deepspeedai / deepspeedai/DeepSpeed
[REQUEST] Supports of backward propagation profiling
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 43.1k
- Forks
- 5k
- Avg merge
- 4d 15h
- Merged PRs (30d)
- 112
Description
Is your feature request related to a problem? Please describe.
Currently DeepSpeed only support the profiling of forward propagation. To calculate the percentage of theoretical peak FLOPs of training, we may need the profiling of backward propagation as well as parameter optimization.
Describe the solution you'd like
The implementation can uses current implementation, while adding the backward analysis.
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 by locating and reading DeepSpeed's existing forward-propagation profiling implementation. Determine how backward propagation and parameter optimization should be measured consistently with it; done means profiling reports the additional training phases well enough to calculate theoretical-peak-FLOPs percentages.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- machine-learning, performance
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 30/100