deepspeedai / deepspeedai/DeepSpeed
[BUG] Does deepspeed work with torch amp autocast?
Open
Nobody has claimed this yet.
bug
training
- Dominant language
- Python
- Stars
- 43.1k
- Forks
- 5k
- Avg merge
- 4d 15h
- Merged PRs (30d)
- 112
Description
During training, fp16 is set to true.
But for some submodules, i want to using fp32.
with torch.cuda.amp.autocast(enabled=False):
x = self.model(x.float())
An error occur, show half vs float type not matching
Contributor guide
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 reported torch.cuda.amp.autocast(enabled=False) block and the DeepSpeed configuration where fp16 is enabled. Reproduce the half-versus-float mismatch with a minimal submodule example, then trace DeepSpeed's mixed-precision handling. Done means establishing whether this combination is supported and documenting or fixing the behavior based on the reproduction.
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
- Needs clarification
- Newbie friendliness
- 25/100