https://github.com/NERSC/pytorch-examples

PyTorch examples for NERSC systems
https://github.com/NERSC/pytorch-examples

Last synced: 27 days ago
JSON representation

PyTorch examples for NERSC systems

• Host: GitHub
• URL: https://github.com/NERSC/pytorch-examples
• Owner: NERSC
• Created: 2018-10-08T19:25:18.000Z (over 5 years ago)
• Default Branch: main
• Last Pushed: 2023-01-25T18:28:32.000Z (over 1 year ago)
• Last Synced: 2024-04-16T03:34:51.540Z (2 months ago)
• Language: Jupyter Notebook
• Size: 370 KB
• Stars: 25
• Watchers: 18
• Forks: 10
• Open Issues: 5
• Metadata Files:
  • Readme: README.md

Lists

• awesome-stars - NERSC/pytorch-examples - PyTorch examples for NERSC systems (Jupyter Notebook)

README

        
# NERSC PyTorch examples

This repository contains some PyTorch example models and training code

with support for distributed training on NERSC systems.

The layout of this package can also serve as a template for PyTorch

projects and the provided BaseTrainer and train.py script can be used to

reduce boiler plate.

## Package layout

The directory layout of this repo is designed to be flexible:

- Configuration files (in YAML format) go in `configs/`

- Dataset specifications using PyTorch's Dataset API go into `datasets/`

- Model implementations go into `models/`

- Trainer implementations go into `trainers/`. Trainers inherit from

  `BaseTrainer` and are responsible for constructing models as well as training

  and evaluating them.

All examples are run with the generic training script, `train.py`.

## Examples

This package currently contains the following examples:

- CIFAR10 classification with ResNet50 or generic CNN model.

- HEP-CNN classification (https://arxiv.org/abs/1711.03573).

- Minimal Hello World example.

## How to run

To run the examples on the Cori supercomputer, you may use the provided

example SLURM batch script. Here's how to run the ResNet50 CIFAR10 example

on 4 Haswell nodes:

`sbatch -N 4 scripts/train_cori.sh configs/cifar10.yaml`