter 46680: loss 1.4416, time 14.55ms, mfu 0.33%\n",
            "iter 46681: loss 1.5183, time 21.31ms, mfu 0.23%\n",
            "iter 46682: loss 1.4497, time 19.66ms, mfu 0.25%\n",
            "iter 46683: loss 1.3490, time 18.05ms, mfu 0.27%\n",
            "iter 46684: loss 1.5211, time 12.05ms, mfu 0.40%\n",
            "iter 46685: loss 1.5823, time 11.99ms, mfu 0.40%\n",
            "iter 46686: loss 1.3852, time 12.17ms, mfu 0.40%\n",
            "iter 46687: loss 1.4079, time 11.99ms, mfu 0.40%\n",
            "iter 46688: loss 1.5592, time 11.99ms, mfu 0.40%\n",
            "iter 46689: loss 1.4675, time 12.02ms, mfu 0.40%\n",
            "iter 46690: loss 1.5149, time 12.65ms, mfu 0.38%\n",
            "iter 46691: loss 1.4299, time 12.28ms, mfu 0.39%\n",
            "iter 46692: loss 1.5495, time 11.98ms, mfu 0.40%\n",
            "iter 46693: loss 1.4656, time 12.02ms, mfu 0.40%\n",
            "iter 46694: loss 1.5143, time 12.07ms, mfu 0.40%\n",
            "iter 46695: loss 1.3764, time 12.01ms, mfu 0.40%\n",
            "iter 46696: loss 1.4561, time 12.23ms, mfu 0.39%\n",
            "iter 46697: loss 1.5527, time 12.29ms, mfu 0.39%\n",
            "iter 46698: loss 1.5597, time 12.24ms, mfu 0.39%\n",
            "iter 46699: loss 1.5215, time 19.46ms, mfu 0.25%\n",
            "iter 46700: loss 1.4657, time 12.42ms, mfu 0.39%\n",
            "iter 46701: loss 1.3095, time 12.02ms, mfu 0.40%\n",
            "iter 46702: loss 1.5065, time 12.03ms, mfu 0.40%\n",
            "iter 46703: loss 1.3646, time 12.06ms, mfu 0.40%\n",
            "iter 46704: loss 1.4626, time 12.14ms, mfu 0.40%\n",
            "iter 46705: loss 1.4359, time 12.08ms, mfu 0.40%\n",
            "iter 46706: loss 1.4243, time 12.52ms, mfu 0.39%\n",
            "iter 46707: loss 1.4327, time 12.70ms, mfu 0.38%\n",
            "iter 46708: loss 1.4955, time 14.52ms, mfu 0.33%\n",
            "iter 46709: loss 1.5021, time 13.37ms, mfu 0.36%\n",
            "iter 46710: loss 1.4256, time 11.72ms, mfu 0.41%\n",
            "iter 46711: loss 1.5179, time 11.61ms, mfu 0.42%\n",
            "iter 46712: loss 1.4113, time 11.63ms, mfu 0.41%\n",
            "iter 46713: loss 1.4776, time 11.62ms, mfu 0.42%\n",
            "iter 46714: loss 1.4466, time 11.95ms, mfu 0.40%\n",
            "iter 46715: loss 1.5448, time 16.98ms, mfu 0.28%\n",
            "iter 46716: loss 1.4051, time 12.13ms, mfu 0.40%\n",
            "iter 46717: loss 1.4876, time 11.57ms, mfu 0.42%\n",
            "iter 46718: loss 1.4031, time 11.68ms, mfu 0.41%\n",
            "iter 46719: loss 1.6543, time 11.52ms, mfu 0.42%\n",
            "iter 46720: loss 1.3528, time 11.66ms, mfu 0.41%\n",
            "iter 46721: loss 1.3629, time 13.48ms, mfu 0.36%\n",
            "iter 46722: loss 1.4948, time 11.58ms, mfu 0.42%\n",
            "iter 46723: loss 1.5269, time 11.65ms, mfu 0.41%\n",
            "iter 46724: loss 1.3870, time 15.54ms, mfu 0.31%\n",
            "iter 46725: loss 1.5583, time 11.58ms, mfu 0.42%\n",
            "iter 46726: loss 1.4884, time 11.55ms, mfu 0.42%\n",
            "iter 46727: loss 1.4575, time 11.62ms, mfu 0.42%\n",
            "iter 46728: loss 1.4182, time 11.55ms, mfu 0.42%\n",
            "iter 46729: loss 1.5058, time 11.94ms, mfu 0.40%\n",
            "iter 46730: loss 1.4099, time 11.86ms, mfu 0.41%\n",
            "iter 46731: loss 1.4131, time 13.75ms, mfu 0.35%\n",
            "iter 46732: loss 1.5074, time 14.33ms, mfu 0.34%\n",
            "iter 46733: loss 1.4920, time 12.39ms, mfu 0.39%\n",
            "iter 46734: loss 1.3507, time 13.52ms, mfu 0.36%\n",
            "iter 46735: loss 1.4794, time 11.58ms, mfu 0.42%\n",
            "iter 46736: loss 1.4264, time 11.50ms, mfu 0.42%\n",
            "iter 46737: loss 1.4754, time 15.07ms, mfu 0.32%\n",
            "iter 46738: loss 1.5470, time 12.64ms, mfu 0.38%\n",
            "iter 46739: loss 1.4565, time 12.20ms, mfu 0.40%\n",
            "iter 46740: loss 1.4468, time 14.51ms, mfu 0.33%\n",
            "iter 46741: loss 1.4053, time 12.22ms, mfu 0.39%\n",
            "iter 46742: loss 1.4893, time 11.66ms, mfu 0.41%\n",
            "iter 46743: loss 1.4478, time 12.64ms, mfu 0.38%\n",
            "iter 46744: loss 1.5718, time 15.41ms, mfu 0.31%\n",
            "iter 46745: loss 1.4383, time 14.18ms, mfu 0.34%\n",
            "iter 46746: loss 1.4346, time 12.32ms, mfu 0.39%\n",
            "iter 46747: loss 1.4743, time 13.27ms, mfu 0.36%\n",
            "iter 46748: loss 1.6261, time 17.15ms, mfu 0.28%\n",
            "iter 46749: loss 1.4584, time 14.90ms, mfu 0.32%\n",
            "iter 46750: loss 1.4989, time 14.76ms, mfu 0.33%\n",
            "iter 46751: loss 1.4490, time 15.06ms, mfu 0.32%\n",
            "iter 46752: loss 1.4501, time 15.67ms, mfu 0.31%\n",
            "iter 46753: loss 1.4227, time 16.41ms, mfu 0.29%\n",
            "iter 46754: loss 1.5044, time 12.43ms, mfu 0.39%\n",
            "iter 46755: loss 1.4481, time 12.39ms, mfu 0.39%\n",
            "iter 46756: loss 1.4094, time 14.00ms, mfu 0.34%\n",
            "iter 46757: loss 1.4531, time 16.93ms, mfu 0.28%\n",
            "iter 46758: loss 1.3874, time 12.92ms, mfu 0.37%\n",
            "iter 46759: loss 1.5817, time 12.06ms, mfu 0.40%\n",
            "iter 46760: loss 1.4746, time 11.82ms, mfu 0.41%\n",
            "iter 46761: loss 1.5679, time 11.77ms, mfu 0.41%\n",
            "iter 46762: loss 1.4954, time 16.59ms, mfu 0.29%\n",
            "iter 46763: loss 1.4723, time 21.08ms, mfu 0.23%\n",
            "iter 46764: loss 1.4749, time 15.80ms, mfu 0.31%\n",
            "iter 46765: loss 1.4528, time 15.67ms, mfu 0.31%\n",
            "iter 46766: loss 1.5953, time 14.94ms, mfu 0.32%\n",
            "iter 46767: loss 1.3075, time 14.97ms, mfu 0.32%\n",
            "iter 46768: loss 1.4181, time 16.47ms, mfu 0.29%\n",
            "iter 46769: loss 1.4151, time 13.96ms, mfu 0.35%\n",
            "iter 46770: loss 1.5715, time 12.81ms, mfu 0.38%\n",
            "iter 46771: loss 1.6064, time 14.47ms, mfu 0.33%\n",
            "iter 46772: loss 1.4889, time 13.72ms, mfu 0.35%\n",
            "iter 46773: loss 1.4067, time 12.23ms, mfu 0.39%\n",
            "iter 46774: loss 1.4949, time 14.22ms, mfu 0.34%\n",
            "iter 46775: loss 1.5278, time 14.69ms, mfu 0.33%\n",
            "iter 46776: loss 1.4683, time 14.53ms, mfu 0.33%\n",
            "iter 46777: loss 1.4304, time 18.25ms, mfu 0.26%\n",
            "iter 46778: loss 1.5154, time 14.97ms, mfu 0.32%\n",
            "iter 46779: loss 1.4865, time 14.89ms, mfu 0.32%\n",
            "iter 46780: loss 1.4417, time 16.97ms, mfu 0.28%\n",
            "iter 46781: loss 1.4914, time 14.40ms, mfu 0.33%\n",
            "iter 46782: loss 1.4533, time 14.15ms, mfu 0.34%\n",
            "iter 46783: loss 1.4926, time 14.05ms, mfu 0.34%\n",
            "iter 46784: loss 1.4663, time 13.78ms, mfu 0.35%\n",
            "iter 46785: loss 1.5166, time 13.60ms, mfu 0.35%\n",
            "iter 46786: loss 1.4791, time 13.72ms, mfu 0.35%\n",
            "iter 46787: loss 1.4461, time 13.90ms, mfu 0.35%\n",
            "iter 46788: loss 1.4404, time 14.03ms, mfu 0.34%\n",
            "iter 46789: loss 1.4945, time 13.93ms, mfu 0.35%\n",
            "iter 46790: loss 1.3697, time 14.01ms, mfu 0.34%\n",
            "iter 46791: loss 1.5167, time 13.91ms, mfu 0.35%\n",
            "iter 46792: loss 1.4108, time 20.74ms, mfu 0.23%\n",
            "iter 46793: loss 1.5341, time 14.46ms, mfu 0.33%\n",
            "iter 46794: loss 1.4247, time 14.99ms, mfu 0.32%\n",
            "iter 46795: loss 1.4699, time 14.79ms, mfu 0.33%\n",
            "iter 46796: loss 1.4916, time 14.85ms, mfu 0.32%\n",
            "iter 46797: loss 1.4957, time 14.77ms, mfu 0.33%\n",
            "iter 46798: loss 1.5009, time 18.79ms, mfu 0.26%\n",
            "iter 46799: loss 1.4156, time 14.75ms, mfu 0.33%\n",
            "iter 46800: loss 1.3863, time 15.60ms, mfu 0.31%\n",
            "iter 46801: loss 1.3832, time 15.17ms, mfu 0.32%\n",
            "iter 46802: loss 1.4403, time 15.70ms, mfu 0.31%\n",
            "iter 46803: loss 1.4360, time 15.99ms, mfu 0.30%\n",
            "iter 46804: loss 1.5062, time 15.95ms, mfu 0.30%\n",
            "iter 46805: loss 1.5118, time 21.58ms, mfu 0.22%\n",
            "iter 46806: loss 1.5459, time 15.98ms, mfu 0.30%\n",
            "iter 46807: loss 1.4576, time 15.24ms, mfu 0.32%\n",
            "iter 46808: loss 1.4446, time 12.26ms, mfu 0.39%\n",
            "iter 46809: loss 1.4100, time 16.35ms, mfu 0.29%\n",
            "iter 46810: loss 1.4761, time 13.20ms, mfu 0.37%\n",
            "iter 46811: loss 1.5751, time 9.77ms, mfu 0.49%\n",
            "iter 46812: loss 1.4479, time 9.83ms, mfu 0.49%\n",
            "iter 46813: loss 1.4659, time 9.69ms, mfu 0.50%\n",
            "iter 46814: loss 1.5072, time 9.78ms, mfu 0.49%\n",
            "iter 46815: loss 1.5177, time 9.44ms, mfu 0.51%\n",
            "iter 46816: loss 1.4247, time 9.57ms, mfu 0.50%\n",
            "iter 46817: loss 1.4863, time 9.52ms, mfu 0.51%\n",
            "iter 46818: loss 1.4792, time 10.22ms, mfu 0.47%\n",
            "iter 46819: loss 1.4821, time 9.54ms, mfu 0.51%\n",
            "iter 46820: loss 1.5333, time 9.85ms, mfu 0.49%\n",
            "iter 46821: loss 1.4517, time 9.92ms, mfu 0.49%\n",
            "iter 46822: loss 1.5134, time 14.39ms, mfu 0.34%\n",
            "iter 46823: loss 1.4472, time 10.17ms, mfu 0.47%\n",
            "iter 46824: loss 1.3746, time 11.38ms, mfu 0.42%\n",
            "iter 46825: loss 1.5591, time 9.33ms, mfu 0.52%\n",
            "iter 46826: loss 1.4130, time 9.39ms, mfu 0.51%\n",
            "iter 46827: loss 1.4136, time 9.45ms, mfu 0.51%\n",
            "iter 46828: loss 1.5270, time 9.63ms, mfu 0.50%\n",
            "iter 46829: loss 1.3969, time 11.96ms, mfu 0.40%\n",
            "iter 46830: loss 1.4015, time 11.81ms, mfu 0.41%\n",
            "iter 46831: loss 1.4087, time 10.11ms, mfu 0.48%\n",
            "iter 46832: loss 1.4768, time 9.62ms, mfu 0.50%\n",
            "iter 46833: loss 1.5098, time 9.27ms, mfu 0.52%\n",
            "iter 46834: loss 1.4501, time 9.40ms, mfu 0.51%\n",
            "iter 46835: loss 1.5624, time 9.47ms, mfu 0.51%\n",
            "iter 46836: loss 1.4598, time 9.14ms, mfu 0.53%\n",
            "iter 46837: loss 1.4564, time 9.28ms, mfu 0.52%\n",
            "iter 46838: loss 1.4584, time 9.48ms, mfu 0.51%\n",
            "iter 46839: loss 1.4122, time 9.23ms, mfu 0.52%\n",
            "iter 46840: loss 1.3790, time 9.19ms, mfu 0.52%\n",
            "iter 46841: loss 1.5371, time 9.33ms, mfu 0.52%\n",
            "iter 46842: loss 1.6120, time 11.37ms, mfu 0.42%\n",
            "iter 46843: loss 1.4273, time 16.29ms, mfu 0.30%\n",
            "iter 46844: loss 1.5158, time 12.46ms, mfu 0.39%\n",
            "iter 46845: loss 1.4574, time 9.53ms, mfu 0.51%\n",
            "iter 46846: loss 1.4841, time 9.34ms, mfu 0.52%\n",
            "iter 46847: loss 1.3605, time 9.49ms, mfu 0.51%\n",
            "iter 46848: loss 1.5473, time 9.44ms, mfu 0.51%\n",
            "iter 46849: loss 1.5292, time 9.41ms, mfu 0.51%\n",
            "iter 46850: loss 1.5208, time 9.26ms, mfu 0.52%\n",
            "iter 46851: loss 1.5424, time 9.58ms, mfu 0.50%\n",
            "iter 46852: loss 1.5538, time 9.22ms, mfu 0.52%\n",
            "iter 46853: loss 1.4591, time 9.22ms, mfu 0.52%\n",
            "iter 46854: loss 1.4657, time 9.57ms, mfu 0.50%\n",
            "iter 46855: loss 1.3798, time 9.27ms, mfu 0.52%\n",
            "iter 46856: loss 1.3633, time 9.12ms, mfu 0.53%\n",
            "iter 46857: loss 1.4513, time 9.22ms, mfu 0.52%\n",
            "iter 46858: loss 1.4895, time 9.30ms, mfu 0.52%\n",
            "iter 46859: loss 1.5531, time 9.14ms, mfu 0.53%\n",
            "iter 46860: loss 1.6516, time 9.17ms, mfu 0.53%\n",
            "iter 46861: loss 1.4957, time 9.40ms, mfu 0.51%\n",
            "iter 46862: loss 1.5232, time 9.22ms, mfu 0.52%\n",
            "iter 46863: loss 1.5510, time 9.50ms, mfu 0.51%\n",
            "iter 46864: loss 1.5357, time 9.66ms, mfu 0.50%\n",
            "iter 46865: loss 1.3814, time 13.21ms, mfu 0.37%\n",
            "iter 46866: loss 1.5278, time 9.46ms, mfu 0.51%\n",
            "iter 46867: loss 1.5605, time 9.34ms, mfu 0.52%\n",
            "iter 46868: loss 1.4397, time 9.44ms, mfu 0.51%\n",
            "iter 46869: loss 1.4764, time 9.28ms, mfu 0.52%\n",
            "iter 46870: loss 1.4625, time 9.25ms, mfu 0.52%\n",
            "iter 46871: loss 1.4543, time 9.43ms, mfu 0.51%\n",
            "iter 46872: loss 1.5811, time 9.81ms, mfu 0.49%\n",
            "iter 46873: loss 1.5528, time 10.27ms, mfu 0.47%\n",
            "iter 46874: loss 1.5108, time 9.17ms, mfu 0.53%\n",
            "iter 46875: loss 1.4899, time 9.38ms, mfu 0.51%\n",
            "iter 46876: loss 1.4902, time 9.15ms, mfu 0.53%\n",
            "iter 46877: loss 1.4382, time 9.18ms, mfu 0.53%\n",
            "iter 46878: loss 1.4021, time 9.46ms, mfu 0.51%\n",
            "iter 46879: loss 1.3704, time 9.28ms, mfu 0.52%\n",
            "iter 46880: loss 1.4524, time 11.93ms, mfu 0.40%\n",
            "iter 46881: loss 1.4117, time 9.74ms, mfu 0.50%\n",
            "iter 46882: loss 1.4950, time 9.47ms, mfu 0.51%\n",
            "iter 46883: loss 1.5438, time 9.23ms, mfu 0.52%\n",
            "iter 46884: loss 1.5506, time 9.44ms, mfu 0.51%\n",
            "iter 46885: loss 1.5123, time 9.67ms, mfu 0.50%\n",
            "iter 46886: loss 1.4693, time 9.18ms, mfu 0.53%\n",
            "iter 46887: loss 1.4512, time 14.40ms, mfu 0.33%\n",
            "iter 46888: loss 1.3475, time 9.64ms, mfu 0.50%\n",
            "iter 46889: loss 1.4352, time 9.44ms, mfu 0.51%\n",
            "iter 46890: loss 1.4086, time 12.19ms, mfu 0.40%\n",
            "iter 46891: loss 1.4969, time 9.47ms, mfu 0.51%\n",
            "iter 46892: loss 1.5292, time 9.11ms, mfu 0.53%\n",
            "iter 46893: loss 1.4667, time 10.63ms, mfu 0.45%\n",
            "iter 46894: loss 1.4741, time 9.30ms, mfu 0.52%\n",
            "iter 46895: loss 1.4928, time 10.47ms, mfu 0.46%\n",
            "iter 46896: loss 1.5164, time 9.47ms, mfu 0.51%\n",
            "iter 46897: loss 1.4063, time 10.14ms, mfu 0.48%\n",
            "iter 46898: loss 1.3156, time 9.71ms, mfu 0.50%\n",
            "iter 46899: loss 1.3667, time 11.66ms, mfu 0.41%\n",
            "iter 46900: loss 1.5065, time 11.54ms, mfu 0.42%\n",
            "iter 46901: loss 1.4910, time 10.03ms, mfu 0.48%\n",
            "iter 46902: loss 1.4336, time 9.57ms, mfu 0.50%\n",
            "iter 46903: loss 1.3925, time 10.34ms, mfu 0.47%\n",
            "iter 46904: loss 1.4219, time 9.55ms, mfu 0.51%\n",
            "iter 46905: loss 1.4105, time 9.57ms, mfu 0.50%\n",
            "iter 46906: loss 1.4646, time 9.46ms, mfu 0.51%\n",
            "iter 46907: loss 1.4331, time 9.67ms, mfu 0.50%\n",
            "iter 46908: loss 1.3447, time 13.23ms, mfu 0.36%\n",
            "iter 46909: loss 1.5001, time 10.06ms, mfu 0.48%\n",
            "iter 46910: loss 1.4662, time 11.23ms, mfu 0.43%\n",
            "iter 46911: loss 1.4420, time 15.03ms, mfu 0.32%\n",
            "iter 46912: loss 1.5027, time 12.40ms, mfu 0.39%\n",
            "iter 46913: loss 1.4365, time 9.79ms, mfu 0.49%\n",
            "iter 46914: loss 1.5790, time 9.41ms, mfu 0.51%\n",
            "iter 46915: loss 1.3811, time 9.64ms, mfu 0.50%\n",
            "iter 46916: loss 1.3751, time 10.01ms, mfu 0.48%\n",
            "iter 46917: loss 1.5436, time 9.76ms, mfu 0.49%\n",
            "iter 46918: loss 1.4866, time 9.46ms, mfu 0.51%\n",
            "iter 46919: loss 1.3731, time 9.48ms, mfu 0.51%\n",
            "iter 46920: loss 1.4332, time 9.52ms, mfu 0.51%\n",
            "iter 46921: loss 1.4578, time 9.35ms, mfu 0.52%\n",
            "iter 46922: loss 1.4882, time 9.93ms, mfu 0.49%\n",
            "iter 46923: loss 1.4402, time 11.00ms, mfu 0.44%\n",
            "iter 46924: loss 1.4150, time 18.67ms, mfu 0.26%\n",
            "iter 46925: loss 1.3752, time 15.11ms, mfu 0.32%\n",
            "iter 46926: loss 1.4177, time 19.74ms, mfu 0.24%\n",
            "iter 46927: loss 1.4730, time 25.13ms, mfu 0.19%\n",
            "iter 46928: loss 1.4693, time 19.97ms, mfu 0.24%\n",
            "iter 46929: loss 1.4925, time 17.21ms, mfu 0.28%\n",
            "iter 46930: loss 1.4489, time 17.85ms, mfu 0.27%\n",
            "iter 46931: loss 1.5360, time 18.81ms, mfu 0.26%\n",
            "iter 46932: loss 1.5330, time 16.28ms, mfu 0.30%\n",
            "iter 46933: loss 1.5377, time 16.56ms, mfu 0.29%\n",
            "iter 46934: loss 1.5942, time 16.83ms, mfu 0.29%\n",
            "iter 46935: loss 1.5255, time 16.69ms, mfu 0.29%\n",
            "iter 46936: loss 1.4274, time 16.02ms, mfu 0.30%\n",
            "iter 46937: loss 1.4239, time 17.46ms, mfu 0.28%\n",
            "iter 46938: loss 1.4392, time 18.30ms, mfu 0.26%\n",
            "iter 46939: loss 1.4852, time 15.28ms, mfu 0.32%\n",
            "iter 46940: loss 1.4544, time 14.38ms, mfu 0.34%\n",
            "iter 46941: loss 1.4107, time 15.57ms, mfu 0.31%\n",
            "iter 46942: loss 1.5270, time 14.94ms, mfu 0.32%\n",
            "iter 46943: loss 1.4545, time 13.54ms, mfu 0.36%\n",
            "iter 46944: loss 1.5604, time 13.85ms, mfu 0.35%\n",
            "iter 46945: loss 1.5465, time 14.52ms, mfu 0.33%\n",
            "iter 46946: loss 1.4259, time 12.34ms, mfu 0.39%\n",
            "iter 46947: loss 1.4026, time 9.42ms, mfu 0.51%\n",
            "iter 46948: loss 1.5665, time 9.62ms, mfu 0.50%\n",
            "iter 46949: loss 1.3584, time 9.60ms, mfu 0.50%\n",
            "iter 46950: loss 1.6323, time 9.24ms, mfu 0.52%\n",
            "iter 46951: loss 1.4765, time 9.64ms, mfu 0.50%\n",
            "iter 46952: loss 1.3863, time 9.61ms, mfu 0.50%\n",
            "iter 46953: loss 1.3951, time 9.28ms, mfu 0.52%\n",
            "iter 46954: loss 1.5075, time 9.52ms, mfu 0.51%\n",
            "iter 46955: loss 1.4490, time 9.69ms, mfu 0.50%\n",
            "iter 46956: loss 1.4525, time 13.81ms, mfu 0.35%\n",
            "iter 46957: loss 1.4351, time 9.58ms, mfu 0.50%\n",
            "iter 46958: loss 1.5278, time 9.48ms, mfu 0.51%\n",
            "iter 46959: loss 1.5301, time 9.34ms, mfu 0.52%\n",
            "iter 46960: loss 1.5658, time 9.19ms, mfu 0.52%\n",
            "iter 46961: loss 1.4025, time 10.53ms, mfu 0.46%\n",
            "iter 46962: loss 1.4449, time 9.27ms, mfu 0.52%\n",
            "iter 46963: loss 1.5221, time 9.65ms, mfu 0.50%\n",
            "iter 46964: loss 1.4376, time 9.29ms, mfu 0.52%\n",
            "iter 46965: loss 1.4801, time 9.48ms, mfu 0.51%\n",
            "iter 46966: loss 1.4497, time 9.31ms, mfu 0.52%\n",
            "iter 46967: loss 1.4748, time 10.45ms, mfu 0.46%\n",
            "iter 46968: loss 1.5704, time 11.42ms, mfu 0.42%\n",
            "iter 46969: loss 1.5582, time 9.99ms, mfu 0.48%\n",
            "iter 46970: loss 1.4684, time 9.21ms, mfu 0.52%\n",
            "iter 46971: loss 1.4455, time 10.00ms, mfu 0.48%\n",
            "iter 46972: loss 1.5087, time 9.22ms, mfu 0.52%\n",
            "iter 46973: loss 1.5885, time 9.37ms, mfu 0.51%\n",
            "iter 46974: loss 1.4387, time 9.40ms, mfu 0.51%\n",
            "iter 46975: loss 1.5268, time 9.59ms, mfu 0.50%\n",
            "iter 46976: loss 1.5853, time 9.65ms, mfu 0.50%\n",
            "iter 46977: loss 1.6247, time 11.78ms, mfu 0.41%\n",
            "iter 46978: loss 1.5246, time 16.22ms, mfu 0.30%\n",
            "iter 46979: loss 1.4162, time 10.66ms, mfu 0.45%\n",
            "iter 46980: loss 1.4077, time 9.38ms, mfu 0.51%\n",
            "iter 46981: loss 1.5212, time 9.73ms, mfu 0.50%\n",
            "iter 46982: loss 1.4315, time 9.88ms, mfu 0.49%\n",
            "iter 46983: loss 1.4788, time 9.39ms, mfu 0.51%\n",
            "iter 46984: loss 1.4619, time 10.88ms, mfu 0.44%\n",
            "iter 46985: loss 1.4059, time 9.47ms, mfu 0.51%\n",
            "iter 46986: loss 1.5442, time 9.50ms, mfu 0.51%\n",
            "iter 46987: loss 1.4969, time 9.74ms, mfu 0.50%\n",
            "iter 46988: loss 1.5797, time 9.61ms, mfu 0.50%\n",
            "iter 46989: loss 1.4548, time 9.50ms, mfu 0.51%\n",
            "iter 46990: loss 1.4434, time 9.63ms, mfu 0.50%\n",
            "iter 46991: loss 1.4436, time 9.51ms, mfu 0.51%\n",
            "iter 46992: loss 1.5468, time 10.34ms, mfu 0.47%\n",
            "iter 46993: loss 1.4525, time 9.68ms, mfu 0.50%\n",
            "iter 46994: loss 1.4831, time 9.67ms, mfu 0.50%\n",
            "iter 46995: loss 1.4161, time 9.48ms, mfu 0.51%\n",
            "iter 46996: loss 1.5167, time 9.31ms, mfu 0.52%\n",
            "iter 46997: loss 1.4621, time 11.36ms, mfu 0.42%\n",
            "iter 46998: loss 1.4425, time 19.14ms, mfu 0.25%\n",
            "iter 46999: loss 1.4055, time 11.82ms, mfu 0.41%\n",
            "step 47000: train loss 1.3549, val loss 1.5210\n",
            "iter 47000: loss 1.4892, time 17.02ms, mfu 0.28%\n",
            "iter 47001: loss 1.4881, time 9.52ms, mfu 0.51%\n",
            "iter 47002: loss 1.3795, time 9.42ms, mfu 0.51%\n",
            "iter 47003: loss 1.4390, time 9.32ms, mfu 0.52%\n",
            "iter 47004: loss 1.4989, time 9.40ms, mfu 0.51%\n",
            "iter 47005: loss 1.4957, time 10.57ms, mfu 0.46%\n",
            "iter 47006: loss 1.5061, time 10.31ms, mfu 0.47%\n",
            "iter 47007: loss 1.3912, time 16.12ms, mfu 0.30%\n",
            "iter 47008: loss 1.4661, time 10.07ms, mfu 0.48%\n",
            "iter 47009: loss 1.5264, time 9.70ms, mfu 0.50%\n",
            "iter 47010: loss 1.4248, time 9.88ms, mfu 0.49%\n",
            "iter 47011: loss 1.3865, time 11.20ms, mfu 0.43%\n",
            "iter 47012: loss 1.4217, time 12.40ms, mfu 0.39%\n",
            "iter 47013: loss 1.4653, time 9.81ms, mfu 0.49%\n",
            "iter 47014: loss 1.3790, time 9.68ms, mfu 0.50%\n",
            "iter 47015: loss 1.4374, time 9.51ms, mfu 0.51%\n",
            "iter 47016: loss 1.5962, time 9.43ms, mfu 0.51%\n",
            "iter 47017: loss 1.5850, time 13.55ms, mfu 0.36%\n",
            "iter 47018: loss 1.5057, time 11.83ms, mfu 0.41%\n",
            "iter 47019: loss 1.4859, time 9.46ms, mfu 0.51%\n",
            "iter 47020: loss 1.5295, time 9.53ms, mfu 0.51%\n",
            "iter 47021: loss 1.5006, time 9.46ms, mfu 0.51%\n",
            "iter 47022: loss 1.4236, time 9.35ms, mfu 0.52%\n",
            "iter 47023: loss 1.4346, time 9.68ms, mfu 0.50%\n",
            "iter 47024: loss 1.4698, time 9.68ms, mfu 0.50%\n",
            "iter 47025: loss 1.4505, time 10.48ms, mfu 0.46%\n",
            "iter 47026: loss 1.4143, time 9.52ms, mfu 0.51%\n",
            "iter 47027: loss 1.4539, time 9.30ms, mfu 0.52%\n",
            "iter 47028: loss 1.4516, time 9.23ms, mfu 0.52%\n",
            "iter 47029: loss 1.3931, time 9.34ms, mfu 0.52%\n",
            "iter 47030: loss 1.6613, time 17.53ms, mfu 0.28%\n",
            "iter 47031: loss 1.4471, time 20.73ms, mfu 0.23%\n",
            "iter 47032: loss 1.5122, time 17.03ms, mfu 0.28%\n",
            "iter 47033: loss 1.6690, time 16.72ms, mfu 0.29%\n",
            "iter 47034: loss 1.4881, time 24.74ms, mfu 0.19%\n",
            "iter 47035: loss 1.4188, time 17.38ms, mfu 0.28%\n",
            "iter 47036: loss 1.5100, time 16.11ms, mfu 0.30%\n",
            "iter 47037: loss 1.5607, time 16.62ms, mfu 0.29%\n",
            "iter 47038: loss 1.4847, time 16.75ms, mfu 0.29%\n",
            "iter 47039: loss 1.4608, time 16.60ms, mfu 0.29%\n",
            "iter 47040: loss 1.5524, time 16.10ms, mfu 0.30%\n",
            "iter 47041: loss 1.4644, time 13.47ms, mfu 0.36%\n",
            "iter 47042: loss 1.4986, time 14.96ms, mfu 0.32%\n",
            "iter 47043: loss 1.5409, time 15.72ms, mfu 0.31%\n",
            "iter 47044: loss 1.4479, time 16.57ms, mfu 0.29%\n",
            "iter 47045: loss 1.4434, time 13.89ms, mfu 0.35%\n",
            "iter 47046: loss 1.4920, time 14.55ms, mfu 0.33%\n",
            "iter 47047: loss 1.5374, time 16.50ms, mfu 0.29%\n",
            "iter 47048: loss 1.5454, time 19.12ms, mfu 0.25%\n",
            "iter 47049: loss 1.4861, time 16.77ms, mfu 0.29%\n",
            "iter 47050: loss 1.4001, time 17.19ms, mfu 0.28%\n",
            "iter 47051: loss 1.4473, time 15.28ms, mfu 0.32%\n",
            "iter 47052: loss 1.5251, time 16.71ms, mfu 0.29%\n",
            "iter 47053: loss 1.3852, time 16.90ms, mfu 0.29%\n",
            "iter 47054: loss 1.4778, time 19.64ms, mfu 0.25%\n",
            "iter 47055: loss 1.4537, time 18.65ms, mfu 0.26%\n",
            "iter 47056: loss 1.4656, time 17.69ms, mfu 0.27%\n",
            "iter 47057: loss 1.5359, time 16.68ms, mfu 0.29%\n",
            "iter 47058: loss 1.4183, time 16.17ms, mfu 0.30%\n",
            "iter 47059: loss 1.5448, time 22.40ms, mfu 0.22%\n",
            "iter 47060: loss 1.3817, time 24.29ms, mfu 0.20%\n",
            "iter 47061: loss 1.5913, time 17.25ms, mfu 0.28%\n",
            "iter 47062: loss 1.3686, time 16.01ms, mfu 0.30%\n",
            "iter 47063: loss 1.4543, time 16.49ms, mfu 0.29%\n",
            "iter 47064: loss 1.4751, time 18.74ms, mfu 0.26%\n",
            "iter 47065: loss 1.3946, time 16.97ms, mfu 0.28%\n",
            "iter 47066: loss 1.4123, time 16.92ms, mfu 0.29%\n",
            "iter 47067: loss 1.4508, time 16.39ms, mfu 0.29%\n",
            "iter 47068: loss 1.5543, time 16.11ms, mfu 0.30%\n",
            "iter 47069: loss 1.4678, time 28.29ms, mfu 0.17%\n",
            "iter 47070: loss 1.5066, time 33.15ms, mfu 0.15%\n",
            "iter 47071: loss 1.3801, time 17.96ms, mfu 0.27%\n",
            "iter 47072: loss 1.5097, time 19.03ms, mfu 0.25%\n",
            "iter 47073: loss 1.3697, time 16.00ms, mfu 0.30%\n",
            "iter 47074: loss 1.4537, time 16.30ms, mfu 0.30%\n",
            "iter 47075: loss 1.5135, time 17.54ms, mfu 0.27%\n",
            "iter 47076: loss 1.4370, time 18.34ms, mfu 0.26%\n",
            "iter 47077: loss 1.4006, time 9.88ms, mfu 0.49%\n",
            "iter 47078: loss 1.5446, time 9.76ms, mfu 0.49%\n",
            "iter 47079: loss 1.4965, time 9.46ms, mfu 0.51%\n",
            "iter 47080: loss 1.4352, time 9.44ms, mfu 0.51%\n",
            "iter 47081: loss 1.5395, time 9.28ms, mfu 0.52%\n",
            "iter 47082: loss 1.6504, time 9.30ms, mfu 0.52%\n",
            "iter 47083: loss 1.5047, time 9.36ms, mfu 0.52%\n",
            "iter 47084: loss 1.6165, time 9.37ms, mfu 0.51%\n",
            "iter 47085: loss 1.3570, time 9.60ms, mfu 0.50%\n",
            "iter 47086: loss 1.5764, time 9.46ms, mfu 0.51%\n",
            "iter 47087: loss 1.4560, time 14.67ms, mfu 0.33%\n",
            "iter 47088: loss 1.4619, time 10.65ms, mfu 0.45%\n",
            "iter 47089: loss 1.6301, time 9.72ms, mfu 0.50%\n",
            "iter 47090: loss 1.5339, time 9.52ms, mfu 0.51%\n",
            "iter 47091: loss 1.4595, time 9.33ms, mfu 0.52%\n",
            "iter 47092: loss 1.4765, time 9.41ms, mfu 0.51%\n",
            "iter 47093: loss 1.5381, time 9.32ms, mfu 0.52%\n",
            "iter 47094: loss 1.5054, time 9.62ms, mfu 0.50%\n",
            "iter 47095: loss 1.4670, time 9.49ms, mfu 0.51%\n",
            "iter 47096: loss 1.4352, time 9.55ms, mfu 0.50%\n",
            "iter 47097: loss 1.3075, time 9.34ms, mfu 0.52%\n",
            "iter 47098: loss 1.4414, time 9.37ms, mfu 0.51%\n",
            "iter 47099: loss 1.5319, time 9.65ms, mfu 0.50%\n",
            "iter 47100: loss 1.5264, time 9.66ms, mfu 0.50%\n",
            "iter 47101: loss 1.4691, time 9.25ms, mfu 0.52%\n",
            "iter 47102: loss 1.3843, time 9.30ms, mfu 0.52%\n",
            "iter 47103: loss 1.5091, time 9.51ms, mfu 0.51%\n",
            "iter 47104: loss 1.4794, time 9.31ms, mfu 0.52%\n",
            "iter 47105: loss 1.5087, time 9.35ms, mfu 0.52%\n",
            "iter 47106: loss 1.5056, time 9.55ms, mfu 0.51%\n",
            "iter 47107: loss 1.4650, time 9.32ms, mfu 0.52%\n",
            "iter 47108: loss 1.6495, time 9.34ms, mfu 0.52%\n",
            "iter 47109: loss 1.3919, time 16.41ms, mfu 0.29%\n",
            "iter 47110: loss 1.5058, time 12.67ms, mfu 0.38%\n",
            "iter 47111: loss 1.5548, time 9.42ms, mfu 0.51%\n",
            "iter 47112: loss 1.2145, time 10.10ms, mfu 0.48%\n",
            "iter 47113: loss 1.3772, time 9.80ms, mfu 0.49%\n",
            "iter 47114: loss 1.3826, time 9.56ms, mfu 0.50%\n",
            "iter 47115: loss 1.3262, time 10.23ms, mfu 0.47%\n",
            "iter 47116: loss 1.4938, time 9.42ms, mfu 0.51%\n",
            "iter 47117: loss 1.4379, time 9.82ms, mfu 0.49%\n",
            "iter 47118: loss 1.4795, time 9.60ms, mfu 0.50%\n",
            "iter 47119: loss 1.4142, time 9.83ms, mfu 0.49%\n",
            "iter 47120: loss 1.3722, time 9.73ms, mfu 0.50%\n",
            "iter 47121: loss 1.3795, time 9.38ms, mfu 0.51%\n",
            "iter 47122: loss 1.5434, time 9.28ms, mfu 0.52%\n",
            "iter 47123: loss 1.5053, time 9.31ms, mfu 0.52%\n",
            "iter 47124: loss 1.4909, time 9.51ms, mfu 0.51%\n",
            "iter 47125: loss 1.5180, time 9.36ms, mfu 0.52%\n",
            "iter 47126: loss 1.4278, time 9.41ms, mfu 0.51%\n",
            "iter 47127: loss 1.4623, time 9.51ms, mfu 0.51%\n",
            "iter 47128: loss 1.6322, time 9.48ms, mfu 0.51%\n",
            "iter 47129: loss 1.4093, time 9.36ms, mfu 0.52%\n",
            "iter 47130: loss 1.5257, time 13.26ms, mfu 0.36%\n",
            "iter 47131: loss 1.4717, time 9.98ms, mfu 0.48%\n",
            "iter 47132: loss 1.5082, time 9.31ms, mfu 0.52%\n",
            "iter 47133: loss 1.5577, time 9.31ms, mfu 0.52%\n",
            "iter 47134: loss 1.4888, time 9.38ms, mfu 0.51%\n",
            "iter 47135: loss 1.5965, time 9.40ms, mfu 0.51%\n",
            "iter 47136: loss 1.5570, time 9.51ms, mfu 0.51%\n",
            "iter 47137: loss 1.4662, time 9.63ms, mfu 0.50%\n",
            "iter 47138: loss 1.5340, time 9.44ms, mfu 0.51%\n",
            "iter 47139: loss 1.4682, time 10.91ms, mfu 0.44%\n",
            "iter 47140: loss 1.5200, time 9.47ms, mfu 0.51%\n",
            "iter 47141: loss 1.4872, time 10.63ms, mfu 0.45%\n",
            "iter 47142: loss 1.5622, time 9.51ms, mfu 0.51%\n",
            "iter 47143: loss 1.3744, time 10.14ms, mfu 0.48%\n",
            "iter 47144: loss 1.5718, time 9.35ms, mfu 0.52%\n",
            "iter 47145: loss 1.4864, time 9.60ms, mfu 0.50%\n",
            "iter 47146: loss 1.4618, time 9.40ms, mfu 0.51%\n",
            "iter 47147: loss 1.4774, time 11.22ms, mfu 0.43%\n",
            "iter 47148: loss 1.4719, time 9.63ms, mfu 0.50%\n",
            "iter 47149: loss 1.4283, time 9.68ms, mfu 0.50%\n",
            "iter 47150: loss 1.4053, time 9.50ms, mfu 0.51%\n",
            "iter 47151: loss 1.4293, time 12.00ms, mfu 0.40%\n",
            "iter 47152: loss 1.3811, time 11.35ms, mfu 0.42%\n",
            "iter 47153: loss 1.4735, time 9.67ms, mfu 0.50%\n",
            "iter 47154: loss 1.5464, time 9.61ms, mfu 0.50%\n",
            "iter 47155: loss 1.5800, time 9.67ms, mfu 0.50%\n",
            "iter 47156: loss 1.5211, time 9.50ms, mfu 0.51%\n",
            "iter 47157: loss 1.5611, time 9.51ms, mfu 0.51%\n",
            "iter 47158: loss 1.4951, time 9.86ms, mfu 0.49%\n",
            "iter 47159: loss 1.4363, time 9.91ms, mfu 0.49%\n",
            "iter 47160: loss 1.6018, time 9.95ms, mfu 0.48%\n",
            "iter 47161: loss 1.4948, time 9.92ms, mfu 0.49%\n",
            "iter 47162: loss 1.6106, time 10.34ms, mfu 0.47%\n",
            "iter 47163: loss 1.5136, time 10.30ms, mfu 0.47%\n",
            "iter 47164: loss 1.5001, time 9.76ms, mfu 0.49%\n",
            "iter 47165: loss 1.4584, time 11.90ms, mfu 0.41%\n",
            "iter 47166: loss 1.5336, time 15.89ms, mfu 0.30%\n",
            "iter 47167: loss 1.5194, time 12.73ms, mfu 0.38%\n",
            "iter 47168: loss 1.5761, time 10.12ms, mfu 0.48%\n",
            "iter 47169: loss 1.4900, time 11.21ms, mfu 0.43%\n",
            "iter 47170: loss 1.3924, time 9.53ms, mfu 0.51%\n",
            "iter 47171: loss 1.4324, time 13.86ms, mfu 0.35%\n",
            "iter 47172: loss 1.5228, time 9.58ms, mfu 0.50%\n",
            "iter 47173: loss 1.4481, time 10.71ms, mfu 0.45%\n",
            "iter 47174: loss 1.4160, time 12.33ms, mfu 0.39%\n",
            "iter 47175: loss 1.4406, time 11.00ms, mfu 0.44%\n",
            "iter 47176: loss 1.4160, time 9.40ms, mfu 0.51%\n",
            "iter 47177: loss 1.5015, time 9.64ms, mfu 0.50%\n",
            "iter 47178: loss 1.5520, time 9.49ms, mfu 0.51%\n",
            "iter 47179: loss 1.4989, time 9.43ms, mfu 0.51%\n",
            "iter 47180: loss 1.4320, time 9.39ms, mfu 0.51%\n",
            "iter 47181: loss 1.5913, time 11.22ms, mfu 0.43%\n",
            "iter 47182: loss 1.5495, time 10.28ms, mfu 0.47%\n",
            "iter 47183: loss 1.4268, time 9.52ms, mfu 0.51%\n",
            "iter 47184: loss 1.4460, time 9.29ms, mfu 0.52%\n",
            "iter 47185: loss 1.5036, time 9.49ms, mfu 0.51%\n",
            "iter 47186: loss 1.4585, time 9.41ms, mfu 0.51%\n",
            "iter 47187: loss 1.3789, time 9.49ms, mfu 0.51%\n",
            "iter 47188: loss 1.4370, time 10.60ms, mfu 0.46%\n",
            "iter 47189: loss 1.3760, time 9.63ms, mfu 0.50%\n",
            "iter 47190: loss 1.4130, time 9.41ms, mfu 0.51%\n",
            "iter 47191: loss 1.4272, time 9.84ms, mfu 0.49%\n",
            "iter 47192: loss 1.5916, time 14.60ms, mfu 0.33%\n",
            "iter 47193: loss 1.4190, time 9.85ms, mfu 0.49%\n",
            "iter 47194: loss 1.4396, time 9.48ms, mfu 0.51%\n",
            "iter 47195: loss 1.4699, time 9.57ms, mfu 0.50%\n",
            "iter 47196: loss 1.3579, time 9.37ms, mfu 0.51%\n",
            "iter 47197: loss 1.6122, time 9.32ms, mfu 0.52%\n",
            "iter 47198: loss 1.3800, time 9.70ms, mfu 0.50%\n",
            "iter 47199: loss 1.4630, time 9.49ms, mfu 0.51%\n",
            "iter 47200: loss 1.5097, time 9.42ms, mfu 0.51%\n",
            "iter 47201: loss 1.5419, time 9.42ms, mfu 0.51%\n",
            "iter 47202: loss 1.3978, time 9.61ms, mfu 0.50%\n",
            "iter 47203: loss 1.4910, time 9.43ms, mfu 0.51%\n",
            "iter 47204: loss 1.5637, time 9.32ms, mfu 0.52%\n",
            "iter 47205: loss 1.3736, time 9.70ms, mfu 0.50%\n",
            "iter 47206: loss 1.3811, time 9.35ms, mfu 0.52%\n",
            "iter 47207: loss 1.4824, time 9.31ms, mfu 0.52%\n",
            "iter 47208: loss 1.4362, time 9.73ms, mfu 0.50%\n",
            "iter 47209: loss 1.4810, time 9.61ms, mfu 0.50%\n",
            "iter 47210: loss 1.4599, time 9.54ms, mfu 0.51%\n",
            "iter 47211: loss 1.3786, time 9.64ms, mfu 0.50%\n",
            "iter 47212: loss 1.5525, time 9.92ms, mfu 0.49%\n",
            "iter 47213: loss 1.5357, time 9.45ms, mfu 0.51%\n",
            "iter 47214: loss 1.4445, time 14.03ms, mfu 0.34%\n",
            "iter 47215: loss 1.4050, time 9.70ms, mfu 0.50%\n",
            "iter 47216: loss 1.5223, time 9.30ms, mfu 0.52%\n",
            "iter 47217: loss 1.4561, time 9.38ms, mfu 0.51%\n",
            "iter 47218: loss 1.5494, time 9.44ms, mfu 0.51%\n",
            "iter 47219: loss 1.5557, time 9.52ms, mfu 0.51%\n",
            "iter 47220: loss 1.4687, time 13.14ms, mfu 0.37%\n",
            "iter 47221: loss 1.5081, time 10.21ms, mfu 0.47%\n",
            "iter 47222: loss 1.6040, time 9.39ms, mfu 0.51%\n",
            "iter 47223: loss 1.4271, time 9.35ms, mfu 0.52%\n",
            "iter 47224: loss 1.5068, time 9.57ms, mfu 0.50%\n",
            "iter 47225: loss 1.4664, time 9.42ms, mfu 0.51%\n",
            "iter 47226: loss 1.3106, time 9.42ms, mfu 0.51%\n",
            "iter 47227: loss 1.4302, time 9.54ms, mfu 0.51%\n",
            "iter 47228: loss 1.3800, time 11.24ms, mfu 0.43%\n",
            "iter 47229: loss 1.5309, time 9.57ms, mfu 0.50%\n",
            "iter 47230: loss 1.4593, time 9.40ms, mfu 0.51%\n",
            "iter 47231: loss 1.5257, time 9.52ms, mfu 0.51%\n",
            "iter 47232: loss 1.5508, time 9.45ms, mfu 0.51%\n",
            "iter 47233: loss 1.4111, time 10.34ms, mfu 0.47%\n",
            "iter 47234: loss 1.5006, time 9.44ms, mfu 0.51%\n",
            "iter 47235: loss 1.4589, time 12.60ms, mfu 0.38%\n",
            "iter 47236: loss 1.3679, time 10.60ms, mfu 0.46%\n",
            "iter 47237: loss 1.4865, time 9.47ms, mfu 0.51%\n",
            "iter 47238: loss 1.5197, time 9.57ms, mfu 0.50%\n",
            "iter 47239: loss 1.4616, time 9.49ms, mfu 0.51%\n",
            "iter 47240: loss 1.5436, time 10.51ms, mfu 0.46%\n",
            "iter 47241: loss 1.5406, time 9.56ms, mfu 0.50%\n",
            "iter 47242: loss 1.5452, time 9.59ms, mfu 0.50%\n",
            "iter 47243: loss 1.4814, time 9.35ms, mfu 0.52%\n",
            "iter 47244: loss 1.5125, time 9.45ms, mfu 0.51%\n",
            "iter 47245: loss 1.4753, time 9.84ms, mfu 0.49%\n",
            "iter 47246: loss 1.3785, time 11.08ms, mfu 0.44%\n",
            "iter 47247: loss 1.5156, time 9.72ms, mfu 0.50%\n",
            "iter 47248: loss 1.5133, time 9.56ms, mfu 0.50%\n",
            "iter 47249: loss 1.4899, time 9.94ms, mfu 0.49%\n",
            "iter 47250: loss 1.4148, time 9.63ms, mfu 0.50%\n",
            "iter 47251: loss 1.4528, time 9.53ms, mfu 0.51%\n",
            "iter 47252: loss 1.4109, time 9.42ms, mfu 0.51%\n",
            "iter 47253: loss 1.4524, time 9.45ms, mfu 0.51%\n",
            "iter 47254: loss 1.4376, time 9.54ms, mfu 0.51%\n",
            "iter 47255: loss 1.4850, time 9.68ms, mfu 0.50%\n",
            "iter 47256: loss 1.4849, time 11.36ms, mfu 0.42%\n",
            "iter 47257: loss 1.5886, time 10.89ms, mfu 0.44%\n",
            "iter 47258: loss 1.4293, time 9.33ms, mfu 0.52%\n",
            "iter 47259: loss 1.4512, time 9.71ms, mfu 0.50%\n",
            "iter 47260: loss 1.3871, time 9.46ms, mfu 0.51%\n",
            "iter 47261: loss 1.5296, time 11.18ms, mfu 0.43%\n",
            "iter 47262: loss 1.5018, time 11.15ms, mfu 0.43%\n",
            "iter 47263: loss 1.4107, time 9.37ms, mfu 0.51%\n",
            "iter 47264: loss 1.4352, time 9.57ms, mfu 0.50%\n",
            "iter 47265: loss 1.4257, time 9.49ms, mfu 0.51%\n",
            "iter 47266: loss 1.4204, time 9.29ms, mfu 0.52%\n",
            "iter 47267: loss 1.5448, time 11.87ms, mfu 0.41%\n",
            "iter 47268: loss 1.4534, time 15.65ms, mfu 0.31%\n",
            "iter 47269: loss 1.4652, time 13.34ms, mfu 0.36%\n",
            "iter 47270: loss 1.4197, time 10.15ms, mfu 0.47%\n",
            "iter 47271: loss 1.5426, time 9.60ms, mfu 0.50%\n",
            "iter 47272: loss 1.4073, time 9.57ms, mfu 0.50%\n",
            "iter 47273: loss 1.4414, time 11.50ms, mfu 0.42%\n",
            "iter 47274: loss 1.4877, time 12.05ms, mfu 0.40%\n",
            "iter 47275: loss 1.5595, time 9.55ms, mfu 0.51%\n",
            "iter 47276: loss 1.5010, time 13.29ms, mfu 0.36%\n",
            "iter 47277: loss 1.5624, time 9.74ms, mfu 0.50%\n",
            "iter 47278: loss 1.5505, time 9.41ms, mfu 0.51%\n",
            "iter 47279: loss 1.5252, time 9.39ms, mfu 0.51%\n",
            "iter 47280: loss 1.5266, time 9.37ms, mfu 0.51%\n",
            "iter 47281: loss 1.5368, time 9.33ms, mfu 0.52%\n",
            "iter 47282: loss 1.4419, time 9.45ms, mfu 0.51%\n",
            "iter 47283: loss 1.3943, time 9.34ms, mfu 0.52%\n",
            "iter 47284: loss 1.4431, time 9.46ms, mfu 0.51%\n",
            "iter 47285: loss 1.5135, time 9.73ms, mfu 0.50%\n",
            "iter 47286: loss 1.5177, time 9.57ms, mfu 0.50%\n",
            "iter 47287: loss 1.3898, time 10.14ms, mfu 0.48%\n",
            "iter 47288: loss 1.3440, time 9.99ms, mfu 0.48%\n",
            "iter 47289: loss 1.4870, time 9.36ms, mfu 0.52%\n",
            "iter 47290: loss 1.4337, time 9.45ms, mfu 0.51%\n",
            "iter 47291: loss 1.3851, time 9.30ms, mfu 0.52%\n",
            "iter 47292: loss 1.4643, time 9.28ms, mfu 0.52%\n",
            "iter 47293: loss 1.5729, time 9.30ms, mfu 0.52%\n",
            "iter 47294: loss 1.5228, time 9.38ms, mfu 0.51%\n",
            "iter 47295: loss 1.4669, time 9.85ms, mfu 0.49%\n",
            "iter 47296: loss 1.5574, time 9.45ms, mfu 0.51%\n",
            "iter 47297: loss 1.3774, time 9.45ms, mfu 0.51%\n",
            "iter 47298: loss 1.4585, time 14.91ms, mfu 0.32%\n",
            "iter 47299: loss 1.3867, time 9.76ms, mfu 0.49%\n",
            "iter 47300: loss 1.5839, time 9.31ms, mfu 0.52%\n",
            "iter 47301: loss 1.5107, time 9.34ms, mfu 0.52%\n",
            "iter 47302: loss 1.4122, time 9.33ms, mfu 0.52%\n",
            "iter 47303: loss 1.3376, time 9.43ms, mfu 0.51%\n",
            "iter 47304: loss 1.5183, time 9.39ms, mfu 0.51%\n",
            "iter 47305: loss 1.4621, time 9.58ms, mfu 0.50%\n",
            "iter 47306: loss 1.5491, time 9.27ms, mfu 0.52%\n",
            "iter 47307: loss 1.4665, time 9.29ms, mfu 0.52%\n",
            "iter 47308: loss 1.5456, time 9.45ms, mfu 0.51%\n",
            "iter 47309: loss 1.5009, time 9.77ms, mfu 0.49%\n",
            "iter 47310: loss 1.5124, time 9.50ms, mfu 0.51%\n",
            "iter 47311: loss 1.5672, time 9.36ms, mfu 0.52%\n",
            "iter 47312: loss 1.3980, time 9.84ms, mfu 0.49%\n",
            "iter 47313: loss 1.4711, time 12.54ms, mfu 0.38%\n",
            "iter 47314: loss 1.4110, time 9.92ms, mfu 0.49%\n",
            "iter 47315: loss 1.4698, time 9.76ms, mfu 0.49%\n",
            "iter 47316: loss 1.3958, time 9.37ms, mfu 0.51%\n",
            "iter 47317: loss 1.4953, time 9.36ms, mfu 0.52%\n",
            "iter 47318: loss 1.5145, time 9.60ms, mfu 0.50%\n",
            "iter 47319: loss 1.4524, time 12.21ms, mfu 0.40%\n",
            "iter 47320: loss 1.4551, time 10.84ms, mfu 0.44%\n",
            "iter 47321: loss 1.4005, time 9.61ms, mfu 0.50%\n",
            "iter 47322: loss 1.5511, time 9.32ms, mfu 0.52%\n",
            "iter 47323: loss 1.3845, time 9.51ms, mfu 0.51%\n",
            "iter 47324: loss 1.4111, time 9.73ms, mfu 0.50%\n",
            "iter 47325: loss 1.5787, time 9.38ms, mfu 0.51%\n",
            "iter 47326: loss 1.3184, time 9.46ms, mfu 0.51%\n",
            "iter 47327: loss 1.5044, time 9.43ms, mfu 0.51%\n",
            "iter 47328: loss 1.3724, time 9.58ms, mfu 0.50%\n",
            "iter 47329: loss 1.3004, time 9.74ms, mfu 0.50%\n",
            "iter 47330: loss 1.5566, time 9.35ms, mfu 0.52%\n",
            "iter 47331: loss 1.4947, time 9.30ms, mfu 0.52%\n",
            "iter 47332: loss 1.4214, time 9.44ms, mfu 0.51%\n",
            "iter 47333: loss 1.5568, time 9.35ms, mfu 0.52%\n",
            "iter 47334: loss 1.4483, time 9.49ms, mfu 0.51%\n",
            "iter 47335: loss 1.4464, time 9.41ms, mfu 0.51%\n",
            "iter 47336: loss 1.4074, time 9.37ms, mfu 0.51%\n",
            "iter 47337: loss 1.4588, time 9.68ms, mfu 0.50%\n",
            "iter 47338: loss 1.5693, time 9.57ms, mfu 0.50%\n",
            "iter 47339: loss 1.5077, time 9.51ms, mfu 0.51%\n",
            "iter 47340: loss 1.5537, time 9.49ms, mfu 0.51%\n",
            "iter 47341: loss 1.4526, time 16.32ms, mfu 0.30%\n",
            "iter 47342: loss 1.4460, time 9.72ms, mfu 0.50%\n",
            "iter 47343: loss 1.4553, time 9.64ms, mfu 0.50%\n",
            "iter 47344: loss 1.5052, time 9.56ms, mfu 0.50%\n",
            "iter 47345: loss 1.4183, time 9.53ms, mfu 0.51%\n",
            "iter 47346: loss 1.4950, time 9.69ms, mfu 0.50%\n",
            "iter 47347: loss 1.4912, time 11.27ms, mfu 0.43%\n",
            "iter 47348: loss 1.4672, time 11.03ms, mfu 0.44%\n",
            "iter 47349: loss 1.3729, time 10.89ms, mfu 0.44%\n",
            "iter 47350: loss 1.4630, time 9.38ms, mfu 0.51%\n",
            "iter 47351: loss 1.4132, time 9.45ms, mfu 0.51%\n",
            "iter 47352: loss 1.4645, time 9.44ms, mfu 0.51%\n",
            "iter 47353: loss 1.5603, time 11.24ms, mfu 0.43%\n",
            "iter 47354: loss 1.4671, time 9.75ms, mfu 0.49%\n",
            "iter 47355: loss 1.4638, time 9.63ms, mfu 0.50%\n",
            "iter 47356: loss 1.5122, time 9.32ms, mfu 0.52%\n",
            "iter 47357: loss 1.4927, time 9.37ms, mfu 0.51%\n",
            "iter 47358: loss 1.4658, time 9.48ms, mfu 0.51%\n",
            "iter 47359: loss 1.4114, time 9.62ms, mfu 0.50%\n",
            "iter 47360: loss 1.4578, time 9.23ms, mfu 0.52%\n",
            "iter 47361: loss 1.4483, time 9.66ms, mfu 0.50%\n",
            "iter 47362: loss 1.4883, time 13.63ms, mfu 0.35%\n",
            "iter 47363: loss 1.5093, time 9.77ms, mfu 0.49%\n",
            "iter 47364: loss 1.4569, time 10.39ms, mfu 0.46%\n",
            "iter 47365: loss 1.4758, time 9.60ms, mfu 0.50%\n",
            "iter 47366: loss 1.5021, time 9.44ms, mfu 0.51%\n",
            "iter 47367: loss 1.4330, time 9.54ms, mfu 0.51%\n",
            "iter 47368: loss 1.6223, time 9.58ms, mfu 0.50%\n",
            "iter 47369: loss 1.5015, time 11.19ms, mfu 0.43%\n",
            "iter 47370: loss 1.5031, time 16.64ms, mfu 0.29%\n",
            "iter 47371: loss 1.4729, time 12.55ms, mfu 0.38%\n",
            "iter 47372: loss 1.4062, time 9.53ms, mfu 0.51%\n",
            "iter 47373: loss 1.3754, time 12.27ms, mfu 0.39%\n",
            "iter 47374: loss 1.5234, time 12.23ms, mfu 0.39%\n",
            "iter 47375: loss 1.4942, time 9.52ms, mfu 0.51%\n",
            "iter 47376: loss 1.3874, time 9.34ms, mfu 0.52%\n",
            "iter 47377: loss 1.4486, time 9.37ms, mfu 0.51%\n",
            "iter 47378: loss 1.5310, time 9.88ms, mfu 0.49%\n",
            "iter 47379: loss 1.3827, time 9.40ms, mfu 0.51%\n",
            "iter 47380: loss 1.3347, time 9.44ms, mfu 0.51%\n",
            "iter 47381: loss 1.4150, time 9.38ms, mfu 0.51%\n",
            "iter 47382: loss 1.4848, time 13.99ms, mfu 0.34%\n",
            "iter 47383: loss 1.4936, time 9.67ms, mfu 0.50%\n",
            "iter 47384: loss 1.5541, time 9.60ms, mfu 0.50%\n",
            "iter 47385: loss 1.5421, time 9.33ms, mfu 0.52%\n",
            "iter 47386: loss 1.5937, time 9.42ms, mfu 0.51%\n",
            "iter 47387: loss 1.5801, time 10.23ms, mfu 0.47%\n",
            "iter 47388: loss 1.4099, time 9.73ms, mfu 0.50%\n",
            "iter 47389: loss 1.4980, time 11.40ms, mfu 0.42%\n",
            "iter 47390: loss 1.3240, time 10.55ms, mfu 0.46%\n",
            "iter 47391: loss 1.4387, time 9.41ms, mfu 0.51%\n",
            "iter 47392: loss 1.4913, time 9.31ms, mfu 0.52%\n",
            "iter 47393: loss 1.4404, time 9.60ms, mfu 0.50%\n",
            "iter 47394: loss 1.2906, time 9.66ms, mfu 0.50%\n",
            "iter 47395: loss 1.4343, time 9.53ms, mfu 0.51%\n",
            "iter 47396: loss 1.5505, time 9.47ms, mfu 0.51%\n",
            "iter 47397: loss 1.4667, time 9.78ms, mfu 0.49%\n",
            "iter 47398: loss 1.4378, time 9.42ms, mfu 0.51%\n",
            "iter 47399: loss 1.3693, time 9.31ms, mfu 0.52%\n",
            "iter 47400: loss 1.5354, time 10.12ms, mfu 0.48%\n",
            "iter 47401: loss 1.5574, time 9.31ms, mfu 0.52%\n",
            "iter 47402: loss 1.4459, time 9.35ms, mfu 0.52%\n",
            "iter 47403: loss 1.4468, time 12.59ms, mfu 0.38%\n",
            "iter 47404: loss 1.4410, time 11.01ms, mfu 0.44%\n",
            "iter 47405: loss 1.3793, time 9.55ms, mfu 0.51%\n",
            "iter 47406: loss 1.5167, time 9.32ms, mfu 0.52%\n",
            "iter 47407: loss 1.3481, time 9.60ms, mfu 0.50%\n",
            "iter 47408: loss 1.4111, time 9.95ms, mfu 0.48%\n",
            "iter 47409: loss 1.5307, time 9.35ms, mfu 0.52%\n",
            "iter 47410: loss 1.5281, time 9.60ms, mfu 0.50%\n",
            "iter 47411: loss 1.5619, time 9.91ms, mfu 0.49%\n",
            "iter 47412: loss 1.4900, time 9.53ms, mfu 0.51%\n",
            "iter 47413: loss 1.5202, time 9.51ms, mfu 0.51%\n",
            "iter 47414: loss 1.4511, time 9.39ms, mfu 0.51%\n",
            "iter 47415: loss 1.4462, time 9.32ms, mfu 0.52%\n",
            "iter 47416: loss 1.4461, time 9.37ms, mfu 0.51%\n",
            "iter 47417: loss 1.4844, time 9.29ms, mfu 0.52%\n",
            "iter 47418: loss 1.5584, time 9.49ms, mfu 0.51%\n",
            "iter 47419: loss 1.5311, time 9.37ms, mfu 0.51%\n",
            "iter 47420: loss 1.4720, time 9.30ms, mfu 0.52%\n",
            "iter 47421: loss 1.4837, time 9.56ms, mfu 0.50%\n",
            "iter 47422: loss 1.5941, time 9.43ms, mfu 0.51%\n",
            "iter 47423: loss 1.5044, time 9.50ms, mfu 0.51%\n",
            "iter 47424: loss 1.4813, time 9.75ms, mfu 0.49%\n",
            "iter 47425: loss 1.5084, time 14.52ms, mfu 0.33%\n",
            "iter 47426: loss 1.4125, time 9.64ms, mfu 0.50%\n",
            "iter 47427: loss 1.4366, time 9.58ms, mfu 0.50%\n",
            "iter 47428: loss 1.4110, time 9.41ms, mfu 0.51%\n",
            "iter 47429: loss 1.4546, time 9.53ms, mfu 0.51%\n",
            "iter 47430: loss 1.4593, time 12.45ms, mfu 0.39%\n",
            "iter 47431: loss 1.4740, time 9.33ms, mfu 0.52%\n",
            "iter 47432: loss 1.6259, time 9.32ms, mfu 0.52%\n",
            "iter 47433: loss 1.5216, time 9.50ms, mfu 0.51%\n",
            "iter 47434: loss 1.5423, time 9.38ms, mfu 0.51%\n",
            "iter 47435: loss 1.4335, time 9.41ms, mfu 0.51%\n",
            "iter 47436: loss 1.4962, time 9.30ms, mfu 0.52%\n",
            "iter 47437: loss 1.5060, time 9.25ms, mfu 0.52%\n",
            "iter 47438: loss 1.5611, time 9.72ms, mfu 0.50%\n",
            "iter 47439: loss 1.4075, time 11.39ms, mfu 0.42%\n",
            "iter 47440: loss 1.4297, time 9.79ms, mfu 0.49%\n",
            "iter 47441: loss 1.4856, time 10.38ms, mfu 0.46%\n",
            "iter 47442: loss 1.3548, time 9.60ms, mfu 0.50%\n",
            "iter 47443: loss 1.5125, time 10.70ms, mfu 0.45%\n",
            "iter 47444: loss 1.4578, time 9.47ms, mfu 0.51%\n",
            "iter 47445: loss 1.4495, time 9.41ms, mfu 0.51%\n",
            "iter 47446: loss 1.3219, time 13.64ms, mfu 0.35%\n",
            "iter 47447: loss 1.4808, time 13.08ms, mfu 0.37%\n",
            "iter 47448: loss 1.4023, time 11.30ms, mfu 0.43%\n",
            "iter 47449: loss 1.5620, time 9.63ms, mfu 0.50%\n",
            "iter 47450: loss 1.3462, time 9.81ms, mfu 0.49%\n",
            "iter 47451: loss 1.5291, time 9.91ms, mfu 0.49%\n",
            "iter 47452: loss 1.3263, time 9.72ms, mfu 0.50%\n",
            "iter 47453: loss 1.4411, time 9.65ms, mfu 0.50%\n",
            "iter 47454: loss 1.4958, time 9.75ms, mfu 0.49%\n",
            "iter 47455: loss 1.5252, time 9.54ms, mfu 0.51%\n",
            "iter 47456: loss 1.4737, time 9.36ms, mfu 0.52%\n",
            "iter 47457: loss 1.4677, time 9.32ms, mfu 0.52%\n",
            "iter 47458: loss 1.5360, time 9.62ms, mfu 0.50%\n",
            "iter 47459: loss 1.3493, time 9.51ms, mfu 0.51%\n",
            "iter 47460: loss 1.4440, time 9.52ms, mfu 0.51%\n",
            "iter 47461: loss 1.5138, time 10.01ms, mfu 0.48%\n",
            "iter 47462: loss 1.3547, time 9.33ms, mfu 0.52%\n",
            "iter 47463: loss 1.4718, time 9.66ms, mfu 0.50%\n",
            "iter 47464: loss 1.4944, time 13.00ms, mfu 0.37%\n",
            "iter 47465: loss 1.4654, time 13.17ms, mfu 0.37%\n",
            "iter 47466: loss 1.3423, time 14.95ms, mfu 0.32%\n",
            "iter 47467: loss 1.5499, time 12.33ms, mfu 0.39%\n",
            "iter 47468: loss 1.6407, time 12.01ms, mfu 0.40%\n",
            "iter 47469: loss 1.3775, time 12.08ms, mfu 0.40%\n",
            "iter 47470: loss 1.5812, time 15.42ms, mfu 0.31%\n",
            "iter 47471: loss 1.5021, time 15.82ms, mfu 0.30%\n",
            "iter 47472: loss 1.4640, time 11.71ms, mfu 0.41%\n",
            "iter 47473: loss 1.5066, time 11.11ms, mfu 0.43%\n",
            "iter 47474: loss 1.4574, time 9.33ms, mfu 0.52%\n",
            "iter 47475: loss 1.4900, time 9.37ms, mfu 0.51%\n",
            "iter 47476: loss 1.4231, time 9.62ms, mfu 0.50%\n",
            "iter 47477: loss 1.5036, time 9.44ms, mfu 0.51%\n",
            "iter 47478: loss 1.5108, time 9.51ms, mfu 0.51%\n",
            "iter 47479: loss 1.5344, time 9.27ms, mfu 0.52%\n",
            "iter 47480: loss 1.5305, time 9.19ms, mfu 0.52%\n",
            "iter 47481: loss 1.4049, time 9.36ms, mfu 0.52%\n",
            "iter 47482: loss 1.4942, time 9.21ms, mfu 0.52%\n",
            "iter 47483: loss 1.4706, time 9.23ms, mfu 0.52%\n",
            "iter 47484: loss 1.4493, time 9.17ms, mfu 0.53%\n",
            "iter 47485: loss 1.5031, time 12.38ms, mfu 0.39%\n",
            "iter 47486: loss 1.4197, time 11.12ms, mfu 0.43%\n",
            "iter 47487: loss 1.4287, time 9.36ms, mfu 0.52%\n",
            "iter 47488: loss 1.4708, time 9.35ms, mfu 0.52%\n",
            "iter 47489: loss 1.4137, time 9.18ms, mfu 0.53%\n",
            "iter 47490: loss 1.4481, time 9.15ms, mfu 0.53%\n",
            "iter 47491: loss 1.4671, time 9.22ms, mfu 0.52%\n",
            "iter 47492: loss 1.5072, time 9.27ms, mfu 0.52%\n",
            "iter 47493: loss 1.5258, time 9.47ms, mfu 0.51%\n",
            "iter 47494: loss 1.4426, time 9.16ms, mfu 0.53%\n",
            "iter 47495: loss 1.4293, time 9.32ms, mfu 0.52%\n",
            "iter 47496: loss 1.4051, time 9.60ms, mfu 0.50%\n",
            "iter 47497: loss 1.4410, time 9.25ms, mfu 0.52%\n",
            "iter 47498: loss 1.6492, time 9.34ms, mfu 0.52%\n",
            "iter 47499: loss 1.3581, time 9.28ms, mfu 0.52%\n",
            "step 47500: train loss 1.3313, val loss 1.6152\n",
            "iter 47500: loss 1.3772, time 16.29ms, mfu 0.30%\n",
            "iter 47501: loss 1.3044, time 9.16ms, mfu 0.53%\n",
            "iter 47502: loss 1.3651, time 9.44ms, mfu 0.51%\n",
            "iter 47503: loss 1.5105, time 9.36ms, mfu 0.52%\n",
            "iter 47504: loss 1.4897, time 9.27ms, mfu 0.52%\n",
            "iter 47505: loss 1.4414, time 9.19ms, mfu 0.52%\n",
            "iter 47506: loss 1.4607, time 9.73ms, mfu 0.50%\n",
            "iter 47507: loss 1.5113, time 15.46ms, mfu 0.31%\n",
            "iter 47508: loss 1.4242, time 10.33ms, mfu 0.47%\n",
            "iter 47509: loss 1.5063, time 9.51ms, mfu 0.51%\n",
            "iter 47510: loss 1.3998, time 9.20ms, mfu 0.52%\n",
            "iter 47511: loss 1.4896, time 9.51ms, mfu 0.51%\n",
            "iter 47512: loss 1.4837, time 11.96ms, mfu 0.40%\n",
            "iter 47513: loss 1.5960, time 9.50ms, mfu 0.51%\n",
            "iter 47514: loss 1.4096, time 9.19ms, mfu 0.52%\n",
            "iter 47515: loss 1.3289, time 9.38ms, mfu 0.51%\n",
            "iter 47516: loss 1.5229, time 9.30ms, mfu 0.52%\n",
            "iter 47517: loss 1.4876, time 9.16ms, mfu 0.53%\n",
            "iter 47518: loss 1.5078, time 9.28ms, mfu 0.52%\n",
            "iter 47519: loss 1.3607, time 9.11ms, mfu 0.53%\n",
            "iter 47520: loss 1.5552, time 10.98ms, mfu 0.44%\n",
            "iter 47521: loss 1.4818, time 10.92ms, mfu 0.44%\n",
            "iter 47522: loss 1.5850, time 9.26ms, mfu 0.52%\n",
            "iter 47523: loss 1.6234, time 9.47ms, mfu 0.51%\n",
            "iter 47524: loss 1.5752, time 13.24ms, mfu 0.36%\n",
            "iter 47525: loss 1.4542, time 12.71ms, mfu 0.38%\n",
            "iter 47526: loss 1.5542, time 11.32ms, mfu 0.43%\n",
            "iter 47527: loss 1.4451, time 9.24ms, mfu 0.52%\n",
            "iter 47528: loss 1.4584, time 13.53ms, mfu 0.36%\n",
            "iter 47529: loss 1.4924, time 9.39ms, mfu 0.51%\n",
            "iter 47530: loss 1.4329, time 10.62ms, mfu 0.45%\n",
            "iter 47531: loss 1.4577, time 9.44ms, mfu 0.51%\n",
            "iter 47532: loss 1.4333, time 9.20ms, mfu 0.52%\n",
            "iter 47533: loss 1.5996, time 9.47ms, mfu 0.51%\n",
            "iter 47534: loss 1.4154, time 9.29ms, mfu 0.52%\n",
            "iter 47535: loss 1.5226, time 9.56ms, mfu 0.50%\n",
            "iter 47536: loss 1.5033, time 9.61ms, mfu 0.50%\n",
            "iter 47537: loss 1.4756, time 15.42ms, mfu 0.31%\n",
            "iter 47538: loss 1.5164, time 16.17ms, mfu 0.30%\n",
            "iter 47539: loss 1.4047, time 16.40ms, mfu 0.29%\n",
            "iter 47540: loss 1.4900, time 14.35ms, mfu 0.34%\n",
            "iter 47541: loss 1.4622, time 16.82ms, mfu 0.29%\n",
            "iter 47542: loss 1.5212, time 19.54ms, mfu 0.25%\n",
            "iter 47543: loss 1.4934, time 16.20ms, mfu 0.30%\n",
            "iter 47544: loss 1.5216, time 21.11ms, mfu 0.23%\n",
            "iter 47545: loss 1.4926, time 16.74ms, mfu 0.29%\n",
            "iter 47546: loss 1.4867, time 16.88ms, mfu 0.29%\n",
            "iter 47547: loss 1.4644, time 16.80ms, mfu 0.29%\n",
            "iter 47548: loss 1.5303, time 16.45ms, mfu 0.29%\n",
            "iter 47549: loss 1.3526, time 14.76ms, mfu 0.33%\n",
            "iter 47550: loss 1.4965, time 14.24ms, mfu 0.34%\n",
            "iter 47551: loss 1.4396, time 15.92ms, mfu 0.30%\n",
            "iter 47552: loss 1.5481, time 16.37ms, mfu 0.29%\n",
            "iter 47553: loss 1.5197, time 16.67ms, mfu 0.29%\n",
            "iter 47554: loss 1.4661, time 18.46ms, mfu 0.26%\n",
            "iter 47555: loss 1.4816, time 17.32ms, mfu 0.28%\n",
            "iter 47556: loss 1.3336, time 16.88ms, mfu 0.29%\n",
            "iter 47557: loss 1.3315, time 33.48ms, mfu 0.14%\n",
            "iter 47558: loss 1.4810, time 15.30ms, mfu 0.32%\n",
            "iter 47559: loss 1.3624, time 25.96ms, mfu 0.19%\n",
            "iter 47560: loss 1.4519, time 15.51ms, mfu 0.31%\n",
            "iter 47561: loss 1.4796, time 16.23ms, mfu 0.30%\n",
            "iter 47562: loss 1.5177, time 15.47ms, mfu 0.31%\n",
            "iter 47563: loss 1.5453, time 16.36ms, mfu 0.29%\n",
            "iter 47564: loss 1.6817, time 16.63ms, mfu 0.29%\n",
            "iter 47565: loss 1.5167, time 16.74ms, mfu 0.29%\n",
            "iter 47566: loss 1.6572, time 16.44ms, mfu 0.29%\n",
            "iter 47567: loss 1.4213, time 16.39ms, mfu 0.29%\n",
            "iter 47568: loss 1.5090, time 16.88ms, mfu 0.29%\n",
            "iter 47569: loss 1.4226, time 17.71ms, mfu 0.27%\n",
            "iter 47570: loss 1.5327, time 20.95ms, mfu 0.23%\n",
            "iter 47571: loss 1.4911, time 14.76ms, mfu 0.33%\n",
            "iter 47572: loss 1.3935, time 14.96ms, mfu 0.32%\n",
            "iter 47573: loss 1.5178, time 18.09ms, mfu 0.27%\n",
            "iter 47574: loss 1.4263, time 15.17ms, mfu 0.32%\n",
            "iter 47575: loss 1.3044, time 16.50ms, mfu 0.29%\n",
            "iter 47576: loss 1.5001, time 12.87ms, mfu 0.37%\n",
            "iter 47577: loss 1.5513, time 16.74ms, mfu 0.29%\n",
            "iter 47578: loss 1.3833, time 13.01ms, mfu 0.37%\n",
            "iter 47579: loss 1.4569, time 13.75ms, mfu 0.35%\n",
            "iter 47580: loss 1.4877, time 9.42ms, mfu 0.51%\n",
            "iter 47581: loss 1.5198, time 9.45ms, mfu 0.51%\n",
            "iter 47582: loss 1.4005, time 9.45ms, mfu 0.51%\n",
            "iter 47583: loss 1.5676, time 9.45ms, mfu 0.51%\n",
            "iter 47584: loss 1.4854, time 11.25ms, mfu 0.43%\n",
            "iter 47585: loss 1.4893, time 13.10ms, mfu 0.37%\n",
            "iter 47586: loss 1.6185, time 9.62ms, mfu 0.50%\n",
            "iter 47587: loss 1.5112, time 9.25ms, mfu 0.52%\n",
            "iter 47588: loss 1.4162, time 9.58ms, mfu 0.50%\n",
            "iter 47589: loss 1.5039, time 9.27ms, mfu 0.52%\n",
            "iter 47590: loss 1.4653, time 11.72ms, mfu 0.41%\n",
            "iter 47591: loss 1.5487, time 16.34ms, mfu 0.30%\n",
            "iter 47592: loss 1.4580, time 24.33ms, mfu 0.20%\n",
            "iter 47593: loss 1.6145, time 35.06ms, mfu 0.14%\n",
            "iter 47594: loss 1.3890, time 15.62ms, mfu 0.31%\n",
            "iter 47595: loss 1.5405, time 22.64ms, mfu 0.21%\n",
            "iter 47596: loss 1.5021, time 16.22ms, mfu 0.30%\n",
            "iter 47597: loss 1.4750, time 19.62ms, mfu 0.25%\n",
            "iter 47598: loss 1.6225, time 17.31ms, mfu 0.28%\n",
            "iter 47599: loss 1.4754, time 19.31ms, mfu 0.25%\n",
            "iter 47600: loss 1.5044, time 16.82ms, mfu 0.29%\n",
            "iter 47601: loss 1.4873, time 17.48ms, mfu 0.28%\n",
            "iter 47602: loss 1.4762, time 19.80ms, mfu 0.24%\n",
            "iter 47603: loss 1.4345, time 21.11ms, mfu 0.23%\n",
            "iter 47604: loss 1.4068, time 14.71ms, mfu 0.33%\n",
            "iter 47605: loss 1.4235, time 14.22ms, mfu 0.34%\n",
            "iter 47606: loss 1.4141, time 16.19ms, mfu 0.30%\n",
            "iter 47607: loss 1.4239, time 14.60ms, mfu 0.33%\n",
            "iter 47608: loss 1.4811, time 16.13ms, mfu 0.30%\n",
            "iter 47609: loss 1.3537, time 18.28ms, mfu 0.26%\n",
            "iter 47610: loss 1.4592, time 19.49ms, mfu 0.25%\n",
            "iter 47611: loss 1.4621, time 17.57ms, mfu 0.27%\n",
            "iter 47612: loss 1.4152, time 16.36ms, mfu 0.29%\n",
            "iter 47613: loss 1.5828, time 16.68ms, mfu 0.29%\n",
            "iter 47614: loss 1.4217, time 16.62ms, mfu 0.29%\n",
            "iter 47615: loss 1.4053, time 17.14ms, mfu 0.28%\n",
            "iter 47616: loss 1.5059, time 16.51ms, mfu 0.29%\n",
            "iter 47617: loss 1.4635, time 15.04ms, mfu 0.32%\n",
            "iter 47618: loss 1.4672, time 24.75ms, mfu 0.19%\n",
            "iter 47619: loss 1.3430, time 17.54ms, mfu 0.27%\n",
            "iter 47620: loss 1.4559, time 14.95ms, mfu 0.32%\n",
            "iter 47621: loss 1.4416, time 23.51ms, mfu 0.21%\n",
            "iter 47622: loss 1.5497, time 30.98ms, mfu 0.16%\n",
            "iter 47623: loss 1.4672, time 13.75ms, mfu 0.35%\n",
            "iter 47624: loss 1.4756, time 12.68ms, mfu 0.38%\n",
            "iter 47625: loss 1.5028, time 13.99ms, mfu 0.34%\n",
            "iter 47626: loss 1.5285, time 35.93ms, mfu 0.13%\n",
            "iter 47627: loss 1.4915, time 34.04ms, mfu 0.14%\n",
            "iter 47628: loss 1.4240, time 35.74ms, mfu 0.13%\n",
            "iter 47629: loss 1.4646, time 27.36ms, mfu 0.18%\n",
            "iter 47630: loss 1.4072, time 38.07ms, mfu 0.13%\n",
            "iter 47631: loss 1.5396, time 18.99ms, mfu 0.25%\n",
            "iter 47632: loss 1.5086, time 29.95ms, mfu 0.16%\n",
            "iter 47633: loss 1.5681, time 27.11ms, mfu 0.18%\n",
            "iter 47634: loss 1.4472, time 16.86ms, mfu 0.29%\n",
            "iter 47635: loss 1.4595, time 34.06ms, mfu 0.14%\n",
            "iter 47636: loss 1.6203, time 29.83ms, mfu 0.16%\n",
            "iter 47637: loss 1.4411, time 33.58ms, mfu 0.14%\n",
            "iter 47638: loss 1.4151, time 26.68ms, mfu 0.18%\n",
            "iter 47639: loss 1.5439, time 26.44ms, mfu 0.18%\n",
            "iter 47640: loss 1.4785, time 25.73ms, mfu 0.19%\n",
            "iter 47641: loss 1.4682, time 17.16ms, mfu 0.28%\n",
            "iter 47642: loss 1.4217, time 29.21ms, mfu 0.17%\n",
            "iter 47643: loss 1.5127, time 28.47ms, mfu 0.17%\n",
            "iter 47644: loss 1.4956, time 34.02ms, mfu 0.14%\n",
            "iter 47645: loss 1.4395, time 31.53ms, mfu 0.15%\n",
            "iter 47646: loss 1.4039, time 40.40ms, mfu 0.12%\n",
            "iter 47647: loss 1.4600, time 41.30ms, mfu 0.12%\n",
            "iter 47648: loss 1.5129, time 29.66ms, mfu 0.16%\n",
            "iter 47649: loss 1.4974, time 35.91ms, mfu 0.13%\n",
            "iter 47650: loss 1.4850, time 45.39ms, mfu 0.11%\n",
            "iter 47651: loss 1.4802, time 50.76ms, mfu 0.10%\n",
            "iter 47652: loss 1.4696, time 34.23ms, mfu 0.14%\n",
            "iter 47653: loss 1.4681, time 17.58ms, mfu 0.27%\n",
            "iter 47654: loss 1.3987, time 52.78ms, mfu 0.09%\n",
            "iter 47655: loss 1.4595, time 25.58ms, mfu 0.19%\n",
            "iter 47656: loss 1.5426, time 34.69ms, mfu 0.14%\n",
            "iter 47657: loss 1.4248, time 37.36ms, mfu 0.13%\n",
            "iter 47658: loss 1.4706, time 32.60ms, mfu 0.15%\n",
            "iter 47659: loss 1.4816, time 31.05ms, mfu 0.16%\n",
            "iter 47660: loss 1.4414, time 35.23ms, mfu 0.14%\n",
            "iter 47661: loss 1.4362, time 48.52ms, mfu 0.10%\n",
            "iter 47662: loss 1.4164, time 21.84ms, mfu 0.22%\n",
            "iter 47663: loss 1.4428, time 32.73ms, mfu 0.15%\n",
            "iter 47664: loss 1.4618, time 34.75ms, mfu 0.14%\n",
            "iter 47665: loss 1.4545, time 22.22ms, mfu 0.22%\n",
            "iter 47666: loss 1.4337, time 30.69ms, mfu 0.16%\n",
            "iter 47667: loss 1.4451, time 21.83ms, mfu 0.22%\n",
            "iter 47668: loss 1.2181, time 25.04ms, mfu 0.19%\n",
            "iter 47669: loss 1.5548, time 24.96ms, mfu 0.19%\n",
            "iter 47670: loss 1.5236, time 23.29ms, mfu 0.21%\n",
            "iter 47671: loss 1.3638, time 19.00ms, mfu 0.25%\n",
            "iter 47672: loss 1.4217, time 17.39ms, mfu 0.28%\n",
            "iter 47673: loss 1.4434, time 23.12ms, mfu 0.21%\n",
            "iter 47674: loss 1.5403, time 35.17ms, mfu 0.14%\n",
            "iter 47675: loss 1.5042, time 35.20ms, mfu 0.14%\n",
            "iter 47676: loss 1.5066, time 28.52ms, mfu 0.17%\n",
            "iter 47677: loss 1.4565, time 19.48ms, mfu 0.25%\n",
            "iter 47678: loss 1.4230, time 13.07ms, mfu 0.37%\n",
            "iter 47679: loss 1.3922, time 13.11ms, mfu 0.37%\n",
            "iter 47680: loss 1.5672, time 17.11ms, mfu 0.28%\n",
            "iter 47681: loss 1.3856, time 13.14ms, mfu 0.37%\n",
            "iter 47682: loss 1.4965, time 12.88ms, mfu 0.37%\n",
            "iter 47683: loss 1.4725, time 14.77ms, mfu 0.33%\n",
            "iter 47684: loss 1.4766, time 22.09ms, mfu 0.22%\n",
            "iter 47685: loss 1.5126, time 22.20ms, mfu 0.22%\n",
            "iter 47686: loss 1.4916, time 25.06ms, mfu 0.19%\n",
            "iter 47687: loss 1.5488, time 19.51ms, mfu 0.25%\n",
            "iter 47688: loss 1.4067, time 18.99ms, mfu 0.25%\n",
            "iter 47689: loss 1.4919, time 18.67ms, mfu 0.26%\n",
            "iter 47690: loss 1.3617, time 23.94ms, mfu 0.20%\n",
            "iter 47691: loss 1.4274, time 18.70ms, mfu 0.26%\n",
            "iter 47692: loss 1.3584, time 29.24ms, mfu 0.16%\n",
            "iter 47693: loss 1.4778, time 18.92ms, mfu 0.25%\n",
            "iter 47694: loss 1.3869, time 24.25ms, mfu 0.20%\n",
            "iter 47695: loss 1.4533, time 20.17ms, mfu 0.24%\n",
            "iter 47696: loss 1.4498, time 33.14ms, mfu 0.15%\n",
            "iter 47697: loss 1.5253, time 12.55ms, mfu 0.38%\n",
            "iter 47698: loss 1.4910, time 14.13ms, mfu 0.34%\n",
            "iter 47699: loss 1.4335, time 15.04ms, mfu 0.32%\n",
            "iter 47700: loss 1.4326, time 21.02ms, mfu 0.23%\n",
            "iter 47701: loss 1.5590, time 20.97ms, mfu 0.23%\n",
            "iter 47702: loss 1.4885, time 14.86ms, mfu 0.32%\n",
            "iter 47703: loss 1.4882, time 12.73ms, mfu 0.38%\n",
            "iter 47704: loss 1.5005, time 14.68ms, mfu 0.33%\n",
            "iter 47705: loss 1.4935, time 12.62ms, mfu 0.38%\n",
            "iter 47706: loss 1.3821, time 12.26ms, mfu 0.39%\n",
            "iter 47707: loss 1.5577, time 13.58ms, mfu 0.36%\n",
            "iter 47708: loss 1.4022, time 18.85ms, mfu 0.26%\n",
            "iter 47709: loss 1.4141, time 14.81ms, mfu 0.33%\n",
            "iter 47710: loss 1.5128, time 14.30ms, mfu 0.34%\n",
            "iter 47711: loss 1.5061, time 14.86ms, mfu 0.32%\n",
            "iter 47712: loss 1.4460, time 15.14ms, mfu 0.32%\n",
            "iter 47713: loss 1.4963, time 13.47ms, mfu 0.36%\n",
            "iter 47714: loss 1.3826, time 12.47ms, mfu 0.39%\n",
            "iter 47715: loss 1.4700, time 12.27ms, mfu 0.39%\n",
            "iter 47716: loss 1.4654, time 12.13ms, mfu 0.40%\n",
            "iter 47717: loss 1.4826, time 13.51ms, mfu 0.36%\n",
            "iter 47718: loss 1.3713, time 13.28ms, mfu 0.36%\n",
            "iter 47719: loss 1.4443, time 13.34ms, mfu 0.36%\n",
            "iter 47720: loss 1.4726, time 12.12ms, mfu 0.40%\n",
            "iter 47721: loss 1.5158, time 13.57ms, mfu 0.36%\n",
            "iter 47722: loss 1.5000, time 11.80ms, mfu 0.41%\n",
            "iter 47723: loss 1.4574, time 11.86ms, mfu 0.41%\n",
            "iter 47724: loss 1.4154, time 19.91ms, mfu 0.24%\n",
            "iter 47725: loss 1.5185, time 15.42ms, mfu 0.31%\n",
            "iter 47726: loss 1.4141, time 12.25ms, mfu 0.39%\n",
            "iter 47727: loss 1.5361, time 13.52ms, mfu 0.36%\n",
            "iter 47728: loss 1.5165, time 13.27ms, mfu 0.36%\n",
            "iter 47729: loss 1.4859, time 12.61ms, mfu 0.38%\n",
            "iter 47730: loss 1.5044, time 13.02ms, mfu 0.37%\n",
            "iter 47731: loss 1.4863, time 12.48ms, mfu 0.39%\n",
            "iter 47732: loss 1.4176, time 16.78ms, mfu 0.29%\n",
            "iter 47733: loss 1.5342, time 24.40ms, mfu 0.20%\n",
            "iter 47734: loss 1.5907, time 18.66ms, mfu 0.26%\n",
            "iter 47735: loss 1.5185, time 18.89ms, mfu 0.26%\n",
            "iter 47736: loss 1.4675, time 21.24ms, mfu 0.23%\n",
            "iter 47737: loss 1.4695, time 31.83ms, mfu 0.15%\n",
            "iter 47738: loss 1.5409, time 30.27ms, mfu 0.16%\n",
            "iter 47739: loss 1.5102, time 18.42ms, mfu 0.26%\n",
            "iter 47740: loss 1.5714, time 19.96ms, mfu 0.24%\n",
            "iter 47741: loss 1.4783, time 21.70ms, mfu 0.22%\n",
            "iter 47742: loss 1.4880, time 18.99ms, mfu 0.25%\n",
            "iter 47743: loss 1.4494, time 19.47ms, mfu 0.25%\n",
            "iter 47744: loss 1.3746, time 15.49ms, mfu 0.31%\n",
            "iter 47745: loss 1.4225, time 35.20ms, mfu 0.14%\n",
            "iter 47746: loss 1.4229, time 28.47ms, mfu 0.17%\n",
            "iter 47747: loss 1.5250, time 35.91ms, mfu 0.13%\n",
            "iter 47748: loss 1.4756, time 22.91ms, mfu 0.21%\n",
            "iter 47749: loss 1.5251, time 24.51ms, mfu 0.20%\n",
            "iter 47750: loss 1.3779, time 18.72ms, mfu 0.26%\n",
            "iter 47751: loss 1.4568, time 17.94ms, mfu 0.27%\n",
            "iter 47752: loss 1.4588, time 16.93ms, mfu 0.28%\n",
            "iter 47753: loss 1.4414, time 26.64ms, mfu 0.18%\n",
            "iter 47754: loss 1.4906, time 15.85ms, mfu 0.30%\n",
            "iter 47755: loss 1.5547, time 27.68ms, mfu 0.17%\n",
            "iter 47756: loss 1.5415, time 16.98ms, mfu 0.28%\n",
            "iter 47757: loss 1.3683, time 33.46ms, mfu 0.14%\n",
            "iter 47758: loss 1.5217, time 22.20ms, mfu 0.22%\n",
            "iter 47759: loss 1.5297, time 21.59ms, mfu 0.22%\n",
            "iter 47760: loss 1.4279, time 25.05ms, mfu 0.19%\n",
            "iter 47761: loss 1.4758, time 11.71ms, mfu 0.41%\n",
            "iter 47762: loss 1.5744, time 21.06ms, mfu 0.23%\n",
            "iter 47763: loss 1.4882, time 21.03ms, mfu 0.23%\n",
            "iter 47764: loss 1.4580, time 18.47ms, mfu 0.26%\n",
            "iter 47765: loss 1.5211, time 19.41ms, mfu 0.25%\n",
            "iter 47766: loss 1.5634, time 22.84ms, mfu 0.21%\n",
            "iter 47767: loss 1.4719, time 32.13ms, mfu 0.15%\n",
            "iter 47768: loss 1.4495, time 31.59ms, mfu 0.15%\n",
            "iter 47769: loss 1.3570, time 19.40ms, mfu 0.25%\n",
            "iter 47770: loss 1.5105, time 23.64ms, mfu 0.20%\n",
            "iter 47771: loss 1.4998, time 21.88ms, mfu 0.22%\n",
            "iter 47772: loss 1.5310, time 21.03ms, mfu 0.23%\n",
            "iter 47773: loss 1.4714, time 22.63ms, mfu 0.21%\n",
            "iter 47774: loss 1.5441, time 19.01ms, mfu 0.25%\n",
            "iter 47775: loss 1.3682, time 18.93ms, mfu 0.25%\n",
            "iter 47776: loss 1.4798, time 23.11ms, mfu 0.21%\n",
            "iter 47777: loss 1.5064, time 23.07ms, mfu 0.21%\n",
            "iter 47778: loss 1.4555, time 13.20ms, mfu 0.37%\n",
            "iter 47779: loss 1.6527, time 14.86ms, mfu 0.32%\n",
            "iter 47780: loss 1.4732, time 15.98ms, mfu 0.30%\n",
            "iter 47781: loss 1.4719, time 13.40ms, mfu 0.36%\n",
            "iter 47782: loss 1.3833, time 15.36ms, mfu 0.31%\n",
            "iter 47783: loss 1.3327, time 23.69ms, mfu 0.20%\n",
            "iter 47784: loss 1.4742, time 18.15ms, mfu 0.27%\n",
            "iter 47785: loss 1.5130, time 27.49ms, mfu 0.18%\n",
            "iter 47786: loss 1.3954, time 28.92ms, mfu 0.17%\n",
            "iter 47787: loss 1.4515, time 29.96ms, mfu 0.16%\n",
            "iter 47788: loss 1.4798, time 26.62ms, mfu 0.18%\n",
            "iter 47789: loss 1.5656, time 30.13ms, mfu 0.16%\n",
            "iter 47790: loss 1.4846, time 25.42ms, mfu 0.19%\n",
            "iter 47791: loss 1.4528, time 38.71ms, mfu 0.12%\n",
            "iter 47792: loss 1.4078, time 35.31ms, mfu 0.14%\n",
            "iter 47793: loss 1.5058, time 38.53ms, mfu 0.13%\n",
            "iter 47794: loss 1.4556, time 22.70ms, mfu 0.21%\n",
            "iter 47795: loss 1.4511, time 24.19ms, mfu 0.20%\n",
            "iter 47796: loss 1.4777, time 15.19ms, mfu 0.32%\n",
            "iter 47797: loss 1.5024, time 14.88ms, mfu 0.32%\n",
            "iter 47798: loss 1.4899, time 15.38ms, mfu 0.31%\n",
            "iter 47799: loss 1.4878, time 14.48ms, mfu 0.33%\n",
            "iter 47800: loss 1.3724, time 14.11ms, mfu 0.34%\n",
            "iter 47801: loss 1.5525, time 13.71ms, mfu 0.35%\n",
            "iter 47802: loss 1.4504, time 13.28ms, mfu 0.36%\n",
            "iter 47803: loss 1.4968, time 13.22ms, mfu 0.36%\n",
            "iter 47804: loss 1.5312, time 13.36ms, mfu 0.36%\n",
            "iter 47805: loss 1.4782, time 13.49ms, mfu 0.36%\n",
            "iter 47806: loss 1.4934, time 13.55ms, mfu 0.36%\n",
            "iter 47807: loss 1.4239, time 13.59ms, mfu 0.35%\n",
            "iter 47808: loss 1.4408, time 19.74ms, mfu 0.24%\n",
            "iter 47809: loss 1.4412, time 20.91ms, mfu 0.23%\n",
            "iter 47810: loss 1.5735, time 14.89ms, mfu 0.32%\n",
            "iter 47811: loss 1.4629, time 13.71ms, mfu 0.35%\n",
            "iter 47812: loss 1.5325, time 14.16ms, mfu 0.34%\n",
            "iter 47813: loss 1.4272, time 15.16ms, mfu 0.32%\n",
            "iter 47814: loss 1.5566, time 14.45ms, mfu 0.33%\n",
            "iter 47815: loss 1.4661, time 14.91ms, mfu 0.32%\n",
            "iter 47816: loss 1.4548, time 14.87ms, mfu 0.32%\n",
            "iter 47817: loss 1.5446, time 14.64ms, mfu 0.33%\n",
            "iter 47818: loss 1.4413, time 14.76ms, mfu 0.33%\n",
            "iter 47819: loss 1.4912, time 14.28ms, mfu 0.34%\n",
            "iter 47820: loss 1.5222, time 14.74ms, mfu 0.33%\n",
            "iter 47821: loss 1.4454, time 14.30ms, mfu 0.34%\n",
            "iter 47822: loss 1.4525, time 21.38ms, mfu 0.23%\n",
            "iter 47823: loss 1.4125, time 24.64ms, mfu 0.20%\n",
            "iter 47824: loss 1.3672, time 16.26ms, mfu 0.30%\n",
            "iter 47825: loss 1.3518, time 11.17ms, mfu 0.43%\n",
            "iter 47826: loss 1.4812, time 11.09ms, mfu 0.43%\n",
            "iter 47827: loss 1.5282, time 11.64ms, mfu 0.41%\n",
            "iter 47828: loss 1.3210, time 11.40ms, mfu 0.42%\n",
            "iter 47829: loss 1.4696, time 9.29ms, mfu 0.52%\n",
            "iter 47830: loss 1.4576, time 9.26ms, mfu 0.52%\n",
            "iter 47831: loss 1.4035, time 9.27ms, mfu 0.52%\n",
            "iter 47832: loss 1.4860, time 12.20ms, mfu 0.40%\n",
            "iter 47833: loss 1.5825, time 9.29ms, mfu 0.52%\n",
            "iter 47834: loss 1.5314, time 9.24ms, mfu 0.52%\n",
            "iter 47835: loss 1.5108, time 9.52ms, mfu 0.51%\n",
            "iter 47836: loss 1.5510, time 9.44ms, mfu 0.51%\n",
            "iter 47837: loss 1.4571, time 9.98ms, mfu 0.48%\n",
            "iter 47838: loss 1.4687, time 11.12ms, mfu 0.43%\n",
            "iter 47839: loss 1.4138, time 9.27ms, mfu 0.52%\n",
            "iter 47840: loss 1.5555, time 10.24ms, mfu 0.47%\n",
            "iter 47841: loss 1.4029, time 14.84ms, mfu 0.32%\n",
            "iter 47842: loss 1.5786, time 10.86ms, mfu 0.44%\n",
            "iter 47843: loss 1.4796, time 10.02ms, mfu 0.48%\n",
            "iter 47844: loss 1.3574, time 9.61ms, mfu 0.50%\n",
            "iter 47845: loss 1.4676, time 9.45ms, mfu 0.51%\n",
            "iter 47846: loss 1.4526, time 9.50ms, mfu 0.51%\n",
            "iter 47847: loss 1.5450, time 9.31ms, mfu 0.52%\n",
            "iter 47848: loss 1.4295, time 9.38ms, mfu 0.51%\n",
            "iter 47849: loss 1.5128, time 11.13ms, mfu 0.43%\n",
            "iter 47850: loss 1.4386, time 9.78ms, mfu 0.49%\n",
            "iter 47851: loss 1.4704, time 9.69ms, mfu 0.50%\n",
            "iter 47852: loss 1.5148, time 9.48ms, mfu 0.51%\n",
            "iter 47853: loss 1.4182, time 9.21ms, mfu 0.52%\n",
            "iter 47854: loss 1.5049, time 9.15ms, mfu 0.53%\n",
            "iter 47855: loss 1.5610, time 9.17ms, mfu 0.53%\n",
            "iter 47856: loss 1.5640, time 9.28ms, mfu 0.52%\n",
            "iter 47857: loss 1.4457, time 10.97ms, mfu 0.44%\n",
            "iter 47858: loss 1.4650, time 11.19ms, mfu 0.43%\n",
            "iter 47859: loss 1.5255, time 9.41ms, mfu 0.51%\n",
            "iter 47860: loss 1.4196, time 9.37ms, mfu 0.51%\n",
            "iter 47861: loss 1.3879, time 9.90ms, mfu 0.49%\n",
            "iter 47862: loss 1.4502, time 15.17ms, mfu 0.32%\n",
            "iter 47863: loss 1.5413, time 9.80ms, mfu 0.49%\n",
            "iter 47864: loss 1.4862, time 9.71ms, mfu 0.50%\n",
            "iter 47865: loss 1.4313, time 10.91ms, mfu 0.44%\n",
            "iter 47866: loss 1.5140, time 10.10ms, mfu 0.48%\n",
            "iter 47867: loss 1.3952, time 9.40ms, mfu 0.51%\n",
            "iter 47868: loss 1.4171, time 9.61ms, mfu 0.50%\n",
            "iter 47869: loss 1.3501, time 9.50ms, mfu 0.51%\n",
            "iter 47870: loss 1.4395, time 11.26ms, mfu 0.43%\n",
            "iter 47871: loss 1.4871, time 36.60ms, mfu 0.13%\n",
            "iter 47872: loss 1.4853, time 22.34ms, mfu 0.22%\n",
            "iter 47873: loss 1.5409, time 14.28ms, mfu 0.34%\n",
            "iter 47874: loss 1.5312, time 13.75ms, mfu 0.35%\n",
            "iter 47875: loss 1.4161, time 14.02ms, mfu 0.34%\n",
            "iter 47876: loss 1.4028, time 14.26ms, mfu 0.34%\n",
            "iter 47877: loss 1.4307, time 15.94ms, mfu 0.30%\n",
            "iter 47878: loss 1.5371, time 17.31ms, mfu 0.28%\n",
            "iter 47879: loss 1.4525, time 12.65ms, mfu 0.38%\n",
            "iter 47880: loss 1.5390, time 16.37ms, mfu 0.29%\n",
            "iter 47881: loss 1.4346, time 15.35ms, mfu 0.31%\n",
            "iter 47882: loss 1.4436, time 16.77ms, mfu 0.29%\n",
            "iter 47883: loss 1.5940, time 17.19ms, mfu 0.28%\n",
            "iter 47884: loss 1.4999, time 15.72ms, mfu 0.31%\n",
            "iter 47885: loss 1.3939, time 17.55ms, mfu 0.27%\n",
            "iter 47886: loss 1.4416, time 16.97ms, mfu 0.28%\n",
            "iter 47887: loss 1.3867, time 15.25ms, mfu 0.32%\n",
            "iter 47888: loss 1.5131, time 18.52ms, mfu 0.26%\n",
            "iter 47889: loss 1.4982, time 28.53ms, mfu 0.17%\n",
            "iter 47890: loss 1.4524, time 25.33ms, mfu 0.19%\n",
            "iter 47891: loss 1.5305, time 16.93ms, mfu 0.28%\n",
            "iter 47892: loss 1.4820, time 15.94ms, mfu 0.30%\n",
            "iter 47893: loss 1.5436, time 15.31ms, mfu 0.31%\n",
            "iter 47894: loss 1.3777, time 16.65ms, mfu 0.29%\n",
            "iter 47895: loss 1.5175, time 16.82ms, mfu 0.29%\n",
            "iter 47896: loss 1.3880, time 16.60ms, mfu 0.29%\n",
            "iter 47897: loss 1.4428, time 18.14ms, mfu 0.27%\n",
            "iter 47898: loss 1.5251, time 14.79ms, mfu 0.33%\n",
            "iter 47899: loss 1.5392, time 15.85ms, mfu 0.30%\n",
            "iter 47900: loss 1.4052, time 17.31ms, mfu 0.28%\n",
            "iter 47901: loss 1.5807, time 27.75ms, mfu 0.17%\n",
            "iter 47902: loss 1.4989, time 17.02ms, mfu 0.28%\n",
            "iter 47903: loss 1.4640, time 14.89ms, mfu 0.32%\n",
            "iter 47904: loss 1.4391, time 15.76ms, mfu 0.31%\n",
            "iter 47905: loss 1.4521, time 16.61ms, mfu 0.29%\n",
            "iter 47906: loss 1.3641, time 16.56ms, mfu 0.29%\n",
            "iter 47907: loss 1.5547, time 16.03ms, mfu 0.30%\n",
            "iter 47908: loss 1.4439, time 16.37ms, mfu 0.29%\n",
            "iter 47909: loss 1.4679, time 16.41ms, mfu 0.29%\n",
            "iter 47910: loss 1.3605, time 16.11ms, mfu 0.30%\n",
            "iter 47911: loss 1.4636, time 16.41ms, mfu 0.29%\n",
            "iter 47912: loss 1.4503, time 17.66ms, mfu 0.27%\n",
            "iter 47913: loss 1.5436, time 16.90ms, mfu 0.29%\n",
            "iter 47914: loss 1.4502, time 21.46ms, mfu 0.22%\n",
            "iter 47915: loss 1.4065, time 18.11ms, mfu 0.27%\n",
            "iter 47916: loss 1.3397, time 16.65ms, mfu 0.29%\n",
            "iter 47917: loss 1.5633, time 28.68ms, mfu 0.17%\n",
            "iter 47918: loss 1.4314, time 14.79ms, mfu 0.33%\n",
            "iter 47919: loss 1.4194, time 15.77ms, mfu 0.31%\n",
            "iter 47920: loss 1.5729, time 16.94ms, mfu 0.28%\n",
            "iter 47921: loss 1.3917, time 16.94ms, mfu 0.28%\n",
            "iter 47922: loss 1.6201, time 13.62ms, mfu 0.35%\n",
            "iter 47923: loss 1.4648, time 13.26ms, mfu 0.36%\n",
            "iter 47924: loss 1.5091, time 12.63ms, mfu 0.38%\n",
            "iter 47925: loss 1.4790, time 18.92ms, mfu 0.25%\n",
            "iter 47926: loss 1.3872, time 36.00ms, mfu 0.13%\n",
            "iter 47927: loss 1.5608, time 20.80ms, mfu 0.23%\n",
            "iter 47928: loss 1.4473, time 16.89ms, mfu 0.29%\n",
            "iter 47929: loss 1.4722, time 24.99ms, mfu 0.19%\n",
            "iter 47930: loss 1.4342, time 22.96ms, mfu 0.21%\n",
            "iter 47931: loss 1.3830, time 22.36ms, mfu 0.22%\n",
            "iter 47932: loss 1.5160, time 29.15ms, mfu 0.17%\n",
            "iter 47933: loss 1.4032, time 30.83ms, mfu 0.16%\n",
            "iter 47934: loss 1.4263, time 24.72ms, mfu 0.20%\n",
            "iter 47935: loss 1.4117, time 44.27ms, mfu 0.11%\n",
            "iter 47936: loss 1.4000, time 26.15ms, mfu 0.18%\n",
            "iter 47937: loss 1.4721, time 24.54ms, mfu 0.20%\n",
            "iter 47938: loss 1.4178, time 17.48ms, mfu 0.28%\n",
            "iter 47939: loss 1.5582, time 23.19ms, mfu 0.21%\n",
            "iter 47940: loss 1.4990, time 33.92ms, mfu 0.14%\n",
            "iter 47941: loss 1.4809, time 33.57ms, mfu 0.14%\n",
            "iter 47942: loss 1.4631, time 21.54ms, mfu 0.22%\n",
            "iter 47943: loss 1.4468, time 31.58ms, mfu 0.15%\n",
            "iter 47944: loss 1.5144, time 13.83ms, mfu 0.35%\n",
            "iter 47945: loss 1.4401, time 9.38ms, mfu 0.51%\n",
            "iter 47946: loss 1.4413, time 9.34ms, mfu 0.52%\n",
            "iter 47947: loss 1.5008, time 9.63ms, mfu 0.50%\n",
            "iter 47948: loss 1.4486, time 9.98ms, mfu 0.48%\n",
            "iter 47949: loss 1.5042, time 9.50ms, mfu 0.51%\n",
            "iter 47950: loss 1.4269, time 9.79ms, mfu 0.49%\n",
            "iter 47951: loss 1.4593, time 9.63ms, mfu 0.50%\n",
            "iter 47952: loss 1.4505, time 11.50ms, mfu 0.42%\n",
            "iter 47953: loss 1.6197, time 9.59ms, mfu 0.50%\n",
            "iter 47954: loss 1.4344, time 9.58ms, mfu 0.50%\n",
            "iter 47955: loss 1.4264, time 9.39ms, mfu 0.51%\n",
            "iter 47956: loss 1.5515, time 9.86ms, mfu 0.49%\n",
            "iter 47957: loss 1.5396, time 9.47ms, mfu 0.51%\n",
            "iter 47958: loss 1.4011, time 9.48ms, mfu 0.51%\n",
            "iter 47959: loss 1.4860, time 9.45ms, mfu 0.51%\n",
            "iter 47960: loss 1.4873, time 9.76ms, mfu 0.49%\n",
            "iter 47961: loss 1.4582, time 9.77ms, mfu 0.49%\n",
            "iter 47962: loss 1.5385, time 9.35ms, mfu 0.52%\n",
            "iter 47963: loss 1.3882, time 9.33ms, mfu 0.52%\n",
            "iter 47964: loss 1.4358, time 13.86ms, mfu 0.35%\n",
            "iter 47965: loss 1.5172, time 9.79ms, mfu 0.49%\n",
            "iter 47966: loss 1.6289, time 9.68ms, mfu 0.50%\n",
            "iter 47967: loss 1.5084, time 11.37ms, mfu 0.42%\n",
            "iter 47968: loss 1.5742, time 10.52ms, mfu 0.46%\n",
            "iter 47969: loss 1.4199, time 9.65ms, mfu 0.50%\n",
            "iter 47970: loss 1.4645, time 9.42ms, mfu 0.51%\n",
            "iter 47971: loss 1.4842, time 10.69ms, mfu 0.45%\n",
            "iter 47972: loss 1.5005, time 9.64ms, mfu 0.50%\n",
            "iter 47973: loss 1.5159, time 9.52ms, mfu 0.51%\n",
            "iter 47974: loss 1.5022, time 9.46ms, mfu 0.51%\n",
            "iter 47975: loss 1.4160, time 9.47ms, mfu 0.51%\n",
            "iter 47976: loss 1.5100, time 9.42ms, mfu 0.51%\n",
            "iter 47977: loss 1.5168, time 9.49ms, mfu 0.51%\n",
            "iter 47978: loss 1.3911, time 11.07ms, mfu 0.44%\n",
            "iter 47979: loss 1.3931, time 9.81ms, mfu 0.49%\n",
            "iter 47980: loss 1.4664, time 9.60ms, mfu 0.50%\n",
            "iter 47981: loss 1.4680, time 9.51ms, mfu 0.51%\n",
            "iter 47982: loss 1.4751, time 9.48ms, mfu 0.51%\n",
            "iter 47983: loss 1.4570, time 9.29ms, mfu 0.52%\n",
            "iter 47984: loss 1.5890, time 9.47ms, mfu 0.51%\n",
            "iter 47985: loss 1.5388, time 14.08ms, mfu 0.34%\n",
            "iter 47986: loss 1.5412, time 10.14ms, mfu 0.48%\n",
            "iter 47987: loss 1.5755, time 10.23ms, mfu 0.47%\n",
            "iter 47988: loss 1.3639, time 11.38ms, mfu 0.42%\n",
            "iter 47989: loss 1.4807, time 9.74ms, mfu 0.50%\n",
            "iter 47990: loss 1.3745, time 9.61ms, mfu 0.50%\n",
            "iter 47991: loss 1.5525, time 9.64ms, mfu 0.50%\n",
            "iter 47992: loss 1.5406, time 9.50ms, mfu 0.51%\n",
            "iter 47993: loss 1.4681, time 10.74ms, mfu 0.45%\n",
            "iter 47994: loss 1.4409, time 9.75ms, mfu 0.49%\n",
            "iter 47995: loss 1.4880, time 9.47ms, mfu 0.51%\n",
            "iter 47996: loss 1.5445, time 9.76ms, mfu 0.49%\n",
            "iter 47997: loss 1.3346, time 9.73ms, mfu 0.50%\n",
            "iter 47998: loss 1.4843, time 11.01ms, mfu 0.44%\n",
            "iter 47999: loss 1.4269, time 12.33ms, mfu 0.39%\n",
            "step 48000: train loss 1.3381, val loss 1.5364\n",
            "iter 48000: loss 1.4895, time 18.95ms, mfu 0.25%\n",
            "iter 48001: loss 1.4109, time 9.39ms, mfu 0.51%\n",
            "iter 48002: loss 1.4587, time 9.56ms, mfu 0.50%\n",
            "iter 48003: loss 1.4057, time 9.26ms, mfu 0.52%\n",
            "iter 48004: loss 1.4042, time 9.55ms, mfu 0.51%\n",
            "iter 48005: loss 1.5426, time 14.31ms, mfu 0.34%\n",
            "iter 48006: loss 1.4963, time 9.52ms, mfu 0.51%\n",
            "iter 48007: loss 1.4571, time 9.45ms, mfu 0.51%\n",
            "iter 48008: loss 1.4617, time 9.15ms, mfu 0.53%\n",
            "iter 48009: loss 1.5342, time 9.28ms, mfu 0.52%\n",
            "iter 48010: loss 1.4876, time 9.50ms, mfu 0.51%\n",
            "iter 48011: loss 1.4871, time 9.32ms, mfu 0.52%\n",
            "iter 48012: loss 1.3829, time 9.93ms, mfu 0.49%\n",
            "iter 48013: loss 1.4785, time 9.37ms, mfu 0.51%\n",
            "iter 48014: loss 1.3662, time 9.34ms, mfu 0.52%\n",
            "iter 48015: loss 1.5750, time 9.39ms, mfu 0.51%\n",
            "iter 48016: loss 1.4101, time 9.19ms, mfu 0.52%\n",
            "iter 48017: loss 1.4209, time 9.19ms, mfu 0.52%\n",
            "iter 48018: loss 1.5174, time 9.26ms, mfu 0.52%\n",
            "iter 48019: loss 1.5327, time 9.19ms, mfu 0.52%\n",
            "iter 48020: loss 1.5815, time 9.32ms, mfu 0.52%\n",
            "iter 48021: loss 1.4868, time 9.31ms, mfu 0.52%\n",
            "iter 48022: loss 1.3572, time 9.31ms, mfu 0.52%\n",
            "iter 48023: loss 1.4132, time 9.53ms, mfu 0.51%\n",
            "iter 48024: loss 1.4039, time 9.29ms, mfu 0.52%\n",
            "iter 48025: loss 1.4137, time 9.19ms, mfu 0.52%\n",
            "iter 48026: loss 1.4674, time 9.24ms, mfu 0.52%\n",
            "iter 48027: loss 1.4605, time 13.26ms, mfu 0.36%\n",
            "iter 48028: loss 1.3841, time 9.42ms, mfu 0.51%\n",
            "iter 48029: loss 1.5291, time 9.37ms, mfu 0.51%\n",
            "iter 48030: loss 1.6473, time 11.74ms, mfu 0.41%\n",
            "iter 48031: loss 1.3092, time 9.36ms, mfu 0.52%\n",
            "iter 48032: loss 1.4575, time 9.24ms, mfu 0.52%\n",
            "iter 48033: loss 1.3337, time 9.58ms, mfu 0.50%\n",
            "iter 48034: loss 1.5648, time 11.00ms, mfu 0.44%\n",
            "iter 48035: loss 1.4622, time 14.48ms, mfu 0.33%\n",
            "iter 48036: loss 1.4235, time 11.95ms, mfu 0.40%\n",
            "iter 48037: loss 1.4040, time 10.85ms, mfu 0.44%\n",
            "iter 48038: loss 1.4458, time 10.50ms, mfu 0.46%\n",
            "iter 48039: loss 1.3922, time 9.46ms, mfu 0.51%\n",
            "iter 48040: loss 1.4189, time 9.39ms, mfu 0.51%\n",
            "iter 48041: loss 1.5328, time 9.31ms, mfu 0.52%\n",
            "iter 48042: loss 1.4955, time 9.33ms, mfu 0.52%\n",
            "iter 48043: loss 1.4321, time 9.33ms, mfu 0.52%\n",
            "iter 48044: loss 1.5211, time 9.15ms, mfu 0.53%\n",
            "iter 48045: loss 1.3892, time 9.23ms, mfu 0.52%\n",
            "iter 48046: loss 1.4272, time 9.58ms, mfu 0.50%\n",
            "iter 48047: loss 1.3838, time 11.30ms, mfu 0.43%\n",
            "iter 48048: loss 1.5193, time 11.53ms, mfu 0.42%\n",
            "iter 48049: loss 1.4568, time 9.32ms, mfu 0.52%\n",
            "iter 48050: loss 1.4109, time 9.54ms, mfu 0.51%\n",
            "iter 48051: loss 1.4541, time 9.19ms, mfu 0.52%\n",
            "iter 48052: loss 1.3973, time 9.26ms, mfu 0.52%\n",
            "iter 48053: loss 1.3927, time 9.33ms, mfu 0.52%\n",
            "iter 48054: loss 1.5287, time 9.18ms, mfu 0.53%\n",
            "iter 48055: loss 1.4573, time 9.21ms, mfu 0.52%\n",
            "iter 48056: loss 1.4721, time 10.75ms, mfu 0.45%\n",
            "iter 48057: loss 1.3914, time 10.41ms, mfu 0.46%\n",
            "iter 48058: loss 1.4112, time 9.30ms, mfu 0.52%\n",
            "iter 48059: loss 1.4808, time 9.60ms, mfu 0.50%\n",
            "iter 48060: loss 1.4989, time 9.38ms, mfu 0.51%\n",
            "iter 48061: loss 1.6062, time 10.66ms, mfu 0.45%\n",
            "iter 48062: loss 1.4670, time 9.26ms, mfu 0.52%\n",
            "iter 48063: loss 1.4701, time 9.23ms, mfu 0.52%\n",
            "iter 48064: loss 1.4716, time 9.29ms, mfu 0.52%\n",
            "iter 48065: loss 1.5625, time 9.67ms, mfu 0.50%\n",
            "iter 48066: loss 1.4427, time 9.49ms, mfu 0.51%\n",
            "iter 48067: loss 1.4189, time 9.40ms, mfu 0.51%\n",
            "iter 48068: loss 1.3993, time 12.39ms, mfu 0.39%\n",
            "iter 48069: loss 1.5265, time 10.95ms, mfu 0.44%\n",
            "iter 48070: loss 1.4385, time 9.26ms, mfu 0.52%\n",
            "iter 48071: loss 1.5015, time 9.31ms, mfu 0.52%\n",
            "iter 48072: loss 1.4823, time 9.49ms, mfu 0.51%\n",
            "iter 48073: loss 1.4632, time 9.48ms, mfu 0.51%\n",
            "iter 48074: loss 1.4667, time 9.49ms, mfu 0.51%\n",
            "iter 48075: loss 1.4814, time 9.16ms, mfu 0.53%\n",
            "iter 48076: loss 1.5817, time 9.33ms, mfu 0.52%\n",
            "iter 48077: loss 1.4932, time 9.56ms, mfu 0.50%\n",
            "iter 48078: loss 1.4464, time 10.06ms, mfu 0.48%\n",
            "iter 48079: loss 1.4688, time 10.14ms, mfu 0.48%\n",
            "iter 48080: loss 1.3183, time 9.44ms, mfu 0.51%\n",
            "iter 48081: loss 1.3782, time 9.25ms, mfu 0.52%\n",
            "iter 48082: loss 1.4031, time 9.46ms, mfu 0.51%\n",
            "iter 48083: loss 1.5137, time 9.36ms, mfu 0.52%\n",
            "iter 48084: loss 1.4714, time 9.29ms, mfu 0.52%\n",
            "iter 48085: loss 1.4484, time 9.26ms, mfu 0.52%\n",
            "iter 48086: loss 1.5514, time 9.23ms, mfu 0.52%\n",
            "iter 48087: loss 1.3742, time 9.52ms, mfu 0.51%\n",
            "iter 48088: loss 1.4923, time 9.52ms, mfu 0.51%\n",
            "iter 48089: loss 1.4139, time 9.22ms, mfu 0.52%\n",
            "iter 48090: loss 1.5040, time 12.65ms, mfu 0.38%\n",
            "iter 48091: loss 1.5773, time 11.68ms, mfu 0.41%\n",
            "iter 48092: loss 1.4970, time 12.14ms, mfu 0.40%\n",
            "iter 48093: loss 1.4831, time 10.97ms, mfu 0.44%\n",
            "iter 48094: loss 1.4104, time 10.77ms, mfu 0.45%\n",
            "iter 48095: loss 1.4088, time 9.44ms, mfu 0.51%\n",
            "iter 48096: loss 1.5787, time 9.21ms, mfu 0.52%\n",
            "iter 48097: loss 1.4864, time 9.32ms, mfu 0.52%\n",
            "iter 48098: loss 1.3558, time 10.48ms, mfu 0.46%\n",
            "iter 48099: loss 1.4031, time 12.25ms, mfu 0.39%\n",
            "iter 48100: loss 1.5307, time 11.89ms, mfu 0.41%\n",
            "iter 48101: loss 1.5710, time 9.73ms, mfu 0.50%\n",
            "iter 48102: loss 1.4528, time 9.36ms, mfu 0.52%\n",
            "iter 48103: loss 1.4614, time 9.46ms, mfu 0.51%\n",
            "iter 48104: loss 1.5613, time 9.48ms, mfu 0.51%\n",
            "iter 48105: loss 1.6360, time 9.31ms, mfu 0.52%\n",
            "iter 48106: loss 1.5791, time 9.32ms, mfu 0.52%\n",
            "iter 48107: loss 1.5235, time 9.31ms, mfu 0.52%\n",
            "iter 48108: loss 1.3468, time 9.46ms, mfu 0.51%\n",
            "iter 48109: loss 1.4535, time 9.45ms, mfu 0.51%\n",
            "iter 48110: loss 1.4914, time 9.27ms, mfu 0.52%\n",
            "iter 48111: loss 1.4696, time 14.29ms, mfu 0.34%\n",
            "iter 48112: loss 1.4067, time 10.97ms, mfu 0.44%\n",
            "iter 48113: loss 1.5163, time 10.01ms, mfu 0.48%\n",
            "iter 48114: loss 1.3983, time 9.67ms, mfu 0.50%\n",
            "iter 48115: loss 1.5229, time 9.45ms, mfu 0.51%\n",
            "iter 48116: loss 1.4586, time 9.75ms, mfu 0.49%\n",
            "iter 48117: loss 1.4307, time 9.72ms, mfu 0.50%\n",
            "iter 48118: loss 1.3968, time 9.30ms, mfu 0.52%\n",
            "iter 48119: loss 1.4102, time 9.30ms, mfu 0.52%\n",
            "iter 48120: loss 1.4375, time 9.50ms, mfu 0.51%\n",
            "iter 48121: loss 1.4866, time 14.00ms, mfu 0.34%\n",
            "iter 48122: loss 1.3626, time 11.69ms, mfu 0.41%\n",
            "iter 48123: loss 1.3719, time 12.47ms, mfu 0.39%\n",
            "iter 48124: loss 1.4772, time 9.66ms, mfu 0.50%\n",
            "iter 48125: loss 1.5094, time 9.44ms, mfu 0.51%\n",
            "iter 48126: loss 1.5237, time 9.39ms, mfu 0.51%\n",
            "iter 48127: loss 1.5290, time 9.29ms, mfu 0.52%\n",
            "iter 48128: loss 1.4788, time 9.35ms, mfu 0.52%\n",
            "iter 48129: loss 1.5213, time 9.61ms, mfu 0.50%\n",
            "iter 48130: loss 1.6187, time 9.31ms, mfu 0.52%\n",
            "iter 48131: loss 1.4230, time 10.47ms, mfu 0.46%\n",
            "iter 48132: loss 1.4143, time 12.65ms, mfu 0.38%\n",
            "iter 48133: loss 1.4452, time 10.28ms, mfu 0.47%\n",
            "iter 48134: loss 1.4187, time 9.36ms, mfu 0.52%\n",
            "iter 48135: loss 1.3735, time 9.64ms, mfu 0.50%\n",
            "iter 48136: loss 1.4864, time 19.47ms, mfu 0.25%\n",
            "iter 48137: loss 1.4405, time 46.86ms, mfu 0.10%\n",
            "iter 48138: loss 1.4315, time 36.80ms, mfu 0.13%\n",
            "iter 48139: loss 1.5347, time 36.54ms, mfu 0.13%\n",
            "iter 48140: loss 1.5587, time 17.80ms, mfu 0.27%\n",
            "iter 48141: loss 1.4252, time 22.18ms, mfu 0.22%\n",
            "iter 48142: loss 1.4882, time 37.08ms, mfu 0.13%\n",
            "iter 48143: loss 1.4339, time 25.34ms, mfu 0.19%\n",
            "iter 48144: loss 1.5216, time 22.63ms, mfu 0.21%\n",
            "iter 48145: loss 1.6270, time 27.18ms, mfu 0.18%\n",
            "iter 48146: loss 1.5230, time 16.40ms, mfu 0.29%\n",
            "iter 48147: loss 1.4582, time 17.41ms, mfu 0.28%\n",
            "iter 48148: loss 1.4570, time 16.97ms, mfu 0.28%\n",
            "iter 48149: loss 1.4950, time 17.37ms, mfu 0.28%\n",
            "iter 48150: loss 1.5471, time 15.36ms, mfu 0.31%\n",
            "iter 48151: loss 1.4206, time 34.15ms, mfu 0.14%\n",
            "iter 48152: loss 1.5957, time 18.16ms, mfu 0.27%\n",
            "iter 48153: loss 1.4416, time 15.10ms, mfu 0.32%\n",
            "iter 48154: loss 1.4356, time 15.84ms, mfu 0.30%\n",
            "iter 48155: loss 1.5220, time 13.82ms, mfu 0.35%\n",
            "iter 48156: loss 1.4189, time 16.05ms, mfu 0.30%\n",
            "iter 48157: loss 1.6135, time 16.10ms, mfu 0.30%\n",
            "iter 48158: loss 1.5249, time 16.57ms, mfu 0.29%\n",
            "iter 48159: loss 1.4506, time 34.16ms, mfu 0.14%\n",
            "iter 48160: loss 1.4648, time 21.16ms, mfu 0.23%\n",
            "iter 48161: loss 1.3487, time 18.61ms, mfu 0.26%\n",
            "iter 48162: loss 1.5221, time 18.31ms, mfu 0.26%\n",
            "iter 48163: loss 1.3176, time 23.00ms, mfu 0.21%\n",
            "iter 48164: loss 1.6295, time 24.74ms, mfu 0.19%\n",
            "iter 48165: loss 1.5414, time 16.59ms, mfu 0.29%\n",
            "iter 48166: loss 1.5118, time 16.57ms, mfu 0.29%\n",
            "iter 48167: loss 1.3572, time 16.36ms, mfu 0.29%\n",
            "iter 48168: loss 1.5012, time 14.69ms, mfu 0.33%\n",
            "iter 48169: loss 1.5233, time 16.83ms, mfu 0.29%\n",
            "iter 48170: loss 1.4714, time 16.49ms, mfu 0.29%\n",
            "iter 48171: loss 1.2865, time 17.97ms, mfu 0.27%\n",
            "iter 48172: loss 1.4655, time 15.90ms, mfu 0.30%\n",
            "iter 48173: loss 1.5274, time 15.63ms, mfu 0.31%\n",
            "iter 48174: loss 1.5545, time 26.15ms, mfu 0.18%\n",
            "iter 48175: loss 1.5928, time 16.93ms, mfu 0.28%\n",
            "iter 48176: loss 1.4573, time 18.42ms, mfu 0.26%\n",
            "iter 48177: loss 1.5043, time 14.68ms, mfu 0.33%\n",
            "iter 48178: loss 1.3152, time 16.95ms, mfu 0.28%\n",
            "iter 48179: loss 1.4425, time 14.88ms, mfu 0.32%\n",
            "iter 48180: loss 1.5014, time 14.65ms, mfu 0.33%\n",
            "iter 48181: loss 1.5340, time 14.28ms, mfu 0.34%\n",
            "iter 48182: loss 1.5055, time 16.03ms, mfu 0.30%\n",
            "iter 48183: loss 1.3406, time 14.76ms, mfu 0.33%\n",
            "iter 48184: loss 1.4044, time 16.15ms, mfu 0.30%\n",
            "iter 48185: loss 1.5608, time 15.37ms, mfu 0.31%\n",
            "iter 48186: loss 1.4897, time 15.13ms, mfu 0.32%\n",
            "iter 48187: loss 1.4779, time 19.69ms, mfu 0.24%\n",
            "iter 48188: loss 1.4990, time 25.07ms, mfu 0.19%\n",
            "iter 48189: loss 1.5291, time 29.48ms, mfu 0.16%\n",
            "iter 48190: loss 1.4014, time 23.52ms, mfu 0.21%\n",
            "iter 48191: loss 1.4557, time 24.87ms, mfu 0.19%\n",
            "iter 48192: loss 1.5030, time 15.15ms, mfu 0.32%\n",
            "iter 48193: loss 1.3581, time 9.39ms, mfu 0.51%\n",
            "iter 48194: loss 1.4519, time 9.89ms, mfu 0.49%\n",
            "iter 48195: loss 1.4542, time 9.38ms, mfu 0.51%\n",
            "iter 48196: loss 1.3963, time 9.41ms, mfu 0.51%\n",
            "iter 48197: loss 1.4915, time 9.38ms, mfu 0.51%\n",
            "iter 48198: loss 1.5328, time 9.45ms, mfu 0.51%\n",
            "iter 48199: loss 1.4542, time 9.35ms, mfu 0.52%\n",
            "iter 48200: loss 1.4145, time 9.55ms, mfu 0.50%\n",
            "iter 48201: loss 1.5005, time 9.80ms, mfu 0.49%\n",
            "iter 48202: loss 1.4503, time 13.01ms, mfu 0.37%\n",
            "iter 48203: loss 1.3890, time 9.92ms, mfu 0.49%\n",
            "iter 48204: loss 1.4520, time 9.42ms, mfu 0.51%\n",
            "iter 48205: loss 1.2957, time 9.53ms, mfu 0.51%\n",
            "iter 48206: loss 1.4606, time 9.65ms, mfu 0.50%\n",
            "iter 48207: loss 1.4879, time 9.41ms, mfu 0.51%\n",
            "iter 48208: loss 1.4492, time 10.92ms, mfu 0.44%\n",
            "iter 48209: loss 1.4617, time 10.68ms, mfu 0.45%\n",
            "iter 48210: loss 1.5678, time 9.69ms, mfu 0.50%\n",
            "iter 48211: loss 1.5034, time 9.44ms, mfu 0.51%\n",
            "iter 48212: loss 1.4836, time 10.10ms, mfu 0.48%\n",
            "iter 48213: loss 1.4124, time 9.75ms, mfu 0.49%\n",
            "iter 48214: loss 1.4732, time 9.78ms, mfu 0.49%\n",
            "iter 48215: loss 1.5420, time 9.36ms, mfu 0.52%\n",
            "iter 48216: loss 1.4089, time 9.83ms, mfu 0.49%\n",
            "iter 48217: loss 1.4424, time 9.76ms, mfu 0.49%\n",
            "iter 48218: loss 1.4460, time 10.80ms, mfu 0.45%\n",
            "iter 48219: loss 1.3211, time 9.66ms, mfu 0.50%\n",
            "iter 48220: loss 1.4913, time 11.47ms, mfu 0.42%\n",
            "iter 48221: loss 1.5557, time 9.80ms, mfu 0.49%\n",
            "iter 48222: loss 1.4995, time 9.43ms, mfu 0.51%\n",
            "iter 48223: loss 1.4501, time 13.67ms, mfu 0.35%\n",
            "iter 48224: loss 1.4420, time 9.88ms, mfu 0.49%\n",
            "iter 48225: loss 1.5238, time 9.86ms, mfu 0.49%\n",
            "iter 48226: loss 1.5856, time 9.46ms, mfu 0.51%\n",
            "iter 48227: loss 1.4202, time 9.63ms, mfu 0.50%\n",
            "iter 48228: loss 1.4452, time 9.64ms, mfu 0.50%\n",
            "iter 48229: loss 1.3427, time 9.72ms, mfu 0.50%\n",
            "iter 48230: loss 1.4274, time 9.31ms, mfu 0.52%\n",
            "iter 48231: loss 1.5524, time 9.27ms, mfu 0.52%\n",
            "iter 48232: loss 1.4610, time 9.67ms, mfu 0.50%\n",
            "iter 48233: loss 1.4927, time 9.70ms, mfu 0.50%\n",
            "iter 48234: loss 1.2986, time 12.85ms, mfu 0.38%\n",
            "iter 48235: loss 1.5085, time 9.71ms, mfu 0.50%\n",
            "iter 48236: loss 1.4096, time 9.52ms, mfu 0.51%\n",
            "iter 48237: loss 1.4780, time 9.46ms, mfu 0.51%\n",
            "iter 48238: loss 1.5441, time 9.85ms, mfu 0.49%\n",
            "iter 48239: loss 1.5370, time 11.79ms, mfu 0.41%\n",
            "iter 48240: loss 1.4856, time 14.11ms, mfu 0.34%\n",
            "iter 48241: loss 1.4422, time 9.65ms, mfu 0.50%\n",
            "iter 48242: loss 1.5503, time 9.32ms, mfu 0.52%\n",
            "iter 48243: loss 1.4100, time 10.43ms, mfu 0.46%\n",
            "iter 48244: loss 1.3754, time 13.46ms, mfu 0.36%\n",
            "iter 48245: loss 1.4807, time 9.37ms, mfu 0.51%\n",
            "iter 48246: loss 1.5672, time 9.33ms, mfu 0.52%\n",
            "iter 48247: loss 1.3973, time 9.48ms, mfu 0.51%\n",
            "iter 48248: loss 1.4736, time 9.96ms, mfu 0.48%\n",
            "iter 48249: loss 1.3987, time 9.63ms, mfu 0.50%\n",
            "iter 48250: loss 1.4688, time 9.43ms, mfu 0.51%\n",
            "iter 48251: loss 1.4749, time 9.68ms, mfu 0.50%\n",
            "iter 48252: loss 1.5480, time 9.60ms, mfu 0.50%\n",
            "iter 48253: loss 1.5919, time 10.40ms, mfu 0.46%\n",
            "iter 48254: loss 1.5335, time 9.60ms, mfu 0.50%\n",
            "iter 48255: loss 1.4922, time 9.44ms, mfu 0.51%\n",
            "iter 48256: loss 1.4294, time 9.44ms, mfu 0.51%\n",
            "iter 48257: loss 1.4932, time 10.97ms, mfu 0.44%\n",
            "iter 48258: loss 1.4566, time 11.96ms, mfu 0.40%\n",
            "iter 48259: loss 1.4591, time 16.66ms, mfu 0.29%\n",
            "iter 48260: loss 1.4332, time 15.11ms, mfu 0.32%\n",
            "iter 48261: loss 1.4744, time 14.17ms, mfu 0.34%\n",
            "iter 48262: loss 1.4045, time 17.04ms, mfu 0.28%\n",
            "iter 48263: loss 1.5418, time 15.77ms, mfu 0.31%\n",
            "iter 48264: loss 1.5349, time 16.87ms, mfu 0.29%\n",
            "iter 48265: loss 1.4920, time 17.53ms, mfu 0.28%\n",
            "iter 48266: loss 1.6098, time 13.85ms, mfu 0.35%\n",
            "iter 48267: loss 1.5990, time 17.18ms, mfu 0.28%\n",
            "iter 48268: loss 1.4515, time 17.59ms, mfu 0.27%\n",
            "iter 48269: loss 1.4959, time 17.14ms, mfu 0.28%\n",
            "iter 48270: loss 1.4182, time 17.15ms, mfu 0.28%\n",
            "iter 48271: loss 1.5289, time 15.79ms, mfu 0.31%\n",
            "iter 48272: loss 1.4976, time 15.02ms, mfu 0.32%\n",
            "iter 48273: loss 1.4084, time 14.12ms, mfu 0.34%\n",
            "iter 48274: loss 1.4517, time 17.18ms, mfu 0.28%\n",
            "iter 48275: loss 1.4459, time 30.80ms, mfu 0.16%\n",
            "iter 48276: loss 1.4905, time 18.27ms, mfu 0.26%\n",
            "iter 48277: loss 1.4954, time 16.85ms, mfu 0.29%\n",
            "iter 48278: loss 1.4230, time 15.56ms, mfu 0.31%\n",
            "iter 48279: loss 1.5680, time 17.01ms, mfu 0.28%\n",
            "iter 48280: loss 1.3755, time 17.24ms, mfu 0.28%\n",
            "iter 48281: loss 1.4588, time 19.23ms, mfu 0.25%\n",
            "iter 48282: loss 1.5575, time 19.66ms, mfu 0.25%\n",
            "iter 48283: loss 1.4247, time 14.07ms, mfu 0.34%\n",
            "iter 48284: loss 1.5311, time 28.78ms, mfu 0.17%\n",
            "iter 48285: loss 1.3697, time 23.07ms, mfu 0.21%\n",
            "iter 48286: loss 1.4608, time 22.36ms, mfu 0.22%\n",
            "iter 48287: loss 1.4401, time 22.76ms, mfu 0.21%\n",
            "iter 48288: loss 1.5524, time 25.03ms, mfu 0.19%\n",
            "iter 48289: loss 1.4533, time 40.65ms, mfu 0.12%\n",
            "iter 48290: loss 1.4241, time 37.36ms, mfu 0.13%\n",
            "iter 48291: loss 1.3576, time 20.51ms, mfu 0.24%\n",
            "iter 48292: loss 1.5407, time 15.84ms, mfu 0.30%\n",
            "iter 48293: loss 1.3709, time 16.76ms, mfu 0.29%\n",
            "iter 48294: loss 1.6066, time 16.86ms, mfu 0.29%\n",
            "iter 48295: loss 1.4752, time 19.44ms, mfu 0.25%\n",
            "iter 48296: loss 1.4181, time 22.58ms, mfu 0.21%\n",
            "iter 48297: loss 1.3864, time 19.77ms, mfu 0.24%\n",
            "iter 48298: loss 1.4542, time 17.02ms, mfu 0.28%\n",
            "iter 48299: loss 1.4689, time 24.00ms, mfu 0.20%\n",
            "iter 48300: loss 1.6003, time 26.29ms, mfu 0.18%\n",
            "iter 48301: loss 1.5869, time 16.97ms, mfu 0.28%\n",
            "iter 48302: loss 1.5010, time 14.72ms, mfu 0.33%\n",
            "iter 48303: loss 1.4978, time 26.19ms, mfu 0.18%\n",
            "iter 48304: loss 1.4201, time 16.91ms, mfu 0.29%\n",
            "iter 48305: loss 1.4370, time 15.95ms, mfu 0.30%\n",
            "iter 48306: loss 1.5406, time 33.92ms, mfu 0.14%\n",
            "iter 48307: loss 1.4889, time 21.39ms, mfu 0.23%\n",
            "iter 48308: loss 1.4955, time 26.12ms, mfu 0.18%\n",
            "iter 48309: loss 1.3889, time 18.15ms, mfu 0.27%\n",
            "iter 48310: loss 1.4903, time 10.21ms, mfu 0.47%\n",
            "iter 48311: loss 1.4634, time 9.37ms, mfu 0.51%\n",
            "iter 48312: loss 1.5497, time 9.38ms, mfu 0.51%\n",
            "iter 48313: loss 1.4863, time 9.36ms, mfu 0.52%\n",
            "iter 48314: loss 1.4066, time 9.45ms, mfu 0.51%\n",
            "iter 48315: loss 1.4228, time 9.59ms, mfu 0.50%\n",
            "iter 48316: loss 1.4616, time 9.44ms, mfu 0.51%\n",
            "iter 48317: loss 1.5116, time 9.54ms, mfu 0.51%\n",
            "iter 48318: loss 1.3495, time 9.44ms, mfu 0.51%\n",
            "iter 48319: loss 1.4416, time 9.59ms, mfu 0.50%\n",
            "iter 48320: loss 1.3900, time 9.51ms, mfu 0.51%\n",
            "iter 48321: loss 1.4853, time 9.48ms, mfu 0.51%\n",
            "iter 48322: loss 1.5398, time 9.45ms, mfu 0.51%\n",
            "iter 48323: loss 1.4304, time 9.86ms, mfu 0.49%\n",
            "iter 48324: loss 1.4261, time 15.37ms, mfu 0.31%\n",
            "iter 48325: loss 1.3988, time 10.80ms, mfu 0.45%\n",
            "iter 48326: loss 1.5370, time 10.01ms, mfu 0.48%\n",
            "iter 48327: loss 1.4531, time 9.45ms, mfu 0.51%\n",
            "iter 48328: loss 1.4191, time 9.37ms, mfu 0.51%\n",
            "iter 48329: loss 1.5266, time 9.36ms, mfu 0.52%\n",
            "iter 48330: loss 1.5132, time 9.73ms, mfu 0.50%\n",
            "iter 48331: loss 1.5394, time 9.49ms, mfu 0.51%\n",
            "iter 48332: loss 1.5000, time 9.30ms, mfu 0.52%\n",
            "iter 48333: loss 1.4997, time 9.44ms, mfu 0.51%\n",
            "iter 48334: loss 1.5139, time 9.57ms, mfu 0.50%\n",
            "iter 48335: loss 1.5514, time 9.39ms, mfu 0.51%\n",
            "iter 48336: loss 1.5290, time 9.30ms, mfu 0.52%\n",
            "iter 48337: loss 1.3925, time 9.42ms, mfu 0.51%\n",
            "iter 48338: loss 1.4905, time 17.73ms, mfu 0.27%\n",
            "iter 48339: loss 1.5263, time 12.67ms, mfu 0.38%\n",
            "iter 48340: loss 1.4731, time 9.70ms, mfu 0.50%\n",
            "iter 48341: loss 1.5596, time 9.79ms, mfu 0.49%\n",
            "iter 48342: loss 1.4423, time 11.28ms, mfu 0.43%\n",
            "iter 48343: loss 1.4733, time 9.42ms, mfu 0.51%\n",
            "iter 48344: loss 1.5538, time 13.68ms, mfu 0.35%\n",
            "iter 48345: loss 1.4943, time 9.76ms, mfu 0.49%\n",
            "iter 48346: loss 1.4656, time 9.43ms, mfu 0.51%\n",
            "iter 48347: loss 1.5001, time 13.08ms, mfu 0.37%\n",
            "iter 48348: loss 1.5861, time 10.39ms, mfu 0.46%\n",
            "iter 48349: loss 1.4523, time 9.41ms, mfu 0.51%\n",
            "iter 48350: loss 1.5591, time 9.44ms, mfu 0.51%\n",
            "iter 48351: loss 1.4033, time 9.52ms, mfu 0.51%\n",
            "iter 48352: loss 1.4561, time 9.55ms, mfu 0.51%\n",
            "iter 48353: loss 1.4920, time 9.30ms, mfu 0.52%\n",
            "iter 48354: loss 1.3893, time 9.53ms, mfu 0.51%\n",
            "iter 48355: loss 1.5175, time 9.63ms, mfu 0.50%\n",
            "iter 48356: loss 1.4054, time 9.78ms, mfu 0.49%\n",
            "iter 48357: loss 1.5225, time 9.82ms, mfu 0.49%\n",
            "iter 48358: loss 1.4403, time 11.24ms, mfu 0.43%\n",
            "iter 48359: loss 1.5856, time 9.46ms, mfu 0.51%\n",
            "iter 48360: loss 1.4888, time 9.44ms, mfu 0.51%\n",
            "iter 48361: loss 1.4052, time 9.57ms, mfu 0.50%\n",
            "iter 48362: loss 1.4245, time 10.02ms, mfu 0.48%\n",
            "iter 48363: loss 1.4706, time 9.30ms, mfu 0.52%\n",
            "iter 48364: loss 1.5053, time 9.53ms, mfu 0.51%\n",
            "iter 48365: loss 1.4491, time 15.00ms, mfu 0.32%\n",
            "iter 48366: loss 1.4585, time 10.94ms, mfu 0.44%\n",
            "iter 48367: loss 1.4551, time 10.19ms, mfu 0.47%\n",
            "iter 48368: loss 1.4483, time 11.61ms, mfu 0.42%\n",
            "iter 48369: loss 1.4577, time 10.01ms, mfu 0.48%\n",
            "iter 48370: loss 1.4798, time 9.49ms, mfu 0.51%\n",
            "iter 48371: loss 1.5997, time 9.58ms, mfu 0.50%\n",
            "iter 48372: loss 1.4195, time 9.40ms, mfu 0.51%\n",
            "iter 48373: loss 1.3793, time 9.45ms, mfu 0.51%\n",
            "iter 48374: loss 1.4943, time 10.15ms, mfu 0.48%\n",
            "iter 48375: loss 1.3950, time 9.88ms, mfu 0.49%\n",
            "iter 48376: loss 1.3558, time 9.81ms, mfu 0.49%\n",
            "iter 48377: loss 1.5252, time 9.70ms, mfu 0.50%\n",
            "iter 48378: loss 1.5369, time 9.30ms, mfu 0.52%\n",
            "iter 48379: loss 1.4206, time 9.50ms, mfu 0.51%\n",
            "iter 48380: loss 1.3781, time 9.36ms, mfu 0.52%\n",
            "iter 48381: loss 1.4255, time 10.18ms, mfu 0.47%\n",
            "iter 48382: loss 1.3770, time 11.46ms, mfu 0.42%\n",
            "iter 48383: loss 1.4169, time 9.66ms, mfu 0.50%\n",
            "iter 48384: loss 1.5153, time 9.54ms, mfu 0.51%\n",
            "iter 48385: loss 1.3908, time 9.40ms, mfu 0.51%\n",
            "iter 48386: loss 1.4801, time 14.29ms, mfu 0.34%\n",
            "iter 48387: loss 1.3616, time 13.73ms, mfu 0.35%\n",
            "iter 48388: loss 1.4997, time 10.89ms, mfu 0.44%\n",
            "iter 48389: loss 1.4649, time 9.53ms, mfu 0.51%\n",
            "iter 48390: loss 1.4549, time 9.50ms, mfu 0.51%\n",
            "iter 48391: loss 1.5292, time 9.63ms, mfu 0.50%\n",
            "iter 48392: loss 1.4028, time 9.31ms, mfu 0.52%\n",
            "iter 48393: loss 1.5152, time 9.32ms, mfu 0.52%\n",
            "iter 48394: loss 1.5232, time 9.31ms, mfu 0.52%\n",
            "iter 48395: loss 1.4291, time 9.53ms, mfu 0.51%\n",
            "iter 48396: loss 1.4856, time 9.42ms, mfu 0.51%\n",
            "iter 48397: loss 1.5316, time 9.60ms, mfu 0.50%\n",
            "iter 48398: loss 1.4565, time 12.08ms, mfu 0.40%\n",
            "iter 48399: loss 1.4533, time 9.66ms, mfu 0.50%\n",
            "iter 48400: loss 1.4709, time 9.79ms, mfu 0.49%\n",
            "iter 48401: loss 1.5069, time 10.24ms, mfu 0.47%\n",
            "iter 48402: loss 1.4991, time 9.43ms, mfu 0.51%\n",
            "iter 48403: loss 1.5390, time 9.45ms, mfu 0.51%\n",
            "iter 48404: loss 1.5340, time 9.36ms, mfu 0.52%\n",
            "iter 48405: loss 1.4413, time 9.65ms, mfu 0.50%\n",
            "iter 48406: loss 1.3072, time 9.47ms, mfu 0.51%\n",
            "iter 48407: loss 1.4874, time 13.57ms, mfu 0.36%\n",
            "iter 48408: loss 1.4071, time 9.93ms, mfu 0.49%\n",
            "iter 48409: loss 1.6497, time 11.87ms, mfu 0.41%\n",
            "iter 48410: loss 1.5035, time 9.34ms, mfu 0.52%\n",
            "iter 48411: loss 1.5012, time 9.57ms, mfu 0.50%\n",
            "iter 48412: loss 1.4498, time 9.48ms, mfu 0.51%\n",
            "iter 48413: loss 1.4135, time 9.83ms, mfu 0.49%\n",
            "iter 48414: loss 1.5974, time 9.45ms, mfu 0.51%\n",
            "iter 48415: loss 1.4419, time 9.44ms, mfu 0.51%\n",
            "iter 48416: loss 1.5166, time 9.73ms, mfu 0.50%\n",
            "iter 48417: loss 1.5073, time 9.53ms, mfu 0.51%\n",
            "iter 48418: loss 1.6183, time 9.79ms, mfu 0.49%\n",
            "iter 48419: loss 1.4148, time 9.37ms, mfu 0.51%\n",
            "iter 48420: loss 1.5499, time 9.82ms, mfu 0.49%\n",
            "iter 48421: loss 1.4345, time 9.81ms, mfu 0.49%\n",
            "iter 48422: loss 1.5587, time 9.55ms, mfu 0.51%\n",
            "iter 48423: loss 1.4031, time 9.35ms, mfu 0.52%\n",
            "iter 48424: loss 1.4808, time 9.57ms, mfu 0.50%\n",
            "iter 48425: loss 1.4755, time 9.64ms, mfu 0.50%\n",
            "iter 48426: loss 1.4786, time 9.40ms, mfu 0.51%\n",
            "iter 48427: loss 1.4515, time 9.60ms, mfu 0.50%\n",
            "iter 48428: loss 1.4144, time 12.81ms, mfu 0.38%\n",
            "iter 48429: loss 1.4594, time 11.21ms, mfu 0.43%\n",
            "iter 48430: loss 1.5477, time 9.28ms, mfu 0.52%\n",
            "iter 48431: loss 1.4094, time 9.39ms, mfu 0.51%\n",
            "iter 48432: loss 1.4925, time 9.33ms, mfu 0.52%\n",
            "iter 48433: loss 1.4382, time 9.58ms, mfu 0.50%\n",
            "iter 48434: loss 1.5199, time 9.44ms, mfu 0.51%\n",
            "iter 48435: loss 1.4513, time 9.33ms, mfu 0.52%\n",
            "iter 48436: loss 1.5098, time 9.47ms, mfu 0.51%\n",
            "iter 48437: loss 1.3907, time 9.31ms, mfu 0.52%\n",
            "iter 48438: loss 1.5126, time 9.34ms, mfu 0.52%\n",
            "iter 48439: loss 1.4839, time 14.48ms, mfu 0.33%\n",
            "iter 48440: loss 1.5487, time 14.10ms, mfu 0.34%\n",
            "iter 48441: loss 1.4324, time 9.90ms, mfu 0.49%\n",
            "iter 48442: loss 1.3890, time 9.36ms, mfu 0.52%\n",
            "iter 48443: loss 1.4484, time 9.32ms, mfu 0.52%\n",
            "iter 48444: loss 1.5116, time 9.31ms, mfu 0.52%\n",
            "iter 48445: loss 1.4616, time 9.28ms, mfu 0.52%\n",
            "iter 48446: loss 1.5346, time 9.52ms, mfu 0.51%\n",
            "iter 48447: loss 1.4576, time 9.32ms, mfu 0.52%\n",
            "iter 48448: loss 1.3730, time 9.42ms, mfu 0.51%\n",
            "iter 48449: loss 1.4066, time 13.72ms, mfu 0.35%\n",
            "iter 48450: loss 1.4438, time 9.77ms, mfu 0.49%\n",
            "iter 48451: loss 1.3821, time 9.41ms, mfu 0.51%\n",
            "iter 48452: loss 1.4762, time 11.98ms, mfu 0.40%\n",
            "iter 48453: loss 1.3642, time 9.60ms, mfu 0.50%\n",
            "iter 48454: loss 1.5008, time 9.35ms, mfu 0.52%\n",
            "iter 48455: loss 1.3760, time 9.41ms, mfu 0.51%\n",
            "iter 48456: loss 1.5714, time 10.56ms, mfu 0.46%\n",
            "iter 48457: loss 1.4834, time 11.71ms, mfu 0.41%\n",
            "iter 48458: loss 1.4614, time 9.86ms, mfu 0.49%\n",
            "iter 48459: loss 1.4721, time 9.38ms, mfu 0.51%\n",
            "iter 48460: loss 1.4714, time 9.59ms, mfu 0.50%\n",
            "iter 48461: loss 1.4296, time 10.69ms, mfu 0.45%\n",
            "iter 48462: loss 1.4776, time 9.30ms, mfu 0.52%\n",
            "iter 48463: loss 1.4329, time 9.41ms, mfu 0.51%\n",
            "iter 48464: loss 1.3238, time 9.51ms, mfu 0.51%\n",
            "iter 48465: loss 1.5720, time 9.58ms, mfu 0.50%\n",
            "iter 48466: loss 1.4634, time 9.65ms, mfu 0.50%\n",
            "iter 48467: loss 1.4387, time 9.61ms, mfu 0.50%\n",
            "iter 48468: loss 1.3770, time 10.83ms, mfu 0.45%\n",
            "iter 48469: loss 1.4860, time 9.78ms, mfu 0.49%\n",
            "iter 48470: loss 1.5123, time 14.61ms, mfu 0.33%\n",
            "iter 48471: loss 1.4111, time 10.10ms, mfu 0.48%\n",
            "iter 48472: loss 1.4605, time 9.98ms, mfu 0.48%\n",
            "iter 48473: loss 1.5637, time 9.31ms, mfu 0.52%\n",
            "iter 48474: loss 1.5199, time 9.83ms, mfu 0.49%\n",
            "iter 48475: loss 1.4595, time 9.64ms, mfu 0.50%\n",
            "iter 48476: loss 1.3622, time 9.71ms, mfu 0.50%\n",
            "iter 48477: loss 1.5165, time 9.39ms, mfu 0.51%\n",
            "iter 48478: loss 1.5739, time 9.22ms, mfu 0.52%\n",
            "iter 48479: loss 1.3732, time 9.31ms, mfu 0.52%\n",
            "iter 48480: loss 1.4181, time 9.37ms, mfu 0.51%\n",
            "iter 48481: loss 1.4871, time 11.58ms, mfu 0.42%\n",
            "iter 48482: loss 1.5790, time 9.26ms, mfu 0.52%\n",
            "iter 48483: loss 1.6302, time 9.60ms, mfu 0.50%\n",
            "iter 48484: loss 1.4183, time 9.27ms, mfu 0.52%\n",
            "iter 48485: loss 1.4704, time 9.33ms, mfu 0.52%\n",
            "iter 48486: loss 1.4315, time 11.21ms, mfu 0.43%\n",
            "iter 48487: loss 1.4696, time 12.00ms, mfu 0.40%\n",
            "iter 48488: loss 1.4208, time 9.37ms, mfu 0.51%\n",
            "iter 48489: loss 1.4518, time 9.26ms, mfu 0.52%\n",
            "iter 48490: loss 1.4585, time 9.27ms, mfu 0.52%\n",
            "iter 48491: loss 1.4112, time 12.90ms, mfu 0.37%\n",
            "iter 48492: loss 1.4704, time 10.99ms, mfu 0.44%\n",
            "iter 48493: loss 1.3962, time 9.42ms, mfu 0.51%\n",
            "iter 48494: loss 1.5185, time 9.20ms, mfu 0.52%\n",
            "iter 48495: loss 1.4540, time 9.38ms, mfu 0.51%\n",
            "iter 48496: loss 1.5100, time 9.19ms, mfu 0.52%\n",
            "iter 48497: loss 1.5590, time 9.24ms, mfu 0.52%\n",
            "iter 48498: loss 1.3767, time 9.24ms, mfu 0.52%\n",
            "iter 48499: loss 1.4787, time 9.43ms, mfu 0.51%\n",
            "step 48500: train loss 1.2934, val loss 1.6323\n",
            "iter 48500: loss 1.4479, time 17.71ms, mfu 0.27%\n",
            "iter 48501: loss 1.3270, time 9.40ms, mfu 0.51%\n",
            "iter 48502: loss 1.4173, time 9.19ms, mfu 0.52%\n",
            "iter 48503: loss 1.4629, time 9.27ms, mfu 0.52%\n",
            "iter 48504: loss 1.3949, time 9.28ms, mfu 0.52%\n",
            "iter 48505: loss 1.4417, time 9.25ms, mfu 0.52%\n",
            "iter 48506: loss 1.4497, time 9.27ms, mfu 0.52%\n",
            "iter 48507: loss 1.5178, time 9.28ms, mfu 0.52%\n",
            "iter 48508: loss 1.5272, time 9.65ms, mfu 0.50%\n",
            "iter 48509: loss 1.3945, time 9.41ms, mfu 0.51%\n",
            "iter 48510: loss 1.5158, time 9.16ms, mfu 0.53%\n",
            "iter 48511: loss 1.5280, time 9.44ms, mfu 0.51%\n",
            "iter 48512: loss 1.4202, time 13.19ms, mfu 0.37%\n",
            "iter 48513: loss 1.4088, time 9.58ms, mfu 0.50%\n",
            "iter 48514: loss 1.4936, time 9.27ms, mfu 0.52%\n",
            "iter 48515: loss 1.4304, time 9.18ms, mfu 0.53%\n",
            "iter 48516: loss 1.4403, time 9.26ms, mfu 0.52%\n",
            "iter 48517: loss 1.5337, time 9.97ms, mfu 0.48%\n",
            "iter 48518: loss 1.5189, time 9.27ms, mfu 0.52%\n",
            "iter 48519: loss 1.4894, time 9.21ms, mfu 0.52%\n",
            "iter 48520: loss 1.4824, time 9.19ms, mfu 0.52%\n",
            "iter 48521: loss 1.4817, time 9.65ms, mfu 0.50%\n",
            "iter 48522: loss 1.5213, time 9.26ms, mfu 0.52%\n",
            "iter 48523: loss 1.5901, time 9.19ms, mfu 0.52%\n",
            "iter 48524: loss 1.5143, time 9.48ms, mfu 0.51%\n",
            "iter 48525: loss 1.4092, time 9.46ms, mfu 0.51%\n",
            "iter 48526: loss 1.3605, time 9.42ms, mfu 0.51%\n",
            "iter 48527: loss 1.4429, time 9.28ms, mfu 0.52%\n",
            "iter 48528: loss 1.4513, time 9.21ms, mfu 0.52%\n",
            "iter 48529: loss 1.4289, time 9.19ms, mfu 0.52%\n",
            "iter 48530: loss 1.5052, time 9.76ms, mfu 0.49%\n",
            "iter 48531: loss 1.3722, time 9.30ms, mfu 0.52%\n",
            "iter 48532: loss 1.4246, time 11.71ms, mfu 0.41%\n",
            "iter 48533: loss 1.3099, time 9.15ms, mfu 0.53%\n",
            "iter 48534: loss 1.5126, time 14.55ms, mfu 0.33%\n",
            "iter 48535: loss 1.4508, time 9.67ms, mfu 0.50%\n",
            "iter 48536: loss 1.4728, time 9.37ms, mfu 0.51%\n",
            "iter 48537: loss 1.4768, time 9.24ms, mfu 0.52%\n",
            "iter 48538: loss 1.4606, time 9.36ms, mfu 0.52%\n",
            "iter 48539: loss 1.5538, time 9.73ms, mfu 0.50%\n",
            "iter 48540: loss 1.5682, time 9.23ms, mfu 0.52%\n",
            "iter 48541: loss 1.4345, time 10.92ms, mfu 0.44%\n",
            "iter 48542: loss 1.4821, time 14.74ms, mfu 0.33%\n",
            "iter 48543: loss 1.4473, time 11.96ms, mfu 0.40%\n",
            "iter 48544: loss 1.5269, time 9.46ms, mfu 0.51%\n",
            "iter 48545: loss 1.4327, time 9.19ms, mfu 0.52%\n",
            "iter 48546: loss 1.5060, time 9.42ms, mfu 0.51%\n",
            "iter 48547: loss 1.4369, time 10.62ms, mfu 0.45%\n",
            "iter 48548: loss 1.4804, time 9.36ms, mfu 0.52%\n",
            "iter 48549: loss 1.5216, time 9.50ms, mfu 0.51%\n",
            "iter 48550: loss 1.4309, time 9.29ms, mfu 0.52%\n",
            "iter 48551: loss 1.4065, time 10.57ms, mfu 0.46%\n",
            "iter 48552: loss 1.4880, time 9.36ms, mfu 0.52%\n",
            "iter 48553: loss 1.4345, time 9.30ms, mfu 0.52%\n",
            "iter 48554: loss 1.4193, time 9.60ms, mfu 0.50%\n",
            "iter 48555: loss 1.6067, time 13.26ms, mfu 0.36%\n",
            "iter 48556: loss 1.5566, time 9.49ms, mfu 0.51%\n",
            "iter 48557: loss 1.5711, time 9.30ms, mfu 0.52%\n",
            "iter 48558: loss 1.5406, time 10.39ms, mfu 0.46%\n",
            "iter 48559: loss 1.5496, time 9.49ms, mfu 0.51%\n",
            "iter 48560: loss 1.3901, time 9.55ms, mfu 0.50%\n",
            "iter 48561: loss 1.4250, time 9.32ms, mfu 0.52%\n",
            "iter 48562: loss 1.4628, time 9.45ms, mfu 0.51%\n",
            "iter 48563: loss 1.4680, time 9.38ms, mfu 0.51%\n",
            "iter 48564: loss 1.5011, time 9.37ms, mfu 0.51%\n",
            "iter 48565: loss 1.4873, time 9.25ms, mfu 0.52%\n",
            "iter 48566: loss 1.5058, time 9.72ms, mfu 0.50%\n",
            "iter 48567: loss 1.5316, time 9.44ms, mfu 0.51%\n",
            "iter 48568: loss 1.5424, time 9.65ms, mfu 0.50%\n",
            "iter 48569: loss 1.5359, time 13.41ms, mfu 0.36%\n",
            "iter 48570: loss 1.3709, time 9.96ms, mfu 0.48%\n",
            "iter 48571: loss 1.5284, time 9.39ms, mfu 0.51%\n",
            "iter 48572: loss 1.4264, time 9.24ms, mfu 0.52%\n",
            "iter 48573: loss 1.4798, time 9.40ms, mfu 0.51%\n",
            "iter 48574: loss 1.4346, time 9.27ms, mfu 0.52%\n",
            "iter 48575: loss 1.5127, time 9.61ms, mfu 0.50%\n",
            "iter 48576: loss 1.4484, time 13.99ms, mfu 0.34%\n",
            "iter 48577: loss 1.4992, time 10.74ms, mfu 0.45%\n",
            "iter 48578: loss 1.5185, time 9.15ms, mfu 0.53%\n",
            "iter 48579: loss 1.4382, time 9.33ms, mfu 0.52%\n",
            "iter 48580: loss 1.5361, time 9.24ms, mfu 0.52%\n",
            "iter 48581: loss 1.4350, time 9.60ms, mfu 0.50%\n",
            "iter 48582: loss 1.5103, time 10.96ms, mfu 0.44%\n",
            "iter 48583: loss 1.5430, time 9.58ms, mfu 0.50%\n",
            "iter 48584: loss 1.5430, time 9.53ms, mfu 0.51%\n",
            "iter 48585: loss 1.4497, time 9.34ms, mfu 0.52%\n",
            "iter 48586: loss 1.4871, time 9.85ms, mfu 0.49%\n",
            "iter 48587: loss 1.3808, time 11.68ms, mfu 0.41%\n",
            "iter 48588: loss 1.5032, time 11.55ms, mfu 0.42%\n",
            "iter 48589: loss 1.4115, time 9.50ms, mfu 0.51%\n",
            "iter 48590: loss 1.5300, time 9.20ms, mfu 0.52%\n",
            "iter 48591: loss 1.4770, time 9.31ms, mfu 0.52%\n",
            "iter 48592: loss 1.4665, time 9.26ms, mfu 0.52%\n",
            "iter 48593: loss 1.5115, time 9.38ms, mfu 0.51%\n",
            "iter 48594: loss 1.4124, time 9.39ms, mfu 0.51%\n",
            "iter 48595: loss 1.4081, time 9.17ms, mfu 0.53%\n",
            "iter 48596: loss 1.5023, time 9.24ms, mfu 0.52%\n",
            "iter 48597: loss 1.4611, time 12.12ms, mfu 0.40%\n",
            "iter 48598: loss 1.4596, time 11.68ms, mfu 0.41%\n",
            "iter 48599: loss 1.4183, time 9.24ms, mfu 0.52%\n",
            "iter 48600: loss 1.4905, time 9.35ms, mfu 0.52%\n",
            "iter 48601: loss 1.5198, time 11.34ms, mfu 0.43%\n",
            "iter 48602: loss 1.4701, time 9.38ms, mfu 0.51%\n",
            "iter 48603: loss 1.4492, time 9.29ms, mfu 0.52%\n",
            "iter 48604: loss 1.4315, time 9.32ms, mfu 0.52%\n",
            "iter 48605: loss 1.2853, time 9.30ms, mfu 0.52%\n",
            "iter 48606: loss 1.5458, time 9.26ms, mfu 0.52%\n",
            "iter 48607: loss 1.4714, time 10.26ms, mfu 0.47%\n",
            "iter 48608: loss 1.5821, time 11.24ms, mfu 0.43%\n",
            "iter 48609: loss 1.5130, time 9.25ms, mfu 0.52%\n",
            "iter 48610: loss 1.4689, time 9.53ms, mfu 0.51%\n",
            "iter 48611: loss 1.4028, time 9.42ms, mfu 0.51%\n",
            "iter 48612: loss 1.4693, time 9.51ms, mfu 0.51%\n",
            "iter 48613: loss 1.3894, time 9.32ms, mfu 0.52%\n",
            "iter 48614: loss 1.4388, time 9.30ms, mfu 0.52%\n",
            "iter 48615: loss 1.3797, time 9.52ms, mfu 0.51%\n",
            "iter 48616: loss 1.3541, time 10.50ms, mfu 0.46%\n",
            "iter 48617: loss 1.5630, time 9.30ms, mfu 0.52%\n",
            "iter 48618: loss 1.4857, time 9.49ms, mfu 0.51%\n",
            "iter 48619: loss 1.5966, time 14.33ms, mfu 0.34%\n",
            "iter 48620: loss 1.4931, time 9.49ms, mfu 0.51%\n",
            "iter 48621: loss 1.5044, time 9.37ms, mfu 0.51%\n",
            "iter 48622: loss 1.4665, time 9.73ms, mfu 0.50%\n",
            "iter 48623: loss 1.4000, time 9.27ms, mfu 0.52%\n",
            "iter 48624: loss 1.4467, time 9.39ms, mfu 0.51%\n",
            "iter 48625: loss 1.5349, time 9.58ms, mfu 0.50%\n",
            "iter 48626: loss 1.4639, time 10.53ms, mfu 0.46%\n",
            "iter 48627: loss 1.4398, time 15.54ms, mfu 0.31%\n",
            "iter 48628: loss 1.4944, time 16.44ms, mfu 0.29%\n",
            "iter 48629: loss 1.4128, time 16.90ms, mfu 0.29%\n",
            "iter 48630: loss 1.4361, time 15.88ms, mfu 0.30%\n",
            "iter 48631: loss 1.3898, time 16.89ms, mfu 0.29%\n",
            "iter 48632: loss 1.5563, time 17.18ms, mfu 0.28%\n",
            "iter 48633: loss 1.4140, time 17.26ms, mfu 0.28%\n",
            "iter 48634: loss 1.3828, time 28.67ms, mfu 0.17%\n",
            "iter 48635: loss 1.5502, time 22.72ms, mfu 0.21%\n",
            "iter 48636: loss 1.4527, time 29.41ms, mfu 0.16%\n",
            "iter 48637: loss 1.4515, time 22.87ms, mfu 0.21%\n",
            "iter 48638: loss 1.4792, time 32.20ms, mfu 0.15%\n",
            "iter 48639: loss 1.4586, time 20.83ms, mfu 0.23%\n",
            "iter 48640: loss 1.4669, time 22.47ms, mfu 0.21%\n",
            "iter 48641: loss 1.5244, time 27.03ms, mfu 0.18%\n",
            "iter 48642: loss 1.4365, time 25.69ms, mfu 0.19%\n",
            "iter 48643: loss 1.4215, time 16.66ms, mfu 0.29%\n",
            "iter 48644: loss 1.4690, time 28.57ms, mfu 0.17%\n",
            "iter 48645: loss 1.5429, time 17.68ms, mfu 0.27%\n",
            "iter 48646: loss 1.4390, time 18.49ms, mfu 0.26%\n",
            "iter 48647: loss 1.4040, time 28.83ms, mfu 0.17%\n",
            "iter 48648: loss 1.3997, time 23.00ms, mfu 0.21%\n",
            "iter 48649: loss 1.4892, time 20.46ms, mfu 0.24%\n",
            "iter 48650: loss 1.5811, time 14.87ms, mfu 0.32%\n",
            "iter 48651: loss 1.4745, time 14.82ms, mfu 0.33%\n",
            "iter 48652: loss 1.4543, time 13.33ms, mfu 0.36%\n",
            "iter 48653: loss 1.3712, time 15.77ms, mfu 0.31%\n",
            "iter 48654: loss 1.3808, time 25.97ms, mfu 0.19%\n",
            "iter 48655: loss 1.4625, time 28.83ms, mfu 0.17%\n",
            "iter 48656: loss 1.3841, time 13.26ms, mfu 0.36%\n",
            "iter 48657: loss 1.5319, time 12.22ms, mfu 0.39%\n",
            "iter 48658: loss 1.4741, time 12.13ms, mfu 0.40%\n",
            "iter 48659: loss 1.3384, time 12.28ms, mfu 0.39%\n",
            "iter 48660: loss 1.5808, time 12.20ms, mfu 0.40%\n",
            "iter 48661: loss 1.5026, time 12.21ms, mfu 0.40%\n",
            "iter 48662: loss 1.5148, time 12.13ms, mfu 0.40%\n",
            "iter 48663: loss 1.4435, time 12.20ms, mfu 0.40%\n",
            "iter 48664: loss 1.4027, time 12.39ms, mfu 0.39%\n",
            "iter 48665: loss 1.4308, time 12.22ms, mfu 0.39%\n",
            "iter 48666: loss 1.4524, time 12.86ms, mfu 0.37%\n",
            "iter 48667: loss 1.4539, time 12.52ms, mfu 0.39%\n",
            "iter 48668: loss 1.4742, time 12.30ms, mfu 0.39%\n",
            "iter 48669: loss 1.5077, time 12.23ms, mfu 0.39%\n",
            "iter 48670: loss 1.4963, time 18.14ms, mfu 0.27%\n",
            "iter 48671: loss 1.3359, time 12.90ms, mfu 0.37%\n",
            "iter 48672: loss 1.3888, time 18.99ms, mfu 0.25%\n",
            "iter 48673: loss 1.3987, time 27.08ms, mfu 0.18%\n",
            "iter 48674: loss 1.2923, time 13.25ms, mfu 0.36%\n",
            "iter 48675: loss 1.4074, time 12.10ms, mfu 0.40%\n",
            "iter 48676: loss 1.4388, time 13.43ms, mfu 0.36%\n",
            "iter 48677: loss 1.4227, time 12.93ms, mfu 0.37%\n",
            "iter 48678: loss 1.3786, time 13.00ms, mfu 0.37%\n",
            "iter 48679: loss 1.5137, time 11.67ms, mfu 0.41%\n",
            "iter 48680: loss 1.4362, time 11.87ms, mfu 0.41%\n",
            "iter 48681: loss 1.5215, time 15.04ms, mfu 0.32%\n",
            "iter 48682: loss 1.5155, time 11.83ms, mfu 0.41%\n",
            "iter 48683: loss 1.5151, time 14.21ms, mfu 0.34%\n",
            "iter 48684: loss 1.5744, time 11.84ms, mfu 0.41%\n",
            "iter 48685: loss 1.4267, time 15.80ms, mfu 0.31%\n",
            "iter 48686: loss 1.3434, time 11.94ms, mfu 0.40%\n",
            "iter 48687: loss 1.5601, time 11.96ms, mfu 0.40%\n",
            "iter 48688: loss 1.4233, time 11.67ms, mfu 0.41%\n",
            "iter 48689: loss 1.4351, time 12.20ms, mfu 0.40%\n",
            "iter 48690: loss 1.5463, time 11.89ms, mfu 0.41%\n",
            "iter 48691: loss 1.4854, time 11.78ms, mfu 0.41%\n",
            "iter 48692: loss 1.4659, time 11.80ms, mfu 0.41%\n",
            "iter 48693: loss 1.5832, time 19.96ms, mfu 0.24%\n",
            "iter 48694: loss 1.4487, time 21.34ms, mfu 0.23%\n",
            "iter 48695: loss 1.4327, time 11.80ms, mfu 0.41%\n",
            "iter 48696: loss 1.5363, time 12.43ms, mfu 0.39%\n",
            "iter 48697: loss 1.3713, time 12.68ms, mfu 0.38%\n",
            "iter 48698: loss 1.4513, time 11.78ms, mfu 0.41%\n",
            "iter 48699: loss 1.4085, time 11.69ms, mfu 0.41%\n",
            "iter 48700: loss 1.5179, time 16.41ms, mfu 0.29%\n",
            "iter 48701: loss 1.4044, time 23.08ms, mfu 0.21%\n",
            "iter 48702: loss 1.4764, time 22.84ms, mfu 0.21%\n",
            "iter 48703: loss 1.5034, time 47.46ms, mfu 0.10%\n",
            "iter 48704: loss 1.6141, time 40.69ms, mfu 0.12%\n",
            "iter 48705: loss 1.4186, time 22.73ms, mfu 0.21%\n",
            "iter 48706: loss 1.4702, time 15.24ms, mfu 0.32%\n",
            "iter 48707: loss 1.4560, time 15.17ms, mfu 0.32%\n",
            "iter 48708: loss 1.4523, time 16.06ms, mfu 0.30%\n",
            "iter 48709: loss 1.4099, time 17.13ms, mfu 0.28%\n",
            "iter 48710: loss 1.5168, time 14.75ms, mfu 0.33%\n",
            "iter 48711: loss 1.4949, time 18.83ms, mfu 0.26%\n",
            "iter 48712: loss 1.4334, time 13.48ms, mfu 0.36%\n",
            "iter 48713: loss 1.5052, time 11.94ms, mfu 0.40%\n",
            "iter 48714: loss 1.6237, time 15.09ms, mfu 0.32%\n",
            "iter 48715: loss 1.5027, time 15.17ms, mfu 0.32%\n",
            "iter 48716: loss 1.5108, time 12.63ms, mfu 0.38%\n",
            "iter 48717: loss 1.4289, time 16.93ms, mfu 0.28%\n",
            "iter 48718: loss 1.4268, time 24.04ms, mfu 0.20%\n",
            "iter 48719: loss 1.5063, time 30.39ms, mfu 0.16%\n",
            "iter 48720: loss 1.5007, time 24.14ms, mfu 0.20%\n",
            "iter 48721: loss 1.4196, time 27.21ms, mfu 0.18%\n",
            "iter 48722: loss 1.5271, time 30.99ms, mfu 0.16%\n",
            "iter 48723: loss 1.5467, time 14.50ms, mfu 0.33%\n",
            "iter 48724: loss 1.5775, time 16.96ms, mfu 0.28%\n",
            "iter 48725: loss 1.4911, time 12.95ms, mfu 0.37%\n",
            "iter 48726: loss 1.4565, time 13.46ms, mfu 0.36%\n",
            "iter 48727: loss 1.5327, time 18.77ms, mfu 0.26%\n",
            "iter 48728: loss 1.4651, time 26.24ms, mfu 0.18%\n",
            "iter 48729: loss 1.4622, time 14.61ms, mfu 0.33%\n",
            "iter 48730: loss 1.4193, time 14.70ms, mfu 0.33%\n",
            "iter 48731: loss 1.5507, time 30.90ms, mfu 0.16%\n",
            "iter 48732: loss 1.4046, time 18.60ms, mfu 0.26%\n",
            "iter 48733: loss 1.4695, time 13.06ms, mfu 0.37%\n",
            "iter 48734: loss 1.5833, time 12.03ms, mfu 0.40%\n",
            "iter 48735: loss 1.4492, time 12.12ms, mfu 0.40%\n",
            "iter 48736: loss 1.4951, time 14.53ms, mfu 0.33%\n",
            "iter 48737: loss 1.3817, time 12.03ms, mfu 0.40%\n",
            "iter 48738: loss 1.4204, time 12.60ms, mfu 0.38%\n",
            "iter 48739: loss 1.4741, time 12.26ms, mfu 0.39%\n",
            "iter 48740: loss 1.4580, time 12.27ms, mfu 0.39%\n",
            "iter 48741: loss 1.5357, time 12.66ms, mfu 0.38%\n",
            "iter 48742: loss 1.4646, time 12.04ms, mfu 0.40%\n",
            "iter 48743: loss 1.5704, time 25.16ms, mfu 0.19%\n",
            "iter 48744: loss 1.5499, time 20.15ms, mfu 0.24%\n",
            "iter 48745: loss 1.5242, time 14.65ms, mfu 0.33%\n",
            "iter 48746: loss 1.4666, time 17.01ms, mfu 0.28%\n",
            "iter 48747: loss 1.6024, time 11.94ms, mfu 0.40%\n",
            "iter 48748: loss 1.5087, time 11.97ms, mfu 0.40%\n",
            "iter 48749: loss 1.6008, time 12.23ms, mfu 0.39%\n",
            "iter 48750: loss 1.3652, time 13.14ms, mfu 0.37%\n",
            "iter 48751: loss 1.4618, time 12.99ms, mfu 0.37%\n",
            "iter 48752: loss 1.4627, time 17.18ms, mfu 0.28%\n",
            "iter 48753: loss 1.4740, time 24.31ms, mfu 0.20%\n",
            "iter 48754: loss 1.5256, time 28.41ms, mfu 0.17%\n",
            "iter 48755: loss 1.4702, time 31.36ms, mfu 0.15%\n",
            "iter 48756: loss 1.3557, time 31.61ms, mfu 0.15%\n",
            "iter 48757: loss 1.4931, time 29.69ms, mfu 0.16%\n",
            "iter 48758: loss 1.5504, time 28.07ms, mfu 0.17%\n",
            "iter 48759: loss 1.5721, time 19.47ms, mfu 0.25%\n",
            "iter 48760: loss 1.4495, time 20.05ms, mfu 0.24%\n",
            "iter 48761: loss 1.4413, time 11.88ms, mfu 0.41%\n",
            "iter 48762: loss 1.5968, time 11.93ms, mfu 0.40%\n",
            "iter 48763: loss 1.4654, time 11.73ms, mfu 0.41%\n",
            "iter 48764: loss 1.4999, time 12.33ms, mfu 0.39%\n",
            "iter 48765: loss 1.5002, time 23.59ms, mfu 0.20%\n",
            "iter 48766: loss 1.5260, time 21.06ms, mfu 0.23%\n",
            "iter 48767: loss 1.4564, time 27.85ms, mfu 0.17%\n",
            "iter 48768: loss 1.4972, time 19.67ms, mfu 0.25%\n",
            "iter 48769: loss 1.5530, time 15.86ms, mfu 0.30%\n",
            "iter 48770: loss 1.3470, time 12.50ms, mfu 0.39%\n",
            "iter 48771: loss 1.5243, time 13.70ms, mfu 0.35%\n",
            "iter 48772: loss 1.5508, time 12.01ms, mfu 0.40%\n",
            "iter 48773: loss 1.3736, time 11.77ms, mfu 0.41%\n",
            "iter 48774: loss 1.4268, time 11.61ms, mfu 0.42%\n",
            "iter 48775: loss 1.4312, time 17.27ms, mfu 0.28%\n",
            "iter 48776: loss 1.4033, time 13.93ms, mfu 0.35%\n",
            "iter 48777: loss 1.4576, time 24.59ms, mfu 0.20%\n",
            "iter 48778: loss 1.5361, time 14.45ms, mfu 0.33%\n",
            "iter 48779: loss 1.3737, time 12.40ms, mfu 0.39%\n",
            "iter 48780: loss 1.4667, time 11.70ms, mfu 0.41%\n",
            "iter 48781: loss 1.4697, time 16.26ms, mfu 0.30%\n",
            "iter 48782: loss 1.6533, time 20.98ms, mfu 0.23%\n",
            "iter 48783: loss 1.4833, time 12.80ms, mfu 0.38%\n",
            "iter 48784: loss 1.4514, time 11.98ms, mfu 0.40%\n",
            "iter 48785: loss 1.5128, time 12.05ms, mfu 0.40%\n",
            "iter 48786: loss 1.4741, time 11.77ms, mfu 0.41%\n",
            "iter 48787: loss 1.5268, time 11.69ms, mfu 0.41%\n",
            "iter 48788: loss 1.4701, time 14.04ms, mfu 0.34%\n",
            "iter 48789: loss 1.6097, time 18.08ms, mfu 0.27%\n",
            "iter 48790: loss 1.4843, time 12.91ms, mfu 0.37%\n",
            "iter 48791: loss 1.4353, time 13.74ms, mfu 0.35%\n",
            "iter 48792: loss 1.5387, time 14.83ms, mfu 0.33%\n",
            "iter 48793: loss 1.5456, time 14.65ms, mfu 0.33%\n",
            "iter 48794: loss 1.3863, time 15.99ms, mfu 0.30%\n",
            "iter 48795: loss 1.4207, time 19.66ms, mfu 0.25%\n",
            "iter 48796: loss 1.5201, time 23.43ms, mfu 0.21%\n",
            "iter 48797: loss 1.4241, time 18.24ms, mfu 0.26%\n",
            "iter 48798: loss 1.4056, time 17.00ms, mfu 0.28%\n",
            "iter 48799: loss 1.4455, time 19.44ms, mfu 0.25%\n",
            "iter 48800: loss 1.3475, time 16.23ms, mfu 0.30%\n",
            "iter 48801: loss 1.4974, time 19.26ms, mfu 0.25%\n",
            "iter 48802: loss 1.4105, time 59.05ms, mfu 0.08%\n",
            "iter 48803: loss 1.4526, time 24.85ms, mfu 0.19%\n",
            "iter 48804: loss 1.4478, time 33.15ms, mfu 0.15%\n",
            "iter 48805: loss 1.3942, time 37.76ms, mfu 0.13%\n",
            "iter 48806: loss 1.4657, time 42.84ms, mfu 0.11%\n",
            "iter 48807: loss 1.3804, time 19.65ms, mfu 0.25%\n",
            "iter 48808: loss 1.3844, time 18.16ms, mfu 0.27%\n",
            "iter 48809: loss 1.4242, time 31.19ms, mfu 0.15%\n",
            "iter 48810: loss 1.5349, time 14.15ms, mfu 0.34%\n",
            "iter 48811: loss 1.5130, time 12.87ms, mfu 0.37%\n",
            "iter 48812: loss 1.4654, time 16.97ms, mfu 0.28%\n",
            "iter 48813: loss 1.5205, time 44.67ms, mfu 0.11%\n",
            "iter 48814: loss 1.4838, time 27.65ms, mfu 0.17%\n",
            "iter 48815: loss 1.4872, time 27.77ms, mfu 0.17%\n",
            "iter 48816: loss 1.5194, time 21.04ms, mfu 0.23%\n",
            "iter 48817: loss 1.5182, time 15.64ms, mfu 0.31%\n",
            "iter 48818: loss 1.5066, time 20.95ms, mfu 0.23%\n",
            "iter 48819: loss 1.5549, time 17.15ms, mfu 0.28%\n",
            "iter 48820: loss 1.5231, time 31.07ms, mfu 0.16%\n",
            "iter 48821: loss 1.4324, time 18.78ms, mfu 0.26%\n",
            "iter 48822: loss 1.5042, time 15.91ms, mfu 0.30%\n",
            "iter 48823: loss 1.4540, time 24.77ms, mfu 0.19%\n",
            "iter 48824: loss 1.5116, time 20.71ms, mfu 0.23%\n",
            "iter 48825: loss 1.4680, time 13.68ms, mfu 0.35%\n",
            "iter 48826: loss 1.6028, time 15.55ms, mfu 0.31%\n",
            "iter 48827: loss 1.5777, time 15.03ms, mfu 0.32%\n",
            "iter 48828: loss 1.5690, time 14.02ms, mfu 0.34%\n",
            "iter 48829: loss 1.5663, time 28.29ms, mfu 0.17%\n",
            "iter 48830: loss 1.4014, time 14.19ms, mfu 0.34%\n",
            "iter 48831: loss 1.5083, time 14.69ms, mfu 0.33%\n",
            "iter 48832: loss 1.5627, time 15.98ms, mfu 0.30%\n",
            "iter 48833: loss 1.5092, time 15.14ms, mfu 0.32%\n",
            "iter 48834: loss 1.4459, time 26.81ms, mfu 0.18%\n",
            "iter 48835: loss 1.5996, time 51.58ms, mfu 0.09%\n",
            "iter 48836: loss 1.4788, time 19.04ms, mfu 0.25%\n",
            "iter 48837: loss 1.4828, time 21.82ms, mfu 0.22%\n",
            "iter 48838: loss 1.5558, time 24.03ms, mfu 0.20%\n",
            "iter 48839: loss 1.5113, time 28.49ms, mfu 0.17%\n",
            "iter 48840: loss 1.4732, time 23.46ms, mfu 0.21%\n",
            "iter 48841: loss 1.4748, time 16.26ms, mfu 0.30%\n",
            "iter 48842: loss 1.4563, time 16.24ms, mfu 0.30%\n",
            "iter 48843: loss 1.4609, time 23.61ms, mfu 0.20%\n",
            "iter 48844: loss 1.4867, time 18.38ms, mfu 0.26%\n",
            "iter 48845: loss 1.3210, time 19.03ms, mfu 0.25%\n",
            "iter 48846: loss 1.4395, time 15.92ms, mfu 0.30%\n",
            "iter 48847: loss 1.3831, time 30.83ms, mfu 0.16%\n",
            "iter 48848: loss 1.5041, time 24.52ms, mfu 0.20%\n",
            "iter 48849: loss 1.6003, time 34.60ms, mfu 0.14%\n",
            "iter 48850: loss 1.3650, time 21.34ms, mfu 0.23%\n",
            "iter 48851: loss 1.4772, time 9.71ms, mfu 0.50%\n",
            "iter 48852: loss 1.4135, time 9.42ms, mfu 0.51%\n",
            "iter 48853: loss 1.4229, time 9.51ms, mfu 0.51%\n",
            "iter 48854: loss 1.4460, time 9.97ms, mfu 0.48%\n",
            "iter 48855: loss 1.4633, time 13.63ms, mfu 0.35%\n",
            "iter 48856: loss 1.4924, time 9.70ms, mfu 0.50%\n",
            "iter 48857: loss 1.4460, time 9.48ms, mfu 0.51%\n",
            "iter 48858: loss 1.4831, time 13.01ms, mfu 0.37%\n",
            "iter 48859: loss 1.5687, time 9.50ms, mfu 0.51%\n",
            "iter 48860: loss 1.5440, time 9.98ms, mfu 0.48%\n",
            "iter 48861: loss 1.6077, time 14.13ms, mfu 0.34%\n",
            "iter 48862: loss 1.4937, time 13.98ms, mfu 0.34%\n",
            "iter 48863: loss 1.3821, time 14.78ms, mfu 0.33%\n",
            "iter 48864: loss 1.3911, time 10.77ms, mfu 0.45%\n",
            "iter 48865: loss 1.5874, time 9.67ms, mfu 0.50%\n",
            "iter 48866: loss 1.4236, time 11.59ms, mfu 0.42%\n",
            "iter 48867: loss 1.5870, time 9.38ms, mfu 0.51%\n",
            "iter 48868: loss 1.4146, time 9.35ms, mfu 0.52%\n",
            "iter 48869: loss 1.4855, time 9.65ms, mfu 0.50%\n",
            "iter 48870: loss 1.4805, time 10.92ms, mfu 0.44%\n",
            "iter 48871: loss 1.4360, time 9.73ms, mfu 0.50%\n",
            "iter 48872: loss 1.4158, time 9.55ms, mfu 0.51%\n",
            "iter 48873: loss 1.5369, time 13.58ms, mfu 0.36%\n",
            "iter 48874: loss 1.4800, time 20.05ms, mfu 0.24%\n",
            "iter 48875: loss 1.5059, time 22.08ms, mfu 0.22%\n",
            "iter 48876: loss 1.4474, time 13.82ms, mfu 0.35%\n",
            "iter 48877: loss 1.4302, time 10.52ms, mfu 0.46%\n",
            "iter 48878: loss 1.4099, time 9.44ms, mfu 0.51%\n",
            "iter 48879: loss 1.3676, time 12.36ms, mfu 0.39%\n",
            "iter 48880: loss 1.4535, time 15.59ms, mfu 0.31%\n",
            "iter 48881: loss 1.4255, time 9.56ms, mfu 0.50%\n",
            "iter 48882: loss 1.4044, time 11.09ms, mfu 0.44%\n",
            "iter 48883: loss 1.4415, time 10.06ms, mfu 0.48%\n",
            "iter 48884: loss 1.4052, time 9.54ms, mfu 0.51%\n",
            "iter 48885: loss 1.5822, time 9.53ms, mfu 0.51%\n",
            "iter 48886: loss 1.5228, time 9.34ms, mfu 0.52%\n",
            "iter 48887: loss 1.4666, time 10.91ms, mfu 0.44%\n",
            "iter 48888: loss 1.5601, time 14.23ms, mfu 0.34%\n",
            "iter 48889: loss 1.4276, time 13.67ms, mfu 0.35%\n",
            "iter 48890: loss 1.4297, time 9.33ms, mfu 0.52%\n",
            "iter 48891: loss 1.4796, time 9.79ms, mfu 0.49%\n",
            "iter 48892: loss 1.4486, time 14.65ms, mfu 0.33%\n",
            "iter 48893: loss 1.5450, time 9.81ms, mfu 0.49%\n",
            "iter 48894: loss 1.4870, time 9.77ms, mfu 0.49%\n",
            "iter 48895: loss 1.5432, time 9.48ms, mfu 0.51%\n",
            "iter 48896: loss 1.5006, time 9.34ms, mfu 0.52%\n",
            "iter 48897: loss 1.4529, time 9.44ms, mfu 0.51%\n",
            "iter 48898: loss 1.3616, time 9.34ms, mfu 0.52%\n",
            "iter 48899: loss 1.5451, time 11.80ms, mfu 0.41%\n",
            "iter 48900: loss 1.5996, time 14.21ms, mfu 0.34%\n",
            "iter 48901: loss 1.4740, time 9.64ms, mfu 0.50%\n",
            "iter 48902: loss 1.4451, time 9.29ms, mfu 0.52%\n",
            "iter 48903: loss 1.4305, time 9.45ms, mfu 0.51%\n",
            "iter 48904: loss 1.4168, time 9.40ms, mfu 0.51%\n",
            "iter 48905: loss 1.5370, time 9.57ms, mfu 0.50%\n",
            "iter 48906: loss 1.4912, time 9.81ms, mfu 0.49%\n",
            "iter 48907: loss 1.4557, time 9.89ms, mfu 0.49%\n",
            "iter 48908: loss 1.4862, time 9.68ms, mfu 0.50%\n",
            "iter 48909: loss 1.4819, time 9.44ms, mfu 0.51%\n",
            "iter 48910: loss 1.4367, time 9.28ms, mfu 0.52%\n",
            "iter 48911: loss 1.3849, time 9.35ms, mfu 0.52%\n",
            "iter 48912: loss 1.5440, time 13.28ms, mfu 0.36%\n",
            "iter 48913: loss 1.4398, time 26.80ms, mfu 0.18%\n",
            "iter 48914: loss 1.5392, time 17.17ms, mfu 0.28%\n",
            "iter 48915: loss 1.5267, time 17.24ms, mfu 0.28%\n",
            "iter 48916: loss 1.5005, time 20.70ms, mfu 0.23%\n",
            "iter 48917: loss 1.4251, time 17.40ms, mfu 0.28%\n",
            "iter 48918: loss 1.4694, time 17.22ms, mfu 0.28%\n",
            "iter 48919: loss 1.3881, time 28.02ms, mfu 0.17%\n",
            "iter 48920: loss 1.4167, time 17.60ms, mfu 0.27%\n",
            "iter 48921: loss 1.4032, time 14.78ms, mfu 0.33%\n",
            "iter 48922: loss 1.4921, time 20.37ms, mfu 0.24%\n",
            "iter 48923: loss 1.4537, time 9.68ms, mfu 0.50%\n",
            "iter 48924: loss 1.4920, time 9.44ms, mfu 0.51%\n",
            "iter 48925: loss 1.4373, time 9.41ms, mfu 0.51%\n",
            "iter 48926: loss 1.4853, time 14.26ms, mfu 0.34%\n",
            "iter 48927: loss 1.4994, time 10.15ms, mfu 0.48%\n",
            "iter 48928: loss 1.5586, time 9.78ms, mfu 0.49%\n",
            "iter 48929: loss 1.4074, time 9.59ms, mfu 0.50%\n",
            "iter 48930: loss 1.4575, time 9.29ms, mfu 0.52%\n",
            "iter 48931: loss 1.3135, time 9.54ms, mfu 0.51%\n",
            "iter 48932: loss 1.4070, time 9.39ms, mfu 0.51%\n",
            "iter 48933: loss 1.4641, time 9.31ms, mfu 0.52%\n",
            "iter 48934: loss 1.4861, time 12.54ms, mfu 0.38%\n",
            "iter 48935: loss 1.4193, time 13.61ms, mfu 0.35%\n",
            "iter 48936: loss 1.4937, time 9.40ms, mfu 0.51%\n",
            "iter 48937: loss 1.4593, time 9.46ms, mfu 0.51%\n",
            "iter 48938: loss 1.5658, time 9.89ms, mfu 0.49%\n",
            "iter 48939: loss 1.5105, time 9.79ms, mfu 0.49%\n",
            "iter 48940: loss 1.4577, time 10.33ms, mfu 0.47%\n",
            "iter 48941: loss 1.5429, time 11.46ms, mfu 0.42%\n",
            "iter 48942: loss 1.4255, time 9.44ms, mfu 0.51%\n",
            "iter 48943: loss 1.6104, time 9.55ms, mfu 0.50%\n",
            "iter 48944: loss 1.5376, time 9.42ms, mfu 0.51%\n",
            "iter 48945: loss 1.4484, time 11.43ms, mfu 0.42%\n",
            "iter 48946: loss 1.4645, time 20.69ms, mfu 0.23%\n",
            "iter 48947: loss 1.5381, time 16.85ms, mfu 0.29%\n",
            "iter 48948: loss 1.4573, time 15.05ms, mfu 0.32%\n",
            "iter 48949: loss 1.5268, time 19.61ms, mfu 0.25%\n",
            "iter 48950: loss 1.5216, time 17.30ms, mfu 0.28%\n",
            "iter 48951: loss 1.4599, time 17.68ms, mfu 0.27%\n",
            "iter 48952: loss 1.5397, time 18.85ms, mfu 0.26%\n",
            "iter 48953: loss 1.4607, time 17.06ms, mfu 0.28%\n",
            "iter 48954: loss 1.4749, time 17.03ms, mfu 0.28%\n",
            "iter 48955: loss 1.5056, time 17.07ms, mfu 0.28%\n",
            "iter 48956: loss 1.4264, time 25.42ms, mfu 0.19%\n",
            "iter 48957: loss 1.5562, time 37.37ms, mfu 0.13%\n",
            "iter 48958: loss 1.4043, time 10.23ms, mfu 0.47%\n",
            "iter 48959: loss 1.5280, time 11.12ms, mfu 0.43%\n",
            "iter 48960: loss 1.4383, time 9.32ms, mfu 0.52%\n",
            "iter 48961: loss 1.4343, time 9.54ms, mfu 0.51%\n",
            "iter 48962: loss 1.4809, time 9.27ms, mfu 0.52%\n",
            "iter 48963: loss 1.4121, time 9.95ms, mfu 0.48%\n",
            "iter 48964: loss 1.5193, time 12.15ms, mfu 0.40%\n",
            "iter 48965: loss 1.5621, time 11.65ms, mfu 0.41%\n",
            "iter 48966: loss 1.4402, time 9.76ms, mfu 0.49%\n",
            "iter 48967: loss 1.4490, time 11.26ms, mfu 0.43%\n",
            "iter 48968: loss 1.4624, time 9.38ms, mfu 0.51%\n",
            "iter 48969: loss 1.4183, time 9.46ms, mfu 0.51%\n",
            "iter 48970: loss 1.3719, time 9.39ms, mfu 0.51%\n",
            "iter 48971: loss 1.4348, time 9.23ms, mfu 0.52%\n",
            "iter 48972: loss 1.4187, time 9.31ms, mfu 0.52%\n",
            "iter 48973: loss 1.5607, time 9.31ms, mfu 0.52%\n",
            "iter 48974: loss 1.4312, time 12.35ms, mfu 0.39%\n",
            "iter 48975: loss 1.4815, time 9.35ms, mfu 0.52%\n",
            "iter 48976: loss 1.4720, time 9.76ms, mfu 0.49%\n",
            "iter 48977: loss 1.6821, time 9.43ms, mfu 0.51%\n",
            "iter 48978: loss 1.4710, time 13.54ms, mfu 0.36%\n",
            "iter 48979: loss 1.3922, time 11.41ms, mfu 0.42%\n",
            "iter 48980: loss 1.4975, time 18.85ms, mfu 0.26%\n",
            "iter 48981: loss 1.4957, time 16.27ms, mfu 0.30%\n",
            "iter 48982: loss 1.4413, time 17.25ms, mfu 0.28%\n",
            "iter 48983: loss 1.6150, time 17.02ms, mfu 0.28%\n",
            "iter 48984: loss 1.5254, time 15.45ms, mfu 0.31%\n",
            "iter 48985: loss 1.4500, time 17.02ms, mfu 0.28%\n",
            "iter 48986: loss 1.4226, time 16.45ms, mfu 0.29%\n",
            "iter 48987: loss 1.4584, time 17.01ms, mfu 0.28%\n",
            "iter 48988: loss 1.4338, time 17.00ms, mfu 0.28%\n",
            "iter 48989: loss 1.5096, time 14.39ms, mfu 0.34%\n",
            "iter 48990: loss 1.5308, time 16.94ms, mfu 0.28%\n",
            "iter 48991: loss 1.4595, time 17.70ms, mfu 0.27%\n",
            "iter 48992: loss 1.4971, time 17.74ms, mfu 0.27%\n",
            "iter 48993: loss 1.4840, time 16.69ms, mfu 0.29%\n",
            "iter 48994: loss 1.4471, time 16.39ms, mfu 0.29%\n",
            "iter 48995: loss 1.4119, time 18.17ms, mfu 0.27%\n",
            "iter 48996: loss 1.4650, time 14.14ms, mfu 0.34%\n",
            "iter 48997: loss 1.4804, time 14.04ms, mfu 0.34%\n",
            "iter 48998: loss 1.4209, time 9.58ms, mfu 0.50%\n",
            "iter 48999: loss 1.4122, time 9.73ms, mfu 0.50%\n",
            "step 49000: train loss 1.3822, val loss 1.5502\n",
            "iter 49000: loss 1.4526, time 20.18ms, mfu 0.24%\n",
            "iter 49001: loss 1.3737, time 9.50ms, mfu 0.51%\n",
            "iter 49002: loss 1.5384, time 9.68ms, mfu 0.50%\n",
            "iter 49003: loss 1.5186, time 9.49ms, mfu 0.51%\n",
            "iter 49004: loss 1.5520, time 11.66ms, mfu 0.41%\n",
            "iter 49005: loss 1.5881, time 17.17ms, mfu 0.28%\n",
            "iter 49006: loss 1.4018, time 11.59ms, mfu 0.42%\n",
            "iter 49007: loss 1.4911, time 11.26ms, mfu 0.43%\n",
            "iter 49008: loss 1.3921, time 9.31ms, mfu 0.52%\n",
            "iter 49009: loss 1.4098, time 10.74ms, mfu 0.45%\n",
            "iter 49010: loss 1.5542, time 9.32ms, mfu 0.52%\n",
            "iter 49011: loss 1.5362, time 9.34ms, mfu 0.52%\n",
            "iter 49012: loss 1.3957, time 9.39ms, mfu 0.51%\n",
            "iter 49013: loss 1.5548, time 9.43ms, mfu 0.51%\n",
            "iter 49014: loss 1.4626, time 9.60ms, mfu 0.50%\n",
            "iter 49015: loss 1.5476, time 9.67ms, mfu 0.50%\n",
            "iter 49016: loss 1.4528, time 9.39ms, mfu 0.51%\n",
            "iter 49017: loss 1.4323, time 9.63ms, mfu 0.50%\n",
            "iter 49018: loss 1.3980, time 13.48ms, mfu 0.36%\n",
            "iter 49019: loss 1.4583, time 14.45ms, mfu 0.33%\n",
            "iter 49020: loss 1.4769, time 15.37ms, mfu 0.31%\n",
            "iter 49021: loss 1.4527, time 28.95ms, mfu 0.17%\n",
            "iter 49022: loss 1.3762, time 17.27ms, mfu 0.28%\n",
            "iter 49023: loss 1.5069, time 21.90ms, mfu 0.22%\n",
            "iter 49024: loss 1.4704, time 20.24ms, mfu 0.24%\n",
            "iter 49025: loss 1.5413, time 17.18ms, mfu 0.28%\n",
            "iter 49026: loss 1.5627, time 21.11ms, mfu 0.23%\n",
            "iter 49027: loss 1.5239, time 26.13ms, mfu 0.18%\n",
            "iter 49028: loss 1.4603, time 26.50ms, mfu 0.18%\n",
            "iter 49029: loss 1.5067, time 19.74ms, mfu 0.24%\n",
            "iter 49030: loss 1.3340, time 25.39ms, mfu 0.19%\n",
            "iter 49031: loss 1.4594, time 19.19ms, mfu 0.25%\n",
            "iter 49032: loss 1.5201, time 16.96ms, mfu 0.28%\n",
            "iter 49033: loss 1.4001, time 25.85ms, mfu 0.19%\n",
            "iter 49034: loss 1.5752, time 26.66ms, mfu 0.18%\n",
            "iter 49035: loss 1.3953, time 15.28ms, mfu 0.32%\n",
            "iter 49036: loss 1.4134, time 15.77ms, mfu 0.31%\n",
            "iter 49037: loss 1.5946, time 16.60ms, mfu 0.29%\n",
            "iter 49038: loss 1.5188, time 18.36ms, mfu 0.26%\n",
            "iter 49039: loss 1.5282, time 28.76ms, mfu 0.17%\n",
            "iter 49040: loss 1.5004, time 16.56ms, mfu 0.29%\n",
            "iter 49041: loss 1.5349, time 17.80ms, mfu 0.27%\n",
            "iter 49042: loss 1.4144, time 15.81ms, mfu 0.31%\n",
            "iter 49043: loss 1.5860, time 16.18ms, mfu 0.30%\n",
            "iter 49044: loss 1.5198, time 23.19ms, mfu 0.21%\n",
            "iter 49045: loss 1.5690, time 16.72ms, mfu 0.29%\n",
            "iter 49046: loss 1.5038, time 15.14ms, mfu 0.32%\n",
            "iter 49047: loss 1.4128, time 16.22ms, mfu 0.30%\n",
            "iter 49048: loss 1.4768, time 16.32ms, mfu 0.30%\n",
            "iter 49049: loss 1.4088, time 16.32ms, mfu 0.30%\n",
            "iter 49050: loss 1.4471, time 18.72ms, mfu 0.26%\n",
            "iter 49051: loss 1.4334, time 15.76ms, mfu 0.31%\n",
            "iter 49052: loss 1.6232, time 18.20ms, mfu 0.26%\n",
            "iter 49053: loss 1.5008, time 36.00ms, mfu 0.13%\n",
            "iter 49054: loss 1.5746, time 25.69ms, mfu 0.19%\n",
            "iter 49055: loss 1.4724, time 34.13ms, mfu 0.14%\n",
            "iter 49056: loss 1.5331, time 17.25ms, mfu 0.28%\n",
            "iter 49057: loss 1.5555, time 10.12ms, mfu 0.48%\n",
            "iter 49058: loss 1.4531, time 9.67ms, mfu 0.50%\n",
            "iter 49059: loss 1.4170, time 9.43ms, mfu 0.51%\n",
            "iter 49060: loss 1.4347, time 10.58ms, mfu 0.46%\n",
            "iter 49061: loss 1.3848, time 9.28ms, mfu 0.52%\n",
            "iter 49062: loss 1.4566, time 9.64ms, mfu 0.50%\n",
            "iter 49063: loss 1.3803, time 9.48ms, mfu 0.51%\n",
            "iter 49064: loss 1.5056, time 9.39ms, mfu 0.51%\n",
            "iter 49065: loss 1.2863, time 11.43ms, mfu 0.42%\n",
            "iter 49066: loss 1.4327, time 10.24ms, mfu 0.47%\n",
            "iter 49067: loss 1.4022, time 9.43ms, mfu 0.51%\n",
            "iter 49068: loss 1.5213, time 9.40ms, mfu 0.51%\n",
            "iter 49069: loss 1.3856, time 9.30ms, mfu 0.52%\n",
            "iter 49070: loss 1.4377, time 9.38ms, mfu 0.51%\n",
            "iter 49071: loss 1.4391, time 9.50ms, mfu 0.51%\n",
            "iter 49072: loss 1.4058, time 9.32ms, mfu 0.52%\n",
            "iter 49073: loss 1.4488, time 9.35ms, mfu 0.52%\n",
            "iter 49074: loss 1.5438, time 9.28ms, mfu 0.52%\n",
            "iter 49075: loss 1.4258, time 9.27ms, mfu 0.52%\n",
            "iter 49076: loss 1.4159, time 12.80ms, mfu 0.38%\n",
            "iter 49077: loss 1.5500, time 11.92ms, mfu 0.40%\n",
            "iter 49078: loss 1.3943, time 9.27ms, mfu 0.52%\n",
            "iter 49079: loss 1.4117, time 9.33ms, mfu 0.52%\n",
            "iter 49080: loss 1.5067, time 9.35ms, mfu 0.52%\n",
            "iter 49081: loss 1.5883, time 9.18ms, mfu 0.53%\n",
            "iter 49082: loss 1.5058, time 9.28ms, mfu 0.52%\n",
            "iter 49083: loss 1.5376, time 9.22ms, mfu 0.52%\n",
            "iter 49084: loss 1.5657, time 10.70ms, mfu 0.45%\n",
            "iter 49085: loss 1.5164, time 10.35ms, mfu 0.47%\n",
            "iter 49086: loss 1.3566, time 9.86ms, mfu 0.49%\n",
            "iter 49087: loss 1.4659, time 9.64ms, mfu 0.50%\n",
            "iter 49088: loss 1.5138, time 9.63ms, mfu 0.50%\n",
            "iter 49089: loss 1.5236, time 9.71ms, mfu 0.50%\n",
            "iter 49090: loss 1.4204, time 9.99ms, mfu 0.48%\n",
            "iter 49091: loss 1.3543, time 9.57ms, mfu 0.50%\n",
            "iter 49092: loss 1.5021, time 9.54ms, mfu 0.51%\n",
            "iter 49093: loss 1.3893, time 9.40ms, mfu 0.51%\n",
            "iter 49094: loss 1.5948, time 9.59ms, mfu 0.50%\n",
            "iter 49095: loss 1.4179, time 9.33ms, mfu 0.52%\n",
            "iter 49096: loss 1.5432, time 9.39ms, mfu 0.51%\n",
            "iter 49097: loss 1.4618, time 12.87ms, mfu 0.37%\n",
            "iter 49098: loss 1.5134, time 10.54ms, mfu 0.46%\n",
            "iter 49099: loss 1.4669, time 9.38ms, mfu 0.51%\n",
            "iter 49100: loss 1.3754, time 11.03ms, mfu 0.44%\n",
            "iter 49101: loss 1.4759, time 9.89ms, mfu 0.49%\n",
            "iter 49102: loss 1.5422, time 9.36ms, mfu 0.52%\n",
            "iter 49103: loss 1.5581, time 9.57ms, mfu 0.50%\n",
            "iter 49104: loss 1.4649, time 9.59ms, mfu 0.50%\n",
            "iter 49105: loss 1.5803, time 15.86ms, mfu 0.30%\n",
            "iter 49106: loss 1.4920, time 13.22ms, mfu 0.36%\n",
            "iter 49107: loss 1.4409, time 9.37ms, mfu 0.51%\n",
            "iter 49108: loss 1.4208, time 9.29ms, mfu 0.52%\n",
            "iter 49109: loss 1.5304, time 10.14ms, mfu 0.48%\n",
            "iter 49110: loss 1.4261, time 12.87ms, mfu 0.37%\n",
            "iter 49111: loss 1.4368, time 11.67ms, mfu 0.41%\n",
            "iter 49112: loss 1.5118, time 9.40ms, mfu 0.51%\n",
            "iter 49113: loss 1.5925, time 9.28ms, mfu 0.52%\n",
            "iter 49114: loss 1.4427, time 9.57ms, mfu 0.50%\n",
            "iter 49115: loss 1.4144, time 12.74ms, mfu 0.38%\n",
            "iter 49116: loss 1.4238, time 14.01ms, mfu 0.34%\n",
            "iter 49117: loss 1.5478, time 11.09ms, mfu 0.44%\n",
            "iter 49118: loss 1.5822, time 9.34ms, mfu 0.52%\n",
            "iter 49119: loss 1.3908, time 9.14ms, mfu 0.53%\n",
            "iter 49120: loss 1.4502, time 9.32ms, mfu 0.52%\n",
            "iter 49121: loss 1.4277, time 9.81ms, mfu 0.49%\n",
            "iter 49122: loss 1.5959, time 9.39ms, mfu 0.51%\n",
            "iter 49123: loss 1.4025, time 10.00ms, mfu 0.48%\n",
            "iter 49124: loss 1.5376, time 9.62ms, mfu 0.50%\n",
            "iter 49125: loss 1.5245, time 9.36ms, mfu 0.52%\n",
            "iter 49126: loss 1.3638, time 9.32ms, mfu 0.52%\n",
            "iter 49127: loss 1.4057, time 9.41ms, mfu 0.51%\n",
            "iter 49128: loss 1.4956, time 9.22ms, mfu 0.52%\n",
            "iter 49129: loss 1.3797, time 9.35ms, mfu 0.52%\n",
            "iter 49130: loss 1.4682, time 9.40ms, mfu 0.51%\n",
            "iter 49131: loss 1.5058, time 10.18ms, mfu 0.47%\n",
            "iter 49132: loss 1.4784, time 9.20ms, mfu 0.52%\n",
            "iter 49133: loss 1.4808, time 9.58ms, mfu 0.50%\n",
            "iter 49134: loss 1.4270, time 9.57ms, mfu 0.50%\n",
            "iter 49135: loss 1.4562, time 9.40ms, mfu 0.51%\n",
            "iter 49136: loss 1.4651, time 9.29ms, mfu 0.52%\n",
            "iter 49137: loss 1.4575, time 9.23ms, mfu 0.52%\n",
            "iter 49138: loss 1.5070, time 14.57ms, mfu 0.33%\n",
            "iter 49139: loss 1.5083, time 10.10ms, mfu 0.48%\n",
            "iter 49140: loss 1.3458, time 9.45ms, mfu 0.51%\n",
            "iter 49141: loss 1.5728, time 9.69ms, mfu 0.50%\n",
            "iter 49142: loss 1.4493, time 9.33ms, mfu 0.52%\n",
            "iter 49143: loss 1.4012, time 9.39ms, mfu 0.51%\n",
            "iter 49144: loss 1.5215, time 9.40ms, mfu 0.51%\n",
            "iter 49145: loss 1.4515, time 9.50ms, mfu 0.51%\n",
            "iter 49146: loss 1.4213, time 9.45ms, mfu 0.51%\n",
            "iter 49147: loss 1.5013, time 9.25ms, mfu 0.52%\n",
            "iter 49148: loss 1.5098, time 9.65ms, mfu 0.50%\n",
            "iter 49149: loss 1.5203, time 9.25ms, mfu 0.52%\n",
            "iter 49150: loss 1.4773, time 9.17ms, mfu 0.53%\n",
            "iter 49151: loss 1.5375, time 9.38ms, mfu 0.51%\n",
            "iter 49152: loss 1.4303, time 9.84ms, mfu 0.49%\n",
            "iter 49153: loss 1.4850, time 9.26ms, mfu 0.52%\n",
            "iter 49154: loss 1.4833, time 9.44ms, mfu 0.51%\n",
            "iter 49155: loss 1.5733, time 9.48ms, mfu 0.51%\n",
            "iter 49156: loss 1.5162, time 9.18ms, mfu 0.53%\n",
            "iter 49157: loss 1.3884, time 9.17ms, mfu 0.53%\n",
            "iter 49158: loss 1.4118, time 9.41ms, mfu 0.51%\n",
            "iter 49159: loss 1.5208, time 9.17ms, mfu 0.53%\n",
            "iter 49160: loss 1.5143, time 14.34ms, mfu 0.34%\n",
            "iter 49161: loss 1.5163, time 9.57ms, mfu 0.50%\n",
            "iter 49162: loss 1.4161, time 9.28ms, mfu 0.52%\n",
            "iter 49163: loss 1.4509, time 9.43ms, mfu 0.51%\n",
            "iter 49164: loss 1.5206, time 9.24ms, mfu 0.52%\n",
            "iter 49165: loss 1.5207, time 9.24ms, mfu 0.52%\n",
            "iter 49166: loss 1.5448, time 9.34ms, mfu 0.52%\n",
            "iter 49167: loss 1.4784, time 9.22ms, mfu 0.52%\n",
            "iter 49168: loss 1.6290, time 9.17ms, mfu 0.53%\n",
            "iter 49169: loss 1.4071, time 12.32ms, mfu 0.39%\n",
            "iter 49170: loss 1.4620, time 9.18ms, mfu 0.53%\n",
            "iter 49171: loss 1.3726, time 9.24ms, mfu 0.52%\n",
            "iter 49172: loss 1.4820, time 9.41ms, mfu 0.51%\n",
            "iter 49173: loss 1.5149, time 9.48ms, mfu 0.51%\n",
            "iter 49174: loss 1.4500, time 9.34ms, mfu 0.52%\n",
            "iter 49175: loss 1.4950, time 9.33ms, mfu 0.52%\n",
            "iter 49176: loss 1.5125, time 9.78ms, mfu 0.49%\n",
            "iter 49177: loss 1.3929, time 11.29ms, mfu 0.43%\n",
            "iter 49178: loss 1.5030, time 9.57ms, mfu 0.50%\n",
            "iter 49179: loss 1.5631, time 9.65ms, mfu 0.50%\n",
            "iter 49180: loss 1.4575, time 10.37ms, mfu 0.47%\n",
            "iter 49181: loss 1.5191, time 12.03ms, mfu 0.40%\n",
            "iter 49182: loss 1.4702, time 12.46ms, mfu 0.39%\n",
            "iter 49183: loss 1.4290, time 10.03ms, mfu 0.48%\n",
            "iter 49184: loss 1.4022, time 9.47ms, mfu 0.51%\n",
            "iter 49185: loss 1.6417, time 10.88ms, mfu 0.44%\n",
            "iter 49186: loss 1.4480, time 9.61ms, mfu 0.50%\n",
            "iter 49187: loss 1.5140, time 9.52ms, mfu 0.51%\n",
            "iter 49188: loss 1.5240, time 9.14ms, mfu 0.53%\n",
            "iter 49189: loss 1.4835, time 9.40ms, mfu 0.51%\n",
            "iter 49190: loss 1.4667, time 9.49ms, mfu 0.51%\n",
            "iter 49191: loss 1.5153, time 9.38ms, mfu 0.51%\n",
            "iter 49192: loss 1.3298, time 9.65ms, mfu 0.50%\n",
            "iter 49193: loss 1.5362, time 9.19ms, mfu 0.53%\n",
            "iter 49194: loss 1.5007, time 9.46ms, mfu 0.51%\n",
            "iter 49195: loss 1.4559, time 9.72ms, mfu 0.50%\n",
            "iter 49196: loss 1.4503, time 9.19ms, mfu 0.52%\n",
            "iter 49197: loss 1.4599, time 9.32ms, mfu 0.52%\n",
            "iter 49198: loss 1.4979, time 9.34ms, mfu 0.52%\n",
            "iter 49199: loss 1.4498, time 9.65ms, mfu 0.50%\n",
            "iter 49200: loss 1.4646, time 9.28ms, mfu 0.52%\n",
            "iter 49201: loss 1.5128, time 11.13ms, mfu 0.43%\n",
            "iter 49202: loss 1.4620, time 12.48ms, mfu 0.39%\n",
            "iter 49203: loss 1.4314, time 10.68ms, mfu 0.45%\n",
            "iter 49204: loss 1.3955, time 9.85ms, mfu 0.49%\n",
            "iter 49205: loss 1.3993, time 9.98ms, mfu 0.48%\n",
            "iter 49206: loss 1.4927, time 11.54ms, mfu 0.42%\n",
            "iter 49207: loss 1.5687, time 11.51ms, mfu 0.42%\n",
            "iter 49208: loss 1.5263, time 14.91ms, mfu 0.32%\n",
            "iter 49209: loss 1.5417, time 12.18ms, mfu 0.40%\n",
            "iter 49210: loss 1.5693, time 12.31ms, mfu 0.39%\n",
            "iter 49211: loss 1.4408, time 12.48ms, mfu 0.39%\n",
            "iter 49212: loss 1.5262, time 10.07ms, mfu 0.48%\n",
            "iter 49213: loss 1.4588, time 9.33ms, mfu 0.52%\n",
            "iter 49214: loss 1.3965, time 9.19ms, mfu 0.53%\n",
            "iter 49215: loss 1.5883, time 9.33ms, mfu 0.52%\n",
            "iter 49216: loss 1.4924, time 9.27ms, mfu 0.52%\n",
            "iter 49217: loss 1.5429, time 9.12ms, mfu 0.53%\n",
            "iter 49218: loss 1.4132, time 9.35ms, mfu 0.52%\n",
            "iter 49219: loss 1.3647, time 9.21ms, mfu 0.52%\n",
            "iter 49220: loss 1.4311, time 11.20ms, mfu 0.43%\n",
            "iter 49221: loss 1.4436, time 11.25ms, mfu 0.43%\n",
            "iter 49222: loss 1.2967, time 12.28ms, mfu 0.39%\n",
            "iter 49223: loss 1.4773, time 9.34ms, mfu 0.52%\n",
            "iter 49224: loss 1.4591, time 10.10ms, mfu 0.48%\n",
            "iter 49225: loss 1.5932, time 9.50ms, mfu 0.51%\n",
            "iter 49226: loss 1.5449, time 9.53ms, mfu 0.51%\n",
            "iter 49227: loss 1.4743, time 9.27ms, mfu 0.52%\n",
            "iter 49228: loss 1.4478, time 9.26ms, mfu 0.52%\n",
            "iter 49229: loss 1.5633, time 9.60ms, mfu 0.50%\n",
            "iter 49230: loss 1.4749, time 9.39ms, mfu 0.51%\n",
            "iter 49231: loss 1.4918, time 9.73ms, mfu 0.50%\n",
            "iter 49232: loss 1.4101, time 9.47ms, mfu 0.51%\n",
            "iter 49233: loss 1.5384, time 9.39ms, mfu 0.51%\n",
            "iter 49234: loss 1.5308, time 9.18ms, mfu 0.53%\n",
            "iter 49235: loss 1.4439, time 9.63ms, mfu 0.50%\n",
            "iter 49236: loss 1.5446, time 9.71ms, mfu 0.50%\n",
            "iter 49237: loss 1.5515, time 9.25ms, mfu 0.52%\n",
            "iter 49238: loss 1.4453, time 9.25ms, mfu 0.52%\n",
            "iter 49239: loss 1.5053, time 9.17ms, mfu 0.53%\n",
            "iter 49240: loss 1.4921, time 9.21ms, mfu 0.52%\n",
            "iter 49241: loss 1.5146, time 9.26ms, mfu 0.52%\n",
            "iter 49242: loss 1.5195, time 9.39ms, mfu 0.51%\n",
            "iter 49243: loss 1.4840, time 12.29ms, mfu 0.39%\n",
            "iter 49244: loss 1.4737, time 10.29ms, mfu 0.47%\n",
            "iter 49245: loss 1.4306, time 9.74ms, mfu 0.50%\n",
            "iter 49246: loss 1.3404, time 9.26ms, mfu 0.52%\n",
            "iter 49247: loss 1.4517, time 9.21ms, mfu 0.52%\n",
            "iter 49248: loss 1.4579, time 10.30ms, mfu 0.47%\n",
            "iter 49249: loss 1.4740, time 9.63ms, mfu 0.50%\n",
            "iter 49250: loss 1.3936, time 9.38ms, mfu 0.51%\n",
            "iter 49251: loss 1.5588, time 9.08ms, mfu 0.53%\n",
            "iter 49252: loss 1.5046, time 9.29ms, mfu 0.52%\n",
            "iter 49253: loss 1.3280, time 9.31ms, mfu 0.52%\n",
            "iter 49254: loss 1.5846, time 9.33ms, mfu 0.52%\n",
            "iter 49255: loss 1.4417, time 9.16ms, mfu 0.53%\n",
            "iter 49256: loss 1.4605, time 9.29ms, mfu 0.52%\n",
            "iter 49257: loss 1.4284, time 9.16ms, mfu 0.53%\n",
            "iter 49258: loss 1.4723, time 9.20ms, mfu 0.52%\n",
            "iter 49259: loss 1.4766, time 9.14ms, mfu 0.53%\n",
            "iter 49260: loss 1.4663, time 9.22ms, mfu 0.52%\n",
            "iter 49261: loss 1.4882, time 9.19ms, mfu 0.52%\n",
            "iter 49262: loss 1.4942, time 9.36ms, mfu 0.52%\n",
            "iter 49263: loss 1.5188, time 9.53ms, mfu 0.51%\n",
            "iter 49264: loss 1.3480, time 9.37ms, mfu 0.51%\n",
            "iter 49265: loss 1.5132, time 12.95ms, mfu 0.37%\n",
            "iter 49266: loss 1.6084, time 11.03ms, mfu 0.44%\n",
            "iter 49267: loss 1.4511, time 10.55ms, mfu 0.46%\n",
            "iter 49268: loss 1.3228, time 10.30ms, mfu 0.47%\n",
            "iter 49269: loss 1.5112, time 9.36ms, mfu 0.52%\n",
            "iter 49270: loss 1.4057, time 9.38ms, mfu 0.51%\n",
            "iter 49271: loss 1.4455, time 9.51ms, mfu 0.51%\n",
            "iter 49272: loss 1.4254, time 10.72ms, mfu 0.45%\n",
            "iter 49273: loss 1.4340, time 9.27ms, mfu 0.52%\n",
            "iter 49274: loss 1.3911, time 9.27ms, mfu 0.52%\n",
            "iter 49275: loss 1.5487, time 9.41ms, mfu 0.51%\n",
            "iter 49276: loss 1.3695, time 12.00ms, mfu 0.40%\n",
            "iter 49277: loss 1.5302, time 9.31ms, mfu 0.52%\n",
            "iter 49278: loss 1.5079, time 9.64ms, mfu 0.50%\n",
            "iter 49279: loss 1.5016, time 9.58ms, mfu 0.50%\n",
            "iter 49280: loss 1.3753, time 9.41ms, mfu 0.51%\n",
            "iter 49281: loss 1.5027, time 9.17ms, mfu 0.53%\n",
            "iter 49282: loss 1.4871, time 9.13ms, mfu 0.53%\n",
            "iter 49283: loss 1.4392, time 9.22ms, mfu 0.52%\n",
            "iter 49284: loss 1.4709, time 9.33ms, mfu 0.52%\n",
            "iter 49285: loss 1.4383, time 9.53ms, mfu 0.51%\n",
            "iter 49286: loss 1.5222, time 12.72ms, mfu 0.38%\n",
            "iter 49287: loss 1.4918, time 10.81ms, mfu 0.45%\n",
            "iter 49288: loss 1.4324, time 9.38ms, mfu 0.51%\n",
            "iter 49289: loss 1.4827, time 9.30ms, mfu 0.52%\n",
            "iter 49290: loss 1.4561, time 9.25ms, mfu 0.52%\n",
            "iter 49291: loss 1.4470, time 9.45ms, mfu 0.51%\n",
            "iter 49292: loss 1.4596, time 9.40ms, mfu 0.51%\n",
            "iter 49293: loss 1.4371, time 9.42ms, mfu 0.51%\n",
            "iter 49294: loss 1.5205, time 9.24ms, mfu 0.52%\n",
            "iter 49295: loss 1.4315, time 9.29ms, mfu 0.52%\n",
            "iter 49296: loss 1.4153, time 9.50ms, mfu 0.51%\n",
            "iter 49297: loss 1.5373, time 9.36ms, mfu 0.52%\n",
            "iter 49298: loss 1.4973, time 9.16ms, mfu 0.53%\n",
            "iter 49299: loss 1.4393, time 9.79ms, mfu 0.49%\n",
            "iter 49300: loss 1.4273, time 9.60ms, mfu 0.50%\n",
            "iter 49301: loss 1.4369, time 9.40ms, mfu 0.51%\n",
            "iter 49302: loss 1.4722, time 9.10ms, mfu 0.53%\n",
            "iter 49303: loss 1.4445, time 11.70ms, mfu 0.41%\n",
            "iter 49304: loss 1.4480, time 9.89ms, mfu 0.49%\n",
            "iter 49305: loss 1.4211, time 9.89ms, mfu 0.49%\n",
            "iter 49306: loss 1.4678, time 9.56ms, mfu 0.50%\n",
            "iter 49307: loss 1.5242, time 13.41ms, mfu 0.36%\n",
            "iter 49308: loss 1.4560, time 11.61ms, mfu 0.42%\n",
            "iter 49309: loss 1.4985, time 9.74ms, mfu 0.50%\n",
            "iter 49310: loss 1.5583, time 9.55ms, mfu 0.50%\n",
            "iter 49311: loss 1.5391, time 15.54ms, mfu 0.31%\n",
            "iter 49312: loss 1.3858, time 17.07ms, mfu 0.28%\n",
            "iter 49313: loss 1.5207, time 10.71ms, mfu 0.45%\n",
            "iter 49314: loss 1.4564, time 9.39ms, mfu 0.51%\n",
            "iter 49315: loss 1.5222, time 9.42ms, mfu 0.51%\n",
            "iter 49316: loss 1.5118, time 9.53ms, mfu 0.51%\n",
            "iter 49317: loss 1.5534, time 9.46ms, mfu 0.51%\n",
            "iter 49318: loss 1.4890, time 10.09ms, mfu 0.48%\n",
            "iter 49319: loss 1.3998, time 9.32ms, mfu 0.52%\n",
            "iter 49320: loss 1.5075, time 9.61ms, mfu 0.50%\n",
            "iter 49321: loss 1.4731, time 9.31ms, mfu 0.52%\n",
            "iter 49322: loss 1.4683, time 9.31ms, mfu 0.52%\n",
            "iter 49323: loss 1.3950, time 9.49ms, mfu 0.51%\n",
            "iter 49324: loss 1.4235, time 9.49ms, mfu 0.51%\n",
            "iter 49325: loss 1.4832, time 10.12ms, mfu 0.48%\n",
            "iter 49326: loss 1.4318, time 9.61ms, mfu 0.50%\n",
            "iter 49327: loss 1.4366, time 13.45ms, mfu 0.36%\n",
            "iter 49328: loss 1.4668, time 13.73ms, mfu 0.35%\n",
            "iter 49329: loss 1.4666, time 10.68ms, mfu 0.45%\n",
            "iter 49330: loss 1.4553, time 9.35ms, mfu 0.52%\n",
            "iter 49331: loss 1.3230, time 9.37ms, mfu 0.51%\n",
            "iter 49332: loss 1.4476, time 9.45ms, mfu 0.51%\n",
            "iter 49333: loss 1.3933, time 9.48ms, mfu 0.51%\n",
            "iter 49334: loss 1.4568, time 9.42ms, mfu 0.51%\n",
            "iter 49335: loss 1.4740, time 9.82ms, mfu 0.49%\n",
            "iter 49336: loss 1.3868, time 9.70ms, mfu 0.50%\n",
            "iter 49337: loss 1.3428, time 9.47ms, mfu 0.51%\n",
            "iter 49338: loss 1.4818, time 9.29ms, mfu 0.52%\n",
            "iter 49339: loss 1.5546, time 9.32ms, mfu 0.52%\n",
            "iter 49340: loss 1.4233, time 9.53ms, mfu 0.51%\n",
            "iter 49341: loss 1.4355, time 9.27ms, mfu 0.52%\n",
            "iter 49342: loss 1.4427, time 9.67ms, mfu 0.50%\n",
            "iter 49343: loss 1.5263, time 9.32ms, mfu 0.52%\n",
            "iter 49344: loss 1.5363, time 9.34ms, mfu 0.52%\n",
            "iter 49345: loss 1.4884, time 9.56ms, mfu 0.50%\n",
            "iter 49346: loss 1.3971, time 9.71ms, mfu 0.50%\n",
            "iter 49347: loss 1.3969, time 9.59ms, mfu 0.50%\n",
            "iter 49348: loss 1.4571, time 12.66ms, mfu 0.38%\n",
            "iter 49349: loss 1.2692, time 12.16ms, mfu 0.40%\n",
            "iter 49350: loss 1.4616, time 9.66ms, mfu 0.50%\n",
            "iter 49351: loss 1.5167, time 10.20ms, mfu 0.47%\n",
            "iter 49352: loss 1.4173, time 10.32ms, mfu 0.47%\n",
            "iter 49353: loss 1.4289, time 9.60ms, mfu 0.50%\n",
            "iter 49354: loss 1.5917, time 9.65ms, mfu 0.50%\n",
            "iter 49355: loss 1.4662, time 10.02ms, mfu 0.48%\n",
            "iter 49356: loss 1.4120, time 9.85ms, mfu 0.49%\n",
            "iter 49357: loss 1.5339, time 9.29ms, mfu 0.52%\n",
            "iter 49358: loss 1.4853, time 11.30ms, mfu 0.43%\n",
            "iter 49359: loss 1.4883, time 9.58ms, mfu 0.50%\n",
            "iter 49360: loss 1.5002, time 9.40ms, mfu 0.51%\n",
            "iter 49361: loss 1.5687, time 9.28ms, mfu 0.52%\n",
            "iter 49362: loss 1.4851, time 10.76ms, mfu 0.45%\n",
            "iter 49363: loss 1.4095, time 9.59ms, mfu 0.50%\n",
            "iter 49364: loss 1.3983, time 9.59ms, mfu 0.50%\n",
            "iter 49365: loss 1.5004, time 9.34ms, mfu 0.52%\n",
            "iter 49366: loss 1.4949, time 9.39ms, mfu 0.51%\n",
            "iter 49367: loss 1.3886, time 9.34ms, mfu 0.52%\n",
            "iter 49368: loss 1.4727, time 9.47ms, mfu 0.51%\n",
            "iter 49369: loss 1.4225, time 13.89ms, mfu 0.35%\n",
            "iter 49370: loss 1.4377, time 9.77ms, mfu 0.49%\n",
            "iter 49371: loss 1.4384, time 9.50ms, mfu 0.51%\n",
            "iter 49372: loss 1.3239, time 9.65ms, mfu 0.50%\n",
            "iter 49373: loss 1.3933, time 10.02ms, mfu 0.48%\n",
            "iter 49374: loss 1.4952, time 9.68ms, mfu 0.50%\n",
            "iter 49375: loss 1.6006, time 10.27ms, mfu 0.47%\n",
            "iter 49376: loss 1.3646, time 9.86ms, mfu 0.49%\n",
            "iter 49377: loss 1.4648, time 9.31ms, mfu 0.52%\n",
            "iter 49378: loss 1.4933, time 9.96ms, mfu 0.48%\n",
            "iter 49379: loss 1.5461, time 9.61ms, mfu 0.50%\n",
            "iter 49380: loss 1.5323, time 9.34ms, mfu 0.52%\n",
            "iter 49381: loss 1.5712, time 12.38ms, mfu 0.39%\n",
            "iter 49382: loss 1.3607, time 9.84ms, mfu 0.49%\n",
            "iter 49383: loss 1.4498, time 9.96ms, mfu 0.48%\n",
            "iter 49384: loss 1.4387, time 9.52ms, mfu 0.51%\n",
            "iter 49385: loss 1.4738, time 9.57ms, mfu 0.50%\n",
            "iter 49386: loss 1.4356, time 10.84ms, mfu 0.44%\n",
            "iter 49387: loss 1.5026, time 9.65ms, mfu 0.50%\n",
            "iter 49388: loss 1.5683, time 9.80ms, mfu 0.49%\n",
            "iter 49389: loss 1.4081, time 9.90ms, mfu 0.49%\n",
            "iter 49390: loss 1.3723, time 14.83ms, mfu 0.33%\n",
            "iter 49391: loss 1.4827, time 9.80ms, mfu 0.49%\n",
            "iter 49392: loss 1.4801, time 9.62ms, mfu 0.50%\n",
            "iter 49393: loss 1.4419, time 9.67ms, mfu 0.50%\n",
            "iter 49394: loss 1.4457, time 9.51ms, mfu 0.51%\n",
            "iter 49395: loss 1.4358, time 9.42ms, mfu 0.51%\n",
            "iter 49396: loss 1.4231, time 9.37ms, mfu 0.51%\n",
            "iter 49397: loss 1.4958, time 9.34ms, mfu 0.52%\n",
            "iter 49398: loss 1.4552, time 9.70ms, mfu 0.50%\n",
            "iter 49399: loss 1.4011, time 9.74ms, mfu 0.50%\n",
            "iter 49400: loss 1.5616, time 9.54ms, mfu 0.51%\n",
            "iter 49401: loss 1.4441, time 9.34ms, mfu 0.52%\n",
            "iter 49402: loss 1.5308, time 11.19ms, mfu 0.43%\n",
            "iter 49403: loss 1.3847, time 9.43ms, mfu 0.51%\n",
            "iter 49404: loss 1.4578, time 9.74ms, mfu 0.50%\n",
            "iter 49405: loss 1.4223, time 9.44ms, mfu 0.51%\n",
            "iter 49406: loss 1.6056, time 11.00ms, mfu 0.44%\n",
            "iter 49407: loss 1.4832, time 9.35ms, mfu 0.52%\n",
            "iter 49408: loss 1.3996, time 9.34ms, mfu 0.52%\n",
            "iter 49409: loss 1.4772, time 9.32ms, mfu 0.52%\n",
            "iter 49410: loss 1.4543, time 9.94ms, mfu 0.49%\n",
            "iter 49411: loss 1.4685, time 16.76ms, mfu 0.29%\n",
            "iter 49412: loss 1.5136, time 11.53ms, mfu 0.42%\n",
            "iter 49413: loss 1.3541, time 14.34ms, mfu 0.34%\n",
            "iter 49414: loss 1.4121, time 12.69ms, mfu 0.38%\n",
            "iter 49415: loss 1.4824, time 9.55ms, mfu 0.50%\n",
            "iter 49416: loss 1.4937, time 9.25ms, mfu 0.52%\n",
            "iter 49417: loss 1.5026, time 9.37ms, mfu 0.51%\n",
            "iter 49418: loss 1.5128, time 9.37ms, mfu 0.51%\n",
            "iter 49419: loss 1.4575, time 9.17ms, mfu 0.53%\n",
            "iter 49420: loss 1.4512, time 9.22ms, mfu 0.52%\n",
            "iter 49421: loss 1.3810, time 9.19ms, mfu 0.52%\n",
            "iter 49422: loss 1.5316, time 9.44ms, mfu 0.51%\n",
            "iter 49423: loss 1.4561, time 9.21ms, mfu 0.52%\n",
            "iter 49424: loss 1.3395, time 9.88ms, mfu 0.49%\n",
            "iter 49425: loss 1.5166, time 9.50ms, mfu 0.51%\n",
            "iter 49426: loss 1.4150, time 9.46ms, mfu 0.51%\n",
            "iter 49427: loss 1.4244, time 9.21ms, mfu 0.52%\n",
            "iter 49428: loss 1.4111, time 9.17ms, mfu 0.53%\n",
            "iter 49429: loss 1.5478, time 9.44ms, mfu 0.51%\n",
            "iter 49430: loss 1.4674, time 9.25ms, mfu 0.52%\n",
            "iter 49431: loss 1.5749, time 9.40ms, mfu 0.51%\n",
            "iter 49432: loss 1.4726, time 12.76ms, mfu 0.38%\n",
            "iter 49433: loss 1.5025, time 13.02ms, mfu 0.37%\n",
            "iter 49434: loss 1.4137, time 9.24ms, mfu 0.52%\n",
            "iter 49435: loss 1.4200, time 9.55ms, mfu 0.50%\n",
            "iter 49436: loss 1.4847, time 9.24ms, mfu 0.52%\n",
            "iter 49437: loss 1.5121, time 9.61ms, mfu 0.50%\n",
            "iter 49438: loss 1.6204, time 9.63ms, mfu 0.50%\n",
            "iter 49439: loss 1.4866, time 9.43ms, mfu 0.51%\n",
            "iter 49440: loss 1.5665, time 9.17ms, mfu 0.53%\n",
            "iter 49441: loss 1.4462, time 9.18ms, mfu 0.53%\n",
            "iter 49442: loss 1.4917, time 9.45ms, mfu 0.51%\n",
            "iter 49443: loss 1.3883, time 9.15ms, mfu 0.53%\n",
            "iter 49444: loss 1.4991, time 9.26ms, mfu 0.52%\n",
            "iter 49445: loss 1.4729, time 9.35ms, mfu 0.52%\n",
            "iter 49446: loss 1.4399, time 9.71ms, mfu 0.50%\n",
            "iter 49447: loss 1.4155, time 9.75ms, mfu 0.49%\n",
            "iter 49448: loss 1.4023, time 9.51ms, mfu 0.51%\n",
            "iter 49449: loss 1.4011, time 9.55ms, mfu 0.51%\n",
            "iter 49450: loss 1.4273, time 9.31ms, mfu 0.52%\n",
            "iter 49451: loss 1.4735, time 9.27ms, mfu 0.52%\n",
            "iter 49452: loss 1.4751, time 9.14ms, mfu 0.53%\n",
            "iter 49453: loss 1.4750, time 9.19ms, mfu 0.52%\n",
            "iter 49454: loss 1.5476, time 13.89ms, mfu 0.35%\n",
            "iter 49455: loss 1.4377, time 9.86ms, mfu 0.49%\n",
            "iter 49456: loss 1.4618, time 9.33ms, mfu 0.52%\n",
            "iter 49457: loss 1.4882, time 9.25ms, mfu 0.52%\n",
            "iter 49458: loss 1.3842, time 9.14ms, mfu 0.53%\n",
            "iter 49459: loss 1.5416, time 9.11ms, mfu 0.53%\n",
            "iter 49460: loss 1.4363, time 9.21ms, mfu 0.52%\n",
            "iter 49461: loss 1.5239, time 9.68ms, mfu 0.50%\n",
            "iter 49462: loss 1.7670, time 9.90ms, mfu 0.49%\n",
            "iter 49463: loss 1.4042, time 9.20ms, mfu 0.52%\n",
            "iter 49464: loss 1.3939, time 9.80ms, mfu 0.49%\n",
            "iter 49465: loss 1.4398, time 9.48ms, mfu 0.51%\n",
            "iter 49466: loss 1.5635, time 9.11ms, mfu 0.53%\n",
            "iter 49467: loss 1.4634, time 9.36ms, mfu 0.52%\n",
            "iter 49468: loss 1.4377, time 9.17ms, mfu 0.53%\n",
            "iter 49469: loss 1.4481, time 9.23ms, mfu 0.52%\n",
            "iter 49470: loss 1.4762, time 9.22ms, mfu 0.52%\n",
            "iter 49471: loss 1.5151, time 9.27ms, mfu 0.52%\n",
            "iter 49472: loss 1.4284, time 9.22ms, mfu 0.52%\n",
            "iter 49473: loss 1.5081, time 9.51ms, mfu 0.51%\n",
            "iter 49474: loss 1.5023, time 9.38ms, mfu 0.51%\n",
            "iter 49475: loss 1.3944, time 9.46ms, mfu 0.51%\n",
            "iter 49476: loss 1.4002, time 14.55ms, mfu 0.33%\n",
            "iter 49477: loss 1.4589, time 9.68ms, mfu 0.50%\n",
            "iter 49478: loss 1.5338, time 9.25ms, mfu 0.52%\n",
            "iter 49479: loss 1.4560, time 11.02ms, mfu 0.44%\n",
            "iter 49480: loss 1.5556, time 9.39ms, mfu 0.51%\n",
            "iter 49481: loss 1.3723, time 9.37ms, mfu 0.51%\n",
            "iter 49482: loss 1.4551, time 9.31ms, mfu 0.52%\n",
            "iter 49483: loss 1.4055, time 10.38ms, mfu 0.46%\n",
            "iter 49484: loss 1.4239, time 9.32ms, mfu 0.52%\n",
            "iter 49485: loss 1.5191, time 9.32ms, mfu 0.52%\n",
            "iter 49486: loss 1.5437, time 9.13ms, mfu 0.53%\n",
            "iter 49487: loss 1.5528, time 10.84ms, mfu 0.44%\n",
            "iter 49488: loss 1.5146, time 9.47ms, mfu 0.51%\n",
            "iter 49489: loss 1.5261, time 9.24ms, mfu 0.52%\n",
            "iter 49490: loss 1.4804, time 9.32ms, mfu 0.52%\n",
            "iter 49491: loss 1.3817, time 9.58ms, mfu 0.50%\n",
            "iter 49492: loss 1.3544, time 9.30ms, mfu 0.52%\n",
            "iter 49493: loss 1.5008, time 9.30ms, mfu 0.52%\n",
            "iter 49494: loss 1.4211, time 9.40ms, mfu 0.51%\n",
            "iter 49495: loss 1.4262, time 9.18ms, mfu 0.53%\n",
            "iter 49496: loss 1.4005, time 9.14ms, mfu 0.53%\n",
            "iter 49497: loss 1.4985, time 9.31ms, mfu 0.52%\n",
            "iter 49498: loss 1.3878, time 12.95ms, mfu 0.37%\n",
            "iter 49499: loss 1.3961, time 9.56ms, mfu 0.50%\n",
            "step 49500: train loss 1.3793, val loss 1.6474\n",
            "iter 49500: loss 1.4404, time 16.77ms, mfu 0.29%\n",
            "iter 49501: loss 1.3634, time 9.55ms, mfu 0.50%\n",
            "iter 49502: loss 1.4772, time 9.87ms, mfu 0.49%\n",
            "iter 49503: loss 1.4415, time 11.11ms, mfu 0.43%\n",
            "iter 49504: loss 1.4527, time 9.69ms, mfu 0.50%\n",
            "iter 49505: loss 1.4216, time 9.51ms, mfu 0.51%\n",
            "iter 49506: loss 1.5489, time 9.73ms, mfu 0.50%\n",
            "iter 49507: loss 1.4571, time 11.38ms, mfu 0.42%\n",
            "iter 49508: loss 1.4759, time 9.85ms, mfu 0.49%\n",
            "iter 49509: loss 1.5058, time 10.02ms, mfu 0.48%\n",
            "iter 49510: loss 1.5761, time 9.88ms, mfu 0.49%\n",
            "iter 49511: loss 1.4522, time 10.23ms, mfu 0.47%\n",
            "iter 49512: loss 1.5245, time 12.42ms, mfu 0.39%\n",
            "iter 49513: loss 1.3338, time 10.11ms, mfu 0.48%\n",
            "iter 49514: loss 1.6162, time 9.67ms, mfu 0.50%\n",
            "iter 49515: loss 1.4051, time 9.44ms, mfu 0.51%\n",
            "iter 49516: loss 1.4586, time 13.45ms, mfu 0.36%\n",
            "iter 49517: loss 1.4587, time 17.19ms, mfu 0.28%\n",
            "iter 49518: loss 1.5455, time 14.03ms, mfu 0.34%\n",
            "iter 49519: loss 1.4814, time 9.77ms, mfu 0.49%\n",
            "iter 49520: loss 1.4562, time 9.30ms, mfu 0.52%\n",
            "iter 49521: loss 1.4031, time 9.72ms, mfu 0.50%\n",
            "iter 49522: loss 1.4612, time 9.65ms, mfu 0.50%\n",
            "iter 49523: loss 1.4260, time 9.45ms, mfu 0.51%\n",
            "iter 49524: loss 1.5205, time 9.41ms, mfu 0.51%\n",
            "iter 49525: loss 1.4279, time 10.44ms, mfu 0.46%\n",
            "iter 49526: loss 1.3736, time 9.57ms, mfu 0.50%\n",
            "iter 49527: loss 1.5404, time 9.61ms, mfu 0.50%\n",
            "iter 49528: loss 1.4565, time 9.63ms, mfu 0.50%\n",
            "iter 49529: loss 1.4336, time 9.51ms, mfu 0.51%\n",
            "iter 49530: loss 1.4707, time 9.69ms, mfu 0.50%\n",
            "iter 49531: loss 1.4331, time 9.40ms, mfu 0.51%\n",
            "iter 49532: loss 1.4706, time 9.59ms, mfu 0.50%\n",
            "iter 49533: loss 1.4755, time 9.27ms, mfu 0.52%\n",
            "iter 49534: loss 1.4443, time 9.37ms, mfu 0.51%\n",
            "iter 49535: loss 1.4735, time 9.44ms, mfu 0.51%\n",
            "iter 49536: loss 1.4665, time 9.51ms, mfu 0.51%\n",
            "iter 49537: loss 1.5495, time 9.39ms, mfu 0.51%\n",
            "iter 49538: loss 1.4442, time 14.23ms, mfu 0.34%\n",
            "iter 49539: loss 1.3650, time 10.78ms, mfu 0.45%\n",
            "iter 49540: loss 1.5217, time 9.27ms, mfu 0.52%\n",
            "iter 49541: loss 1.4497, time 9.34ms, mfu 0.52%\n",
            "iter 49542: loss 1.4156, time 9.66ms, mfu 0.50%\n",
            "iter 49543: loss 1.4682, time 9.35ms, mfu 0.52%\n",
            "iter 49544: loss 1.5021, time 9.41ms, mfu 0.51%\n",
            "iter 49545: loss 1.4654, time 9.53ms, mfu 0.51%\n",
            "iter 49546: loss 1.3951, time 9.62ms, mfu 0.50%\n",
            "iter 49547: loss 1.4604, time 9.48ms, mfu 0.51%\n",
            "iter 49548: loss 1.6326, time 9.81ms, mfu 0.49%\n",
            "iter 49549: loss 1.5023, time 9.78ms, mfu 0.49%\n",
            "iter 49550: loss 1.4627, time 9.39ms, mfu 0.51%\n",
            "iter 49551: loss 1.4372, time 9.30ms, mfu 0.52%\n",
            "iter 49552: loss 1.5216, time 9.34ms, mfu 0.52%\n",
            "iter 49553: loss 1.4576, time 9.84ms, mfu 0.49%\n",
            "iter 49554: loss 1.5560, time 12.80ms, mfu 0.38%\n",
            "iter 49555: loss 1.3567, time 9.59ms, mfu 0.50%\n",
            "iter 49556: loss 1.3461, time 9.63ms, mfu 0.50%\n",
            "iter 49557: loss 1.5053, time 9.38ms, mfu 0.51%\n",
            "iter 49558: loss 1.4139, time 9.52ms, mfu 0.51%\n",
            "iter 49559: loss 1.5741, time 11.50ms, mfu 0.42%\n",
            "iter 49560: loss 1.5547, time 11.07ms, mfu 0.44%\n",
            "iter 49561: loss 1.4911, time 9.33ms, mfu 0.52%\n",
            "iter 49562: loss 1.5132, time 9.34ms, mfu 0.52%\n",
            "iter 49563: loss 1.4920, time 9.42ms, mfu 0.51%\n",
            "iter 49564: loss 1.4048, time 9.68ms, mfu 0.50%\n",
            "iter 49565: loss 1.4211, time 9.38ms, mfu 0.51%\n",
            "iter 49566: loss 1.4532, time 9.82ms, mfu 0.49%\n",
            "iter 49567: loss 1.4905, time 9.46ms, mfu 0.51%\n",
            "iter 49568: loss 1.3826, time 9.67ms, mfu 0.50%\n",
            "iter 49569: loss 1.4793, time 11.63ms, mfu 0.41%\n",
            "iter 49570: loss 1.4336, time 9.97ms, mfu 0.48%\n",
            "iter 49571: loss 1.6419, time 9.83ms, mfu 0.49%\n",
            "iter 49572: loss 1.4020, time 9.51ms, mfu 0.51%\n",
            "iter 49573: loss 1.4640, time 10.75ms, mfu 0.45%\n",
            "iter 49574: loss 1.4296, time 9.46ms, mfu 0.51%\n",
            "iter 49575: loss 1.4195, time 9.38ms, mfu 0.51%\n",
            "iter 49576: loss 1.5665, time 9.40ms, mfu 0.51%\n",
            "iter 49577: loss 1.4024, time 9.51ms, mfu 0.51%\n",
            "iter 49578: loss 1.5304, time 9.40ms, mfu 0.51%\n",
            "iter 49579: loss 1.4747, time 9.88ms, mfu 0.49%\n",
            "iter 49580: loss 1.4734, time 13.89ms, mfu 0.35%\n",
            "iter 49581: loss 1.4100, time 11.33ms, mfu 0.43%\n",
            "iter 49582: loss 1.3039, time 9.57ms, mfu 0.50%\n",
            "iter 49583: loss 1.4436, time 9.72ms, mfu 0.50%\n",
            "iter 49584: loss 1.4432, time 9.56ms, mfu 0.50%\n",
            "iter 49585: loss 1.3621, time 9.35ms, mfu 0.52%\n",
            "iter 49586: loss 1.4414, time 10.63ms, mfu 0.45%\n",
            "iter 49587: loss 1.3877, time 9.46ms, mfu 0.51%\n",
            "iter 49588: loss 1.3578, time 9.33ms, mfu 0.52%\n",
            "iter 49589: loss 1.4683, time 9.33ms, mfu 0.52%\n",
            "iter 49590: loss 1.4300, time 9.84ms, mfu 0.49%\n",
            "iter 49591: loss 1.5338, time 13.07ms, mfu 0.37%\n",
            "iter 49592: loss 1.4980, time 9.82ms, mfu 0.49%\n",
            "iter 49593: loss 1.5051, time 9.79ms, mfu 0.49%\n",
            "iter 49594: loss 1.4638, time 9.21ms, mfu 0.52%\n",
            "iter 49595: loss 1.4532, time 9.45ms, mfu 0.51%\n",
            "iter 49596: loss 1.4844, time 9.44ms, mfu 0.51%\n",
            "iter 49597: loss 1.3587, time 9.22ms, mfu 0.52%\n",
            "iter 49598: loss 1.5109, time 10.12ms, mfu 0.48%\n",
            "iter 49599: loss 1.6042, time 9.99ms, mfu 0.48%\n",
            "iter 49600: loss 1.4804, time 9.33ms, mfu 0.52%\n",
            "iter 49601: loss 1.5133, time 12.73ms, mfu 0.38%\n",
            "iter 49602: loss 1.4813, time 12.30ms, mfu 0.39%\n",
            "iter 49603: loss 1.4402, time 10.10ms, mfu 0.48%\n",
            "iter 49604: loss 1.5059, time 9.45ms, mfu 0.51%\n",
            "iter 49605: loss 1.4736, time 9.52ms, mfu 0.51%\n",
            "iter 49606: loss 1.5112, time 11.47ms, mfu 0.42%\n",
            "iter 49607: loss 1.5742, time 9.47ms, mfu 0.51%\n",
            "iter 49608: loss 1.5209, time 9.66ms, mfu 0.50%\n",
            "iter 49609: loss 1.4705, time 9.48ms, mfu 0.51%\n",
            "iter 49610: loss 1.3898, time 10.31ms, mfu 0.47%\n",
            "iter 49611: loss 1.4642, time 11.89ms, mfu 0.41%\n",
            "iter 49612: loss 1.4624, time 11.10ms, mfu 0.43%\n",
            "iter 49613: loss 1.4360, time 9.35ms, mfu 0.52%\n",
            "iter 49614: loss 1.5151, time 9.60ms, mfu 0.50%\n",
            "iter 49615: loss 1.4273, time 9.54ms, mfu 0.51%\n",
            "iter 49616: loss 1.5096, time 9.57ms, mfu 0.50%\n",
            "iter 49617: loss 1.4561, time 9.32ms, mfu 0.52%\n",
            "iter 49618: loss 1.4970, time 9.38ms, mfu 0.51%\n",
            "iter 49619: loss 1.5576, time 14.72ms, mfu 0.33%\n",
            "iter 49620: loss 1.5771, time 14.70ms, mfu 0.33%\n",
            "iter 49621: loss 1.4055, time 14.81ms, mfu 0.33%\n",
            "iter 49622: loss 1.3488, time 9.82ms, mfu 0.49%\n",
            "iter 49623: loss 1.3680, time 9.83ms, mfu 0.49%\n",
            "iter 49624: loss 1.4981, time 10.27ms, mfu 0.47%\n",
            "iter 49625: loss 1.5147, time 9.98ms, mfu 0.48%\n",
            "iter 49626: loss 1.4389, time 9.33ms, mfu 0.52%\n",
            "iter 49627: loss 1.5387, time 9.52ms, mfu 0.51%\n",
            "iter 49628: loss 1.5073, time 9.59ms, mfu 0.50%\n",
            "iter 49629: loss 1.4796, time 9.55ms, mfu 0.50%\n",
            "iter 49630: loss 1.5450, time 9.47ms, mfu 0.51%\n",
            "iter 49631: loss 1.4634, time 9.72ms, mfu 0.50%\n",
            "iter 49632: loss 1.4176, time 9.42ms, mfu 0.51%\n",
            "iter 49633: loss 1.4099, time 9.47ms, mfu 0.51%\n",
            "iter 49634: loss 1.4901, time 9.52ms, mfu 0.51%\n",
            "iter 49635: loss 1.5591, time 9.49ms, mfu 0.51%\n",
            "iter 49636: loss 1.5288, time 9.28ms, mfu 0.52%\n",
            "iter 49637: loss 1.4099, time 9.89ms, mfu 0.49%\n",
            "iter 49638: loss 1.4995, time 9.51ms, mfu 0.51%\n",
            "iter 49639: loss 1.4853, time 9.70ms, mfu 0.50%\n",
            "iter 49640: loss 1.4048, time 9.54ms, mfu 0.51%\n",
            "iter 49641: loss 1.5173, time 9.88ms, mfu 0.49%\n",
            "iter 49642: loss 1.5367, time 12.26ms, mfu 0.39%\n",
            "iter 49643: loss 1.4112, time 12.28ms, mfu 0.39%\n",
            "iter 49644: loss 1.4515, time 9.79ms, mfu 0.49%\n",
            "iter 49645: loss 1.4592, time 10.00ms, mfu 0.48%\n",
            "iter 49646: loss 1.5335, time 9.31ms, mfu 0.52%\n",
            "iter 49647: loss 1.5399, time 9.59ms, mfu 0.50%\n",
            "iter 49648: loss 1.4844, time 9.76ms, mfu 0.49%\n",
            "iter 49649: loss 1.5580, time 9.45ms, mfu 0.51%\n",
            "iter 49650: loss 1.4655, time 9.27ms, mfu 0.52%\n",
            "iter 49651: loss 1.5263, time 9.39ms, mfu 0.51%\n",
            "iter 49652: loss 1.4044, time 9.33ms, mfu 0.52%\n",
            "iter 49653: loss 1.5435, time 9.29ms, mfu 0.52%\n",
            "iter 49654: loss 1.5222, time 9.35ms, mfu 0.52%\n",
            "iter 49655: loss 1.4403, time 9.41ms, mfu 0.51%\n",
            "iter 49656: loss 1.5615, time 9.54ms, mfu 0.51%\n",
            "iter 49657: loss 1.4605, time 9.43ms, mfu 0.51%\n",
            "iter 49658: loss 1.4445, time 11.48ms, mfu 0.42%\n",
            "iter 49659: loss 1.4794, time 9.39ms, mfu 0.51%\n",
            "iter 49660: loss 1.4118, time 9.32ms, mfu 0.52%\n",
            "iter 49661: loss 1.4435, time 9.86ms, mfu 0.49%\n",
            "iter 49662: loss 1.4591, time 10.87ms, mfu 0.44%\n",
            "iter 49663: loss 1.4475, time 11.96ms, mfu 0.40%\n",
            "iter 49664: loss 1.4935, time 10.79ms, mfu 0.45%\n",
            "iter 49665: loss 1.4747, time 9.47ms, mfu 0.51%\n",
            "iter 49666: loss 1.4235, time 9.52ms, mfu 0.51%\n",
            "iter 49667: loss 1.3364, time 9.28ms, mfu 0.52%\n",
            "iter 49668: loss 1.4004, time 9.59ms, mfu 0.50%\n",
            "iter 49669: loss 1.4276, time 9.33ms, mfu 0.52%\n",
            "iter 49670: loss 1.5050, time 9.38ms, mfu 0.51%\n",
            "iter 49671: loss 1.3780, time 9.45ms, mfu 0.51%\n",
            "iter 49672: loss 1.4571, time 9.45ms, mfu 0.51%\n",
            "iter 49673: loss 1.4612, time 9.81ms, mfu 0.49%\n",
            "iter 49674: loss 1.4989, time 9.51ms, mfu 0.51%\n",
            "iter 49675: loss 1.5121, time 9.47ms, mfu 0.51%\n",
            "iter 49676: loss 1.3998, time 9.54ms, mfu 0.51%\n",
            "iter 49677: loss 1.4357, time 9.46ms, mfu 0.51%\n",
            "iter 49678: loss 1.5346, time 10.68ms, mfu 0.45%\n",
            "iter 49679: loss 1.5287, time 9.43ms, mfu 0.51%\n",
            "iter 49680: loss 1.4900, time 9.24ms, mfu 0.52%\n",
            "iter 49681: loss 1.4332, time 9.50ms, mfu 0.51%\n",
            "iter 49682: loss 1.3773, time 9.33ms, mfu 0.52%\n",
            "iter 49683: loss 1.5346, time 9.42ms, mfu 0.51%\n",
            "iter 49684: loss 1.4098, time 9.91ms, mfu 0.49%\n",
            "iter 49685: loss 1.5388, time 14.74ms, mfu 0.33%\n",
            "iter 49686: loss 1.5573, time 10.88ms, mfu 0.44%\n",
            "iter 49687: loss 1.4041, time 9.71ms, mfu 0.50%\n",
            "iter 49688: loss 1.5557, time 9.57ms, mfu 0.50%\n",
            "iter 49689: loss 1.4016, time 9.66ms, mfu 0.50%\n",
            "iter 49690: loss 1.5204, time 9.53ms, mfu 0.51%\n",
            "iter 49691: loss 1.4263, time 9.67ms, mfu 0.50%\n",
            "iter 49692: loss 1.5630, time 9.50ms, mfu 0.51%\n",
            "iter 49693: loss 1.4586, time 12.77ms, mfu 0.38%\n",
            "iter 49694: loss 1.4486, time 9.58ms, mfu 0.50%\n",
            "iter 49695: loss 1.4622, time 10.00ms, mfu 0.48%\n",
            "iter 49696: loss 1.5372, time 9.54ms, mfu 0.51%\n",
            "iter 49697: loss 1.4678, time 9.66ms, mfu 0.50%\n",
            "iter 49698: loss 1.5486, time 12.80ms, mfu 0.38%\n",
            "iter 49699: loss 1.3820, time 10.63ms, mfu 0.45%\n",
            "iter 49700: loss 1.5210, time 9.73ms, mfu 0.50%\n",
            "iter 49701: loss 1.4256, time 9.61ms, mfu 0.50%\n",
            "iter 49702: loss 1.4263, time 11.08ms, mfu 0.44%\n",
            "iter 49703: loss 1.4347, time 10.05ms, mfu 0.48%\n",
            "iter 49704: loss 1.4525, time 10.30ms, mfu 0.47%\n",
            "iter 49705: loss 1.4647, time 15.28ms, mfu 0.32%\n",
            "iter 49706: loss 1.4095, time 10.49ms, mfu 0.46%\n",
            "iter 49707: loss 1.5069, time 9.51ms, mfu 0.51%\n",
            "iter 49708: loss 1.5579, time 9.59ms, mfu 0.50%\n",
            "iter 49709: loss 1.6214, time 10.86ms, mfu 0.44%\n",
            "iter 49710: loss 1.5535, time 12.79ms, mfu 0.38%\n",
            "iter 49711: loss 1.5585, time 10.20ms, mfu 0.47%\n",
            "iter 49712: loss 1.4070, time 9.47ms, mfu 0.51%\n",
            "iter 49713: loss 1.4442, time 9.44ms, mfu 0.51%\n",
            "iter 49714: loss 1.4742, time 10.15ms, mfu 0.48%\n",
            "iter 49715: loss 1.5200, time 9.54ms, mfu 0.51%\n",
            "iter 49716: loss 1.3886, time 9.36ms, mfu 0.52%\n",
            "iter 49717: loss 1.4779, time 9.67ms, mfu 0.50%\n",
            "iter 49718: loss 1.5194, time 9.52ms, mfu 0.51%\n",
            "iter 49719: loss 1.4379, time 9.33ms, mfu 0.52%\n",
            "iter 49720: loss 1.5580, time 11.79ms, mfu 0.41%\n",
            "iter 49721: loss 1.5517, time 15.00ms, mfu 0.32%\n",
            "iter 49722: loss 1.5111, time 11.48ms, mfu 0.42%\n",
            "iter 49723: loss 1.5750, time 10.72ms, mfu 0.45%\n",
            "iter 49724: loss 1.3958, time 9.58ms, mfu 0.50%\n",
            "iter 49725: loss 1.4564, time 14.88ms, mfu 0.32%\n",
            "iter 49726: loss 1.4801, time 9.43ms, mfu 0.51%\n",
            "iter 49727: loss 1.5170, time 9.28ms, mfu 0.52%\n",
            "iter 49728: loss 1.4460, time 9.25ms, mfu 0.52%\n",
            "iter 49729: loss 1.5233, time 9.94ms, mfu 0.49%\n",
            "iter 49730: loss 1.3492, time 10.01ms, mfu 0.48%\n",
            "iter 49731: loss 1.4285, time 9.54ms, mfu 0.51%\n",
            "iter 49732: loss 1.2587, time 9.28ms, mfu 0.52%\n",
            "iter 49733: loss 1.4949, time 9.59ms, mfu 0.50%\n",
            "iter 49734: loss 1.4597, time 9.38ms, mfu 0.51%\n",
            "iter 49735: loss 1.4002, time 9.29ms, mfu 0.52%\n",
            "iter 49736: loss 1.5085, time 9.28ms, mfu 0.52%\n",
            "iter 49737: loss 1.5115, time 9.34ms, mfu 0.52%\n",
            "iter 49738: loss 1.3718, time 9.79ms, mfu 0.49%\n",
            "iter 49739: loss 1.3509, time 9.44ms, mfu 0.51%\n",
            "iter 49740: loss 1.5332, time 9.71ms, mfu 0.50%\n",
            "iter 49741: loss 1.4952, time 9.34ms, mfu 0.52%\n",
            "iter 49742: loss 1.3673, time 9.40ms, mfu 0.51%\n",
            "iter 49743: loss 1.4239, time 9.60ms, mfu 0.50%\n",
            "iter 49744: loss 1.4280, time 13.57ms, mfu 0.36%\n",
            "iter 49745: loss 1.5378, time 15.31ms, mfu 0.31%\n",
            "iter 49746: loss 1.3925, time 17.60ms, mfu 0.27%\n",
            "iter 49747: loss 1.6151, time 14.58ms, mfu 0.33%\n",
            "iter 49748: loss 1.5978, time 14.58ms, mfu 0.33%\n",
            "iter 49749: loss 1.4707, time 15.42ms, mfu 0.31%\n",
            "iter 49750: loss 1.4639, time 14.94ms, mfu 0.32%\n",
            "iter 49751: loss 1.4107, time 15.21ms, mfu 0.32%\n",
            "iter 49752: loss 1.5201, time 13.52ms, mfu 0.36%\n",
            "iter 49753: loss 1.4451, time 12.22ms, mfu 0.39%\n",
            "iter 49754: loss 1.3606, time 14.21ms, mfu 0.34%\n",
            "iter 49755: loss 1.4374, time 12.50ms, mfu 0.39%\n",
            "iter 49756: loss 1.5133, time 12.56ms, mfu 0.38%\n",
            "iter 49757: loss 1.5067, time 11.76ms, mfu 0.41%\n",
            "iter 49758: loss 1.4239, time 13.43ms, mfu 0.36%\n",
            "iter 49759: loss 1.4472, time 15.03ms, mfu 0.32%\n",
            "iter 49760: loss 1.5614, time 15.32ms, mfu 0.31%\n",
            "iter 49761: loss 1.4707, time 20.98ms, mfu 0.23%\n",
            "iter 49762: loss 1.4101, time 13.54ms, mfu 0.36%\n",
            "iter 49763: loss 1.4848, time 13.00ms, mfu 0.37%\n",
            "iter 49764: loss 1.4827, time 14.77ms, mfu 0.33%\n",
            "iter 49765: loss 1.5139, time 12.81ms, mfu 0.38%\n",
            "iter 49766: loss 1.5139, time 12.35ms, mfu 0.39%\n",
            "iter 49767: loss 1.3541, time 14.16ms, mfu 0.34%\n",
            "iter 49768: loss 1.3886, time 12.17ms, mfu 0.40%\n",
            "iter 49769: loss 1.4852, time 12.15ms, mfu 0.40%\n",
            "iter 49770: loss 1.4499, time 15.39ms, mfu 0.31%\n",
            "iter 49771: loss 1.3294, time 12.31ms, mfu 0.39%\n",
            "iter 49772: loss 1.4422, time 12.72ms, mfu 0.38%\n",
            "iter 49773: loss 1.4895, time 13.11ms, mfu 0.37%\n",
            "iter 49774: loss 1.3527, time 12.52ms, mfu 0.39%\n",
            "iter 49775: loss 1.5524, time 12.30ms, mfu 0.39%\n",
            "iter 49776: loss 1.5423, time 12.53ms, mfu 0.38%\n",
            "iter 49777: loss 1.4534, time 14.84ms, mfu 0.32%\n",
            "iter 49778: loss 1.4541, time 12.36ms, mfu 0.39%\n",
            "iter 49779: loss 1.4657, time 12.13ms, mfu 0.40%\n",
            "iter 49780: loss 1.4427, time 12.28ms, mfu 0.39%\n",
            "iter 49781: loss 1.4030, time 12.14ms, mfu 0.40%\n",
            "iter 49782: loss 1.4793, time 12.63ms, mfu 0.38%\n",
            "iter 49783: loss 1.4756, time 12.55ms, mfu 0.38%\n",
            "iter 49784: loss 1.4355, time 20.62ms, mfu 0.23%\n",
            "iter 49785: loss 1.4957, time 16.21ms, mfu 0.30%\n",
            "iter 49786: loss 1.5443, time 18.16ms, mfu 0.27%\n",
            "iter 49787: loss 1.5204, time 15.99ms, mfu 0.30%\n",
            "iter 49788: loss 1.4597, time 13.44ms, mfu 0.36%\n",
            "iter 49789: loss 1.4688, time 13.26ms, mfu 0.36%\n",
            "iter 49790: loss 1.5345, time 16.80ms, mfu 0.29%\n",
            "iter 49791: loss 1.4741, time 15.35ms, mfu 0.31%\n",
            "iter 49792: loss 1.4202, time 19.72ms, mfu 0.24%\n",
            "iter 49793: loss 1.4693, time 19.22ms, mfu 0.25%\n",
            "iter 49794: loss 1.5179, time 13.78ms, mfu 0.35%\n",
            "iter 49795: loss 1.5413, time 12.00ms, mfu 0.40%\n",
            "iter 49796: loss 1.4676, time 12.28ms, mfu 0.39%\n",
            "iter 49797: loss 1.4593, time 11.81ms, mfu 0.41%\n",
            "iter 49798: loss 1.4516, time 12.15ms, mfu 0.40%\n",
            "iter 49799: loss 1.5275, time 15.67ms, mfu 0.31%\n",
            "iter 49800: loss 1.4824, time 14.79ms, mfu 0.33%\n",
            "iter 49801: loss 1.5084, time 15.27ms, mfu 0.32%\n",
            "iter 49802: loss 1.4516, time 14.44ms, mfu 0.33%\n",
            "iter 49803: loss 1.5562, time 11.92ms, mfu 0.40%\n",
            "iter 49804: loss 1.4751, time 11.80ms, mfu 0.41%\n",
            "iter 49805: loss 1.4989, time 11.91ms, mfu 0.40%\n",
            "iter 49806: loss 1.4973, time 17.25ms, mfu 0.28%\n",
            "iter 49807: loss 1.5675, time 14.83ms, mfu 0.33%\n",
            "iter 49808: loss 1.5706, time 12.71ms, mfu 0.38%\n",
            "iter 49809: loss 1.4822, time 11.68ms, mfu 0.41%\n",
            "iter 49810: loss 1.4409, time 12.21ms, mfu 0.39%\n",
            "iter 49811: loss 1.5446, time 12.05ms, mfu 0.40%\n",
            "iter 49812: loss 1.4608, time 11.77ms, mfu 0.41%\n",
            "iter 49813: loss 1.4307, time 11.75ms, mfu 0.41%\n",
            "iter 49814: loss 1.4740, time 11.65ms, mfu 0.41%\n",
            "iter 49815: loss 1.4614, time 11.74ms, mfu 0.41%\n",
            "iter 49816: loss 1.4977, time 11.85ms, mfu 0.41%\n",
            "iter 49817: loss 1.5077, time 12.04ms, mfu 0.40%\n",
            "iter 49818: loss 1.4790, time 14.95ms, mfu 0.32%\n",
            "iter 49819: loss 1.5020, time 15.52ms, mfu 0.31%\n",
            "iter 49820: loss 1.4516, time 12.84ms, mfu 0.38%\n",
            "iter 49821: loss 1.4447, time 14.74ms, mfu 0.33%\n",
            "iter 49822: loss 1.3588, time 19.53ms, mfu 0.25%\n",
            "iter 49823: loss 1.5674, time 20.17ms, mfu 0.24%\n",
            "iter 49824: loss 1.4602, time 14.87ms, mfu 0.32%\n",
            "iter 49825: loss 1.4934, time 13.35ms, mfu 0.36%\n",
            "iter 49826: loss 1.4445, time 14.83ms, mfu 0.33%\n",
            "iter 49827: loss 1.4245, time 12.35ms, mfu 0.39%\n",
            "iter 49828: loss 1.4243, time 12.63ms, mfu 0.38%\n",
            "iter 49829: loss 1.3980, time 13.33ms, mfu 0.36%\n",
            "iter 49830: loss 1.4414, time 14.77ms, mfu 0.33%\n",
            "iter 49831: loss 1.4447, time 13.49ms, mfu 0.36%\n",
            "iter 49832: loss 1.5049, time 15.96ms, mfu 0.30%\n",
            "iter 49833: loss 1.4672, time 15.20ms, mfu 0.32%\n",
            "iter 49834: loss 1.4719, time 14.76ms, mfu 0.33%\n",
            "iter 49835: loss 1.3845, time 14.47ms, mfu 0.33%\n",
            "iter 49836: loss 1.4725, time 15.69ms, mfu 0.31%\n",
            "iter 49837: loss 1.4039, time 19.84ms, mfu 0.24%\n",
            "iter 49838: loss 1.5374, time 12.33ms, mfu 0.39%\n",
            "iter 49839: loss 1.5335, time 14.74ms, mfu 0.33%\n",
            "iter 49840: loss 1.4508, time 12.93ms, mfu 0.37%\n",
            "iter 49841: loss 1.4585, time 11.89ms, mfu 0.41%\n",
            "iter 49842: loss 1.5177, time 11.93ms, mfu 0.40%\n",
            "iter 49843: loss 1.4552, time 11.91ms, mfu 0.40%\n",
            "iter 49844: loss 1.4615, time 12.31ms, mfu 0.39%\n",
            "iter 49845: loss 1.4554, time 13.03ms, mfu 0.37%\n",
            "iter 49846: loss 1.4001, time 15.34ms, mfu 0.31%\n",
            "iter 49847: loss 1.5037, time 12.38ms, mfu 0.39%\n",
            "iter 49848: loss 1.5954, time 12.33ms, mfu 0.39%\n",
            "iter 49849: loss 1.4439, time 12.24ms, mfu 0.39%\n",
            "iter 49850: loss 1.5234, time 12.11ms, mfu 0.40%\n",
            "iter 49851: loss 1.4964, time 12.14ms, mfu 0.40%\n",
            "iter 49852: loss 1.5074, time 16.92ms, mfu 0.29%\n",
            "iter 49853: loss 1.4662, time 14.85ms, mfu 0.32%\n",
            "iter 49854: loss 1.4683, time 12.22ms, mfu 0.39%\n",
            "iter 49855: loss 1.4462, time 12.41ms, mfu 0.39%\n",
            "iter 49856: loss 1.3690, time 12.12ms, mfu 0.40%\n",
            "iter 49857: loss 1.4370, time 14.53ms, mfu 0.33%\n",
            "iter 49858: loss 1.5316, time 13.34ms, mfu 0.36%\n",
            "iter 49859: loss 1.5543, time 13.62ms, mfu 0.35%\n",
            "iter 49860: loss 1.4676, time 12.48ms, mfu 0.39%\n",
            "iter 49861: loss 1.4684, time 12.22ms, mfu 0.39%\n",
            "iter 49862: loss 1.4526, time 13.00ms, mfu 0.37%\n",
            "iter 49863: loss 1.4346, time 14.29ms, mfu 0.34%\n",
            "iter 49864: loss 1.3696, time 12.35ms, mfu 0.39%\n",
            "iter 49865: loss 1.4564, time 12.11ms, mfu 0.40%\n",
            "iter 49866: loss 1.5041, time 12.43ms, mfu 0.39%\n",
            "iter 49867: loss 1.5645, time 12.24ms, mfu 0.39%\n",
            "iter 49868: loss 1.5377, time 15.06ms, mfu 0.32%\n",
            "iter 49869: loss 1.5334, time 13.54ms, mfu 0.36%\n",
            "iter 49870: loss 1.4917, time 12.10ms, mfu 0.40%\n",
            "iter 49871: loss 1.4309, time 12.10ms, mfu 0.40%\n",
            "iter 49872: loss 1.3540, time 12.13ms, mfu 0.40%\n",
            "iter 49873: loss 1.4483, time 12.09ms, mfu 0.40%\n",
            "iter 49874: loss 1.5249, time 13.87ms, mfu 0.35%\n",
            "iter 49875: loss 1.6458, time 14.18ms, mfu 0.34%\n",
            "iter 49876: loss 1.4541, time 13.27ms, mfu 0.36%\n",
            "iter 49877: loss 1.3844, time 16.16ms, mfu 0.30%\n",
            "iter 49878: loss 1.4897, time 13.93ms, mfu 0.35%\n",
            "iter 49879: loss 1.3971, time 14.64ms, mfu 0.33%\n",
            "iter 49880: loss 1.6191, time 12.97ms, mfu 0.37%\n",
            "iter 49881: loss 1.5788, time 11.75ms, mfu 0.41%\n",
            "iter 49882: loss 1.4335, time 11.71ms, mfu 0.41%\n",
            "iter 49883: loss 1.5510, time 11.79ms, mfu 0.41%\n",
            "iter 49884: loss 1.5194, time 16.16ms, mfu 0.30%\n",
            "iter 49885: loss 1.4558, time 11.86ms, mfu 0.41%\n",
            "iter 49886: loss 1.4951, time 11.77ms, mfu 0.41%\n",
            "iter 49887: loss 1.5177, time 11.78ms, mfu 0.41%\n",
            "iter 49888: loss 1.4624, time 11.65ms, mfu 0.41%\n",
            "iter 49889: loss 1.5346, time 12.02ms, mfu 0.40%\n",
            "iter 49890: loss 1.5294, time 11.88ms, mfu 0.41%\n",
            "iter 49891: loss 1.3446, time 11.76ms, mfu 0.41%\n",
            "iter 49892: loss 1.6541, time 12.12ms, mfu 0.40%\n",
            "iter 49893: loss 1.4535, time 11.71ms, mfu 0.41%\n",
            "iter 49894: loss 1.4918, time 11.73ms, mfu 0.41%\n",
            "iter 49895: loss 1.5127, time 11.70ms, mfu 0.41%\n",
            "iter 49896: loss 1.4674, time 11.61ms, mfu 0.42%\n",
            "iter 49897: loss 1.3955, time 11.63ms, mfu 0.41%\n",
            "iter 49898: loss 1.5244, time 11.76ms, mfu 0.41%\n",
            "iter 49899: loss 1.4811, time 11.71ms, mfu 0.41%\n",
            "iter 49900: loss 1.4754, time 11.64ms, mfu 0.41%\n",
            "iter 49901: loss 1.4511, time 18.27ms, mfu 0.26%\n",
            "iter 49902: loss 1.3461, time 12.48ms, mfu 0.39%\n",
            "iter 49903: loss 1.2911, time 11.63ms, mfu 0.41%\n",
            "iter 49904: loss 1.5319, time 14.02ms, mfu 0.34%\n",
            "iter 49905: loss 1.4730, time 11.79ms, mfu 0.41%\n",
            "iter 49906: loss 1.3271, time 11.70ms, mfu 0.41%\n",
            "iter 49907: loss 1.5310, time 11.92ms, mfu 0.40%\n",
            "iter 49908: loss 1.3840, time 12.28ms, mfu 0.39%\n",
            "iter 49909: loss 1.3758, time 14.47ms, mfu 0.33%\n",
            "iter 49910: loss 1.5094, time 15.14ms, mfu 0.32%\n",
            "iter 49911: loss 1.4972, time 12.94ms, mfu 0.37%\n",
            "iter 49912: loss 1.4300, time 12.85ms, mfu 0.38%\n",
            "iter 49913: loss 1.3966, time 14.64ms, mfu 0.33%\n",
            "iter 49914: loss 1.4305, time 15.80ms, mfu 0.31%\n",
            "iter 49915: loss 1.4915, time 15.15ms, mfu 0.32%\n",
            "iter 49916: loss 1.3156, time 17.64ms, mfu 0.27%\n",
            "iter 49917: loss 1.4608, time 15.16ms, mfu 0.32%\n",
            "iter 49918: loss 1.4718, time 14.99ms, mfu 0.32%\n",
            "iter 49919: loss 1.4526, time 15.31ms, mfu 0.31%\n",
            "iter 49920: loss 1.4385, time 13.02ms, mfu 0.37%\n",
            "iter 49921: loss 1.4385, time 15.11ms, mfu 0.32%\n",
            "iter 49922: loss 1.5176, time 13.84ms, mfu 0.35%\n",
            "iter 49923: loss 1.4040, time 13.53ms, mfu 0.36%\n",
            "iter 49924: loss 1.3687, time 11.77ms, mfu 0.41%\n",
            "iter 49925: loss 1.5231, time 11.72ms, mfu 0.41%\n",
            "iter 49926: loss 1.3687, time 12.07ms, mfu 0.40%\n",
            "iter 49927: loss 1.3588, time 14.98ms, mfu 0.32%\n",
            "iter 49928: loss 1.4406, time 14.90ms, mfu 0.32%\n",
            "iter 49929: loss 1.3387, time 14.83ms, mfu 0.33%\n",
            "iter 49930: loss 1.5683, time 14.99ms, mfu 0.32%\n",
            "iter 49931: loss 1.4506, time 17.49ms, mfu 0.28%\n",
            "iter 49932: loss 1.4932, time 15.10ms, mfu 0.32%\n",
            "iter 49933: loss 1.5855, time 16.73ms, mfu 0.29%\n",
            "iter 49934: loss 1.4887, time 16.60ms, mfu 0.29%\n",
            "iter 49935: loss 1.4096, time 13.35ms, mfu 0.36%\n",
            "iter 49936: loss 1.4619, time 12.98ms, mfu 0.37%\n",
            "iter 49937: loss 1.4384, time 15.11ms, mfu 0.32%\n",
            "iter 49938: loss 1.4881, time 13.33ms, mfu 0.36%\n",
            "iter 49939: loss 1.4694, time 13.49ms, mfu 0.36%\n",
            "iter 49940: loss 1.4075, time 14.63ms, mfu 0.33%\n",
            "iter 49941: loss 1.4829, time 14.54ms, mfu 0.33%\n",
            "iter 49942: loss 1.5082, time 14.15ms, mfu 0.34%\n",
            "iter 49943: loss 1.3665, time 15.68ms, mfu 0.31%\n",
            "iter 49944: loss 1.4042, time 14.64ms, mfu 0.33%\n",
            "iter 49945: loss 1.4981, time 16.95ms, mfu 0.28%\n",
            "iter 49946: loss 1.4231, time 15.69ms, mfu 0.31%\n",
            "iter 49947: loss 1.3945, time 15.29ms, mfu 0.32%\n",
            "iter 49948: loss 1.4219, time 14.04ms, mfu 0.34%\n",
            "iter 49949: loss 1.4116, time 13.76ms, mfu 0.35%\n",
            "iter 49950: loss 1.4902, time 13.37ms, mfu 0.36%\n",
            "iter 49951: loss 1.4017, time 13.42ms, mfu 0.36%\n",
            "iter 49952: loss 1.4891, time 13.38ms, mfu 0.36%\n",
            "iter 49953: loss 1.4029, time 13.63ms, mfu 0.35%\n",
            "iter 49954: loss 1.3195, time 15.11ms, mfu 0.32%\n",
            "iter 49955: loss 1.4173, time 15.11ms, mfu 0.32%\n",
            "iter 49956: loss 1.3749, time 14.81ms, mfu 0.33%\n",
            "iter 49957: loss 1.4581, time 13.96ms, mfu 0.35%\n",
            "iter 49958: loss 1.3604, time 14.03ms, mfu 0.34%\n",
            "iter 49959: loss 1.4603, time 14.03ms, mfu 0.34%\n",
            "iter 49960: loss 1.4678, time 15.96ms, mfu 0.30%\n",
            "iter 49961: loss 1.4845, time 14.37ms, mfu 0.34%\n",
            "iter 49962: loss 1.5109, time 14.06ms, mfu 0.34%\n",
            "iter 49963: loss 1.4130, time 15.83ms, mfu 0.30%\n",
            "iter 49964: loss 1.4745, time 16.14ms, mfu 0.30%\n",
            "iter 49965: loss 1.5230, time 15.00ms, mfu 0.32%\n",
            "iter 49966: loss 1.4423, time 15.35ms, mfu 0.31%\n",
            "iter 49967: loss 1.4709, time 15.21ms, mfu 0.32%\n",
            "iter 49968: loss 1.4915, time 14.58ms, mfu 0.33%\n",
            "iter 49969: loss 1.4801, time 14.65ms, mfu 0.33%\n",
            "iter 49970: loss 1.5406, time 14.56ms, mfu 0.33%\n",
            "iter 49971: loss 1.4510, time 14.66ms, mfu 0.33%\n",
            "iter 49972: loss 1.5687, time 15.72ms, mfu 0.31%\n",
            "iter 49973: loss 1.5986, time 18.50ms, mfu 0.26%\n",
            "iter 49974: loss 1.6054, time 16.84ms, mfu 0.29%\n",
            "iter 49975: loss 1.5047, time 15.39ms, mfu 0.31%\n",
            "iter 49976: loss 1.3826, time 12.63ms, mfu 0.38%\n",
            "iter 49977: loss 1.5235, time 10.51ms, mfu 0.46%\n",
            "iter 49978: loss 1.5000, time 9.73ms, mfu 0.50%\n",
            "iter 49979: loss 1.3961, time 9.34ms, mfu 0.52%\n",
            "iter 49980: loss 1.4775, time 9.49ms, mfu 0.51%\n",
            "iter 49981: loss 1.4419, time 9.73ms, mfu 0.50%\n",
            "iter 49982: loss 1.4183, time 9.32ms, mfu 0.52%\n",
            "iter 49983: loss 1.4141, time 9.43ms, mfu 0.51%\n",
            "iter 49984: loss 1.4345, time 9.80ms, mfu 0.49%\n",
            "iter 49985: loss 1.5366, time 9.30ms, mfu 0.52%\n",
            "iter 49986: loss 1.5579, time 9.76ms, mfu 0.49%\n",
            "iter 49987: loss 1.3589, time 10.83ms, mfu 0.45%\n",
            "iter 49988: loss 1.4892, time 9.53ms, mfu 0.51%\n",
            "iter 49989: loss 1.4669, time 9.54ms, mfu 0.51%\n",
            "iter 49990: loss 1.4252, time 9.32ms, mfu 0.52%\n",
            "iter 49991: loss 1.5797, time 9.53ms, mfu 0.51%\n",
            "iter 49992: loss 1.4024, time 10.82ms, mfu 0.45%\n",
            "iter 49993: loss 1.4325, time 9.62ms, mfu 0.50%\n",
            "iter 49994: loss 1.6250, time 15.16ms, mfu 0.32%\n",
            "iter 49995: loss 1.4964, time 11.74ms, mfu 0.41%\n",
            "iter 49996: loss 1.4437, time 9.34ms, mfu 0.52%\n",
            "iter 49997: loss 1.5436, time 10.00ms, mfu 0.48%\n",
            "iter 49998: loss 1.4740, time 9.75ms, mfu 0.49%\n",
            "iter 49999: loss 1.5647, time 9.80ms, mfu 0.49%\n",
            "-----------------------------------------\n",
            "Sample from DIFF Transformer\n",
            "-----------------------------------------\n",
            "\n",
            "For bring a sweet's his vingrrelier and father\n",
            "By some daughtercy into fy than is gaint.\n",
            "\n",
            "JOHN OF CAPULET:\n",
            "So do but my father, and will agazen to\n",
            "that my duked foamon nothy. I cannothing of deperish\n",
            "In dear her souls made cusure to on, set in to goin;\n",
            "But the goes war how viptle from Sox--it it before;\n",
            "Or, but your entery mother, the fee, and fore\n",
            "I bhind, I am the soee in your heap, hark and\n",
            "hath hear qjow hoker'd in frialt to heart the tofd:\n",
            "Rives I will unvil awarew to umight begets,\n",
            "Wought \n",
            "-----------------------------------------\n",
            "Sample from Standard Transformer\n",
            "-----------------------------------------\n",
            "\n",
            "A friar I swall,\n",
            "Ir eyes from live fellowinged, at live were lengues.\n",
            "\n",
            "ANTHARI:\n",
            "I may too, if you well.\n",
            "\n",
            "GLORIZEDWARD:\n",
            "Soon to help'd means his vine! here at the was\n",
            "mote faraitied, igrant of my abune?\n",
            "\n",
            "TORGCALUS:\n",
            "What so, as he noble at servant they maken'd; she ghings from answer:\n",
            "My cheeks lord; not cannot a giveron'd, take.\n",
            "\n",
            "CORIOLANT:\n",
            "Why, we are for I will take my warrabon; for, and repain.\n",
            "\n",
            "LADY GREY:\n",
            "good with furnave more me fround away;\n",
            "What deared, he shall each to draw it that?\n",
            "At we\n"
          ]
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 1000x600 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 1000x600 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "import os\n",
        "import time\n",
        "import math\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.nn.functional as F\n",
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "\n",
        "# Hyperparameters\n",
        "batch_size = 12\n",
        "block_size = 64\n",
        "max_iters = 50000\n",
        "eval_interval = 500\n",
        "learning_rate = 3e-4\n",
        "device = 'cuda' if torch.cuda.is_available() else 'cpu'\n",
        "eval_iters = 1\n",
        "n_embd = 128\n",
        "n_head = 4\n",
        "n_layer = 4\n",
        "dropout = 0.2\n",
        "\n",
        "# Load and preprocess data\n",
        "with open('input.txt', 'r', encoding='utf-8') as f:\n",
        "    text = f.read()\n",
        "\n",
        "chars = sorted(list(set(text)))\n",
        "vocab_size = len(chars)\n",
        "stoi = {ch: i for i, ch in enumerate(chars)}\n",
        "itos = {i: ch for i, ch in enumerate(chars)}\n",
        "encode = lambda s: [stoi[c] for c in s]\n",
        "decode = lambda l: ''.join([itos[i] for i in l])\n",
        "\n",
        "data = torch.tensor(encode(text), dtype=torch.long)\n",
        "n = int(0.9 * len(data))\n",
        "train_data = data[:n]\n",
        "val_data = data[n:]\n",
        "\n",
        "\n",
        "def get_batch(split):\n",
        "    data_split = train_data if split == 'train' else val_data\n",
        "    ix = torch.randint(len(data_split) - block_size, (batch_size,))\n",
        "    x = torch.stack([data_split[i:i + block_size] for i in ix])\n",
        "    y = torch.stack([data_split[i + 1:i + block_size + 1] for i in ix])\n",
        "    return x.to(device), y.to(device)\n",
        "\n",
        "\n",
        "def lambda_init(depth):\n",
        "    return 0.8 - 0.6 * math.exp(-0.3 * (depth - 1))\n",
        "\n",
        "\n",
        "# Multi-head Differential Attention\n",
        "class MultiHeadDiffAttention(nn.Module):\n",
        "    def __init__(self, n_embd, n_head, layer_idx):\n",
        "        super().__init__()\n",
        "        assert n_embd % n_head == 0\n",
        "        self.n_head = n_head\n",
        "        self.head_size = n_embd // n_head\n",
        "        self.lambda_init = lambda_init(layer_idx) \n",
        "\n",
        "        # split qkv\n",
        "        self.q1_proj = nn.Linear(n_embd, n_embd, bias=False)\n",
        "        self.q2_proj = nn.Linear(n_embd, n_embd, bias=False)\n",
        "        self.k1_proj = nn.Linear(n_embd, n_embd, bias=False)\n",
        "        self.k2_proj = nn.Linear(n_embd, n_embd, bias=False)\n",
        "        self.v_proj = nn.Linear(n_embd, 2 * n_embd, bias=False)  # V projects to 2 * n_embd\n",
        "\n",
        "        self.c_proj = nn.Linear(2 * n_embd, n_embd, bias=False)\n",
        "        self.attn_dropout = nn.Dropout(dropout)\n",
        "        self.resid_dropout = nn.Dropout(dropout)\n",
        "        self.subln = nn.LayerNorm(2 * self.head_size, elementwise_affine=False)\n",
        "\n",
        "        # Init λ across heads\n",
        "        self.lambda_q1 = nn.Parameter(torch.randn(n_head, self.head_size) * 0.1)\n",
        "        self.lambda_k1 = nn.Parameter(torch.randn(n_head, self.head_size) * 0.1)\n",
        "        self.lambda_q2 = nn.Parameter(torch.randn(n_head, self.head_size) * 0.1)\n",
        "        self.lambda_k2 = nn.Parameter(torch.randn(n_head, self.head_size) * 0.1)\n",
        "\n",
        "    def forward(self, x):\n",
        "        B, T, C = x.shape\n",
        "\n",
        "        # Project x to get q1, q2, k1, k2, v\n",
        "        q1 = self.q1_proj(x).view(B, T, self.n_head, self.head_size).transpose(1, 2)\n",
        "        q2 = self.q2_proj(x).view(B, T, self.n_head, self.head_size).transpose(1, 2)\n",
        "        k1 = self.k1_proj(x).view(B, T, self.n_head, self.head_size).transpose(1, 2)\n",
        "        k2 = self.k2_proj(x).view(B, T, self.n_head, self.head_size).transpose(1, 2)\n",
        "        v = self.v_proj(x).view(B, T, self.n_head, 2 * self.head_size).transpose(1, 2)\n",
        "\n",
        "        scale = 1.0 / math.sqrt(self.head_size)\n",
        "        att1 = torch.matmul(q1, k1.transpose(-2, -1)) * scale\n",
        "        att2 = torch.matmul(q2, k2.transpose(-2, -1)) * scale\n",
        "\n",
        "        attn_mask = torch.tril(torch.ones(T, T, device=x.device)).unsqueeze(0).unsqueeze(0)\n",
        "        att1 = att1.masked_fill(attn_mask == 0, float('-inf'))\n",
        "        att2 = att2.masked_fill(attn_mask == 0, float('-inf'))\n",
        "\n",
        "        att1 = F.softmax(att1, dim=-1)\n",
        "        att2 = F.softmax(att2, dim=-1)\n",
        "\n",
        "        # Compute λ for each head separately\n",
        "        lambda_1 = torch.exp(torch.sum(self.lambda_q1 * self.lambda_k1, dim=-1)).unsqueeze(-1).unsqueeze(-1)\n",
        "        lambda_2 = torch.exp(torch.sum(self.lambda_q2 * self.lambda_k2, dim=-1)).unsqueeze(-1).unsqueeze(-1)\n",
        "        lambda_full = lambda_1 - lambda_2 + self.lambda_init\n",
        "\n",
        "        att = att1 - lambda_full * att2\n",
        "        att = self.attn_dropout(att)\n",
        "\n",
        "        y = torch.matmul(att, v)  # [B, n_head, T, 2 * head_size]\n",
        "        y = self.subln(y)\n",
        "        y = y * (1 - self.lambda_init)\n",
        "\n",
        "        y = y.transpose(1, 2).contiguous().view(B, T, 2 * C)\n",
        "        y = self.resid_dropout(self.c_proj(y))\n",
        "        return y\n",
        "\n",
        "\n",
        "# Traditional Multi-head Attention\n",
        "class MultiHeadAttention(nn.Module):\n",
        "    def __init__(self, n_embd, n_head, layer_idx=None):\n",
        "        super().__init__()\n",
        "        assert n_embd % n_head == 0\n",
        "        self.n_head = n_head\n",
        "        self.head_size = n_embd // n_head\n",
        "        self.c_attn = nn.Linear(n_embd, 3 * n_embd, bias=False)\n",
        "        self.c_proj = nn.Linear(n_embd, n_embd, bias=False)\n",
        "        self.attn_dropout = nn.Dropout(dropout)\n",
        "        self.resid_dropout = nn.Dropout(dropout)\n",
        "\n",
        "    def forward(self, x):\n",
        "        B, T, C = x.shape\n",
        "        qkv = self.c_attn(x)\n",
        "        q, k, v = qkv.split(C, dim=2)\n",
        "        q = q.view(B, T, self.n_head, self.head_size).transpose(1, 2)\n",
        "        k = k.view(B, T, self.n_head, self.head_size).transpose(1, 2)\n",
        "        v = v.view(B, T, self.n_head, self.head_size).transpose(1, 2)\n",
        "\n",
        "        att = (q @ k.transpose(-2, -1)) * (1.0 / math.sqrt(self.head_size))\n",
        "        attn_mask = torch.tril(torch.ones(T, T, device=x.device)).unsqueeze(0).unsqueeze(0)\n",
        "        att = att.masked_fill(attn_mask == 0, float('-inf'))\n",
        "        att = F.softmax(att, dim=-1)\n",
        "        att = self.attn_dropout(att)\n",
        "\n",
        "        y = att @ v\n",
        "        y = y.transpose(1, 2).contiguous().view(B, T, C)\n",
        "        y = self.resid_dropout(self.c_proj(y))\n",
        "        return y\n",
        "\n",
        "\n",
        "# MLP\n",
        "class MLP(nn.Module):\n",
        "    def __init__(self, n_embd):\n",
        "        super().__init__()\n",
        "        self.c_fc = nn.Linear(n_embd, 4 * n_embd, bias=False)\n",
        "        self.c_proj = nn.Linear(4 * n_embd, n_embd, bias=False)\n",
        "        self.dropout = nn.Dropout(dropout)\n",
        "\n",
        "    def forward(self, x):\n",
        "        x = self.c_fc(x)\n",
        "        x = F.gelu(x)\n",
        "        x = self.c_proj(x)\n",
        "        x = self.dropout(x)\n",
        "        return x\n",
        "\n",
        "\n",
        "# Transformer Block\n",
        "class Block(nn.Module):\n",
        "    def __init__(self, n_embd, n_head, attention_class, layer_idx):\n",
        "        super().__init__()\n",
        "        self.ln_1 = nn.LayerNorm(n_embd)\n",
        "        self.attn = attention_class(n_embd, n_head, layer_idx)\n",
        "        self.ln_2 = nn.LayerNorm(n_embd)\n",
        "        self.mlp = MLP(n_embd)\n",
        "\n",
        "    def forward(self, x):\n",
        "        x = x + self.attn(self.ln_1(x))\n",
        "        x = x + self.mlp(self.ln_2(x))\n",
        "        return x\n",
        "\n",
        "\n",
        "# Transformer Model\n",
        "class nanoGPT(nn.Module):\n",
        "    def __init__(self, vocab_size, n_embd, n_head, n_layer, block_size, attention_class):\n",
        "        super().__init__()\n",
        "        self.block_size = block_size\n",
        "        self.transformer = nn.ModuleDict(dict(\n",
        "            wte=nn.Embedding(vocab_size, n_embd),\n",
        "            wpe=nn.Embedding(block_size, n_embd),\n",
        "            drop=nn.Dropout(dropout),\n",
        "            h=nn.ModuleList([Block(n_embd, n_head, attention_class, layer_idx=i + 1) for i in range(n_layer)]),\n",
        "            ln_f=nn.LayerNorm(n_embd),\n",
        "        ))\n",
        "        self.lm_head = nn.Linear(n_embd, vocab_size, bias=False)\n",
        "        self.n_layer = n_layer\n",
        "        self.n_embd = n_embd\n",
        "\n",
        "    def forward(self, idx, targets=None):\n",
        "        device = idx.device\n",
        "        b, t = idx.size()\n",
        "        pos = torch.arange(0, t, dtype=torch.long, device=device).unsqueeze(0)\n",
        "\n",
        "        tok_emb = self.transformer.wte(idx)\n",
        "        pos_emb = self.transformer.wpe(pos)\n",
        "        x = self.transformer.drop(tok_emb + pos_emb)\n",
        "        for block in self.transformer.h:\n",
        "            x = block(x)\n",
        "        x = self.transformer.ln_f(x)\n",
        "        logits = self.lm_head(x)\n",
        "\n",
        "        loss = None\n",
        "        if targets is not None:\n",
        "            loss = F.cross_entropy(\n",
        "                logits.view(-1, logits.size(-1)), targets.view(-1), ignore_index=-1\n",
        "            )\n",
        "\n",
        "        return logits, loss\n",
        "\n",
        "    @torch.no_grad()\n",
        "    def generate(self, idx, max_new_tokens, temperature=1.0, top_k=None):\n",
        "        for _ in range(max_new_tokens):\n",
        "            idx_cond = idx[:, -self.block_size:]\n",
        "            logits, _ = self(idx_cond)\n",
        "            logits = logits[:, -1, :] / temperature\n",
        "            if top_k is not None:\n",
        "                v, _ = torch.topk(logits, min(top_k, logits.size(-1)))\n",
        "                logits[logits < v[:, [-1]]] = -float('Inf')\n",
        "            probs = F.softmax(logits, dim=-1)\n",
        "            idx_next = torch.multinomial(probs, num_samples=1)\n",
        "            idx = torch.cat((idx, idx_next), dim=1)\n",
        "        return idx\n",
        "\n",
        "\n",
        "# MFU\n",
        "def compute_mfu(model, fwdbwd_per_iter, dt):\n",
        "    N = sum(p.numel() for p in model.parameters())\n",
        "    L, H, Q, T = model.n_layer, model.n_embd, model.n_embd, model.block_size\n",
        "    flops_per_token = 6 * N + 12 * L * H * T\n",
        "    flops_per_fwdbwd = flops_per_token * T\n",
        "    flops_per_iter = flops_per_fwdbwd * fwdbwd_per_iter\n",
        "    flops_achieved = flops_per_iter * (1.0 / dt)  # per second\n",
        "    flops_promised = 7e12  # A100 GPU\n",
        "    mfu = flops_achieved / flops_promised\n",
        "    return mfu\n",
        "\n",
        "\n",
        "# Train and eval\n",
        "@torch.no_grad()\n",
        "def estimate_loss(model):\n",
        "    out = {}\n",
        "    model.eval()\n",
        "    for split in ['train', 'val']:\n",
        "        losses = torch.zeros(eval_iters)\n",
        "        for k in range(eval_iters):\n",
        "            X, Y = get_batch(split)\n",
        "            logits, loss = model(X, Y)\n",
        "            losses[k] = loss.item()\n",
        "        out[split] = losses.mean()\n",
        "    model.train()\n",
        "    return out\n",
        "\n",
        "\n",
        "def train_model(model, model_type):\n",
        "    optimizer = torch.optim.AdamW(model.parameters(), lr=learning_rate)\n",
        "\n",
        "    losses = []\n",
        "    val_losses = []\n",
        "\n",
        "    t0 = time.time()\n",
        "    for iter in range(max_iters):\n",
        "        if iter % eval_interval == 0:\n",
        "            loss_dict = estimate_loss(model)\n",
        "            print(f\"step {iter}: train loss {loss_dict['train']:.4f}, val loss {loss_dict['val']:.4f}\")\n",
        "            losses.append(loss_dict['train'])\n",
        "            val_losses.append(loss_dict['val'])\n",
        "\n",
        "        xb, yb = get_batch('train')\n",
        "\n",
        "        logits, loss = model(xb, yb)\n",
        "        optimizer.zero_grad(set_to_none=True)\n",
        "        loss.backward()\n",
        "        optimizer.step()\n",
        "\n",
        "        if iter % 1 == 0:\n",
        "            t1 = time.time()\n",
        "            dt = t1 - t0\n",
        "            t0 = t1\n",
        "            lossf = loss.item()\n",
        "            mfu = compute_mfu(model, 1, dt)\n",
        "            print(f\"iter {iter}: loss {lossf:.4f}, time {dt * 1000:.2f}ms, mfu {mfu * 100:.2f}%\")\n",
        "\n",
        "    # Save checkpoint\n",
        "    checkpoint_dir = 'checkpoints'\n",
        "    os.makedirs(checkpoint_dir, exist_ok=True)\n",
        "    torch.save(model.state_dict(), os.path.join(checkpoint_dir, f\"{model_type}_model.pt\"))\n",
        "\n",
        "    return losses, val_losses\n",
        "\n",
        "\n",
        "# Init models\n",
        "diff_model = nanoGPT(\n",
        "    vocab_size, n_embd, n_head, n_layer, block_size, MultiHeadDiffAttention\n",
        ").to(device)\n",
        "standard_model = nanoGPT(\n",
        "    vocab_size, n_embd, n_head, n_layer, block_size, MultiHeadAttention\n",
        ").to(device)\n",
        "\n",
        "# Train\n",
        "print(\"Training With DIFF Transformer\")\n",
        "diff_losses, diff_val_losses = train_model(diff_model, \"DIFF\")\n",
        "\n",
        "print(\"\\nTraining With Standard Transformer\")\n",
        "standard_losses, standard_val_losses = train_model(standard_model, \"Standard\")\n",
        "\n",
        "# Sample\n",
        "def sample(model, model_type):\n",
        "    context = torch.zeros((1, 1), dtype=torch.long, device=device)\n",
        "    print(f\"-----------------------------------------\")\n",
        "    print(f\"Sample from {model_type} Transformer\")\n",
        "    print(f\"-----------------------------------------\")\n",
        "    print(decode(model.generate(context, max_new_tokens=500)[0].tolist()))\n",
        "\n",
        "\n",
        "def plot_nll_vs_position():\n",
        "    plt.figure(figsize=(10, 6))\n",
        "    for model, model_type in [(diff_model, 'DIFF'), (standard_model, 'Standard Transformer')]:\n",
        "        nll = []\n",
        "        for pos in range(block_size):\n",
        "            x, y = get_batch('val')\n",
        "            with torch.no_grad():\n",
        "                logits, _ = model(x)\n",
        "                loss = F.cross_entropy(logits[:, pos, :], y[:, pos], reduction='mean')\n",
        "            nll.append(loss.item())\n",
        "        plt.plot(range(1, block_size + 1), nll, label=model_type)\n",
        "    plt.xlabel('Sequence Position')\n",
        "    plt.ylabel('Negative Log-Likelihood')\n",
        "    plt.title('NLL vs Sequence Position')\n",
        "    plt.legend()\n",
        "    plt.show()\n",
        "    plt.close()\n",
        "\n",
        "\n",
        "def plot_training_curves():\n",
        "    plt.figure(figsize=(10, 6))\n",
        "    plt.plot(range(0, max_iters, eval_interval), diff_val_losses, label='DIFF (val)')\n",
        "    plt.plot(range(0, max_iters, eval_interval), standard_val_losses, label='Standard (val)')\n",
        "    plt.xlabel('Steps')\n",
        "    plt.ylabel('Loss')\n",
        "    plt.title('Validation Loss vs Iterations')\n",
        "    plt.legend()\n",
        "    plt.show()\n",
        "    plt.close()\n",
        "\n",
        "\n",
        "sample(diff_model, \"DIFF\")\n",
        "sample(standard_model, \"Standard\")\n",
        "plot_training_curves()\n",
        "plot_nll_vs_position()\n"
      ]
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "gpuType": "T4",
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}