Multi-node training with SyncBN
Open
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 9k
- Forks
- 1.5k
- Avg merge
- 2d 4h
- Merged PRs (30d)
- 3
Description
Hi, does SyncBN in apex support multi-node training?
When training FPN (based on mmdetection) using SyncBN with two nodes (4 GPUs per node), the final results drop nearly 4 mAP compared to a single node (8 GPUs).
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start with the SyncBN implementation and its distributed-training entry points, then reproduce the reported FPN/mmdetection comparison across one node with 8 GPUs and two nodes with 4 GPUs each. The issue is done when multi-node support is confirmed or the cause of the mAP difference is identified and documented.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python, pytorch
- Domain
- distributed-systems, machine-learning
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 20/100