Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

improve flow.load for better error message #10138

Merged
merged 28 commits into from
Apr 23, 2023
Merged

improve flow.load for better error message #10138

merged 28 commits into from
Apr 23, 2023

Conversation

daquexian
Copy link
Contributor

Signed-off-by: daquexian <daquexian566@gmail.com>
Signed-off-by: daquexian <daquexian566@gmail.com>
@daquexian daquexian requested a review from BBuf as a code owner April 14, 2023 10:31
@daquexian daquexian changed the title improve flow.load improve flow.load for better error message Apr 14, 2023
@github-actions
Copy link
Contributor

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
Copy link
Contributor

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

@github-actions
Copy link
Contributor

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

@github-actions
Copy link
Contributor

Speed stats:

@Alokia Alokia self-requested a review April 17, 2023 03:31
Signed-off-by: daquexian <daquexian566@gmail.com>
Signed-off-by: daquexian <daquexian566@gmail.com>
@github-actions
Copy link
Contributor

Speed stats:
GPU Name: GeForce GTX 1080 

❌ OneFlow resnet50 time: 141.3ms (= 14125.1ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 146.3ms (= 14634.6ms / 100, input_shape=[16, 3, 224, 224])
❌ Relative speed: 1.04 (= 146.3ms / 141.3ms)

OneFlow resnet50 time: 82.7ms (= 8273.8ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 93.2ms (= 9316.2ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.13 (= 93.2ms / 82.7ms)

OneFlow resnet50 time: 51.2ms (= 10243.8ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 71.4ms (= 14278.7ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.39 (= 71.4ms / 51.2ms)

OneFlow resnet50 time: 34.4ms (= 6876.7ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 65.0ms (= 12997.0ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.89 (= 65.0ms / 34.4ms)

OneFlow resnet50 time: 26.8ms (= 5352.0ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 64.1ms (= 12814.3ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 2.39 (= 64.1ms / 26.8ms)

OneFlow swin dataloader time: 0.240s (= 47.922s / 200, num_workers=1)
PyTorch swin dataloader time: 0.149s (= 29.786s / 200, num_workers=1)
Relative speed: 0.622 (= 0.149s / 0.240s)

OneFlow swin dataloader time: 0.069s (= 13.829s / 200, num_workers=4)
PyTorch swin dataloader time: 0.041s (= 8.139s / 200, num_workers=4)
Relative speed: 0.589 (= 0.041s / 0.069s)

OneFlow swin dataloader time: 0.039s (= 7.884s / 200, num_workers=8)
PyTorch swin dataloader time: 0.023s (= 4.544s / 200, num_workers=8)
Relative speed: 0.576 (= 0.023s / 0.039s)

❌ OneFlow resnet50 time: 153.8ms (= 15381.8ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 165.4ms (= 16544.9ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.08 (= 165.4ms / 153.8ms)

OneFlow resnet50 time: 94.3ms (= 9430.5ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 104.2ms (= 10418.4ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.10 (= 104.2ms / 94.3ms)

OneFlow resnet50 time: 62.0ms (= 12392.2ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 80.9ms (= 16189.8ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.31 (= 80.9ms / 62.0ms)

OneFlow resnet50 time: 44.3ms (= 8857.5ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 71.9ms (= 14385.1ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.62 (= 71.9ms / 44.3ms)

OneFlow resnet50 time: 38.2ms (= 7642.1ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 67.4ms (= 13489.8ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.77 (= 67.4ms / 38.2ms)

@github-actions
Copy link
Contributor

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/10138/

@github-actions
Copy link
Contributor

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

Signed-off-by: daquexian <daquexian566@gmail.com>
@github-actions
Copy link
Contributor

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/10138/

@github-actions
Copy link
Contributor

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

@github-actions
Copy link
Contributor

Speed stats:
GPU Name: GeForce GTX 1080 

❌ OneFlow resnet50 time: 141.3ms (= 14133.2ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 147.1ms (= 14708.3ms / 100, input_shape=[16, 3, 224, 224])
❌ Relative speed: 1.04 (= 147.1ms / 141.3ms)

OneFlow resnet50 time: 82.9ms (= 8285.2ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 93.2ms (= 9323.6ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.13 (= 93.2ms / 82.9ms)

OneFlow resnet50 time: 51.5ms (= 10303.1ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 64.7ms (= 12944.4ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.26 (= 64.7ms / 51.5ms)

OneFlow resnet50 time: 33.8ms (= 6767.8ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 68.2ms (= 13637.6ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 2.02 (= 68.2ms / 33.8ms)

OneFlow resnet50 time: 27.4ms (= 5479.8ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 63.8ms (= 12759.7ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 2.33 (= 63.8ms / 27.4ms)

OneFlow swin dataloader time: 0.243s (= 48.536s / 200, num_workers=1)
PyTorch swin dataloader time: 0.148s (= 29.618s / 200, num_workers=1)
Relative speed: 0.610 (= 0.148s / 0.243s)

OneFlow swin dataloader time: 0.073s (= 14.513s / 200, num_workers=4)
PyTorch swin dataloader time: 0.044s (= 8.747s / 200, num_workers=4)
Relative speed: 0.603 (= 0.044s / 0.073s)

OneFlow swin dataloader time: 0.044s (= 8.809s / 200, num_workers=8)
PyTorch swin dataloader time: 0.023s (= 4.528s / 200, num_workers=8)
Relative speed: 0.514 (= 0.023s / 0.044s)

❌ OneFlow resnet50 time: 154.4ms (= 15439.2ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 165.5ms (= 16551.5ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.07 (= 165.5ms / 154.4ms)

OneFlow resnet50 time: 94.2ms (= 9421.5ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 103.7ms (= 10373.4ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.10 (= 103.7ms / 94.2ms)

OneFlow resnet50 time: 61.8ms (= 12360.4ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 79.6ms (= 15925.9ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.29 (= 79.6ms / 61.8ms)

OneFlow resnet50 time: 44.1ms (= 8816.6ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 67.1ms (= 13411.4ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.52 (= 67.1ms / 44.1ms)

OneFlow resnet50 time: 38.6ms (= 7725.9ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 70.1ms (= 14010.1ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.81 (= 70.1ms / 38.6ms)

@github-actions
Copy link
Contributor

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/10138/

@github-actions
Copy link
Contributor

Static analysis with clang failed. PR label automerge has been removed

@github-actions
Copy link
Contributor

Speed stats:
GPU Name: GeForce GTX 1080 

❌ OneFlow resnet50 time: 141.2ms (= 14122.3ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 146.0ms (= 14603.0ms / 100, input_shape=[16, 3, 224, 224])
❌ Relative speed: 1.03 (= 146.0ms / 141.2ms)

OneFlow resnet50 time: 81.9ms (= 8185.2ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 93.6ms (= 9362.0ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.14 (= 93.6ms / 81.9ms)

OneFlow resnet50 time: 51.6ms (= 10313.0ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 70.6ms (= 14122.2ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.37 (= 70.6ms / 51.6ms)

OneFlow resnet50 time: 34.4ms (= 6881.0ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 65.5ms (= 13100.3ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.90 (= 65.5ms / 34.4ms)

OneFlow resnet50 time: 26.4ms (= 5280.4ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 64.3ms (= 12854.1ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 2.43 (= 64.3ms / 26.4ms)

OneFlow swin dataloader time: 0.236s (= 47.263s / 200, num_workers=1)
PyTorch swin dataloader time: 0.150s (= 29.957s / 200, num_workers=1)
Relative speed: 0.634 (= 0.150s / 0.236s)

OneFlow swin dataloader time: 0.066s (= 13.293s / 200, num_workers=4)
PyTorch swin dataloader time: 0.042s (= 8.301s / 200, num_workers=4)
Relative speed: 0.625 (= 0.042s / 0.066s)

OneFlow swin dataloader time: 0.041s (= 8.206s / 200, num_workers=8)
PyTorch swin dataloader time: 0.022s (= 4.481s / 200, num_workers=8)
Relative speed: 0.546 (= 0.022s / 0.041s)

❌ OneFlow resnet50 time: 153.1ms (= 15306.3ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 165.3ms (= 16528.3ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.08 (= 165.3ms / 153.1ms)

OneFlow resnet50 time: 93.6ms (= 9359.1ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 104.4ms (= 10440.5ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.12 (= 104.4ms / 93.6ms)

OneFlow resnet50 time: 61.7ms (= 12338.2ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 79.7ms (= 15945.6ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.29 (= 79.7ms / 61.7ms)

OneFlow resnet50 time: 43.9ms (= 8782.6ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 72.3ms (= 14463.8ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.65 (= 72.3ms / 43.9ms)

OneFlow resnet50 time: 38.7ms (= 7733.6ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 68.6ms (= 13722.9ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.77 (= 68.6ms / 38.7ms)

@github-actions
Copy link
Contributor

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/10138/

@github-actions
Copy link
Contributor

CI failed when running job: cpu-misc. PR label automerge has been removed

@github-actions
Copy link
Contributor

Speed stats:
GPU Name: NVIDIA GeForce GTX 1080 

❌ OneFlow resnet50 time: 141.4ms (= 14145.0ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 146.3ms (= 14629.2ms / 100, input_shape=[16, 3, 224, 224])
❌ Relative speed: 1.03 (= 146.3ms / 141.4ms)

OneFlow resnet50 time: 82.1ms (= 8212.3ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 93.7ms (= 9371.7ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.14 (= 93.7ms / 82.1ms)

OneFlow resnet50 time: 52.4ms (= 10481.6ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 72.3ms (= 14466.9ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.38 (= 72.3ms / 52.4ms)

OneFlow resnet50 time: 34.2ms (= 6838.8ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 74.3ms (= 14866.1ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 2.17 (= 74.3ms / 34.2ms)

OneFlow resnet50 time: 26.5ms (= 5299.0ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 65.3ms (= 13052.8ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 2.46 (= 65.3ms / 26.5ms)

OneFlow swin dataloader time: 0.247s (= 49.370s / 200, num_workers=1)
PyTorch swin dataloader time: 0.151s (= 30.230s / 200, num_workers=1)
Relative speed: 0.612 (= 0.151s / 0.247s)

OneFlow swin dataloader time: 0.067s (= 13.456s / 200, num_workers=4)
PyTorch swin dataloader time: 0.040s (= 8.060s / 200, num_workers=4)
Relative speed: 0.599 (= 0.040s / 0.067s)

OneFlow swin dataloader time: 0.039s (= 7.853s / 200, num_workers=8)
PyTorch swin dataloader time: 0.022s (= 4.358s / 200, num_workers=8)
Relative speed: 0.555 (= 0.022s / 0.039s)

❌ OneFlow resnet50 time: 164.9ms (= 16491.8ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 174.1ms (= 17413.0ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.06 (= 174.1ms / 164.9ms)

OneFlow resnet50 time: 103.9ms (= 10391.4ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 114.7ms (= 11473.3ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.10 (= 114.7ms / 103.9ms)

OneFlow resnet50 time: 71.4ms (= 14282.0ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 87.2ms (= 17437.8ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.22 (= 87.2ms / 71.4ms)

OneFlow resnet50 time: 59.6ms (= 11925.2ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 77.6ms (= 15528.1ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.30 (= 77.6ms / 59.6ms)

OneFlow resnet50 time: 52.4ms (= 10470.2ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 69.7ms (= 13944.8ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.33 (= 69.7ms / 52.4ms)

@github-actions
Copy link
Contributor

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/10138/

@mergify mergify bot merged commit 57b0d52 into master Apr 23, 2023
@mergify mergify bot deleted the improve_load branch April 23, 2023 15:40
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants