Reduce arg combinations of sgd/adam/adamw unittest to prevent timeout #9886

marigoold · 2023-02-22T02:35:22Z

sgd / adam / adamw 的单测中参数组合爆炸导致 ci 容易超时，这里对部分参数进行了采样来减少组合数。

修改之后，slowest top 50 已经没有这几个单测了。

github-actions · 2023-02-22T02:39:31Z

Code got formatted by CI. Please request CI again if you still want to have this PR merged. If the PR is from a forked repo, please download the patch files from the GitHub Actions web page and apply them locally.

github-actions · 2023-02-22T02:40:42Z

Speed stats:

daquexian · 2023-02-22T02:45:15Z

python/oneflow/test/modules/test_optim_sgd.py

        arg_dict["tensor_num"] = [1, 4]
        for arg in GenArgDict(arg_dict):
            compare_with_numpy_sgd(test_case, **arg)

+    @autotest(n=10)


这个是不是也不用加，遍历 arg_dict 本身就循环快 300 次了

这个是不是也不用加，遍历 arg_dict 本身就循环快 300 次了

这个我也比较纠结；这里加上是我考虑因为参数里面本身就有随机量，多跑几次能多取一些值，覆盖得稍微全面一些

感觉 300 次足够了，很多参数是正交的，不必把每种组合都跑到

感觉 300 次足够了，很多参数是正交的，不必把每种组合都跑到

ok，那我这里去掉

github-actions · 2023-02-22T02:57:47Z

Speed stats:

GPU Name: GeForce GTX 1080 

❌ OneFlow resnet50 time: 141.0ms (= 14098.3ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 143.3ms (= 14327.4ms / 100, input_shape=[16, 3, 224, 224])
❌ Relative speed: 1.02 (= 143.3ms / 141.0ms)

OneFlow resnet50 time: 80.8ms (= 8077.4ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 86.1ms (= 8608.7ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.07 (= 86.1ms / 80.8ms)

OneFlow resnet50 time: 50.0ms (= 9995.8ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 57.3ms (= 11454.0ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.15 (= 57.3ms / 50.0ms)

OneFlow resnet50 time: 33.4ms (= 6680.8ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 46.5ms (= 9291.2ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.39 (= 46.5ms / 33.4ms)

OneFlow resnet50 time: 25.5ms (= 5099.3ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 42.6ms (= 8529.5ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.67 (= 42.6ms / 25.5ms)

OneFlow swin dataloader time: 0.235s (= 47.095s / 200, num_workers=1)
PyTorch swin dataloader time: 0.150s (= 29.980s / 200, num_workers=1)
Relative speed: 0.637 (= 0.150s / 0.235s)

OneFlow swin dataloader time: 0.064s (= 12.766s / 200, num_workers=4)
PyTorch swin dataloader time: 0.040s (= 8.071s / 200, num_workers=4)
Relative speed: 0.632 (= 0.040s / 0.064s)

OneFlow swin dataloader time: 0.036s (= 7.147s / 200, num_workers=8)
PyTorch swin dataloader time: 0.023s (= 4.538s / 200, num_workers=8)
Relative speed: 0.635 (= 0.023s / 0.036s)

❌ OneFlow resnet50 time: 152.7ms (= 15272.5ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 162.7ms (= 16266.5ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.07 (= 162.7ms / 152.7ms)

OneFlow resnet50 time: 91.9ms (= 9192.6ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 103.4ms (= 10335.9ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.12 (= 103.4ms / 91.9ms)

OneFlow resnet50 time: 60.3ms (= 12063.7ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 78.7ms (= 15743.0ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.30 (= 78.7ms / 60.3ms)

OneFlow resnet50 time: 42.2ms (= 8445.0ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 72.4ms (= 14487.0ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.72 (= 72.4ms / 42.2ms)

OneFlow resnet50 time: 36.1ms (= 7220.9ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 67.5ms (= 13496.8ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.87 (= 67.5ms / 36.1ms)

github-actions · 2023-02-22T03:01:57Z

CI failed when running job: cuda-module. PR label automerge has been removed

github-actions · 2023-02-22T03:02:06Z

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9886/

…adam unittest

…om/Oneflow-Inc/oneflow into fix_reduce_arglist_of_sgd_unittest

github-actions · 2023-02-22T03:10:27Z

Code got formatted by CI. Please request CI again if you still want to have this PR merged. If the PR is from a forked repo, please download the patch files from the GitHub Actions web page and apply them locally.

…adam unittest

…om/Oneflow-Inc/oneflow into fix_reduce_arglist_of_sgd_unittest

github-actions · 2023-02-22T03:14:02Z

Speed stats:

github-actions · 2023-02-22T03:17:01Z

Code got formatted by CI. Please request CI again if you still want to have this PR merged. If the PR is from a forked repo, please download the patch files from the GitHub Actions web page and apply them locally.

github-actions · 2023-02-22T03:17:46Z

Speed stats:

github-actions · 2023-02-22T03:34:31Z

Speed stats:

GPU Name: GeForce GTX 1080 

❌ OneFlow resnet50 time: 141.1ms (= 14106.7ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 142.2ms (= 14216.6ms / 100, input_shape=[16, 3, 224, 224])
❌ Relative speed: 1.01 (= 142.2ms / 141.1ms)

OneFlow resnet50 time: 81.2ms (= 8118.3ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 84.2ms (= 8416.1ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.04 (= 84.2ms / 81.2ms)

OneFlow resnet50 time: 50.8ms (= 10151.1ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 55.4ms (= 11083.9ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.09 (= 55.4ms / 50.8ms)

OneFlow resnet50 time: 33.6ms (= 6711.7ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 42.3ms (= 8451.4ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.26 (= 42.3ms / 33.6ms)

OneFlow resnet50 time: 24.9ms (= 4983.9ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 36.3ms (= 7250.7ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.45 (= 36.3ms / 24.9ms)

OneFlow swin dataloader time: 0.235s (= 46.925s / 200, num_workers=1)
PyTorch swin dataloader time: 0.149s (= 29.809s / 200, num_workers=1)
Relative speed: 0.635 (= 0.149s / 0.235s)

OneFlow swin dataloader time: 0.063s (= 12.673s / 200, num_workers=4)
PyTorch swin dataloader time: 0.041s (= 8.246s / 200, num_workers=4)
Relative speed: 0.651 (= 0.041s / 0.063s)

OneFlow swin dataloader time: 0.036s (= 7.236s / 200, num_workers=8)
PyTorch swin dataloader time: 0.022s (= 4.420s / 200, num_workers=8)
Relative speed: 0.611 (= 0.022s / 0.036s)

❌ OneFlow resnet50 time: 152.8ms (= 15277.9ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 162.1ms (= 16211.1ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.06 (= 162.1ms / 152.8ms)

OneFlow resnet50 time: 92.4ms (= 9239.9ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 104.1ms (= 10411.5ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.13 (= 104.1ms / 92.4ms)

OneFlow resnet50 time: 60.0ms (= 11996.9ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 79.0ms (= 15809.9ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.32 (= 79.0ms / 60.0ms)

OneFlow resnet50 time: 42.0ms (= 8398.8ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 71.2ms (= 14245.1ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.70 (= 71.2ms / 42.0ms)

OneFlow resnet50 time: 36.4ms (= 7286.1ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 69.9ms (= 13979.3ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.92 (= 69.9ms / 36.4ms)

github-actions · 2023-02-22T03:45:37Z

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9886/

github-actions · 2023-02-22T04:53:56Z

Speed stats:

GPU Name: GeForce GTX 1080 

❌ OneFlow resnet50 time: 141.1ms (= 14112.3ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 146.2ms (= 14620.1ms / 100, input_shape=[16, 3, 224, 224])
❌ Relative speed: 1.04 (= 146.2ms / 141.1ms)

OneFlow resnet50 time: 81.7ms (= 8167.0ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 88.1ms (= 8809.9ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.08 (= 88.1ms / 81.7ms)

OneFlow resnet50 time: 50.7ms (= 10135.3ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 56.7ms (= 11349.0ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.12 (= 56.7ms / 50.7ms)

OneFlow resnet50 time: 34.1ms (= 6827.8ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 44.6ms (= 8921.1ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.31 (= 44.6ms / 34.1ms)

OneFlow resnet50 time: 25.0ms (= 5000.3ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 41.5ms (= 8290.1ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.66 (= 41.5ms / 25.0ms)

OneFlow swin dataloader time: 0.239s (= 47.713s / 200, num_workers=1)
PyTorch swin dataloader time: 0.151s (= 30.139s / 200, num_workers=1)
Relative speed: 0.632 (= 0.151s / 0.239s)

OneFlow swin dataloader time: 0.068s (= 13.564s / 200, num_workers=4)
PyTorch swin dataloader time: 0.042s (= 8.311s / 200, num_workers=4)
Relative speed: 0.613 (= 0.042s / 0.068s)

OneFlow swin dataloader time: 0.038s (= 7.543s / 200, num_workers=8)
PyTorch swin dataloader time: 0.023s (= 4.500s / 200, num_workers=8)
Relative speed: 0.597 (= 0.023s / 0.038s)

❌ OneFlow resnet50 time: 152.7ms (= 15272.6ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 163.4ms (= 16335.4ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.07 (= 163.4ms / 152.7ms)

OneFlow resnet50 time: 92.7ms (= 9266.4ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 108.3ms (= 10827.2ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.17 (= 108.3ms / 92.7ms)

OneFlow resnet50 time: 60.6ms (= 12128.6ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 80.7ms (= 16143.2ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.33 (= 80.7ms / 60.6ms)

OneFlow resnet50 time: 42.8ms (= 8567.3ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 71.8ms (= 14354.5ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.68 (= 71.8ms / 42.8ms)

OneFlow resnet50 time: 35.6ms (= 7119.2ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 67.6ms (= 13512.4ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.90 (= 67.6ms / 35.6ms)

github-actions · 2023-02-22T05:55:20Z

Speed stats:

GPU Name: GeForce GTX 1080 

❌ OneFlow resnet50 time: 141.2ms (= 14119.1ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 141.2ms (= 14120.4ms / 100, input_shape=[16, 3, 224, 224])
❌ Relative speed: 1.00 (= 141.2ms / 141.2ms)

OneFlow resnet50 time: 82.1ms (= 8205.7ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 88.2ms (= 8825.0ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.08 (= 88.2ms / 82.1ms)

OneFlow resnet50 time: 51.1ms (= 10222.2ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 57.5ms (= 11498.8ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.12 (= 57.5ms / 51.1ms)

OneFlow resnet50 time: 33.5ms (= 6697.8ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 43.4ms (= 8683.9ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.30 (= 43.4ms / 33.5ms)

OneFlow resnet50 time: 25.6ms (= 5120.7ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 42.6ms (= 8529.0ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.67 (= 42.6ms / 25.6ms)

OneFlow swin dataloader time: 0.241s (= 48.253s / 200, num_workers=1)
PyTorch swin dataloader time: 0.151s (= 30.107s / 200, num_workers=1)
Relative speed: 0.624 (= 0.151s / 0.241s)

OneFlow swin dataloader time: 0.068s (= 13.593s / 200, num_workers=4)
PyTorch swin dataloader time: 0.042s (= 8.408s / 200, num_workers=4)
Relative speed: 0.619 (= 0.042s / 0.068s)

OneFlow swin dataloader time: 0.039s (= 7.885s / 200, num_workers=8)
PyTorch swin dataloader time: 0.022s (= 4.302s / 200, num_workers=8)
Relative speed: 0.546 (= 0.022s / 0.039s)

❌ OneFlow resnet50 time: 153.4ms (= 15341.8ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 163.6ms (= 16357.1ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.07 (= 163.6ms / 153.4ms)

OneFlow resnet50 time: 92.9ms (= 9294.7ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 103.3ms (= 10330.0ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.11 (= 103.3ms / 92.9ms)

OneFlow resnet50 time: 60.3ms (= 12066.7ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 78.7ms (= 15735.0ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.30 (= 78.7ms / 60.3ms)

OneFlow resnet50 time: 42.8ms (= 8551.8ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 79.0ms (= 15796.5ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.85 (= 79.0ms / 42.8ms)

OneFlow resnet50 time: 37.0ms (= 7409.0ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 69.4ms (= 13882.8ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.87 (= 69.4ms / 37.0ms)

github-actions · 2023-02-22T06:27:53Z

Speed stats:

GPU Name: GeForce GTX 1080 

❌ OneFlow resnet50 time: 141.3ms (= 14125.4ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 143.3ms (= 14328.9ms / 100, input_shape=[16, 3, 224, 224])
❌ Relative speed: 1.01 (= 143.3ms / 141.3ms)

OneFlow resnet50 time: 82.1ms (= 8214.1ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 88.9ms (= 8888.6ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.08 (= 88.9ms / 82.1ms)

OneFlow resnet50 time: 51.0ms (= 10192.5ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 57.1ms (= 11428.0ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.12 (= 57.1ms / 51.0ms)

OneFlow resnet50 time: 33.5ms (= 6709.8ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 43.2ms (= 8648.2ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.29 (= 43.2ms / 33.5ms)

OneFlow resnet50 time: 25.9ms (= 5177.0ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 41.8ms (= 8364.9ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.62 (= 41.8ms / 25.9ms)

OneFlow swin dataloader time: 0.237s (= 47.392s / 200, num_workers=1)
PyTorch swin dataloader time: 0.149s (= 29.781s / 200, num_workers=1)
Relative speed: 0.628 (= 0.149s / 0.237s)

OneFlow swin dataloader time: 0.068s (= 13.590s / 200, num_workers=4)
PyTorch swin dataloader time: 0.044s (= 8.836s / 200, num_workers=4)
Relative speed: 0.650 (= 0.044s / 0.068s)

OneFlow swin dataloader time: 0.040s (= 7.955s / 200, num_workers=8)
PyTorch swin dataloader time: 0.022s (= 4.329s / 200, num_workers=8)
Relative speed: 0.544 (= 0.022s / 0.040s)

❌ OneFlow resnet50 time: 152.8ms (= 15283.8ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 164.1ms (= 16413.3ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.07 (= 164.1ms / 152.8ms)

OneFlow resnet50 time: 92.8ms (= 9284.6ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 103.4ms (= 10343.6ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.11 (= 103.4ms / 92.8ms)

OneFlow resnet50 time: 60.6ms (= 12113.1ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 79.1ms (= 15812.4ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.31 (= 79.1ms / 60.6ms)

OneFlow resnet50 time: 42.6ms (= 8512.4ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 71.1ms (= 14217.1ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.67 (= 71.1ms / 42.6ms)

OneFlow resnet50 time: 36.6ms (= 7326.2ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 70.2ms (= 14034.2ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.92 (= 70.2ms / 36.6ms)

github-actions · 2023-02-22T06:28:19Z

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9886/

github-actions · 2023-02-22T06:47:15Z

Speed stats:

github-actions · 2023-02-22T07:04:23Z

Speed stats:

GPU Name: GeForce GTX 1080 

❌ OneFlow resnet50 time: 141.4ms (= 14135.4ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 146.0ms (= 14597.6ms / 100, input_shape=[16, 3, 224, 224])
❌ Relative speed: 1.03 (= 146.0ms / 141.4ms)

OneFlow resnet50 time: 82.8ms (= 8282.7ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 89.1ms (= 8909.4ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.08 (= 89.1ms / 82.8ms)

OneFlow resnet50 time: 51.1ms (= 10217.7ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 60.5ms (= 12092.4ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.18 (= 60.5ms / 51.1ms)

OneFlow resnet50 time: 34.6ms (= 6928.2ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 47.1ms (= 9410.7ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.36 (= 47.1ms / 34.6ms)

OneFlow resnet50 time: 25.5ms (= 5102.9ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 37.6ms (= 7529.7ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.48 (= 37.6ms / 25.5ms)

OneFlow swin dataloader time: 0.240s (= 48.075s / 200, num_workers=1)
PyTorch swin dataloader time: 0.149s (= 29.750s / 200, num_workers=1)
Relative speed: 0.619 (= 0.149s / 0.240s)

OneFlow swin dataloader time: 0.067s (= 13.351s / 200, num_workers=4)
PyTorch swin dataloader time: 0.040s (= 8.028s / 200, num_workers=4)
Relative speed: 0.601 (= 0.040s / 0.067s)

OneFlow swin dataloader time: 0.041s (= 8.127s / 200, num_workers=8)
PyTorch swin dataloader time: 0.023s (= 4.543s / 200, num_workers=8)
Relative speed: 0.559 (= 0.023s / 0.041s)

❌ OneFlow resnet50 time: 153.6ms (= 15355.7ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 164.2ms (= 16424.1ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.07 (= 164.2ms / 153.6ms)

OneFlow resnet50 time: 93.4ms (= 9335.9ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 103.8ms (= 10384.1ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.11 (= 103.8ms / 93.4ms)

OneFlow resnet50 time: 61.7ms (= 12346.2ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 79.1ms (= 15825.0ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.28 (= 79.1ms / 61.7ms)

OneFlow resnet50 time: 43.1ms (= 8620.4ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 69.9ms (= 13985.0ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.62 (= 69.9ms / 43.1ms)

OneFlow resnet50 time: 36.2ms (= 7231.1ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 67.9ms (= 13576.2ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.88 (= 67.9ms / 36.2ms)

github-actions · 2023-02-22T07:49:39Z

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9886/

github-actions · 2023-02-22T08:31:11Z

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9886/

replace some args by random_bool / random_device / random.sample

b819d9c

marigoold requested review from BBuf and daquexian as code owners February 22, 2023 02:35

marigoold requested a review from oneflow-ci-bot February 22, 2023 02:37

marigoold added enhancement automerge test labels Feb 22, 2023

auto format by CI

2365e68

marigoold requested review from oneflow-ci-bot and removed request for oneflow-ci-bot February 22, 2023 02:40

daquexian reviewed Feb 22, 2023

View reviewed changes

github-actions bot removed the automerge label Feb 22, 2023

replace some args by random_bool / random_device / random.sample for …

269032a

…adam unittest

marigoold changed the title ~~Reduce arg combinations of sgd unittest to prevent timeout~~ Reduce arg combinations of sgd/adam/adamw unittest to prevent timeout Feb 22, 2023

marigoold and others added 2 commits February 22, 2023 11:07

Merge branch 'fix_reduce_arglist_of_sgd_unittest' of https://github.c…

04500c8

…om/Oneflow-Inc/oneflow into fix_reduce_arglist_of_sgd_unittest

auto format by CI

7ecefa0

marigoold added 2 commits February 22, 2023 11:12

replace some args by random_bool / random_device / random.sample for …

48454c7

…adam unittest

Merge branch 'fix_reduce_arglist_of_sgd_unittest' of https://github.c…

9250d4f

…om/Oneflow-Inc/oneflow into fix_reduce_arglist_of_sgd_unittest

auto format by CI

234ce16

marigoold requested review from oneflow-ci-bot and removed request for oneflow-ci-bot February 22, 2023 03:17

mosout approved these changes Feb 22, 2023

View reviewed changes

Flowingsun007 approved these changes Feb 22, 2023

View reviewed changes

Merge branch 'master' into fix_reduce_arglist_of_sgd_unittest

4283636

marigoold added the automerge label Feb 22, 2023

marigoold enabled auto-merge (squash) February 22, 2023 04:39

marigoold merged commit 3887640 into master Feb 22, 2023

marigoold deleted the fix_reduce_arglist_of_sgd_unittest branch February 22, 2023 09:11

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Reduce arg combinations of sgd/adam/adamw unittest to prevent timeout #9886

Reduce arg combinations of sgd/adam/adamw unittest to prevent timeout #9886

marigoold commented Feb 22, 2023 •

edited

Loading

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

daquexian Feb 22, 2023

marigoold Feb 22, 2023

daquexian Feb 22, 2023 •

edited

Loading

marigoold Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

Reduce arg combinations of sgd/adam/adamw unittest to prevent timeout #9886

Reduce arg combinations of sgd/adam/adamw unittest to prevent timeout #9886

Conversation

marigoold commented Feb 22, 2023 • edited Loading

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

daquexian Feb 22, 2023

Choose a reason for hiding this comment

marigoold Feb 22, 2023

Choose a reason for hiding this comment

daquexian Feb 22, 2023 • edited Loading

Choose a reason for hiding this comment

marigoold Feb 22, 2023

Choose a reason for hiding this comment

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

github-actions bot commented Feb 22, 2023

marigoold commented Feb 22, 2023 •

edited

Loading

daquexian Feb 22, 2023 •

edited

Loading