Skip to content

[adaptive_]max_pool{1,2,3}d: handle edge case when input is filled with -inf #40665

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
Closed
Baranowski wants to merge 14 commits into from
Closed

[adaptive_]max_pool{1,2,3}d: handle edge case when input is filled with -inf #40665

Baranowski wants to merge 14 commits into from

Conversation

@Baranowski
Copy link
Contributor

@Baranowski Baranowski commented Jun 27, 2020

Fixes #40131

@dr-ci
Copy link

dr-ci bot commented Jun 27, 2020 •
edited
Loading

💊 CI failures summary and remediations

As of commit ffd89d8 (more details on the Dr. CI page):

💚 💚 Looks good so far! There are no failures yet. 💚 💚

This comment was automatically generated by Dr. CI (expand for details).Follow this link to opt-out of these comments for your Pull Requests.

Please report bugs/suggestions on the GitHub issue tracker or post in the (internal) Dr. CI Users group.

See how this bot performed.

This comment has been revised 47 times.

@Baranowski Baranowski force-pushed the wbaranowski-max_pool_nan-40131 branch 2 times, most recently from 1d51752 to 53dfb79 Compare July 2, 2020 07:41
@Baranowski Baranowski marked this pull request as ready for review July 4, 2020 16:45
@Baranowski Baranowski requested a review from xwang233 July 4, 2020 16:46
@Baranowski
Copy link
Contributor Author

Baranowski commented Jul 4, 2020

@xwang233, I didn't know who would be appropriate so I have marked you as a reviewer.

I will create the XLA issue after the first review.

xwang233
xwang233 reviewed Jul 4, 2020
View reviewed changes
Copy link
Collaborator

@xwang233 xwang233 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for the fix! Overall it looks good. Only the default max index needs to be changed.

@xwang233
Copy link
Collaborator

xwang233 commented Jul 4, 2020

I'm not familiar with the XLA test. If XLA failure is real, we can temporarily disable test on XLA and leave a TODO there. Seems like @onlyOnCPUAndCUDA does the work.

@Baranowski
Copy link
Contributor Author

Baranowski commented Jul 5, 2020

Good call, @xwang233. Fixed

xwang233
xwang233 reviewed Jul 5, 2020
View reviewed changes
@Baranowski
Copy link
Contributor Author

Baranowski commented Jul 9, 2020

Thanks for the quick and patient reviews, @xwang233. I won't have the time to work on this after Sunday next week (July 19th) so it would be brilliant to get this PR merged before then.

@xwang233
Copy link
Collaborator

xwang233 commented Jul 9, 2020

@Baranowski Thanks. I don't work at FB, so I can't merge this for you.

@xwang233 xwang233 requested a review from ezyang July 9, 2020 18:15
@ezyang
Copy link
Contributor

ezyang commented Jul 9, 2020

It would be good to get a performance comparison before and after here.

facebook-github-bot
facebook-github-bot reviewed Jul 9, 2020
View reviewed changes
Copy link
Contributor

@facebook-github-bot facebook-github-bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@ezyang is landing this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

@Baranowski
Copy link
Contributor Author

Baranowski commented Jul 11, 2020

I have added benchmarks in the most recent commit. I don't see significant differences.

Master:

Details 
(pytorch-cuda-dev) wbaranowski@qgpu2:~/git/Quansight/vanilla/benchmarks/operator_benchmark$ python -m pt.pool_test --tag all
# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,1]_stride[2,1]_N1_C16_H32_W32_cpu
# Input: kernel: [3, 1], stride: [2, 1], N: 1, C: 16, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 43.277

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,2]_stride[2,2]_N8_C32_H32_W32_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 8, C: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 74.007

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,2]_stride[2,2]_N8_C32_H32_W64_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 8, C: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 103.433

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,2]_stride[2,2]_N8_C32_H64_W32_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 8, C: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 101.649

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,2]_stride[2,2]_N8_C32_H64_W64_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 8, C: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 158.317

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,2]_stride[2,2]_N16_C32_H32_W32_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 16, C: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 77.478

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,2]_stride[2,2]_N16_C32_H32_W64_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 16, C: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 116.714

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,2]_stride[2,2]_N16_C32_H64_W32_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 16, C: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 116.661

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,2]_stride[2,2]_N16_C32_H64_W64_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 16, C: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 158.889

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,3]_stride[2,2]_N8_C32_H32_W32_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 8, C: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 83.278

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,3]_stride[2,2]_N8_C32_H32_W64_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 8, C: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 126.931

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,3]_stride[2,2]_N8_C32_H64_W32_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 8, C: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 113.920

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,3]_stride[2,2]_N8_C32_H64_W64_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 8, C: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 182.647

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,3]_stride[2,2]_N16_C32_H32_W32_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 16, C: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 98.402

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,3]_stride[2,2]_N16_C32_H32_W64_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 16, C: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 111.891

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,3]_stride[2,2]_N16_C32_H64_W32_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 16, C: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 117.598

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,3]_stride[2,2]_N16_C32_H64_W64_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 16, C: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 241.957

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,1]_stride[2,1]_N1_C16_H32_W32_cpu
# Input: kernel: [3, 1], stride: [2, 1], N: 1, C: 16, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 39.386

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,2]_stride[2,2]_N8_C32_H32_W32_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 8, C: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 54.870

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,2]_stride[2,2]_N8_C32_H32_W64_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 8, C: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 54.467

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,2]_stride[2,2]_N8_C32_H64_W32_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 8, C: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 54.567

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,2]_stride[2,2]_N8_C32_H64_W64_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 8, C: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 54.727

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,2]_stride[2,2]_N16_C32_H32_W32_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 16, C: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 64.878

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,2]_stride[2,2]_N16_C32_H32_W64_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 16, C: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 63.164

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,2]_stride[2,2]_N16_C32_H64_W32_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 16, C: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 63.417

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,2]_stride[2,2]_N16_C32_H64_W64_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 16, C: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 62.819

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,3]_stride[2,2]_N8_C32_H32_W32_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 8, C: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 55.215

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,3]_stride[2,2]_N8_C32_H32_W64_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 8, C: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 54.947

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,3]_stride[2,2]_N8_C32_H64_W32_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 8, C: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 55.018

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,3]_stride[2,2]_N8_C32_H64_W64_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 8, C: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 58.489

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,3]_stride[2,2]_N16_C32_H32_W32_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 16, C: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 67.708

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,3]_stride[2,2]_N16_C32_H32_W64_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 16, C: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 64.613

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,3]_stride[2,2]_N16_C32_H64_W32_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 16, C: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 64.880

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,3]_stride[2,2]_N16_C32_H64_W64_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 16, C: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 65.123

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,1,3]_stride[2,1,2]_N1_C16_D16_H32_W32_cpu
# Input: kernel: [3, 1, 3], stride: [2, 1, 2], N: 1, C: 16, D: 16, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 324.005

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N8_C32_D32_H32_W32_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 2255.940

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N8_C32_D32_H32_W64_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 4465.090

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N8_C32_D32_H64_W32_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 4151.426

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N8_C32_D32_H64_W64_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 10130.852

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N16_C32_D32_H32_W32_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 2403.098

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N16_C32_D32_H32_W64_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 7265.930

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N16_C32_D32_H64_W32_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 6345.975

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N16_C32_D32_H64_W64_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 13842.293

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N8_C32_D32_H32_W32_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 1804.096

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N8_C32_D32_H32_W64_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 4145.099

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N8_C32_D32_H64_W32_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 3979.806

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N8_C32_D32_H64_W64_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 10293.466

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N16_C32_D32_H32_W32_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 2501.046

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N16_C32_D32_H32_W64_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 7369.280

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N16_C32_D32_H64_W32_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 6874.013

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N16_C32_D32_H64_W64_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 14725.793

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,1,3]_stride[2,1,2]_N1_C16_D16_H32_W32_cpu
# Input: kernel: [3, 1, 3], stride: [2, 1, 2], N: 1, C: 16, D: 16, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 39.616

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N8_C32_D32_H32_W32_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 78.150

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N8_C32_D32_H32_W64_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 74.670

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N8_C32_D32_H64_W32_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 71.217

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N8_C32_D32_H64_W64_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 77.309

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N16_C32_D32_H32_W32_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 88.279

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N16_C32_D32_H32_W64_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 85.710

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N16_C32_D32_H64_W32_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 87.579

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N16_C32_D32_H64_W64_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 139.081

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N8_C32_D32_H32_W32_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 74.413

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N8_C32_D32_H32_W64_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 83.784

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N8_C32_D32_H64_W32_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 77.414

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N8_C32_D32_H64_W64_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 100.437

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N16_C32_D32_H32_W32_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 90.312

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N16_C32_D32_H32_W64_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 94.083

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N16_C32_D32_H64_W32_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 93.086

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N16_C32_D32_H64_W64_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 145.713

(pytorch-cuda-dev) wbaranowski@qgpu2:~/git/Quansight/vanilla/benchmarks/operator_benchmark$

Modified version:

Details 
(pytorch-cuda-dev) wbaranowski@qgpu2:~/git/Quansight/pytorch/benchmarks/operator_benchmark$ python -m pt.
pool_test --tag all
# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,1]_stride[2,1]_N1_C16_H32_W32_cpu
# Input: kernel: [3, 1], stride: [2, 1], N: 1, C: 16, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 42.339

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,2]_stride[2,2]_N8_C32_H32_W32_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 8, C: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 74.987

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,2]_stride[2,2]_N8_C32_H32_W64_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 8, C: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 103.705

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,2]_stride[2,2]_N8_C32_H64_W32_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 8, C: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 102.937

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,2]_stride[2,2]_N8_C32_H64_W64_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 8, C: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 178.586

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,2]_stride[2,2]_N16_C32_H32_W32_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 16, C: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 94.048

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,2]_stride[2,2]_N16_C32_H32_W64_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 16, C: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 106.972

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,2]_stride[2,2]_N16_C32_H64_W32_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 16, C: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 112.960

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,2]_stride[2,2]_N16_C32_H64_W64_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 16, C: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 159.002

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,3]_stride[2,2]_N8_C32_H32_W32_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 8, C: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 83.364

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,3]_stride[2,2]_N8_C32_H32_W64_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 8, C: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 110.988

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,3]_stride[2,2]_N8_C32_H64_W32_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 8, C: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 115.071

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,3]_stride[2,2]_N8_C32_H64_W64_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 8, C: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 169.171

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,3]_stride[2,2]_N16_C32_H32_W32_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 16, C: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 83.448

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,3]_stride[2,2]_N16_C32_H32_W64_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 16, C: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 112.358

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,3]_stride[2,2]_N16_C32_H64_W32_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 16, C: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 114.631

# Benchmarking PyTorch: AdaptiveMaxPool2d
# Mode: Eager
# Name: AdaptiveMaxPool2d_kernel[3,3]_stride[2,2]_N16_C32_H64_W64_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 16, C: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 168.563

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,1]_stride[2,1]_N1_C16_H32_W32_cpu
# Input: kernel: [3, 1], stride: [2, 1], N: 1, C: 16, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 38.105

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,2]_stride[2,2]_N8_C32_H32_W32_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 8, C: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 53.678

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,2]_stride[2,2]_N8_C32_H32_W64_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 8, C: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 58.790

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,2]_stride[2,2]_N8_C32_H64_W32_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 8, C: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 52.641

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,2]_stride[2,2]_N8_C32_H64_W64_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 8, C: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 54.202

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,2]_stride[2,2]_N16_C32_H32_W32_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 16, C: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 64.891

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,2]_stride[2,2]_N16_C32_H32_W64_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 16, C: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 61.271

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,2]_stride[2,2]_N16_C32_H64_W32_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 16, C: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 62.074

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,2]_stride[2,2]_N16_C32_H64_W64_cpu
# Input: kernel: [3, 2], stride: [2, 2], N: 16, C: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 62.022

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,3]_stride[2,2]_N8_C32_H32_W32_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 8, C: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 55.512

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,3]_stride[2,2]_N8_C32_H32_W64_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 8, C: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 56.149

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,3]_stride[2,2]_N8_C32_H64_W32_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 8, C: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 59.689

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,3]_stride[2,2]_N8_C32_H64_W64_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 8, C: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 59.210

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,3]_stride[2,2]_N16_C32_H32_W32_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 16, C: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 69.061

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,3]_stride[2,2]_N16_C32_H32_W64_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 16, C: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 67.673

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,3]_stride[2,2]_N16_C32_H64_W32_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 16, C: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 67.671

# Benchmarking PyTorch: FractionalMaxPool2d
# Mode: Eager
# Name: FractionalMaxPool2d_kernel[3,3]_stride[2,2]_N16_C32_H64_W64_cpu
# Input: kernel: [3, 3], stride: [2, 2], N: 16, C: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 67.627

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,1,3]_stride[2,1,2]_N1_C16_D16_H32_W32_cpu
# Input: kernel: [3, 1, 3], stride: [2, 1, 2], N: 1, C: 16, D: 16, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 338.750

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N8_C32_D32_H32_W32_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 1580.897

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N8_C32_D32_H32_W64_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 4679.794

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N8_C32_D32_H64_W32_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 3772.246

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N8_C32_D32_H64_W64_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 9960.207

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N16_C32_D32_H32_W32_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 2765.616

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N16_C32_D32_H32_W64_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 6874.791

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N16_C32_D32_H64_W32_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 6816.528

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N16_C32_D32_H64_W64_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 14006.464

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N8_C32_D32_H32_W32_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 1693.150

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N8_C32_D32_H32_W64_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 4250.506

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N8_C32_D32_H64_W32_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 4019.097

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N8_C32_D32_H64_W64_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 9947.090

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N16_C32_D32_H32_W32_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 2356.209

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N16_C32_D32_H32_W64_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 7095.732

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N16_C32_D32_H64_W32_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 6791.053

# Benchmarking PyTorch: AdaptiveMaxPool3d
# Mode: Eager
# Name: AdaptiveMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N16_C32_D32_H64_W64_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 14499.389

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,1,3]_stride[2,1,2]_N1_C16_D16_H32_W32_cpu
# Input: kernel: [3, 1, 3], stride: [2, 1, 2], N: 1, C: 16, D: 16, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 40.421

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N8_C32_D32_H32_W32_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 73.324

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N8_C32_D32_H32_W64_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 76.495

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N8_C32_D32_H64_W32_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 75.037

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N8_C32_D32_H64_W64_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 76.399

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N16_C32_D32_H32_W32_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 104.319

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N16_C32_D32_H32_W64_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 98.106

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N16_C32_D32_H64_W32_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 96.371

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,2,3]_stride[2,2,2]_N16_C32_D32_H64_W64_cpu
# Input: kernel: [3, 2, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 96.063

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N8_C32_D32_H32_W32_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 80.156

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N8_C32_D32_H32_W64_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 81.476

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N8_C32_D32_H64_W32_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 81.561

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N8_C32_D32_H64_W64_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 8, C: 32, D: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 81.320

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N16_C32_D32_H32_W32_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 32, W: 32, device: cpu
Forward Execution Time (us) : 100.050

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N16_C32_D32_H32_W64_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 32, W: 64, device: cpu
Forward Execution Time (us) : 96.316

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N16_C32_D32_H64_W32_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 64, W: 32, device: cpu
Forward Execution Time (us) : 93.940

# Benchmarking PyTorch: FractionalMaxPool3d
# Mode: Eager
# Name: FractionalMaxPool3d_kernel[3,3,3]_stride[2,2,2]_N16_C32_D32_H64_W64_cpu
# Input: kernel: [3, 3, 3], stride: [2, 2, 2], N: 16, C: 32, D: 32, H: 64, W: 64, device: cpu
Forward Execution Time (us) : 94.695

(pytorch-cuda-dev) wbaranowski@qgpu2:~/git/Quansight/pytorch/benchmarks/operator_benchmark$

@Baranowski Baranowski force-pushed the wbaranowski-max_pool_nan-40131 branch from fe71925 to ffd89d8 Compare July 11, 2020 08:21
facebook-github-bot
facebook-github-bot reviewed Jul 13, 2020
View reviewed changes
Copy link
Contributor

@facebook-github-bot facebook-github-bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@ezyang has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

@facebook-github-bot
Copy link
Contributor

facebook-github-bot commented Jul 15, 2020

@ezyang merged this pull request in 20f3051.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@facebook-github-bot facebook-github-bot facebook-github-bot left review comments

@ezyang ezyang ezyang approved these changes

@xwang233 xwang233 xwang233 approved these changes

Assignees

No one assigned

Labels

Merged open source

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

CPU {adaptive,fractional, } max_pool {1,2,3}d will crash with nan and -inf values

6 participants

@Baranowski @xwang233 @ezyang @facebook-github-bot @pytorchbot @mruberry