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.

Sign up for GitHub

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

Jump to bottom

disable large tensor in autotest #8177

Merged
merged 42 commits into from
May 14, 2022
Merged

disable large tensor in autotest #8177

merged 42 commits into from
May 14, 2022

Conversation

daquexian
Copy link
Contributor

去掉 autotest 里的 gc.collect()，改为禁止创建过大的 tensor，避免 gc.collect() 使 ci 速度变慢的问题

@daquexian
disable large tensor in autotest
1952317 
Signed-off-by: daquexian <daquexian566@gmail.com>
@daquexian daquexian requested a review from BBuf as a code owner May 10, 2022 01:37
@daquexian daquexian requested a review from oneflow-ci-bot May 10, 2022 01:42
@daquexian
rename
9e3de7f 
Signed-off-by: daquexian <daquexian566@gmail.com>
@jackalcooper jackalcooper enabled auto-merge (squash) May 10, 2022 02:27
@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:

@github-actions
Copy link
Contributor

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

@github-actions
Copy link
Contributor

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/8177/

@github-actions
Copy link
Contributor

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

daquexian added 2 commits May 14, 2022 11:18
@daquexian
fix consistent addcmul
46ad813 
Signed-off-by: daquexian <daquexian566@gmail.com>
@daquexian
Merge branch 'disable_large_tensor_autotest' of github.com:Oneflow-In…
78766c6 
…c/oneflow into disable_large_tensor_autotest
@daquexian
limit 20 -> 32
7bec88b 
Signed-off-by: daquexian <daquexian566@gmail.com>
@github-actions
Copy link
Contributor

Speed stats: 
GPU Name: NVIDIA GeForce GTX 1080 

❌ OneFlow resnet50 time: 129.5ms (= 12946.8ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 142.9ms (= 14286.2ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.10 (= 142.9ms / 129.5ms)

OneFlow resnet50 time: 77.3ms (= 7731.1ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 84.9ms (= 8492.8ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.10 (= 84.9ms / 77.3ms)

OneFlow resnet50 time: 52.0ms (= 10400.3ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 59.1ms (= 11815.1ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.14 (= 59.1ms / 52.0ms)

OneFlow resnet50 time: 44.1ms (= 8821.3ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 41.3ms (= 8258.5ms / 200, input_shape=[2, 3, 224, 224])
❌ Relative speed: 0.94 (= 41.3ms / 44.1ms)

OneFlow resnet50 time: 35.3ms (= 7061.7ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 37.6ms (= 7525.4ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.07 (= 37.6ms / 35.3ms)

OneFlow swin dataloader time: 0.250s (= 49.912s / 200, num_workers=1)
PyTorch swin dataloader time: 0.151s (= 30.161s / 200, num_workers=1)
Relative speed: 0.604 (= 0.151s / 0.250s)

OneFlow swin dataloader time: 0.105s (= 20.926s / 200, num_workers=4)
PyTorch swin dataloader time: 0.041s (= 8.170s / 200, num_workers=4)
Relative speed: 0.390 (= 0.041s / 0.105s)

OneFlow swin dataloader time: 0.034s (= 6.780s / 200, num_workers=8)
PyTorch swin dataloader time: 0.023s (= 4.516s / 200, num_workers=8)
Relative speed: 0.666 (= 0.023s / 0.034s)

❌ OneFlow resnet50 time: 144.7ms (= 14467.7ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 168.5ms (= 16853.7ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.16 (= 168.5ms / 144.7ms)

OneFlow resnet50 time: 102.0ms (= 10198.8ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 120.7ms (= 12066.1ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.18 (= 120.7ms / 102.0ms)

OneFlow resnet50 time: 75.7ms (= 15139.0ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 87.2ms (= 17436.9ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.15 (= 87.2ms / 75.7ms)

OneFlow resnet50 time: 64.2ms (= 12848.4ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 75.5ms (= 15105.8ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.18 (= 75.5ms / 64.2ms)

OneFlow resnet50 time: 55.6ms (= 11125.6ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 68.7ms (= 13730.5ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.23 (= 68.7ms / 55.6ms)

@github-actions
Copy link
Contributor

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

@github-actions
Copy link
Contributor

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

@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-misc. PR label automerge has been removed

jackalcooper and others added 3 commits May 14, 2022 14:51
@jackalcooper
mv file
83e86f5
@jackalcooper
Merge branch 'master' of https://github.com/Oneflow-Inc/oneflow into …
461d18c 
…disable_large_tensor_autotest
@daquexian
fix consistent normalize
e9d4544 
Signed-off-by: daquexian <daquexian566@gmail.com>
@oneflow-ci-bot oneflow-ci-bot requested review from oneflow-ci-bot and removed request for oneflow-ci-bot May 14, 2022 20:35
@github-actions
Copy link
Contributor

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/8177/

@github-actions
Copy link
Contributor

Speed stats: 
GPU Name: GeForce GTX 1080 

✔️ OneFlow resnet50 time: 128.3ms (= 12825.4ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 145.7ms (= 14566.2ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.14 (= 145.7ms / 128.3ms)

OneFlow resnet50 time: 80.9ms (= 8086.6ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 82.4ms (= 8243.4ms / 100, input_shape=[8, 3, 224, 224])
❌ Relative speed: 1.02 (= 82.4ms / 80.9ms)

OneFlow resnet50 time: 51.7ms (= 10339.8ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 52.7ms (= 10547.4ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.02 (= 52.7ms / 51.7ms)

OneFlow resnet50 time: 42.0ms (= 8403.6ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 48.1ms (= 9616.8ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.14 (= 48.1ms / 42.0ms)

OneFlow resnet50 time: 37.3ms (= 7469.6ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 37.2ms (= 7439.8ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.00 (= 37.2ms / 37.3ms)

OneFlow swin dataloader time: 0.243s (= 48.618s / 200, num_workers=1)
PyTorch swin dataloader time: 0.147s (= 29.446s / 200, num_workers=1)
Relative speed: 0.606 (= 0.147s / 0.243s)

OneFlow swin dataloader time: 0.066s (= 13.166s / 200, num_workers=4)
PyTorch swin dataloader time: 0.040s (= 7.992s / 200, num_workers=4)
Relative speed: 0.607 (= 0.040s / 0.066s)

OneFlow swin dataloader time: 0.036s (= 7.175s / 200, num_workers=8)
PyTorch swin dataloader time: 0.022s (= 4.330s / 200, num_workers=8)
Relative speed: 0.603 (= 0.022s / 0.036s)

❌ OneFlow resnet50 time: 138.2ms (= 13823.4ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 159.9ms (= 15993.0ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.16 (= 159.9ms / 138.2ms)

OneFlow resnet50 time: 89.6ms (= 8961.5ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 101.0ms (= 10099.3ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.13 (= 101.0ms / 89.6ms)

OneFlow resnet50 time: 64.7ms (= 12944.4ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 77.5ms (= 15508.0ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.20 (= 77.5ms / 64.7ms)

OneFlow resnet50 time: 52.4ms (= 10488.2ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 68.5ms (= 13704.4ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.31 (= 68.5ms / 52.4ms)

OneFlow resnet50 time: 49.1ms (= 9819.8ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 75.8ms (= 15157.6ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.54 (= 75.8ms / 49.1ms)

@github-actions
Copy link
Contributor

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

@github-actions
Copy link
Contributor

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

@github-actions
Copy link
Contributor

Speed stats: 
GPU Name: NVIDIA GeForce GTX 1080 

❌ OneFlow resnet50 time: 129.3ms (= 12934.9ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 142.8ms (= 14280.3ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.10 (= 142.8ms / 129.3ms)

OneFlow resnet50 time: 81.5ms (= 8153.6ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 84.6ms (= 8461.8ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.04 (= 84.6ms / 81.5ms)

OneFlow resnet50 time: 52.5ms (= 10497.2ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 58.4ms (= 11680.1ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.11 (= 58.4ms / 52.5ms)

OneFlow resnet50 time: 43.4ms (= 8684.3ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 43.1ms (= 8615.8ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 0.99 (= 43.1ms / 43.4ms)

OneFlow resnet50 time: 37.0ms (= 7407.4ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 38.7ms (= 7745.5ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.05 (= 38.7ms / 37.0ms)

OneFlow swin dataloader time: 0.389s (= 77.844s / 200, num_workers=1)
PyTorch swin dataloader time: 0.149s (= 29.857s / 200, num_workers=1)
Relative speed: 0.384 (= 0.149s / 0.389s)

OneFlow swin dataloader time: 0.066s (= 13.112s / 200, num_workers=4)
PyTorch swin dataloader time: 0.040s (= 8.038s / 200, num_workers=4)
Relative speed: 0.613 (= 0.040s / 0.066s)

OneFlow swin dataloader time: 0.036s (= 7.262s / 200, num_workers=8)
PyTorch swin dataloader time: 0.021s (= 4.287s / 200, num_workers=8)
Relative speed: 0.590 (= 0.021s / 0.036s)

❌ OneFlow resnet50 time: 145.8ms (= 14576.1ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 168.7ms (= 16870.2ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.16 (= 168.7ms / 145.8ms)

OneFlow resnet50 time: 97.2ms (= 9724.6ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 112.1ms (= 11211.4ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.15 (= 112.1ms / 97.2ms)

OneFlow resnet50 time: 75.2ms (= 15039.6ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 87.1ms (= 17419.9ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.16 (= 87.1ms / 75.2ms)

OneFlow resnet50 time: 64.2ms (= 12838.8ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 75.8ms (= 15154.5ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.18 (= 75.8ms / 64.2ms)

OneFlow resnet50 time: 57.6ms (= 11516.2ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 75.7ms (= 15133.7ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.31 (= 75.7ms / 57.6ms)

@jackalcooper jackalcooper merged commit 3939e20 into master May 14, 2022
@jackalcooper jackalcooper deleted the disable_large_tensor_autotest branch May 14, 2022 23:04
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@jackalcooper jackalcooper jackalcooper approved these changes

@Flowingsun007 Flowingsun007 Flowingsun007 approved these changes

@BBuf BBuf Awaiting requested review from BBuf

@oneflow-ci-bot oneflow-ci-bot Awaiting requested review from oneflow-ci-bot

Assignees
No one assigned
Labels
automerge ci enhancement need-all-tests-even-fail
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@daquexian @jackalcooper @Flowingsun007 @oneflow-ci-bot