2019-10-05 18:21:12 +00:00
|
|
|
# Multi-node example
|
2019-09-14 13:55:42 +00:00
|
|
|
|
2019-10-05 18:28:08 +00:00
|
|
|
To run this demo which launches a single job that trains on 2 nodes (2 gpus per node), do the following:
|
2019-09-14 13:55:42 +00:00
|
|
|
|
2019-10-05 18:28:08 +00:00
|
|
|
1. Log into the jumphost node of your SLURM-managed cluster.
|
|
|
|
|
2. Create a conda environment with Lightning and a GPU PyTorch version.
|
|
|
|
|
3. Submit this script.
|
2019-09-08 22:17:33 +00:00
|
|
|
```bash
|
2019-10-05 18:48:22 +00:00
|
|
|
sbatch job_submit.sh --env=YourEnv
|
|
|
|
|
```
|
