lucidrains / lucidrains/speculative-decoding

About batch size > 1

Open
#4 2 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
308
Forks
25
PR merge metrics
No merged PRs in 30d

Description

First of all, thank you for open-sourcing the implementation of speculative decoding at batch size > 1. I would like to ask if it is possible to adapt directly to the models downloaded by huggingface instead of customizing their framework code. Because I tried to use this with codegen, but the generated content is messy. Hope you can answer my confusion at your convenience.

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

The issue names no files or tests. Start by tracing the repository’s speculative-decoding entry point and how it accepts models, then compare that path with Hugging Face’s model-loading path and reproduce the reported CodeGen output; done means supporting downloaded Hugging Face models at batch size >1 without framework customization.

Written by the indexing model from the issue text.

Assessment

Tech stack
huggingface, python
Domain
ai, machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
20/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.