deepspeedai / deepspeedai/DeepSpeed
Inference with an autoregressive Pipeline parallel model
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 43.1k
- Forks
- 5k
- Avg merge
- 4d 15h
- Merged PRs (30d)
- 112
Description
It seems currently that inference with an autoregressive pipeline parallel model isn't implemented in the Pipeline Engine.
I Imagine it's not so straightforward since it would require communicating the inputs from the final stage back to the first stage at every step. Is the Deepspeed library planning anything in this regard, or is there any example code available for how one might acheive this communication step?
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
No file, test, or entry point is named. Start by reading the DeepSpeed Pipeline Engine implementation and its inference and pipeline-parallel examples, then determine whether autoregressive inference and final-stage-to-first-stage communication are supported; done means a documented implementation or example exists for this workflow.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- distributed-systems, machine-learning
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 20/100