deepspeedai / deepspeedai/DeepSpeed
[BUG]Why ZeroOneAdam is easy to OOM compared to Adam optimizer?
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
When using Adam optimizer, the GPU memory used is about 14G for a 16G V100 GPU. However, after using ZeroOneAdam, it's always OOM for me, although I try to cut half of my GPU memory.
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
The report names ZeroOneAdam, Adam, and a 16G V100 GPU but provides no files, tests, versions, or reproduction steps. Start by locating the two optimizer implementations and reproducing the memory comparison; done means the excess memory use is explained or a focused failure case is captured.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- machine-learning, performance
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100