NVIDIA / NVIDIA/apex

minor warm-up step miss

Open
#609 0 comments 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

Minor comment about warm up in https://github.com/NVIDIA/apex/blob/master/examples/imagenet/main_amp.py line 504:

lr = lr*float(1 + step + epoch*len_epoch)/(5.*len_epoch)

Your step starts at 1 and not 0, making a slight offset in this range.

Might be nice to also add Nesterov as an optional argument (since it's used as baseline in many papers).

And maybe also to, optionally, add

transforms.Resize(256),  # resize smaller edge to 256
transforms.RandomResizedCrop(size=224, scale=(0.08, 1.0), ratio=(3. / 4., 4. / 3.)),

to the training loader instead of the random crop (again, baseline argument).

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 in examples/imagenet/main_amp.py around line 504 and inspect the warm-up calculation and training loader. Check the step indexing against the reported offset, then review the requested optional Nesterov and image-transform changes. Done means the intended warm-up behavior is corrected and the requested options are clearly implemented or scoped.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
machine-learning
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.