microsoft / microsoft/TransformerCompression

How to finetune with multi-gpus under data parallel setting?

Open
#167 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
460
Forks
58
Avg merge
8h 40m
Merged PRs (30d)
1

Description

Many thanks for sharing the amazing work!

I'm trying to finetune sliced 7b model on some large dataset with millions of samples.
But the distribute-model seems to be model parallel.
How can we finetune on the model with, let's say 8 gpus, under data parallel setting ?

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

Start by reading the distribute-model entry point and any surrounding fine-tuning documentation to determine how its model-parallel behavior differs from the requested data-parallel setup. The issue does not name files or tests; completion would require an agreed, reproducible multi-GPU fine-tuning path or documentation explaining whether it is supported.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
distributed-systems, machine-learning
Issue type
Documentation
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.