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

Staging #311

Merged
merged 16 commits into from
Nov 22, 2024
Merged

Staging #311

merged 16 commits into from
Nov 22, 2024

Conversation

Delaunay
Copy link
Collaborator

No description provided.

Your Name and others added 11 commits October 30, 2024 13:31
* HPU changes

* HPU pins

* Use HPU device

* Gaudi Tweaks

* Remove pinnning

* call prepare again

* Add docker image for HPU

* Tweaks

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

* -

---------

Co-authored-by: Your Name <you@example.com>
* ROCm changes

* Update ping

* -

* Cleanup the rocm script

* use rocm branch

* -

* Update run_rocm.sh

---------

Co-authored-by: Your Name <you@example.com>
* ROCm changes

* Update ping

* -

* Cleanup the rocm script

* use rocm branch

* -

* Ne wmulti run system

* multinode tweaks

* make sure system config is applied before running

* Update matrix run

* Tweaks

---------

Co-authored-by: Your Name <you@example.com>
@Delaunay Delaunay changed the title Tweaks Staging Nov 21, 2024
@Delaunay
Copy link
Collaborator Author

=================
Benchmark results
=================

System
------
cpu:      AMD EPYC 9654 96-Core Processor
n_cpu:    192
product:  NVIDIA H100 80GB HBM3
n_gpu:    8
memory:   81559.0

Breakdown
---------
bench                    | fail |   n | ngpu |       perf |   sem% |   std% | peak_memory |      score | weight
brax                     |    0 |   1 |    8 |  880265.71 |   0.1% |   0.4% |        5101 |  880265.71 |   1.00
diffusion-gpus           |    0 |   1 |    8 |     395.88 |   0.1% |   1.0% |       62009 |     395.88 |   1.00
diffusion-single         |    0 |   8 |    1 |      52.17 |   0.4% |   9.0% |       53581 |     421.81 |   1.00
dimenet                  |    1 |   8 |    1 |     484.88 |   0.8% |  16.0% |         nan |    3003.97 |   1.00
dinov2-giant-gpus        |    0 |   1 |    8 |     865.35 |   0.4% |   3.2% |       71779 |     865.35 |   1.00
dinov2-giant-single      |    0 |   8 |    1 |     103.60 |   0.4% |   9.6% |       74309 |     837.15 |   1.00
dqn                      |    8 |   8 |    1 |        nan |   nan% |   nan% |         nan |        nan |   1.00
bf16                     |    0 |   8 |    1 |     784.66 |   0.2% |   5.5% |         nan |    6308.71 |   0.00
fp16                     |    0 |   8 |    1 |     799.62 |   0.2% |   6.0% |        1313 |    6433.98 |   0.00
fp32                     |    0 |   8 |    1 |      52.00 |   0.0% |   0.7% |        1697 |     416.22 |   0.00
tf32                     |    0 |   8 |    1 |     415.14 |   0.2% |   5.1% |        1697 |    3326.47 |   0.00
bert-fp16                |    0 |   8 |    1 |     459.73 |   1.0% |  15.4% |         nan |    3770.96 |   0.00
bert-fp32                |    0 |   8 |    1 |     111.45 |   0.4% |   6.9% |       20771 |     902.03 |   0.00
bert-tf32                |    0 |   8 |    1 |     243.95 |   0.7% |  11.1% |         nan |    1988.27 |   0.00
bert-tf32-fp16           |    0 |   8 |    1 |     459.55 |   1.0% |  15.1% |         nan |    3768.04 |   3.00
reformer                 |    0 |   8 |    1 |     101.43 |   0.3% |   6.9% |       13021 |     818.06 |   1.00
t5                       |    0 |   8 |    1 |      89.57 |   0.4% |   8.2% |         251 |     723.55 |   2.00
whisper                  |    0 |   8 |    1 |     885.56 |   0.5% |  10.8% |         791 |    7169.14 |   1.00
lightning                |    0 |   8 |    1 |    1059.89 |   0.3% |   8.6% |       26787 |    8547.25 |   1.00
lightning-gpus           |    0 |   1 |    8 |    8240.39 |   0.0% |   0.1% |       30877 |    8240.39 |   1.00
llava-single             |    0 |   8 |    1 |       3.99 |   0.4% |   9.3% |       72379 |      32.22 |   1.00
llama                    |    0 |   8 |    1 |     771.59 |   4.7% |  85.8% |       27373 |    5810.84 |   1.00
llm-full-mp-gpus         |    1 |   1 |    8 |        nan |   nan% |   nan% |         nan |        nan |   1.00
llm-lora-ddp-gpus        |    0 |   1 |    8 |   27507.41 |   0.4% |   2.2% |       38921 |   27507.41 |   1.00
llm-lora-mp-gpus         |    0 |   1 |    8 |    3452.54 |   1.6% |   8.6% |       54730 |    3452.54 |   1.00
llm-lora-single          |    0 |   8 |    1 |    5097.69 |   0.2% |   3.7% |       49497 |   40818.91 |   1.00
pna                      |    0 |   8 |    1 |    7041.34 |   0.5% |  10.5% |       39319 |   56443.17 |   1.00
ppo                      |    8 |   8 |    1 |        nan |   nan% |   nan% |         nan |        nan |   1.00
recursiongfn             |    0 |   8 |    1 |   11835.51 |   1.2% |  25.5% |       12017 |   95407.38 |   1.00
rlhf-gpus                |    0 |   1 |    8 |    9275.39 |   1.2% |   6.4% |       24001 |    9275.39 |   1.00
rlhf-single              |    0 |   8 |    1 |    1831.11 |   0.6% |  14.3% |       19037 |   14642.37 |   1.00
focalnet                 |    0 |   8 |    1 |     651.81 |   0.6% |  12.6% |       22063 |    5283.09 |   2.00
torchatari               |    0 |   8 |    1 |    6755.71 |   0.7% |  14.3% |        3455 |   53817.14 |   1.00
convnext_large-fp16      |    0 |   8 |    1 |     662.70 |   1.0% |  15.2% |         nan |    5437.04 |   0.00
convnext_large-fp32      |    0 |   8 |    1 |     129.45 |   0.6% |   9.8% |       46935 |    1052.70 |   0.00
convnext_large-tf32      |    0 |   8 |    1 |     239.76 |   0.8% |  12.5% |       49271 |    1959.09 |   0.00
convnext_large-tf32-fp16 |    0 |   8 |    1 |     666.11 |   1.0% |  15.2% |         nan |    5465.52 |   3.00
regnet_y_128gf           |    0 |   8 |    1 |     187.60 |   0.4% |   9.1% |       28505 |    1517.15 |   2.00
resnet152-ddp-gpus       |    0 |   1 |    8 |    8182.16 |   0.0% |   0.3% |       30729 |    8182.16 |   0.00
resnet50                 |    0 |   8 |    1 |    1858.21 |   0.6% |  12.4% |       14297 |   15059.52 |   1.00
resnet50-noio            |    0 |   8 |    1 |    2053.26 |   0.0% |   2.8% |       27119 |   16444.85 |   0.00
vjepa-gpus               |    0 |   1 |    8 |     260.50 |   5.0% |  39.3% |       66791 |     260.50 |   1.00
vjepa-single             |    0 |   8 |    1 |      41.82 |   0.7% |  15.8% |       55811 |     338.22 |   1.00

Scores
------
Failure rate:       6.57% (FAIL)
Score:            2093.20

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

@Delaunay Delaunay marked this pull request as ready for review November 22, 2024 00:43
@Delaunay Delaunay merged commit 6c15d81 into master Nov 22, 2024
3 of 6 checks passed
@Delaunay Delaunay deleted the staging branch November 22, 2024 00:43
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