facebookresearch / facebookresearch/SlowFast
How to train SlowFast on multi GPU nodes machine
Open
- Dominant language
- Python
- Stars
- 7.4k
- Forks
- 1.3k
- PR merge metrics
- No merged PRs in 30d
Description
I'm looking for some guidance on how to run tools/train_net.py for training SlowFast on multi GPUs nodes job on slurm.
I try to change NUM_SHARDS from 1 to N corresponding to a number of GPU nodes and submit a job via sbatch but it still training on one GPU node.
Could any guide or provide some examples to do it?
Thanks!
Contributor guide
Assessment
This issue has not been assessed yet.