NVIDIA / NVIDIA/apex

Why scaling lr with batch_size*world_size/256?

Open
#407 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
9k
Forks
1.5k
Avg merge
2d 4h
Merged PRs (30d)
3

Description

https://github.com/NVIDIA/apex/blob/master/examples/imagenet/main_amp.py#L143

Does it mean that args.lr is the learning rate based on batch size 256?
Can I use it to scale the lr of Adam?

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start at examples/imagenet/main_amp.py around line 143, the location linked by the issue. Read how batch_size, world_size, args.lr, and Adam are used, then clarify the intended learning-rate scaling and its scope in the example documentation or comments. Done means a newcomer can understand the 256-based scaling and whether it applies to Adam.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
distributed-systems, machine-learning
Issue type
Documentation
Difficulty
2/5
Estimated time
1-3 hours
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
30/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.