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

Update cuda #170

Merged
merged 36 commits into from
Oct 23, 2023
Merged

Update cuda #170

merged 36 commits into from
Oct 23, 2023

Conversation

Delaunay
Copy link
Collaborator

No description provided.

pierre.delaunay added 5 commits October 20, 2023 16:01
@Delaunay
Copy link
Collaborator Author

Source: /Tmp/slurm.3767130.0/base/runs/fuvibeve.2023-10-23_12:18:22.214873
=================
Benchmark results
=================
                         fail   n       perf   sem%   std% peak_memory          score weight
bert-fp16                   0   1      59.87   1.8%   9.9%       23976      59.872190   0.00
bert-fp32                   0   1      22.37   0.0%   0.2%       30946      22.372731   0.00
bert-tf32                   0   1      22.41   0.0%   0.2%       30946      22.409891   0.00
bert-tf32-fp16              0   1      59.07   1.9%  10.3%       23976      59.069720   3.00
convnext_large-fp16         0   1     137.50   2.3%  12.3%       26656     137.503359   0.00
convnext_large-fp32         0   1      33.08   0.3%   1.7%       46524      33.083616   0.00
convnext_large-tf32         0   1      33.01   0.3%   1.5%       46524      33.014063   0.00
convnext_large-tf32-fp16    0   1     136.07   2.5%  13.4%       26656     136.069068   3.00
davit_large                 0   1     118.89   0.8%   6.4%       32398     118.891062   1.00
davit_large-multi           0   1     119.36   0.9%   6.6%       32418     119.364233   5.00
dlrm                        0   1  214279.91   0.5%   3.9%        3282  214279.910462   1.00
focalnet                    0   1     169.49   0.5%   4.2%       24378     169.489190   2.00
opt-1_3b                    1   1        NaN    NaN    NaN          -1            NaN   5.00
opt-1_3b-multinode        NaN NaN        NaN    NaN    NaN         NaN            NaN  10.00
opt-6_7b                    1   1        NaN    NaN    NaN          -1            NaN   5.00
opt-6_7b-multinode        NaN NaN        NaN    NaN    NaN         NaN            NaN  10.00
reformer                    0   1      12.16   0.0%   0.1%       24780      12.160582   1.00
regnet_y_128gf              0   1      34.86   0.3%   2.7%       30772      34.856563   2.00
resnet152                   0   1     263.66   1.8%  13.9%       30526     263.659564   1.00
resnet152-multi             0   1     261.95   1.8%  13.7%       30514     261.948591   5.00
resnet50                    0   1     518.95   3.2%  24.4%        4190     518.950184   1.00
rwkv                        1   1        NaN    NaN    NaN        2044            NaN   1.00
stargan                     0   1      12.01   4.0%  30.8%       36306      12.006805   1.00
super-slomo                 0   1      13.12   0.0%   0.2%       36388      13.119863   1.00
t5                          0   1      17.90   1.1%   8.5%       34818      17.903359   2.00
whisper                     0   1     111.12   0.0%   0.2%       35992     111.123644   1.00

Scores
------
Failure rate:      12.50% (FAIL)
Score:              10.43

Errors
------
3 errors, details in HTML report.

@Delaunay
Copy link
Collaborator Author

Source: /Tmp/slurm.3767245.0/base/runs/puridefe.2023-10-23_13:33:56.401976
=================
Benchmark results
=================
                         fail   n       perf   sem%   std% peak_memory          score weight
bert-fp16                   0   1      58.41   2.1%  11.1%          -1      58.407083   0.00
bert-fp32                   0   1      24.19   0.0%   0.2%          -1      24.186551   0.00
bert-tf32                   0   1      24.20   0.0%   0.2%          -1      24.203025   0.00
bert-tf32-fp16              0   1      58.45   2.1%  11.6%          -1      58.450135   3.00
convnext_large-fp16         0   1     149.42   2.1%  11.5%          -1     149.418560   0.00
convnext_large-fp32         0   1      36.41   0.3%   1.4%          -1      36.410330   0.00
convnext_large-tf32         0   1      36.28   0.1%   0.7%          -1      36.278462   0.00
convnext_large-tf32-fp16    0   1     150.57   2.2%  11.8%          -1     150.573622   3.00
davit_large                 0   1     126.59   1.5%  11.4%          -1     126.592700   1.00
davit_large-multi           0   1     126.98   1.5%  11.6%          -1     126.977999   5.00
dlrm                        0   1  246516.01   0.4%   3.4%          -1  246516.013754   1.00
focalnet                    0   1     177.99   0.5%   3.5%          -1     177.992342   2.00
opt-1_3b                    1   1        NaN    NaN    NaN          -1            NaN   5.00
opt-1_3b-multinode        NaN NaN        NaN    NaN    NaN         NaN            NaN  10.00
opt-6_7b                    1   1        NaN    NaN    NaN       13646            NaN   5.00
opt-6_7b-multinode        NaN NaN        NaN    NaN    NaN         NaN            NaN  10.00
reformer                    0   1      12.37   0.0%   0.1%          -1      12.367892   1.00
regnet_y_128gf              0   1      36.74   0.2%   1.9%          -1      36.742808   2.00
resnet152                   0   1     272.77   1.7%  13.2%          -1     272.767066   1.00
resnet152-multi             0   1     270.69   1.7%  13.0%          -1     270.686143   5.00
resnet50                    0   1     565.28   2.7%  20.7%          -1     565.275976   1.00
rwkv                        0   1     118.11   0.1%   0.9%          -1     118.113206   1.00
stargan                     0   1      12.84   4.1%  31.8%          -1      12.840037   1.00
super-slomo                 0   1      13.79   0.0%   0.3%          -1      13.787189   1.00
t5                          0   1      17.99   1.1%   8.7%          -1      17.986944   2.00
whisper                     0   1     113.82   0.2%   1.6%          -1     113.824003   1.00

Scores
------
Failure rate:       8.33% (FAIL)
Score:              11.55

Errors
------
2 errors, details in HTML report.

@Delaunay
Copy link
Collaborator Author

opt-1_3b fail because there is a single GPU, opt-6_7b fails with OOM

@Delaunay Delaunay merged commit 31fdc14 into master Oct 23, 2023
1 of 2 checks passed
@Delaunay Delaunay deleted the update_pytorch branch October 23, 2023 18:23
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

1 participant