NVIDIA / NVIDIA/apex

FusedAdam bias_correction argument incompatible with torch.optim.Adam

Open
#577 1 comment 15 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

The docs suggest that 'apex.optimizers.FusedAdam may be used as a drop-in replacement for torch.optim.Adam'. However, when I load an optimizer state_dict generated by Adam, then resume training, I get:

File "/opt/conda/lib/python3.7/site-packages/apex/optimizers/fused_adam.py", line 105, in step bias_correction = 1 if group['bias_correction'] else 0 KeyError: 'bias_correction'

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 apex/optimizers/fused_adam.py at line 105 and reproduce loading an optimizer state_dict generated by torch.optim.Adam before resuming training. Check the documented drop-in replacement behavior and verify that resuming from an Adam state does not raise the reported KeyError.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
machine-learning
Issue type
Bug
Difficulty
3/5
Estimated time
1-2 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
48/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.