Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/mlcommons/hpc
Reference implementations of MLPerf™ HPC training benchmarks
https://github.com/mlcommons/hpc
Last synced: 11 days ago
JSON representation
Reference implementations of MLPerf™ HPC training benchmarks
- Host: GitHub
- URL: https://github.com/mlcommons/hpc
- Owner: mlcommons
- License: apache-2.0
- Created: 2020-10-12T17:34:48.000Z (about 4 years ago)
- Default Branch: main
- Last Pushed: 2024-05-24T21:55:20.000Z (6 months ago)
- Last Synced: 2024-08-01T16:46:43.320Z (3 months ago)
- Language: Jupyter Notebook
- Homepage: https://mlcommons.org/en/groups/training-hpc/
- Size: 2.85 MB
- Stars: 37
- Watchers: 16
- Forks: 15
- Open Issues: 13
-
Metadata Files:
- Readme: README.md
- Contributing: CONTRIBUTING.md
- License: LICENSE.md
- Codeowners: .github/CODEOWNERS
Awesome Lists containing this project
README
# MLPerf™ HPC reference implementations
This is a repository of reference implementations for the MLPerf HPC benchmarks.
General format should follow https://github.com/mlperf/training.
## Rules
The MLPerf HPC rules are based on the MLPerf Training rules with
some adjustments.
The MLPerf Training rules are available at [training\_rules](https://github.com/mlcommons/training_policies/blob/master/training_rules.adoc).
The MLPerf HPC specific rules are at [hpc\_training\_rules](https://github.com/mlcommons/training_policies/blob/master/hpc_training_rules.adoc).
## Compliance
The MLPerf logging package implements logging and compliance-checking utilities. This is available in hpc-1.0-branch of the MLPerf logging repository (https://github.com/mlcommons/logging/tree/hpc-1.0-branch).
These work for the HPC v2.0 submissions as well.
To install and test compliance of your runs/submissions:
```
# Install the package into your python environment.
# A development install (-e) is recommended for now so you can pull new updates.
git clone -b hpc-1.0-branch https://github.com/mlcommons/logging mlperf-logging
pip install [--user] -e mlperf-logging
# Test a full submission folder
python3 -m mlperf_logging.package_checker hpc 1.0.0
```
There is also a script that performs compliance checks and summarizes the results. From the mlperf-logging directory (https://github.com/mlcommons/logging), use
```
./scripts/verify_for_v1.0_hpc.sh
```