sgl-project / sgl-project/SpecForge

Any plan to support training draft model with target model's kv cache?

Open
#119 0 comments 1 reaction 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
1.2k
Forks
347
Avg merge
4d 1h
Merged PRs (30d)
41

Description

Thanks for your excellent project! I am wondering whether it is possible to support training the draft model with target model's kv cache instead of hidden states?

Contributor guide

No contributing guide indexed for this repository

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

No file, test, or entry point is named. Start by locating the training path that currently uses the target model's hidden states and review how its inputs and KV cache are handled. Done would mean a defined, tested approach for training the draft model from the target model's KV cache.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.