testbed.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## save my precious data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/alekseev/dialogue-augmentation/.venv/lib/python3.8/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
      "  from .autonotebook import tqdm as notebook_tqdm\n"
     ]
    }
   ],
   "source": [
    "from datasets import DatasetDict, load_from_disk\n",
    "\n",
    "dataset = DatasetDict({\n",
    "    'trivial': load_from_disk('data/train/trivial'),\n",
    "    'advanced': load_from_disk('data/train/advanced'),\n",
    "    'crazy': load_from_disk('data/train/crazy')\n",
    "})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Creating parquet from Arrow format: 100%|██████████| 145/145 [00:02<00:00, 50.14ba/s]\n",
      "Creating parquet from Arrow format: 100%|██████████| 145/145 [00:02<00:00, 52.03ba/s]\n",
      "Creating parquet from Arrow format: 100%|██████████| 145/145 [00:02<00:00, 50.83ba/s]\n",
      "Uploading the dataset shards: 100%|██████████| 3/3 [01:16<00:00, 25.58s/it]\n",
      "Creating parquet from Arrow format: 100%|██████████| 109/109 [00:02<00:00, 42.92ba/s]\n",
      "Creating parquet from Arrow format: 100%|██████████| 109/109 [00:02<00:00, 43.12ba/s]\n",
      "Creating parquet from Arrow format: 100%|██████████| 109/109 [00:02<00:00, 43.06ba/s]\n",
      "Creating parquet from Arrow format: 100%|██████████| 109/109 [00:02<00:00, 43.43ba/s]\n",
      "Uploading the dataset shards: 100%|██████████| 4/4 [01:26<00:00, 21.55s/it]\n",
      "Creating parquet from Arrow format: 100%|██████████| 73/73 [00:02<00:00, 30.07ba/s]\n",
      "Creating parquet from Arrow format: 100%|██████████| 73/73 [00:02<00:00, 30.21ba/s]\n",
      "Creating parquet from Arrow format: 100%|██████████| 73/73 [00:02<00:00, 30.35ba/s]\n",
      "Creating parquet from Arrow format: 100%|██████████| 73/73 [00:02<00:00, 31.02ba/s]\n",
      "Creating parquet from Arrow format: 100%|██████████| 73/73 [00:02<00:00, 30.90ba/s]\n",
      "Creating parquet from Arrow format: 100%|██████████| 73/73 [00:02<00:00, 30.83ba/s]\n",
      "Uploading the dataset shards: 100%|██████████| 6/6 [01:58<00:00, 19.68s/it]\n",
      "README.md: 100%|██████████| 657/657 [00:00<00:00, 90.6kB/s]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "CommitInfo(commit_url='https://huggingface.co/datasets/voorhs/augmented/commit/8ca9f8af7b776888a7b762ee58266ad92e58dedb', commit_message='Upload dataset', commit_description='', oid='8ca9f8af7b776888a7b762ee58266ad92e58dedb', pr_url=None, pr_revision=None, pr_num=None)"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset.push_to_hub('voorhs/augmented')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## testing multiwoz-1-domain"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "8437"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from mylib.datasets import MultiWOZServiceClfDataset\n",
    "\n",
    "dataset = MultiWOZServiceClfDataset(path='data/multiwoz', split='train')\n",
    "len(dataset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1000"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset = MultiWOZServiceClfDataset(path='data/multiwoz', split='validation')\n",
    "len(dataset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "3250"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset = MultiWOZServiceClfDataset(path='data/multiwoz-1-domain', split='train')\n",
    "len(dataset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "209"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset = MultiWOZServiceClfDataset(path='data/multiwoz-1-domain', split='validation')\n",
    "len(dataset)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## backbone list"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "hf_models = [\n",
    "    'google-bert/bert-base-uncased',\n",
    "    'Shitao/RetroMAE',\n",
    "    'FacebookAI/roberta-base',\n",
    "    'WhereIsAI/UAE-Large-V1'\n",
    "]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "google-bert/bert-base-uncased ['[UNK]', '[SEP]', '[PAD]', '[CLS]', '[MASK]']\n",
      "Shitao/RetroMAE ['[UNK]', '[SEP]', '[PAD]', '[CLS]', '[MASK]']\n",
      "FacebookAI/roberta-base ['<s>', '</s>', '<unk>', '<pad>', '<mask>']\n",
      "WhereIsAI/UAE-Large-V1 ['[UNK]', '[SEP]', '[PAD]', '[CLS]', '[MASK]']\n"
     ]
    }
   ],
   "source": [
    "from transformers import AutoTokenizer\n",
    "\n",
    "for m in hf_models:\n",
    "    tokenizer = AutoTokenizer.from_pretrained(m)\n",
    "    print(m, tokenizer.all_special_tokens)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'<s>'"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "t = AutoTokenizer.from_pretrained('FacebookAI/roberta-base')\n",
    "t.cls_token"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## prune demo"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "\n",
    "dias = json.load(open('data/augmented/prune-test/00000.json', 'r'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'content': [{'speaker': 0, 'utterance': 'Hello.'},\n",
       "  {'speaker': 0,\n",
       "   'utterance': 'I am looking for a flight from JFK to MSP, can you help me in booking?'},\n",
       "  {'speaker': 0, 'utterance': 'Myself Katherine Jones.'},\n",
       "  {'speaker': 0,\n",
       "   'utterance': 'I would like to return my journey for the morning flight.'},\n",
       "  {'speaker': 0,\n",
       "   'utterance': \"That's ok, no problem. Thank you for the information.\"}],\n",
       " 'source_dataset_name': 'AirDialogue',\n",
       " 'idx_within_source': 268933,\n",
       " 'id': 443619}"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dias[2]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## domain datasets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "DatasetDict({\n",
       "    train: Dataset({\n",
       "        features: ['content', 'source_dataset_name', 'idx_within_source', 'services'],\n",
       "        num_rows: 5403\n",
       "    })\n",
       "    validation: Dataset({\n",
       "        features: ['content', 'source_dataset_name', 'idx_within_source', 'services'],\n",
       "        num_rows: 836\n",
       "    })\n",
       "    test: Dataset({\n",
       "        features: ['content', 'source_dataset_name', 'idx_within_source', 'services'],\n",
       "        num_rows: 1806\n",
       "    })\n",
       "})"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from datasets import load_from_disk\n",
    "\n",
    "dataset = load_from_disk('data/benchmarks/SGD')\n",
    "dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# from mylib.datasets import DomainDataset\n",
    "\n",
    "# dataset = DomainDataset('data/benchmarks-one-domain-filtered/sgd', 'train')\n",
    "# len(dataset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'content': [{'speaker': 0,\n",
       "   'utterance': \"I'm starving! Can you help me find a restaurant that serves Breakfast in Fairfield?\"},\n",
       "  {'speaker': 1,\n",
       "   'utterance': \"There is 1 restaurant that meets your needs. Mimi's Cafe in Fairfield serves breakfast.\"},\n",
       "  {'speaker': 0, 'utterance': 'Is there a band playing there?'},\n",
       "  {'speaker': 1,\n",
       "   'utterance': \"No, unfortunately, they don't offer live music.\"},\n",
       "  {'speaker': 0, 'utterance': \"That's okay. It will work anyway.\"},\n",
       "  {'speaker': 1, 'utterance': 'Do you wish to make a reservation?'},\n",
       "  {'speaker': 0,\n",
       "   'utterance': \"Yes, I'll need to make a reservation for a table for 2.\"},\n",
       "  {'speaker': 1, 'utterance': 'What time will you be visiting?'},\n",
       "  {'speaker': 0,\n",
       "   'utterance': 'Please make it for half past 5 in the evening.'},\n",
       "  {'speaker': 1,\n",
       "   'utterance': \"You'd like a table for 2 today at Mimi's Cafe in Fairfield at 5:30 pm. Is this correct?\"},\n",
       "  {'speaker': 0,\n",
       "   'utterance': \"I'm sorry, I need to change that. I'd like it on the 9th and it will be for 3 people.\"},\n",
       "  {'speaker': 1,\n",
       "   'utterance': \"Let's review those changes. You'd like the reservation on March 9th for 3. Have I got it right?\"},\n",
       "  {'speaker': 0,\n",
       "   'utterance': \"I'm sorry. I need to make it on Monday next week at half past 12 in the afternoon.\"},\n",
       "  {'speaker': 1,\n",
       "   'utterance': \"You'd like your reservation for 3 next Monday at 12:30 pm. Is that correct?\"},\n",
       "  {'speaker': 0,\n",
       "   'utterance': \"Yes, I'm sorry for the confusion. Can you tell me their phone number please?\"},\n",
       "  {'speaker': 1,\n",
       "   'utterance': 'The phone number is 707-421-0835. Your reservation is confirmed.'},\n",
       "  {'speaker': 0, 'utterance': 'Great! Thank you so much for your help.'},\n",
       "  {'speaker': 1, 'utterance': \"You're welcome. Have a nice day!\"}],\n",
       " 'source_dataset_name': 'SGD/train',\n",
       " 'idx_within_source': 10,\n",
       " 'services': [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0]}"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset['train'][10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2978"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from mylib.datasets import DomainDataset\n",
    "\n",
    "dataset = DomainDataset('data/benchmarks-bert/multiwoz', split='train')\n",
    "len(dataset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(dataset[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## fixing problems :("
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Dataset({\n",
       "    features: ['content', 'source_dataset_name', 'idx_within_source', 'id'],\n",
       "    num_rows: 450750\n",
       "})"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from datasets import load_from_disk\n",
    "\n",
    "dataset = load_from_disk('data/augmented/prune-collected')\n",
    "dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[None, -inf]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import json\n",
    "\n",
    "json.loads('[null, -Infinity]')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## filtering by source"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/alekseev/dialogue-augmentation/.venv/lib/python3.8/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
      "  from .autonotebook import tqdm as notebook_tqdm\n"
     ]
    }
   ],
   "source": [
    "from datasets import load_from_disk\n",
    "\n",
    "path_in = 'data/train-unfair/trivial'\n",
    "in_dataset = load_from_disk(path_in)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'source_dataset_name': 'AirDialogue',\n",
       " 'idx_within_source': 185688,\n",
       " 'id': 360925,\n",
       " 'pos': [{'augmentation': 'back-translate',\n",
       "   'content': [{'speaker': 0, 'utterance': 'Hey.'},\n",
       "    {'speaker': 1, 'utterance': 'Hey, what can I do for you?'},\n",
       "    {'speaker': 0,\n",
       "     'utterance': \"I'm very happy to visit Ellenville Faw Ice Pavees in New York.\"},\n",
       "    {'speaker': 1,\n",
       "     'utterance': 'Of course, can I know the dates of the trip?'},\n",
       "    {'speaker': 0, 'utterance': 'My trip to the 28th and 30th of April.'},\n",
       "    {'speaker': 1, 'utterance': 'Do you have any contact restrictions?'},\n",
       "    {'speaker': 0, 'utterance': \"I'm fine with maximum communications.\"},\n",
       "    {'speaker': 1, 'utterance': 'May I ask your name?'},\n",
       "    {'speaker': 0, 'utterance': 'My name is Gary Evans.'},\n",
       "    {'speaker': 1, 'utterance': 'Wait a minute, please.'},\n",
       "    {'speaker': 0, 'utterance': 'Please make sure I want to leave tonight.'},\n",
       "    {'speaker': 1,\n",
       "     'utterance': 'Sorry, no flights with the dates you choose.'},\n",
       "    {'speaker': 0, 'utterance': \"It's okay, thanks for the information.\"},\n",
       "    {'speaker': 1, 'utterance': 'Thank you for your help.'}]},\n",
       "  {'augmentation': 'insert',\n",
       "   'content': [{'speaker': 0, 'utterance': 'Hello Hello.'},\n",
       "    {'speaker': 1, 'utterance': 'Hello, How can I assist you?'},\n",
       "    {'speaker': 0,\n",
       "     'utterance': 'I am very much excited to visit Ellenville Fault ed Ice Caves in New York - EWR, can you please reserve a flight from Houston - IAH?'},\n",
       "    {'speaker': 1, 'utterance': 'Sure, May I know the travelling dates?'},\n",
       "    {'speaker': 0, 'utterance': 'PS My travel dates are Apr 28th and 30th.'},\n",
       "    {'speaker': 1, 'utterance': 'Do ing you have any connection limit?'},\n",
       "    {'speaker': 0,\n",
       "     'utterance': 'Yes I am fine with maximum one GB connection.'},\n",
       "    {'speaker': 1, 'utterance': 'May I know your name?'},\n",
       "    {'speaker': 0, 'utterance': 'My name is Gary Evans.'},\n",
       "    {'speaker': 1, 'utterance': 'Â Please wait a minute.'},\n",
       "    {'speaker': 0,\n",
       "     'utterance': 'Q Please make sure, I want to depart in the evening.'},\n",
       "    {'speaker': 1,\n",
       "     'utterance': 'Sorry, No flights found with your chosen dates.'},\n",
       "    {'speaker': 0, 'utterance': 'No problem, thank you for the information.'},\n",
       "    {'speaker': 1, 'utterance': 'Thank ing you for availing us.'}]},\n",
       "  {'augmentation': 'replace',\n",
       "   'content': [{'speaker': 0, 'utterance': 'Hello.'},\n",
       "    {'speaker': 1, 'utterance': 'Hello, How can I help you?'},\n",
       "    {'speaker': 0,\n",
       "     'utterance': 'I am very oooo ative to visit Ellenville Fault Ice Caves in New AR - EWR, can you please ate a flight from Houston - IAH?'},\n",
       "    {'speaker': 1, 'utterance': 'Sure, May I know the travelling dates?'},\n",
       "    {'speaker': 0, 'utterance': 'My travel days are Apr 28th and 30th.'},\n",
       "    {'speaker': 1, 'utterance': 'Do you have any connection limit?'},\n",
       "    {'speaker': 0, 'utterance': 'I am fine with maximum one connection.'},\n",
       "    {'speaker': 1, 'utterance': 'May I know your name?'},\n",
       "    {'speaker': 0, 'utterance': 'My Name is Gary Evans.'},\n",
       "    {'speaker': 1, 'utterance': 'Please wait a minute.'},\n",
       "    {'speaker': 0,\n",
       "     'utterance': 'Please make sure, I want to depart in the evening.'},\n",
       "    {'speaker': 1, 'utterance': 'Sorry, No ont ups with your chosen dates.'},\n",
       "    {'speaker': 0, 'utterance': 'No problem, thank you for the information.'},\n",
       "    {'speaker': 1, 'utterance': 'Thank you for availing us.'}]}],\n",
       " 'orig': {'augmentation': None,\n",
       "  'content': [{'speaker': 0, 'utterance': 'Hello.'},\n",
       "   {'speaker': 1, 'utterance': 'Hello, How can I assist you?'},\n",
       "   {'speaker': 0,\n",
       "    'utterance': 'I am very much excited to visit Ellenville Fault Ice Caves in New York - EWR, can you please reserve a flight from Houston - IAH?'},\n",
       "   {'speaker': 1, 'utterance': 'Sure, May I know the travelling dates?'},\n",
       "   {'speaker': 0, 'utterance': 'My travel dates are Apr 28th and 30th.'},\n",
       "   {'speaker': 1, 'utterance': 'Do you have any connection limit?'},\n",
       "   {'speaker': 0, 'utterance': 'I am fine with maximum one connection.'},\n",
       "   {'speaker': 1, 'utterance': 'May I know your name?'},\n",
       "   {'speaker': 0, 'utterance': 'My name is Gary Evans.'},\n",
       "   {'speaker': 1, 'utterance': 'Please wait a minute.'},\n",
       "   {'speaker': 0,\n",
       "    'utterance': 'Please make sure, I want to depart in the evening.'},\n",
       "   {'speaker': 1,\n",
       "    'utterance': 'Sorry, No flights found with your chosen dates.'},\n",
       "   {'speaker': 0, 'utterance': 'No problem, thank you for the information.'},\n",
       "   {'speaker': 1, 'utterance': 'Thank you for availing us.'}]}}"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "in_dataset[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## torch max"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([3, 4])"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import torch\n",
    "\n",
    "a = torch.tensor([[1,2],[3,4]])\n",
    "\n",
    "torch.max(a, dim=0)[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## testing datasets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "433340"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from datasets import load_from_disk\n",
    "\n",
    "dataset = load_from_disk('data/train-retromae/trivial')\n",
    "len(dataset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['AirDialogue',\n",
       " 'SimJointGEN',\n",
       " 'Disambiguation',\n",
       " 'ABCD',\n",
       " 'MetaLWOZ',\n",
       " 'Taskmaster1',\n",
       " 'MS-DC',\n",
       " 'KETOD',\n",
       " 'FRAMES']"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset.unique('source_dataset_name')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "([{'speaker': 0, 'utterance': 'Hello.'},\n",
       "  {'speaker': 1, 'utterance': 'Hello, how may I help you?'},\n",
       "  {'speaker': 0,\n",
       "   'utterance': 'Can you please help me in reserving a flight ticket to ATL from JFK?'},\n",
       "  {'speaker': 1, 'utterance': 'Sure, help me with the trip dates?'},\n",
       "  {'speaker': 0, 'utterance': 'My trip dates are from Aug 25 to Aug 27.'},\n",
       "  {'speaker': 1, 'utterance': 'Do you have any preferences?'},\n",
       "  {'speaker': 0,\n",
       "   'utterance': 'I need a business class ticket. I would like to travel in connecting flight.'},\n",
       "  {'speaker': 1, 'utterance': 'May I know your name please?'},\n",
       "  {'speaker': 0, 'utterance': 'I am George Turner.'},\n",
       "  {'speaker': 1,\n",
       "   'utterance': 'Sorry, we did not find any flights running in your desired routes.'},\n",
       "  {'speaker': 0, 'utterance': 'Ok, no problem. Thank you.'},\n",
       "  {'speaker': 1, 'utterance': 'Thank you for reaching us.'}],\n",
       " [{'speaker': 0, 'utterance': 'Hello.'},\n",
       "  {'speaker': 1, 'utterance': 'Hello, how may I help you?'},\n",
       "  {'speaker': 0,\n",
       "   'utterance': 'Can you please help me in reserving a flight ticket to ATL from JFK?'},\n",
       "  {'speaker': 1, 'utterance': 'Sure, help me with the trip dates?'},\n",
       "  {'speaker': 0, 'utterance': 'My trip dates are from Aug 25 to Aug 27.'},\n",
       "  {'speaker': 0,\n",
       "   'utterance': 'I need a business class ticket. I would like to travel in connecting flight.'},\n",
       "  {'speaker': 1, 'utterance': 'May I know your name please?'},\n",
       "  {'speaker': 0, 'utterance': 'I am George Turner.'},\n",
       "  {'speaker': 1,\n",
       "   'utterance': 'Sorry, we did not find any flights running in your desired routes.'},\n",
       "  {'speaker': 0, 'utterance': 'Ok, no problem. Thank you.'},\n",
       "  {'speaker': 1, 'utterance': 'Thank you for reaching us.'}])"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "from torch.utils.data import DataLoader\n",
    "\n",
    "def collate_fn(batch):\n",
    "    origs =  [o for o, p in batch]\n",
    "    pos =  [p for o, p in batch]\n",
    "    return origs, pos\n",
    "    \n",
    "loader = DataLoader(\n",
    "    dataset=dataset,\n",
    "    batch_size=3,\n",
    "    shuffle=False,\n",
    "    num_workers=0,\n",
    "    collate_fn=collate_fn,\n",
    "    drop_last=False\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(3, 3)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "origs, pos = next(iter(loader))\n",
    "len(origs), len(pos)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[{'speaker': 0, 'utterance': 'Hello.'},\n",
       " {'speaker': 1, 'utterance': 'Hello, how may I help you?'},\n",
       " {'speaker': 0,\n",
       "  'utterance': 'Can you please help me in reserving a flight ticket to ATL from JFK?'},\n",
       " {'speaker': 1, 'utterance': 'Sure, help me with the trip dates?'},\n",
       " {'speaker': 0, 'utterance': 'My trip dates are from Aug 25 to Aug 27.'},\n",
       " {'speaker': 1, 'utterance': 'Do you have any preferences?'},\n",
       " {'speaker': 0,\n",
       "  'utterance': 'I need a business class ticket. I would like to travel in connecting flight.'},\n",
       " {'speaker': 1, 'utterance': 'May I know your name please?'},\n",
       " {'speaker': 0, 'utterance': 'I am George Turner.'},\n",
       " {'speaker': 1,\n",
       "  'utterance': 'Sorry, we did not find any flights running in your desired routes.'},\n",
       " {'speaker': 0, 'utterance': 'Ok, no problem. Thank you.'},\n",
       " {'speaker': 1, 'utterance': 'Thank you for reaching us.'}]"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "origs[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[{'speaker': 0, 'utterance': 'Q Hello.'},\n",
       " {'speaker': 1, 'utterance': 'Hello, how may haps I help you?'},\n",
       " {'speaker': 0,\n",
       "  'utterance': 'Can you please help me in reserving a flight ticket to ATL X from JFK?'},\n",
       " {'speaker': 1, 'utterance': 'Sure, help me with the trip dates?'},\n",
       " {'speaker': 0, 'utterance': 'My trip dates are from Aug 25 to Aug 27.'},\n",
       " {'speaker': 1, 'utterance': 'Do you have any preferences?'},\n",
       " {'speaker': 0,\n",
       "  'utterance': 'I need a business class ticket. I would like to travel in connecting flight.'},\n",
       " {'speaker': 1, 'utterance': 'May may I know your name please?'},\n",
       " {'speaker': 0, 'utterance': 'Hello I am George Turner.'},\n",
       " {'speaker': 1,\n",
       "  'utterance': 'Sorry, we did not find any flights running in your desired routes.'},\n",
       " {'speaker': 0, 'utterance': 'Q Ok, no problem. Â Thank you.'},\n",
       " {'speaker': 1, 'utterance': 'Â Thank you for reaching us.'}]"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pos[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## dataset statistics"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Описание одного датасета:\n",
    "- число семплов\n",
    "- сколько каких аугментаций\n",
    "- распределение числа токенов в диалоге\n",
    "- распределение числа утерансов в диалоге\n",
    "- сколько отфильтровано для bert, roberta, retromae"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/alekseev/dialogue-augmentation/.venv/lib/python3.8/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
      "  from .autonotebook import tqdm as notebook_tqdm\n"
     ]
    }
   ],
   "source": [
    "from datasets import load_from_disk\n",
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "sns.set_style('whitegrid')\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "\n",
    "def print_report(path, aug_names):\n",
    "    stats = load_from_disk(path)\n",
    "    \n",
    "    print('TOTAL SAMPLES:', len(stats))\n",
    "\n",
    "    df = stats.to_pandas()\n",
    "    \n",
    "    print('\\nSAMPLES OF EACH AUGMENTATION:')\n",
    "    \n",
    "    print(df[aug_names].sum())\n",
    "\n",
    "    n_toks_bert = [f'n-toks-bert-{name}' for name in aug_names]\n",
    "    n_toks_roberta = [f'n-toks-roberta-{name}' for name in aug_names]\n",
    "    n_toks_retromae = [f'n-toks-retromae-{name}' for name in aug_names]\n",
    "\n",
    "    print('\\nSAMPLES WITH EXCEEDING TOKENS:')\n",
    "    \n",
    "    print('\\nBERT:')\n",
    "    print((df[n_toks_bert] > 512).mean())\n",
    "    print('\\nRoBERTa:')\n",
    "    print((df[n_toks_roberta] > 512).mean())\n",
    "    print('\\nRetroMAE:')\n",
    "    print((df[n_toks_retromae] > 512).mean())\n",
    "\n",
    "    fig, ax = plt.subplots(1, 2, figsize=(8, 4))\n",
    "    df[n_toks_bert].plot(kind='hist', logy=True, alpha=0.5, ax=ax[0])\n",
    "    \n",
    "    n_uts_names = [f'n-uts-{name}' for name in aug_names]\n",
    "    sns.histplot(df[n_uts_names], bins=20, ax=ax[1])\n",
    "\n",
    "    plt.tight_layout()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TOTAL SAMPLES: 434094\n",
      "\n",
      "SAMPLES OF EACH AUGMENTATION:\n",
      "back-translate    434094\n",
      "insert            434094\n",
      "replace           434094\n",
      "dtype: int64\n",
      "\n",
      "SAMPLES WITH EXCEEDING TOKENS:\n",
      "\n",
      "BERT:\n",
      "n-toks-bert-back-translate    0.000521\n",
      "n-toks-bert-insert            0.001226\n",
      "n-toks-bert-replace           0.000069\n",
      "dtype: float64\n",
      "\n",
      "RoBERTa:\n",
      "n-toks-roberta-back-translate    0.000498\n",
      "n-toks-roberta-insert            0.000707\n",
      "n-toks-roberta-replace           0.000055\n",
      "dtype: float64\n",
      "\n",
      "RetroMAE:\n",
      "n-toks-retromae-back-translate    0.000521\n",
      "n-toks-retromae-insert            0.001226\n",
      "n-toks-retromae-replace           0.000069\n",
      "dtype: float64\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 800x400 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print_report('data/reports/trivial', aug_names=['back-translate', 'insert', 'replace'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TOTAL SAMPLES: 434094\n",
      "\n",
      "SAMPLES OF EACH AUGMENTATION:\n",
      "replace    434094\n",
      "prune      430995\n",
      "shuffle    275715\n",
      "dtype: int64\n",
      "\n",
      "SAMPLES WITH EXCEEDING TOKENS:\n",
      "\n",
      "BERT:\n",
      "n-toks-bert-replace    0.000069\n",
      "n-toks-bert-prune      0.000002\n",
      "n-toks-bert-shuffle    0.000051\n",
      "dtype: float64\n",
      "\n",
      "RoBERTa:\n",
      "n-toks-roberta-replace    0.000055\n",
      "n-toks-roberta-prune      0.000005\n",
      "n-toks-roberta-shuffle    0.000046\n",
      "dtype: float64\n",
      "\n",
      "RetroMAE:\n",
      "n-toks-retromae-replace    0.000069\n",
      "n-toks-retromae-prune      0.000002\n",
      "n-toks-retromae-shuffle    0.000051\n",
      "dtype: float64\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 800x400 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print_report('data/reports/advanced', aug_names=['replace', 'prune', 'shuffle'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TOTAL SAMPLES: 434094\n",
      "\n",
      "SAMPLES OF EACH AUGMENTATION:\n",
      "replace                 434094\n",
      "back-translate-prune    430771\n",
      "prune-insert            430995\n",
      "prune-replace           430995\n",
      "dtype: int64\n",
      "\n",
      "SAMPLES WITH EXCEEDING TOKENS:\n",
      "\n",
      "BERT:\n",
      "n-toks-bert-replace                 0.000069\n",
      "n-toks-bert-back-translate-prune    0.000009\n",
      "n-toks-bert-prune-insert            0.000435\n",
      "n-toks-bert-prune-replace           0.000002\n",
      "dtype: float64\n",
      "\n",
      "RoBERTa:\n",
      "n-toks-roberta-replace                 0.000055\n",
      "n-toks-roberta-back-translate-prune    0.000009\n",
      "n-toks-roberta-prune-insert            0.000071\n",
      "n-toks-roberta-prune-replace           0.000005\n",
      "dtype: float64\n",
      "\n",
      "RetroMAE:\n",
      "n-toks-retromae-replace                 0.000069\n",
      "n-toks-retromae-back-translate-prune    0.000009\n",
      "n-toks-retromae-prune-insert            0.000435\n",
      "n-toks-retromae-prune-replace           0.000002\n",
      "dtype: float64\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 800x400 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print_report('data/reports/crazy', aug_names=['replace', 'back-translate-prune', 'prune-insert', 'prune-replace'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TOTAL SAMPLES: 434094\n",
      "\n",
      "SAMPLES OF EACH AUGMENTATION:\n",
      "replace            434094\n",
      "shuffle-insert     275715\n",
      "shuffle-replace    275715\n",
      "dtype: int64\n",
      "\n",
      "SAMPLES WITH EXCEEDING TOKENS:\n",
      "\n",
      "BERT:\n",
      "n-toks-bert-replace            0.000069\n",
      "n-toks-bert-shuffle-insert     0.004326\n",
      "n-toks-bert-shuffle-replace    0.000065\n",
      "dtype: float64\n",
      "\n",
      "RoBERTa:\n",
      "n-toks-roberta-replace            0.000055\n",
      "n-toks-roberta-shuffle-insert     0.004144\n",
      "n-toks-roberta-shuffle-replace    0.000051\n",
      "dtype: float64\n",
      "\n",
      "RetroMAE:\n",
      "n-toks-retromae-replace            0.000069\n",
      "n-toks-retromae-shuffle-insert     0.004326\n",
      "n-toks-retromae-shuffle-replace    0.000065\n",
      "dtype: float64\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 800x400 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print_report('data/reports/crazy', aug_names=['replace', 'shuffle-insert', 'shuffle-replace'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## bench"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/alekseev/dialogue-augmentation/.venv/lib/python3.8/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
      "  from .autonotebook import tqdm as notebook_tqdm\n"
     ]
    },
    {
     "ename": "ModuleNotFoundError",
     "evalue": "No module named 'src'",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mModuleNotFoundError\u001b[0m                       Traceback (most recent call last)",
      "Cell \u001b[0;32mIn[1], line 1\u001b[0m\n\u001b[0;32m----> 1\u001b[0m \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;21;01mruconveval\u001b[39;00m\u001b[38;5;21;01m.\u001b[39;00m\u001b[38;5;21;01msrc\u001b[39;00m\u001b[38;5;21;01m.\u001b[39;00m\u001b[38;5;21;01mtask\u001b[39;00m \u001b[38;5;28;01mimport\u001b[39;00m Task, TaskConfig, TaskType, DatasetConfig\n\u001b[1;32m      3\u001b[0m dataset_config \u001b[38;5;241m=\u001b[39m DatasetConfig(\n\u001b[1;32m      4\u001b[0m     dataset\u001b[38;5;241m=\u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mmassive\u001b[39m\u001b[38;5;124m'\u001b[39m,\n\u001b[1;32m      5\u001b[0m     language\u001b[38;5;241m=\u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;124men\u001b[39m\u001b[38;5;124m'\u001b[39m\n\u001b[1;32m      6\u001b[0m )\n\u001b[1;32m      8\u001b[0m task_config \u001b[38;5;241m=\u001b[39m TaskConfig(\n\u001b[1;32m      9\u001b[0m     task\u001b[38;5;241m=\u001b[39mTaskType\u001b[38;5;241m.\u001b[39mclassification,\n\u001b[1;32m     10\u001b[0m     dataset_config\u001b[38;5;241m=\u001b[39mdataset_config\n\u001b[1;32m     11\u001b[0m )\n",
      "File \u001b[0;32m/cephfs/home/alekseev/dialogue-augmentation/ruconveval/src/task.py:10\u001b[0m\n\u001b[1;32m      8\u001b[0m \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;21;01m.\u001b[39;00m\u001b[38;5;21;01mdata\u001b[39;00m\u001b[38;5;21;01m.\u001b[39;00m\u001b[38;5;21;01mbase\u001b[39;00m \u001b[38;5;28;01mimport\u001b[39;00m BaseDataset\n\u001b[1;32m      9\u001b[0m \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;21;01m.\u001b[39;00m\u001b[38;5;21;01mdata\u001b[39;00m\u001b[38;5;21;01m.\u001b[39;00m\u001b[38;5;21;01mclassification\u001b[39;00m \u001b[38;5;28;01mimport\u001b[39;00m Banking77, Clinc150, Massive, Minds14\n\u001b[0;32m---> 10\u001b[0m \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;21;01m.\u001b[39;00m\u001b[38;5;21;01mdata\u001b[39;00m\u001b[38;5;21;01m.\u001b[39;00m\u001b[38;5;21;01moos_detection\u001b[39;00m \u001b[38;5;28;01mimport\u001b[39;00m Clinc150OOSDetection\n\u001b[1;32m     11\u001b[0m \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;21;01m.\u001b[39;00m\u001b[38;5;21;01mdata\u001b[39;00m\u001b[38;5;21;01m.\u001b[39;00m\u001b[38;5;21;01mresponse_selection\u001b[39;00m \u001b[38;5;28;01mimport\u001b[39;00m AmazonQA, Ubuntu\n\u001b[1;32m     12\u001b[0m \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;21;01m.\u001b[39;00m\u001b[38;5;21;01mevaluation\u001b[39;00m\u001b[38;5;21;01m.\u001b[39;00m\u001b[38;5;21;01mbase\u001b[39;00m \u001b[38;5;28;01mimport\u001b[39;00m BaseEvaluator\n",
      "File \u001b[0;32m/cephfs/home/alekseev/dialogue-augmentation/ruconveval/src/data/oos_detection.py:1\u001b[0m\n\u001b[0;32m----> 1\u001b[0m \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;21;01msrc\u001b[39;00m\u001b[38;5;21;01m.\u001b[39;00m\u001b[38;5;21;01mdata\u001b[39;00m\u001b[38;5;21;01m.\u001b[39;00m\u001b[38;5;21;01mbase\u001b[39;00m \u001b[38;5;28;01mimport\u001b[39;00m ClassificationDatasetSplit\n\u001b[1;32m      2\u001b[0m \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;21;01msrc\u001b[39;00m\u001b[38;5;21;01m.\u001b[39;00m\u001b[38;5;21;01mdata\u001b[39;00m\u001b[38;5;21;01m.\u001b[39;00m\u001b[38;5;21;01mclassification\u001b[39;00m \u001b[38;5;28;01mimport\u001b[39;00m Clinc150\n\u001b[1;32m      5\u001b[0m \u001b[38;5;28;01mclass\u001b[39;00m \u001b[38;5;21;01mClinc150OOSDetection\u001b[39;00m(Clinc150):\n",
      "\u001b[0;31mModuleNotFoundError\u001b[0m: No module named 'src'"
     ]
    }
   ],
   "source": [
    "from ruconveval.src.task import Task, TaskConfig, TaskType, DatasetConfig\n",
    "\n",
    "dataset_config = DatasetConfig(\n",
    "    dataset='massive',\n",
    "    language='en'\n",
    ")\n",
    "\n",
    "task_config = TaskConfig(\n",
    "    task=TaskType.classification,\n",
    "    dataset_config=dataset_config\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## save model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from mylib.learners import DialogueEncoderLearner as Learner, DialogueEncoderLearnerConfig as LearnerConfig\n",
    "from mylib.modeling.dialogue import BaselineDialogueEncoder as Model, BaselineDialogueEncoderConfig as ModelConfig\n",
    "\n",
    "model_config = ModelConfig(hf_model='google-bert/bert-base-uncased')\n",
    "model = Model(model_config)\n",
    "\n",
    "ckpt_path = 'logs/comet/dialogue-encoder/ba6388e710684d94af81ebab88c2ff5a/checkpoints/epoch=4-step=8460.ckpt'\n",
    "learner_config = LearnerConfig()\n",
    "learner = Learner.load_from_checkpoint(\n",
    "    checkpoint_path=ckpt_path,\n",
    "    map_location='cpu',\n",
    "    model=model,\n",
    "    config=learner_config\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "transformers.models.bert.modeling_bert.BertModel"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model = learner.model.model\n",
    "type(model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.save_pretrained('pretrained/bert-fair-trivial')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "('pretrained/bert-fair-trivial/tokenizer_config.json',\n",
       " 'pretrained/bert-fair-trivial/special_tokens_map.json',\n",
       " 'pretrained/bert-fair-trivial/vocab.txt',\n",
       " 'pretrained/bert-fair-trivial/added_tokens.json',\n",
       " 'pretrained/bert-fair-trivial/tokenizer.json')"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "learner.model.tokenizer.save_pretrained('pretrained/bert-fair-trivial')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## filter by source"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "15abc1f12f464e9b88fa720568f19163",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Loading dataset from disk:   0%|          | 0/64 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "433340"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from datasets import load_from_disk\n",
    "\n",
    "dataset = load_from_disk('data/train-bert/trivial')\n",
    "len(dataset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['back-translate', 'insert', 'replace']"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "augs = [aug['augmentation'] for aug in dataset[0]['pos']]\n",
    "augs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "17656626b475450d994acfd4981795e8",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Map:   0%|          | 0/433340 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "mapped = dataset.map(\n",
    "    lambda pos: {'pos': [aug for aug in pos if aug['augmentation'] != 'back-translate']},\n",
    "    input_columns='pos',\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'source_dataset_name': 'AirDialogue',\n",
       " 'idx_within_source': 185688,\n",
       " 'id': 360925,\n",
       " 'pos': [{'augmentation': 'insert',\n",
       "   'content': [{'speaker': 0, 'utterance': 'Hello Hello.'},\n",
       "    {'speaker': 1, 'utterance': 'Hello, How can I assist you?'},\n",
       "    {'speaker': 0,\n",
       "     'utterance': 'I am very much excited to visit Ellenville Fault ed Ice Caves in New York - EWR, can you please reserve a flight from Houston - IAH?'},\n",
       "    {'speaker': 1, 'utterance': 'Sure, May I know the travelling dates?'},\n",
       "    {'speaker': 0, 'utterance': 'PS My travel dates are Apr 28th and 30th.'},\n",
       "    {'speaker': 1, 'utterance': 'Do ing you have any connection limit?'},\n",
       "    {'speaker': 0,\n",
       "     'utterance': 'Yes I am fine with maximum one GB connection.'},\n",
       "    {'speaker': 1, 'utterance': 'May I know your name?'},\n",
       "    {'speaker': 0, 'utterance': 'My name is Gary Evans.'},\n",
       "    {'speaker': 1, 'utterance': 'Â Please wait a minute.'},\n",
       "    {'speaker': 0,\n",
       "     'utterance': 'Q Please make sure, I want to depart in the evening.'},\n",
       "    {'speaker': 1,\n",
       "     'utterance': 'Sorry, No flights found with your chosen dates.'},\n",
       "    {'speaker': 0, 'utterance': 'No problem, thank you for the information.'},\n",
       "    {'speaker': 1, 'utterance': 'Thank ing you for availing us.'}]},\n",
       "  {'augmentation': 'replace',\n",
       "   'content': [{'speaker': 0, 'utterance': 'Hello.'},\n",
       "    {'speaker': 1, 'utterance': 'Hello, How can I help you?'},\n",
       "    {'speaker': 0,\n",
       "     'utterance': 'I am very oooo ative to visit Ellenville Fault Ice Caves in New AR - EWR, can you please ate a flight from Houston - IAH?'},\n",
       "    {'speaker': 1, 'utterance': 'Sure, May I know the travelling dates?'},\n",
       "    {'speaker': 0, 'utterance': 'My travel days are Apr 28th and 30th.'},\n",
       "    {'speaker': 1, 'utterance': 'Do you have any connection limit?'},\n",
       "    {'speaker': 0, 'utterance': 'I am fine with maximum one connection.'},\n",
       "    {'speaker': 1, 'utterance': 'May I know your name?'},\n",
       "    {'speaker': 0, 'utterance': 'My Name is Gary Evans.'},\n",
       "    {'speaker': 1, 'utterance': 'Please wait a minute.'},\n",
       "    {'speaker': 0,\n",
       "     'utterance': 'Please make sure, I want to depart in the evening.'},\n",
       "    {'speaker': 1, 'utterance': 'Sorry, No ont ups with your chosen dates.'},\n",
       "    {'speaker': 0, 'utterance': 'No problem, thank you for the information.'},\n",
       "    {'speaker': 1, 'utterance': 'Thank you for availing us.'}]}],\n",
       " 'orig': {'augmentation': None,\n",
       "  'content': [{'speaker': 0, 'utterance': 'Hello.'},\n",
       "   {'speaker': 1, 'utterance': 'Hello, How can I assist you?'},\n",
       "   {'speaker': 0,\n",
       "    'utterance': 'I am very much excited to visit Ellenville Fault Ice Caves in New York - EWR, can you please reserve a flight from Houston - IAH?'},\n",
       "   {'speaker': 1, 'utterance': 'Sure, May I know the travelling dates?'},\n",
       "   {'speaker': 0, 'utterance': 'My travel dates are Apr 28th and 30th.'},\n",
       "   {'speaker': 1, 'utterance': 'Do you have any connection limit?'},\n",
       "   {'speaker': 0, 'utterance': 'I am fine with maximum one connection.'},\n",
       "   {'speaker': 1, 'utterance': 'May I know your name?'},\n",
       "   {'speaker': 0, 'utterance': 'My name is Gary Evans.'},\n",
       "   {'speaker': 1, 'utterance': 'Please wait a minute.'},\n",
       "   {'speaker': 0,\n",
       "    'utterance': 'Please make sure, I want to depart in the evening.'},\n",
       "   {'speaker': 1,\n",
       "    'utterance': 'Sorry, No flights found with your chosen dates.'},\n",
       "   {'speaker': 0, 'utterance': 'No problem, thank you for the information.'},\n",
       "   {'speaker': 1, 'utterance': 'Thank you for availing us.'}]}}"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mapped[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## lightning to hugging face"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from mylib.modeling.dialogue import BaselineDialogueEncoderConfig, BaselineDialogueEncoder\n",
    "\n",
    "\n",
    "def get_raw_model(model_config: BaselineDialogueEncoderConfig):\n",
    "    model = BaselineDialogueEncoder(model_config)\n",
    "    model.requires_grad_(False)\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from mylib.learners import DialogueEncoderLearner, DialogueEncoderLearnerConfig\n",
    "\n",
    "\n",
    "def get_pretrained_model(model, init_from):\n",
    "    return DialogueEncoderLearner.load_from_checkpoint(\n",
    "        checkpoint_path=init_from,\n",
    "        model=model,\n",
    "        config=DialogueEncoderLearnerConfig()\n",
    "    ).model.model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "checkpoint_codenames ={\n",
    "    'google-bert/bert-base-uncased':\n",
    "    {\n",
    "        'trivial': \"ba6388e710684d94af81ebab88c2ff5a\",\n",
    "        'advanced': \"0d0b24e2d3864cf8a6c8b7d5e15b1c83\",\n",
    "        'crazy' :\"d7e24b7b434e486d866f35d70ac0c503\"\n",
    "    },\n",
    "    'FacebookAI/roberta-base':\n",
    "    {\n",
    "        'trivial': \"e4fa2d98923e4c8ca4fbe20c7ccc4c67\",\n",
    "        'advanced': \"849fa15859484683a4341837e7d1dd1d\",\n",
    "        'crazy': \"523da04c3c5a41379acc298ee09c7a0a\"\n",
    "    },\n",
    "    'Shitao/RetroMAE':\n",
    "    {\n",
    "        'trivial': \"408af9ac5d114a9da1fa007277343f15\",\n",
    "        'advanced': \"e2de548066f14753bdf9a646e992c02f\",\n",
    "        'crazy' :\"ca66a8cd07c4418daea805093d00cdd3\"\n",
    "    }\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['logs/comet/dialogue-encoder/ca66a8cd07c4418daea805093d00cdd3/checkpoints/epoch=2-step=5055.ckpt',\n",
       " 'logs/comet/dialogue-encoder/ca66a8cd07c4418daea805093d00cdd3/checkpoints/epoch=3-step=6740.ckpt',\n",
       " 'logs/comet/dialogue-encoder/ca66a8cd07c4418daea805093d00cdd3/checkpoints/epoch=0-step=1685.ckpt',\n",
       " 'logs/comet/dialogue-encoder/ca66a8cd07c4418daea805093d00cdd3/checkpoints/epoch=4-step=8425.ckpt',\n",
       " 'logs/comet/dialogue-encoder/ca66a8cd07c4418daea805093d00cdd3/checkpoints/epoch=1-step=3370.ckpt']"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from glob import glob\n",
    "\n",
    "def get_files(ckpt_code):\n",
    "    return glob(f\"logs/comet/dialogue-encoder/{ckpt_code}/checkpoints/*.ckpt\")\n",
    "\n",
    "get_files('ca66a8cd07c4418daea805093d00cdd3')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "from transformers import AutoTokenizer\n",
    "\n",
    "def to_hugging_face(ckpt_path, model_config, path_out):\n",
    "    model = get_raw_model(model_config)\n",
    "    model = get_pretrained_model(model, ckpt_path)\n",
    "    model.save_pretrained(path_out)\n",
    "    tok = AutoTokenizer.from_pretrained(model_config.hf_model)\n",
    "    tok.save_pretrained(path_out)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of RobertaModel were not initialized from the model checkpoint at FacebookAI/roberta-base and are newly initialized: ['roberta.pooler.dense.bias', 'roberta.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of RobertaModel were not initialized from the model checkpoint at FacebookAI/roberta-base and are newly initialized: ['roberta.pooler.dense.bias', 'roberta.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of RobertaModel were not initialized from the model checkpoint at FacebookAI/roberta-base and are newly initialized: ['roberta.pooler.dense.bias', 'roberta.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of RobertaModel were not initialized from the model checkpoint at FacebookAI/roberta-base and are newly initialized: ['roberta.pooler.dense.bias', 'roberta.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of RobertaModel were not initialized from the model checkpoint at FacebookAI/roberta-base and are newly initialized: ['roberta.pooler.dense.bias', 'roberta.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of RobertaModel were not initialized from the model checkpoint at FacebookAI/roberta-base and are newly initialized: ['roberta.pooler.dense.bias', 'roberta.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of RobertaModel were not initialized from the model checkpoint at FacebookAI/roberta-base and are newly initialized: ['roberta.pooler.dense.bias', 'roberta.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of RobertaModel were not initialized from the model checkpoint at FacebookAI/roberta-base and are newly initialized: ['roberta.pooler.dense.bias', 'roberta.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of RobertaModel were not initialized from the model checkpoint at FacebookAI/roberta-base and are newly initialized: ['roberta.pooler.dense.bias', 'roberta.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of RobertaModel were not initialized from the model checkpoint at FacebookAI/roberta-base and are newly initialized: ['roberta.pooler.dense.bias', 'roberta.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of RobertaModel were not initialized from the model checkpoint at FacebookAI/roberta-base and are newly initialized: ['roberta.pooler.dense.bias', 'roberta.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of RobertaModel were not initialized from the model checkpoint at FacebookAI/roberta-base and are newly initialized: ['roberta.pooler.dense.bias', 'roberta.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of RobertaModel were not initialized from the model checkpoint at FacebookAI/roberta-base and are newly initialized: ['roberta.pooler.dense.bias', 'roberta.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of RobertaModel were not initialized from the model checkpoint at FacebookAI/roberta-base and are newly initialized: ['roberta.pooler.dense.bias', 'roberta.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of RobertaModel were not initialized from the model checkpoint at FacebookAI/roberta-base and are newly initialized: ['roberta.pooler.dense.bias', 'roberta.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of BertModel were not initialized from the model checkpoint at Shitao/RetroMAE and are newly initialized: ['bert.pooler.dense.bias', 'bert.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of BertModel were not initialized from the model checkpoint at Shitao/RetroMAE and are newly initialized: ['bert.pooler.dense.bias', 'bert.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of BertModel were not initialized from the model checkpoint at Shitao/RetroMAE and are newly initialized: ['bert.pooler.dense.bias', 'bert.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of BertModel were not initialized from the model checkpoint at Shitao/RetroMAE and are newly initialized: ['bert.pooler.dense.bias', 'bert.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of BertModel were not initialized from the model checkpoint at Shitao/RetroMAE and are newly initialized: ['bert.pooler.dense.bias', 'bert.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of BertModel were not initialized from the model checkpoint at Shitao/RetroMAE and are newly initialized: ['bert.pooler.dense.bias', 'bert.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of BertModel were not initialized from the model checkpoint at Shitao/RetroMAE and are newly initialized: ['bert.pooler.dense.bias', 'bert.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of BertModel were not initialized from the model checkpoint at Shitao/RetroMAE and are newly initialized: ['bert.pooler.dense.bias', 'bert.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of BertModel were not initialized from the model checkpoint at Shitao/RetroMAE and are newly initialized: ['bert.pooler.dense.bias', 'bert.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of BertModel were not initialized from the model checkpoint at Shitao/RetroMAE and are newly initialized: ['bert.pooler.dense.bias', 'bert.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of BertModel were not initialized from the model checkpoint at Shitao/RetroMAE and are newly initialized: ['bert.pooler.dense.bias', 'bert.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of BertModel were not initialized from the model checkpoint at Shitao/RetroMAE and are newly initialized: ['bert.pooler.dense.bias', 'bert.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of BertModel were not initialized from the model checkpoint at Shitao/RetroMAE and are newly initialized: ['bert.pooler.dense.bias', 'bert.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of BertModel were not initialized from the model checkpoint at Shitao/RetroMAE and are newly initialized: ['bert.pooler.dense.bias', 'bert.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "Some weights of BertModel were not initialized from the model checkpoint at Shitao/RetroMAE and are newly initialized: ['bert.pooler.dense.bias', 'bert.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n"
     ]
    }
   ],
   "source": [
    "for backbone_name in checkpoint_codenames.keys():\n",
    "    model_config = BaselineDialogueEncoderConfig(hf_model=backbone_name)\n",
    "    for aug_set in ['trivial', 'advanced', 'crazy']:\n",
    "        for ckpt in get_files(checkpoint_codenames[backbone_name][aug_set]):\n",
    "            backbone_for_path = backbone_name.replace('/', '-')\n",
    "            ckpt_for_path = ckpt.split('/')[-1].split('-')[0]\n",
    "            to_hugging_face(ckpt, model_config, f'pretrained/{backbone_for_path}/{aug_set}/{ckpt_for_path}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## inspecting sgd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "DatasetDict({\n",
       "    train: Dataset({\n",
       "        features: ['content', 'source_dataset_name', 'idx_within_source', 'services'],\n",
       "        num_rows: 2979\n",
       "    })\n",
       "    validation: Dataset({\n",
       "        features: ['content', 'source_dataset_name', 'idx_within_source', 'services'],\n",
       "        num_rows: 204\n",
       "    })\n",
       "    test: Dataset({\n",
       "        features: ['content', 'source_dataset_name', 'idx_within_source', 'services'],\n",
       "        num_rows: 226\n",
       "    })\n",
       "})"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from datasets import load_from_disk\n",
    "\n",
    "dataset = load_from_disk('data/benchmarks/MultiWOZ_2.1')\n",
    "dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 127,  287,  513, 1199,  326,  282])"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import numpy as np\n",
    "\n",
    "a = dataset['train']['services']\n",
    "np.sum(a, axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([12,  0, 67, 62, 52, 33])"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "a = dataset['test']['services']\n",
    "np.sum(a, axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([11,  0, 56, 50, 57, 30])"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "a = dataset['validation']['services']\n",
    "np.sum(a, axis=0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## SFR Embeddings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8906ed8a2c2f469db2052a8f4f181af8",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Loading checkpoint shards:   0%|          | 0/3 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8e627eec902f4c1c948f9a420f1b9d84",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Map:   0%|          | 0/6911 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "Cell \u001b[0;32mIn[2], line 59\u001b[0m\n\u001b[1;32m     56\u001b[0m \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;21;01mtorch\u001b[39;00m \u001b[38;5;28;01mimport\u001b[39;00m no_grad\n\u001b[1;32m     58\u001b[0m \u001b[38;5;28;01mwith\u001b[39;00m no_grad():\n\u001b[0;32m---> 59\u001b[0m     dataset \u001b[38;5;241m=\u001b[39m \u001b[43mdataset\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mmap\u001b[49m\u001b[43m(\u001b[49m\n\u001b[1;32m     60\u001b[0m \u001b[43m        \u001b[49m\u001b[38;5;28;43;01mlambda\u001b[39;49;00m\u001b[43m \u001b[49m\u001b[43mbatch\u001b[49m\u001b[43m:\u001b[49m\u001b[43m \u001b[49m\u001b[43m{\u001b[49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[38;5;124;43membedding\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[43m:\u001b[49m\u001b[43m \u001b[49m\u001b[43mencoder\u001b[49m\u001b[43m(\u001b[49m\u001b[43mbatch\u001b[49m\u001b[43m)\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mdetach\u001b[49m\u001b[43m(\u001b[49m\u001b[43m)\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mcpu\u001b[49m\u001b[43m(\u001b[49m\u001b[43m)\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mnumpy\u001b[49m\u001b[43m(\u001b[49m\u001b[43m)\u001b[49m\u001b[43m}\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m     61\u001b[0m \u001b[43m        \u001b[49m\u001b[43minput_columns\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[38;5;124;43mcontent\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[43m,\u001b[49m\n\u001b[1;32m     62\u001b[0m \u001b[43m        \u001b[49m\u001b[43mbatched\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[38;5;28;43;01mTrue\u001b[39;49;00m\u001b[43m,\u001b[49m\n\u001b[1;32m     63\u001b[0m \u001b[43m        \u001b[49m\u001b[43mbatch_size\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43margs\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mbatch_size\u001b[49m\n\u001b[1;32m     64\u001b[0m \u001b[43m    \u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m     66\u001b[0m \u001b[38;5;66;03m# save to disk\u001b[39;00m\n\u001b[1;32m     67\u001b[0m dataset\u001b[38;5;241m.\u001b[39msave_to_disk(args\u001b[38;5;241m.\u001b[39mpath_out)\n",
      "File \u001b[0;32m~/dialogue-augmentation/.venv/lib/python3.8/site-packages/datasets/arrow_dataset.py:593\u001b[0m, in \u001b[0;36mtransmit_tasks.<locals>.wrapper\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m    591\u001b[0m     \u001b[38;5;28mself\u001b[39m: \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mDataset\u001b[39m\u001b[38;5;124m\"\u001b[39m \u001b[38;5;241m=\u001b[39m kwargs\u001b[38;5;241m.\u001b[39mpop(\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mself\u001b[39m\u001b[38;5;124m\"\u001b[39m)\n\u001b[1;32m    592\u001b[0m \u001b[38;5;66;03m# apply actual function\u001b[39;00m\n\u001b[0;32m--> 593\u001b[0m out: Union[\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mDataset\u001b[39m\u001b[38;5;124m\"\u001b[39m, \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mDatasetDict\u001b[39m\u001b[38;5;124m\"\u001b[39m] \u001b[38;5;241m=\u001b[39m \u001b[43mfunc\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    594\u001b[0m datasets: List[\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mDataset\u001b[39m\u001b[38;5;124m\"\u001b[39m] \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mlist\u001b[39m(out\u001b[38;5;241m.\u001b[39mvalues()) \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28misinstance\u001b[39m(out, \u001b[38;5;28mdict\u001b[39m) \u001b[38;5;28;01melse\u001b[39;00m [out]\n\u001b[1;32m    595\u001b[0m \u001b[38;5;28;01mfor\u001b[39;00m dataset \u001b[38;5;129;01min\u001b[39;00m datasets:\n\u001b[1;32m    596\u001b[0m     \u001b[38;5;66;03m# Remove task templates if a column mapping of the template is no longer valid\u001b[39;00m\n",
      "File \u001b[0;32m~/dialogue-augmentation/.venv/lib/python3.8/site-packages/datasets/arrow_dataset.py:558\u001b[0m, in \u001b[0;36mtransmit_format.<locals>.wrapper\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m    551\u001b[0m self_format \u001b[38;5;241m=\u001b[39m {\n\u001b[1;32m    552\u001b[0m     \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mtype\u001b[39m\u001b[38;5;124m\"\u001b[39m: \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_format_type,\n\u001b[1;32m    553\u001b[0m     \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mformat_kwargs\u001b[39m\u001b[38;5;124m\"\u001b[39m: \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_format_kwargs,\n\u001b[1;32m    554\u001b[0m     \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mcolumns\u001b[39m\u001b[38;5;124m\"\u001b[39m: \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_format_columns,\n\u001b[1;32m    555\u001b[0m     \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124moutput_all_columns\u001b[39m\u001b[38;5;124m\"\u001b[39m: \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_output_all_columns,\n\u001b[1;32m    556\u001b[0m }\n\u001b[1;32m    557\u001b[0m \u001b[38;5;66;03m# apply actual function\u001b[39;00m\n\u001b[0;32m--> 558\u001b[0m out: Union[\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mDataset\u001b[39m\u001b[38;5;124m\"\u001b[39m, \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mDatasetDict\u001b[39m\u001b[38;5;124m\"\u001b[39m] \u001b[38;5;241m=\u001b[39m \u001b[43mfunc\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    559\u001b[0m datasets: List[\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mDataset\u001b[39m\u001b[38;5;124m\"\u001b[39m] \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mlist\u001b[39m(out\u001b[38;5;241m.\u001b[39mvalues()) \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28misinstance\u001b[39m(out, \u001b[38;5;28mdict\u001b[39m) \u001b[38;5;28;01melse\u001b[39;00m [out]\n\u001b[1;32m    560\u001b[0m \u001b[38;5;66;03m# re-apply format to the output\u001b[39;00m\n",
      "File \u001b[0;32m~/dialogue-augmentation/.venv/lib/python3.8/site-packages/datasets/arrow_dataset.py:3105\u001b[0m, in \u001b[0;36mDataset.map\u001b[0;34m(self, function, with_indices, with_rank, input_columns, batched, batch_size, drop_last_batch, remove_columns, keep_in_memory, load_from_cache_file, cache_file_name, writer_batch_size, features, disable_nullable, fn_kwargs, num_proc, suffix_template, new_fingerprint, desc)\u001b[0m\n\u001b[1;32m   3099\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m transformed_dataset \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m:\n\u001b[1;32m   3100\u001b[0m     \u001b[38;5;28;01mwith\u001b[39;00m hf_tqdm(\n\u001b[1;32m   3101\u001b[0m         unit\u001b[38;5;241m=\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m examples\u001b[39m\u001b[38;5;124m\"\u001b[39m,\n\u001b[1;32m   3102\u001b[0m         total\u001b[38;5;241m=\u001b[39mpbar_total,\n\u001b[1;32m   3103\u001b[0m         desc\u001b[38;5;241m=\u001b[39mdesc \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mMap\u001b[39m\u001b[38;5;124m\"\u001b[39m,\n\u001b[1;32m   3104\u001b[0m     ) \u001b[38;5;28;01mas\u001b[39;00m pbar:\n\u001b[0;32m-> 3105\u001b[0m         \u001b[38;5;28;01mfor\u001b[39;00m rank, done, content \u001b[38;5;129;01min\u001b[39;00m Dataset\u001b[38;5;241m.\u001b[39m_map_single(\u001b[38;5;241m*\u001b[39m\u001b[38;5;241m*\u001b[39mdataset_kwargs):\n\u001b[1;32m   3106\u001b[0m             \u001b[38;5;28;01mif\u001b[39;00m done:\n\u001b[1;32m   3107\u001b[0m                 shards_done \u001b[38;5;241m+\u001b[39m\u001b[38;5;241m=\u001b[39m \u001b[38;5;241m1\u001b[39m\n",
      "File \u001b[0;32m~/dialogue-augmentation/.venv/lib/python3.8/site-packages/datasets/arrow_dataset.py:3482\u001b[0m, in \u001b[0;36mDataset._map_single\u001b[0;34m(shard, function, with_indices, with_rank, input_columns, batched, batch_size, drop_last_batch, remove_columns, keep_in_memory, cache_file_name, writer_batch_size, features, disable_nullable, fn_kwargs, new_fingerprint, rank, offset)\u001b[0m\n\u001b[1;32m   3478\u001b[0m indices \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mlist\u001b[39m(\n\u001b[1;32m   3479\u001b[0m     \u001b[38;5;28mrange\u001b[39m(\u001b[38;5;241m*\u001b[39m(\u001b[38;5;28mslice\u001b[39m(i, i \u001b[38;5;241m+\u001b[39m batch_size)\u001b[38;5;241m.\u001b[39mindices(shard\u001b[38;5;241m.\u001b[39mnum_rows)))\n\u001b[1;32m   3480\u001b[0m )  \u001b[38;5;66;03m# Something simpler?\u001b[39;00m\n\u001b[1;32m   3481\u001b[0m \u001b[38;5;28;01mtry\u001b[39;00m:\n\u001b[0;32m-> 3482\u001b[0m     batch \u001b[38;5;241m=\u001b[39m \u001b[43mapply_function_on_filtered_inputs\u001b[49m\u001b[43m(\u001b[49m\n\u001b[1;32m   3483\u001b[0m \u001b[43m        \u001b[49m\u001b[43mbatch\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m   3484\u001b[0m \u001b[43m        \u001b[49m\u001b[43mindices\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m   3485\u001b[0m \u001b[43m        \u001b[49m\u001b[43mcheck_same_num_examples\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[38;5;28;43mlen\u001b[39;49m\u001b[43m(\u001b[49m\u001b[43mshard\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mlist_indexes\u001b[49m\u001b[43m(\u001b[49m\u001b[43m)\u001b[49m\u001b[43m)\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m>\u001b[39;49m\u001b[43m \u001b[49m\u001b[38;5;241;43m0\u001b[39;49m\u001b[43m,\u001b[49m\n\u001b[1;32m   3486\u001b[0m \u001b[43m        \u001b[49m\u001b[43moffset\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43moffset\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m   3487\u001b[0m \u001b[43m    \u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m   3488\u001b[0m \u001b[38;5;28;01mexcept\u001b[39;00m NumExamplesMismatchError:\n\u001b[1;32m   3489\u001b[0m     \u001b[38;5;28;01mraise\u001b[39;00m DatasetTransformationNotAllowedError(\n\u001b[1;32m   3490\u001b[0m         \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mUsing `.map` in batched mode on a dataset with attached indexes is allowed only if it doesn\u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mt create or remove existing examples. You can first run `.drop_index() to remove your index and then re-add it.\u001b[39m\u001b[38;5;124m\"\u001b[39m\n\u001b[1;32m   3491\u001b[0m     ) \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m\n",
      "File \u001b[0;32m~/dialogue-augmentation/.venv/lib/python3.8/site-packages/datasets/arrow_dataset.py:3361\u001b[0m, in \u001b[0;36mDataset._map_single.<locals>.apply_function_on_filtered_inputs\u001b[0;34m(pa_inputs, indices, check_same_num_examples, offset)\u001b[0m\n\u001b[1;32m   3359\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m with_rank:\n\u001b[1;32m   3360\u001b[0m     additional_args \u001b[38;5;241m+\u001b[39m\u001b[38;5;241m=\u001b[39m (rank,)\n\u001b[0;32m-> 3361\u001b[0m processed_inputs \u001b[38;5;241m=\u001b[39m \u001b[43mfunction\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mfn_args\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43madditional_args\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mfn_kwargs\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m   3362\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28misinstance\u001b[39m(processed_inputs, LazyDict):\n\u001b[1;32m   3363\u001b[0m     processed_inputs \u001b[38;5;241m=\u001b[39m {\n\u001b[1;32m   3364\u001b[0m         k: v \u001b[38;5;28;01mfor\u001b[39;00m k, v \u001b[38;5;129;01min\u001b[39;00m processed_inputs\u001b[38;5;241m.\u001b[39mdata\u001b[38;5;241m.\u001b[39mitems() \u001b[38;5;28;01mif\u001b[39;00m k \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;129;01min\u001b[39;00m processed_inputs\u001b[38;5;241m.\u001b[39mkeys_to_format\n\u001b[1;32m   3365\u001b[0m     }\n",
      "Cell \u001b[0;32mIn[2], line 60\u001b[0m, in \u001b[0;36m<lambda>\u001b[0;34m(batch)\u001b[0m\n\u001b[1;32m     56\u001b[0m \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;21;01mtorch\u001b[39;00m \u001b[38;5;28;01mimport\u001b[39;00m no_grad\n\u001b[1;32m     58\u001b[0m \u001b[38;5;28;01mwith\u001b[39;00m no_grad():\n\u001b[1;32m     59\u001b[0m     dataset \u001b[38;5;241m=\u001b[39m dataset\u001b[38;5;241m.\u001b[39mmap(\n\u001b[0;32m---> 60\u001b[0m         \u001b[38;5;28;01mlambda\u001b[39;00m batch: {\u001b[38;5;124m'\u001b[39m\u001b[38;5;124membedding\u001b[39m\u001b[38;5;124m'\u001b[39m: \u001b[43mencoder\u001b[49m\u001b[43m(\u001b[49m\u001b[43mbatch\u001b[49m\u001b[43m)\u001b[49m\u001b[38;5;241m.\u001b[39mdetach()\u001b[38;5;241m.\u001b[39mcpu()\u001b[38;5;241m.\u001b[39mnumpy()},\n\u001b[1;32m     61\u001b[0m         input_columns\u001b[38;5;241m=\u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mcontent\u001b[39m\u001b[38;5;124m'\u001b[39m,\n\u001b[1;32m     62\u001b[0m         batched\u001b[38;5;241m=\u001b[39m\u001b[38;5;28;01mTrue\u001b[39;00m,\n\u001b[1;32m     63\u001b[0m         batch_size\u001b[38;5;241m=\u001b[39margs\u001b[38;5;241m.\u001b[39mbatch_size\n\u001b[1;32m     64\u001b[0m     )\n\u001b[1;32m     66\u001b[0m \u001b[38;5;66;03m# save to disk\u001b[39;00m\n\u001b[1;32m     67\u001b[0m dataset\u001b[38;5;241m.\u001b[39msave_to_disk(args\u001b[38;5;241m.\u001b[39mpath_out)\n",
      "File \u001b[0;32m~/dialogue-augmentation/.venv/lib/python3.8/site-packages/torch/nn/modules/module.py:1511\u001b[0m, in \u001b[0;36mModule._wrapped_call_impl\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m   1509\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_compiled_call_impl(\u001b[38;5;241m*\u001b[39margs, \u001b[38;5;241m*\u001b[39m\u001b[38;5;241m*\u001b[39mkwargs)  \u001b[38;5;66;03m# type: ignore[misc]\u001b[39;00m\n\u001b[1;32m   1510\u001b[0m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[0;32m-> 1511\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43m_call_impl\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n",
      "File \u001b[0;32m~/dialogue-augmentation/.venv/lib/python3.8/site-packages/torch/nn/modules/module.py:1520\u001b[0m, in \u001b[0;36mModule._call_impl\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m   1515\u001b[0m \u001b[38;5;66;03m# If we don't have any hooks, we want to skip the rest of the logic in\u001b[39;00m\n\u001b[1;32m   1516\u001b[0m \u001b[38;5;66;03m# this function, and just call forward.\u001b[39;00m\n\u001b[1;32m   1517\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m (\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_backward_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_backward_pre_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_forward_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_forward_pre_hooks\n\u001b[1;32m   1518\u001b[0m         \u001b[38;5;129;01mor\u001b[39;00m _global_backward_pre_hooks \u001b[38;5;129;01mor\u001b[39;00m _global_backward_hooks\n\u001b[1;32m   1519\u001b[0m         \u001b[38;5;129;01mor\u001b[39;00m _global_forward_hooks \u001b[38;5;129;01mor\u001b[39;00m _global_forward_pre_hooks):\n\u001b[0;32m-> 1520\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mforward_call\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m   1522\u001b[0m \u001b[38;5;28;01mtry\u001b[39;00m:\n\u001b[1;32m   1523\u001b[0m     result \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;01mNone\u001b[39;00m\n",
      "File \u001b[0;32m/cephfs/home/alekseev/dialogue-augmentation/mylib/modeling/dialogue/baseline_dialogue_encoder.py:80\u001b[0m, in \u001b[0;36mBaselineDialogueEncoder.forward\u001b[0;34m(self, batch)\u001b[0m\n\u001b[1;32m     70\u001b[0m \u001b[38;5;250m\u001b[39m\u001b[38;5;124;03m\"\"\"\u001b[39;00m\n\u001b[1;32m     71\u001b[0m \u001b[38;5;124;03minput: batch of dialogues\u001b[39;00m\n\u001b[1;32m     72\u001b[0m \u001b[38;5;124;03moutput: (B,H)\u001b[39;00m\n\u001b[1;32m     73\u001b[0m \u001b[38;5;124;03m\"\"\"\u001b[39;00m\n\u001b[1;32m     74\u001b[0m inputs \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_tokenize(\n\u001b[1;32m     75\u001b[0m     \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mtokenizer,\n\u001b[1;32m     76\u001b[0m     batch,\n\u001b[1;32m     77\u001b[0m     truncation\u001b[38;5;241m=\u001b[39m\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mconfig\u001b[38;5;241m.\u001b[39mtruncation,\n\u001b[1;32m     78\u001b[0m     max_length\u001b[38;5;241m=\u001b[39m\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mconfig\u001b[38;5;241m.\u001b[39mmax_length\n\u001b[1;32m     79\u001b[0m )\u001b[38;5;241m.\u001b[39mto(\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mdevice)\n\u001b[0;32m---> 80\u001b[0m outputs \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mmodel\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43minputs\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m     81\u001b[0m encodings \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mpooler(outputs\u001b[38;5;241m.\u001b[39mlast_hidden_state, inputs\u001b[38;5;241m.\u001b[39mattention_mask)\n\u001b[1;32m     82\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m encodings\n",
      "File \u001b[0;32m~/dialogue-augmentation/.venv/lib/python3.8/site-packages/torch/nn/modules/module.py:1511\u001b[0m, in \u001b[0;36mModule._wrapped_call_impl\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m   1509\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_compiled_call_impl(\u001b[38;5;241m*\u001b[39margs, \u001b[38;5;241m*\u001b[39m\u001b[38;5;241m*\u001b[39mkwargs)  \u001b[38;5;66;03m# type: ignore[misc]\u001b[39;00m\n\u001b[1;32m   1510\u001b[0m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[0;32m-> 1511\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43m_call_impl\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n",
      "File \u001b[0;32m~/dialogue-augmentation/.venv/lib/python3.8/site-packages/torch/nn/modules/module.py:1520\u001b[0m, in \u001b[0;36mModule._call_impl\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m   1515\u001b[0m \u001b[38;5;66;03m# If we don't have any hooks, we want to skip the rest of the logic in\u001b[39;00m\n\u001b[1;32m   1516\u001b[0m \u001b[38;5;66;03m# this function, and just call forward.\u001b[39;00m\n\u001b[1;32m   1517\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m (\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_backward_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_backward_pre_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_forward_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_forward_pre_hooks\n\u001b[1;32m   1518\u001b[0m         \u001b[38;5;129;01mor\u001b[39;00m _global_backward_pre_hooks \u001b[38;5;129;01mor\u001b[39;00m _global_backward_hooks\n\u001b[1;32m   1519\u001b[0m         \u001b[38;5;129;01mor\u001b[39;00m _global_forward_hooks \u001b[38;5;129;01mor\u001b[39;00m _global_forward_pre_hooks):\n\u001b[0;32m-> 1520\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mforward_call\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m   1522\u001b[0m \u001b[38;5;28;01mtry\u001b[39;00m:\n\u001b[1;32m   1523\u001b[0m     result \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;01mNone\u001b[39;00m\n",
      "File \u001b[0;32m~/dialogue-augmentation/.venv/lib/python3.8/site-packages/transformers/models/mistral/modeling_mistral.py:1039\u001b[0m, in \u001b[0;36mMistralModel.forward\u001b[0;34m(self, input_ids, attention_mask, position_ids, past_key_values, inputs_embeds, use_cache, output_attentions, output_hidden_states, return_dict)\u001b[0m\n\u001b[1;32m   1029\u001b[0m     layer_outputs \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_gradient_checkpointing_func(\n\u001b[1;32m   1030\u001b[0m         decoder_layer\u001b[38;5;241m.\u001b[39m\u001b[38;5;21m__call__\u001b[39m,\n\u001b[1;32m   1031\u001b[0m         hidden_states,\n\u001b[0;32m   (...)\u001b[0m\n\u001b[1;32m   1036\u001b[0m         use_cache,\n\u001b[1;32m   1037\u001b[0m     )\n\u001b[1;32m   1038\u001b[0m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[0;32m-> 1039\u001b[0m     layer_outputs \u001b[38;5;241m=\u001b[39m \u001b[43mdecoder_layer\u001b[49m\u001b[43m(\u001b[49m\n\u001b[1;32m   1040\u001b[0m \u001b[43m        \u001b[49m\u001b[43mhidden_states\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m   1041\u001b[0m \u001b[43m        \u001b[49m\u001b[43mattention_mask\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mattention_mask\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m   1042\u001b[0m \u001b[43m        \u001b[49m\u001b[43mposition_ids\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mposition_ids\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m   1043\u001b[0m \u001b[43m        \u001b[49m\u001b[43mpast_key_value\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mpast_key_values\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m   1044\u001b[0m \u001b[43m        \u001b[49m\u001b[43moutput_attentions\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43moutput_attentions\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m   1045\u001b[0m \u001b[43m        \u001b[49m\u001b[43muse_cache\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43muse_cache\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m   1046\u001b[0m \u001b[43m    \u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m   1048\u001b[0m hidden_states \u001b[38;5;241m=\u001b[39m layer_outputs[\u001b[38;5;241m0\u001b[39m]\n\u001b[1;32m   1050\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m use_cache:\n",
      "File \u001b[0;32m~/dialogue-augmentation/.venv/lib/python3.8/site-packages/torch/nn/modules/module.py:1511\u001b[0m, in \u001b[0;36mModule._wrapped_call_impl\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m   1509\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_compiled_call_impl(\u001b[38;5;241m*\u001b[39margs, \u001b[38;5;241m*\u001b[39m\u001b[38;5;241m*\u001b[39mkwargs)  \u001b[38;5;66;03m# type: ignore[misc]\u001b[39;00m\n\u001b[1;32m   1510\u001b[0m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[0;32m-> 1511\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43m_call_impl\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n",
      "File \u001b[0;32m~/dialogue-augmentation/.venv/lib/python3.8/site-packages/torch/nn/modules/module.py:1520\u001b[0m, in \u001b[0;36mModule._call_impl\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m   1515\u001b[0m \u001b[38;5;66;03m# If we don't have any hooks, we want to skip the rest of the logic in\u001b[39;00m\n\u001b[1;32m   1516\u001b[0m \u001b[38;5;66;03m# this function, and just call forward.\u001b[39;00m\n\u001b[1;32m   1517\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m (\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_backward_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_backward_pre_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_forward_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_forward_pre_hooks\n\u001b[1;32m   1518\u001b[0m         \u001b[38;5;129;01mor\u001b[39;00m _global_backward_pre_hooks \u001b[38;5;129;01mor\u001b[39;00m _global_backward_hooks\n\u001b[1;32m   1519\u001b[0m         \u001b[38;5;129;01mor\u001b[39;00m _global_forward_hooks \u001b[38;5;129;01mor\u001b[39;00m _global_forward_pre_hooks):\n\u001b[0;32m-> 1520\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mforward_call\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m   1522\u001b[0m \u001b[38;5;28;01mtry\u001b[39;00m:\n\u001b[1;32m   1523\u001b[0m     result \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;01mNone\u001b[39;00m\n",
      "File \u001b[0;32m~/dialogue-augmentation/.venv/lib/python3.8/site-packages/transformers/models/mistral/modeling_mistral.py:754\u001b[0m, in \u001b[0;36mMistralDecoderLayer.forward\u001b[0;34m(self, hidden_states, attention_mask, position_ids, past_key_value, output_attentions, use_cache, **kwargs)\u001b[0m\n\u001b[1;32m    751\u001b[0m hidden_states \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39minput_layernorm(hidden_states)\n\u001b[1;32m    753\u001b[0m \u001b[38;5;66;03m# Self Attention\u001b[39;00m\n\u001b[0;32m--> 754\u001b[0m hidden_states, self_attn_weights, present_key_value \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mself_attn\u001b[49m\u001b[43m(\u001b[49m\n\u001b[1;32m    755\u001b[0m \u001b[43m    \u001b[49m\u001b[43mhidden_states\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mhidden_states\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    756\u001b[0m \u001b[43m    \u001b[49m\u001b[43mattention_mask\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mattention_mask\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    757\u001b[0m \u001b[43m    \u001b[49m\u001b[43mposition_ids\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mposition_ids\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    758\u001b[0m \u001b[43m    \u001b[49m\u001b[43mpast_key_value\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mpast_key_value\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    759\u001b[0m \u001b[43m    \u001b[49m\u001b[43moutput_attentions\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43moutput_attentions\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    760\u001b[0m \u001b[43m    \u001b[49m\u001b[43muse_cache\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43muse_cache\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    761\u001b[0m \u001b[43m\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    762\u001b[0m hidden_states \u001b[38;5;241m=\u001b[39m residual \u001b[38;5;241m+\u001b[39m hidden_states\n\u001b[1;32m    764\u001b[0m \u001b[38;5;66;03m# Fully Connected\u001b[39;00m\n",
      "File \u001b[0;32m~/dialogue-augmentation/.venv/lib/python3.8/site-packages/torch/nn/modules/module.py:1511\u001b[0m, in \u001b[0;36mModule._wrapped_call_impl\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m   1509\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_compiled_call_impl(\u001b[38;5;241m*\u001b[39margs, \u001b[38;5;241m*\u001b[39m\u001b[38;5;241m*\u001b[39mkwargs)  \u001b[38;5;66;03m# type: ignore[misc]\u001b[39;00m\n\u001b[1;32m   1510\u001b[0m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[0;32m-> 1511\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43m_call_impl\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n",
      "File \u001b[0;32m~/dialogue-augmentation/.venv/lib/python3.8/site-packages/torch/nn/modules/module.py:1520\u001b[0m, in \u001b[0;36mModule._call_impl\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m   1515\u001b[0m \u001b[38;5;66;03m# If we don't have any hooks, we want to skip the rest of the logic in\u001b[39;00m\n\u001b[1;32m   1516\u001b[0m \u001b[38;5;66;03m# this function, and just call forward.\u001b[39;00m\n\u001b[1;32m   1517\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m (\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_backward_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_backward_pre_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_forward_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_forward_pre_hooks\n\u001b[1;32m   1518\u001b[0m         \u001b[38;5;129;01mor\u001b[39;00m _global_backward_pre_hooks \u001b[38;5;129;01mor\u001b[39;00m _global_backward_hooks\n\u001b[1;32m   1519\u001b[0m         \u001b[38;5;129;01mor\u001b[39;00m _global_forward_hooks \u001b[38;5;129;01mor\u001b[39;00m _global_forward_pre_hooks):\n\u001b[0;32m-> 1520\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mforward_call\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m   1522\u001b[0m \u001b[38;5;28;01mtry\u001b[39;00m:\n\u001b[1;32m   1523\u001b[0m     result \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;01mNone\u001b[39;00m\n",
      "File \u001b[0;32m~/dialogue-augmentation/.venv/lib/python3.8/site-packages/transformers/models/mistral/modeling_mistral.py:663\u001b[0m, in \u001b[0;36mMistralSdpaAttention.forward\u001b[0;34m(self, hidden_states, attention_mask, position_ids, past_key_value, output_attentions, use_cache)\u001b[0m\n\u001b[1;32m    660\u001b[0m     kv_seq_len \u001b[38;5;241m+\u001b[39m\u001b[38;5;241m=\u001b[39m past_key_value\u001b[38;5;241m.\u001b[39mget_usable_length(kv_seq_len, \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mlayer_idx)\n\u001b[1;32m    661\u001b[0m cos, sin \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mrotary_emb(value_states, seq_len\u001b[38;5;241m=\u001b[39mkv_seq_len)\n\u001b[0;32m--> 663\u001b[0m query_states, key_states \u001b[38;5;241m=\u001b[39m \u001b[43mapply_rotary_pos_emb\u001b[49m\u001b[43m(\u001b[49m\u001b[43mquery_states\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mkey_states\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mcos\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43msin\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mposition_ids\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    665\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m past_key_value \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m:\n\u001b[1;32m    666\u001b[0m     cache_kwargs \u001b[38;5;241m=\u001b[39m {\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124msin\u001b[39m\u001b[38;5;124m\"\u001b[39m: sin, \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mcos\u001b[39m\u001b[38;5;124m\"\u001b[39m: cos}  \u001b[38;5;66;03m# Specific to RoPE models\u001b[39;00m\n",
      "File \u001b[0;32m~/dialogue-augmentation/.venv/lib/python3.8/site-packages/transformers/models/mistral/modeling_mistral.py:160\u001b[0m, in \u001b[0;36mapply_rotary_pos_emb\u001b[0;34m(q, k, cos, sin, position_ids, unsqueeze_dim)\u001b[0m\n\u001b[1;32m    158\u001b[0m cos \u001b[38;5;241m=\u001b[39m cos[position_ids]\u001b[38;5;241m.\u001b[39munsqueeze(unsqueeze_dim)\n\u001b[1;32m    159\u001b[0m sin \u001b[38;5;241m=\u001b[39m sin[position_ids]\u001b[38;5;241m.\u001b[39munsqueeze(unsqueeze_dim)\n\u001b[0;32m--> 160\u001b[0m q_embed \u001b[38;5;241m=\u001b[39m (q \u001b[38;5;241m*\u001b[39m cos) \u001b[38;5;241m+\u001b[39m (\u001b[43mrotate_half\u001b[49m\u001b[43m(\u001b[49m\u001b[43mq\u001b[49m\u001b[43m)\u001b[49m \u001b[38;5;241m*\u001b[39m sin)\n\u001b[1;32m    161\u001b[0m k_embed \u001b[38;5;241m=\u001b[39m (k \u001b[38;5;241m*\u001b[39m cos) \u001b[38;5;241m+\u001b[39m (rotate_half(k) \u001b[38;5;241m*\u001b[39m sin)\n\u001b[1;32m    162\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m q_embed, k_embed\n",
      "File \u001b[0;32m~/dialogue-augmentation/.venv/lib/python3.8/site-packages/transformers/models/mistral/modeling_mistral.py:133\u001b[0m, in \u001b[0;36mrotate_half\u001b[0;34m(x)\u001b[0m\n\u001b[1;32m    131\u001b[0m x1 \u001b[38;5;241m=\u001b[39m x[\u001b[38;5;241m.\u001b[39m\u001b[38;5;241m.\u001b[39m\u001b[38;5;241m.\u001b[39m, : x\u001b[38;5;241m.\u001b[39mshape[\u001b[38;5;241m-\u001b[39m\u001b[38;5;241m1\u001b[39m] \u001b[38;5;241m/\u001b[39m\u001b[38;5;241m/\u001b[39m \u001b[38;5;241m2\u001b[39m]\n\u001b[1;32m    132\u001b[0m x2 \u001b[38;5;241m=\u001b[39m x[\u001b[38;5;241m.\u001b[39m\u001b[38;5;241m.\u001b[39m\u001b[38;5;241m.\u001b[39m, x\u001b[38;5;241m.\u001b[39mshape[\u001b[38;5;241m-\u001b[39m\u001b[38;5;241m1\u001b[39m] \u001b[38;5;241m/\u001b[39m\u001b[38;5;241m/\u001b[39m \u001b[38;5;241m2\u001b[39m :]\n\u001b[0;32m--> 133\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m torch\u001b[38;5;241m.\u001b[39mcat((\u001b[38;5;241;43m-\u001b[39;49m\u001b[43mx2\u001b[49m, x1), dim\u001b[38;5;241m=\u001b[39m\u001b[38;5;241m-\u001b[39m\u001b[38;5;241m1\u001b[39m)\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "\n",
    "# from argparse import ArgumentParser\n",
    "\n",
    "# ap = ArgumentParser()\n",
    "# ap.add_argument('--seed', default=0, type=int)\n",
    "# ap.add_argument('--path-in', required=True)\n",
    "# ap.add_argument('--path-out', required=True)\n",
    "# ap.add_argument('--model', required=True)\n",
    "# ap.add_argument('--pooling', choices=['cls', 'avg', 'last'], required=True)\n",
    "# ap.add_argument('--batch-size', required=True, type=int)\n",
    "# ap.add_argument('--cuda', required=True)\n",
    "# args = ap.parse_args()\n",
    "\n",
    "from dataclasses import dataclass\n",
    "\n",
    "@dataclass\n",
    "class Args:\n",
    "    seed=0\n",
    "    path_in='data/benchmarks-bert/sgd/train'\n",
    "    path_out='data/scatter-analysis/sfrtest-embedded/sgd'\n",
    "    model='Salesforce/SFR-Embedding-Mistral'\n",
    "    pooling='last'\n",
    "    batch_size=16\n",
    "    cuda=\"0\"\n",
    "\n",
    "args = Args()\n",
    "\n",
    "\n",
    "from mylib.utils.training import init_environment\n",
    "init_environment(args)\n",
    "\n",
    "# load model\n",
    "from mylib.modeling.dialogue import BaselineDialogueEncoder, BaselineDialogueEncoderConfig\n",
    "\n",
    "config = BaselineDialogueEncoderConfig(\n",
    "    hf_model=args.model,\n",
    "    pooling=args.pooling,\n",
    "    truncation=True,\n",
    "    max_length=4096 if args.pooling == 'last' else 512\n",
    ")\n",
    "if args.pooling == 'last':\n",
    "    # LLM\n",
    "    kwargs = dict()\n",
    "else:\n",
    "    # BERT-like\n",
    "    kwargs = {}\n",
    "encoder = BaselineDialogueEncoder(config, **kwargs).cuda().eval()\n",
    "\n",
    "# load data\n",
    "from datasets import load_from_disk, disable_caching\n",
    "disable_caching()\n",
    "\n",
    "dataset = load_from_disk(args.path_in)\n",
    "\n",
    "# feed to model and collect embeddings\n",
    "from torch import no_grad\n",
    "\n",
    "with no_grad():\n",
    "    dataset = dataset.map(\n",
    "        lambda batch: {'embedding': encoder(batch).detach().cpu().numpy()},\n",
    "        input_columns='content',\n",
    "        batched=True,\n",
    "        batch_size=args.batch_size\n",
    "    )\n",
    "\n",
    "# save to disk\n",
    "dataset.save_to_disk(args.path_out)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "batch_dict = batch_dict.to('cuda')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "542dd1f3d7b6433c9a5f24470200bb67",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Loading checkpoint shards:   0%|          | 0/3 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "model = AutoModel.from_pretrained(\n",
    "    \"Salesforce/SFR-Embedding-Mistral\",\n",
    "    device_map='auto',\n",
    "    load_in_8bit=True\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[86.5625, 36.75], [34.84375, 82.0625]]\n"
     ]
    }
   ],
   "source": [
    "\n",
    "outputs = model(**batch_dict)\n",
    "embeddings = last_token_pool(outputs.last_hidden_state, batch_dict[\"attention_mask\"])\n",
    "\n",
    "# normalize embeddings\n",
    "embeddings = F.normalize(embeddings, p=2, dim=1)\n",
    "scores = (embeddings[:2] @ embeddings[2:].T) * 100\n",
    "print(scores.tolist())\n",
    "# [[86.7153549194336, 36.64569091796875], [35.00493621826172, 82.0738525390625]]\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "VENV",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}