lm-sys / lm-sys/FastChat

Vicuna-7b for full-parameter fine-tuning based on DeepSpeed

Open
#2,631 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
39.5k
Forks
4.8k
PR merge metrics
No merged PRs in 30d

Description

Issue #2249
I changed the train_lora.sh script and try to do full-parameter fine-tuning, but i found that no mater how i change the ds_config file(which btw, gradient checkpoint or offload, or zero1/zero2), the gpu mem used for my training didn't change. i'm so confused...... and i checked transformer source code, i found that it's accelerate package which intergrates some feature of deepspeed that transformer actually working on deepspeed config. so, it's useless? i need some help

I have the same issue when I am using the deepspeed parameter (previously commented fsdp) in the "train_vicuna_7b.sh" file

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 with train_lora.sh and train_vicuna_7b.sh, then inspect the referenced ds_config and how the deepspeed parameter is passed into training. Compare the configured options with the observed GPU memory behavior. Done means determining whether the configuration is being applied and documenting or correcting the cause.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, shell
Domain
distributed-systems, machine-learning
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.