deepspeedai / deepspeedai/DeepSpeed
how to understand "DeepSpeed uses gradient accumulation to extract pipeline parallelism"
Open
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 43.1k
- Forks
- 5k
- Avg merge
- 4d 15h
- Merged PRs (30d)
- 112
Description
hi, I am a little bit confused about the relationship of gradient accumulation stages and pipeline stages when I read your pipeline source code "total_steps = 2 * (self.micro_batches + self.stages - 1)", could you describe it detailed?
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 pipeline source code around total_steps = 2 * (self.micro_batches + self.stages - 1) and trace how micro_batches and stages are used. Document the relationship between gradient accumulation and pipeline stages, including a clear explanation of how the expression determines the schedule length.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- distributed-systems, machine-learning
- Issue type
- Documentation
- Difficulty
- 3/5
- Estimated time
- 1-2 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 25/100