deepspeedai / deepspeedai/DeepSpeed

[REQUEST] Make DeepSpeed FastGen support tokens as input?

Open
#5,261 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

enhancement
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.
Current examples and documents I found can only support text as input:

import mii
pipe = mii.pipeline("mistralai/Mistral-7B-v0.1")
response = pipe(["DeepSpeed is", "Seattle is"], max_new_tokens=128)
print(response)

For example, If I want to use the chat function in mistralai/Mixtral-8x7B-Instruct-v0.1, theoretically, I have to make the following formats
image

Describe the solution you'd like
If I can only use text as input, I cannot really do that.

Describe alternatives you've considered
VLLM supports tokens as input

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with the MII pipeline example in the issue and compare the requested token-input behavior with the vLLM alternative mentioned. Done means DeepSpeed FastGen accepts token inputs for inference, including the chat use case described.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
30/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.