Lightning-AI / Lightning-AI/lit-llama

Adapter finetuning do not run on two cards (A100 40G)

Open
#459 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
6.1k
Forks
517
PR merge metrics
No merged PRs in 30d

Description

I am try to finetune on two cards but the adapter.py hang in there after this output, I waited so long (1 hour, went for food). what is the reason?

from jsonargparse.cli import CLI
[2023-10-02 11:27:18,651] [INFO] [real_accelerator.py:158:get_accelerator] Setting ds_accelerator to cuda (auto detect)
initializing deepspeed distributed: GLOBAL_RANK: 1, MEMBER: 2/2
Enabling DeepSpeed BF16. Model parameters and inputs will be cast to bfloat16.
[rank: 0] Seed set to 1337
[rank: 1] Seed set to 1338
Number of trainable parameters: 1229760
Number of trainable parameters: 1229760

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 adapter.py and reproduce the two-card run using the reported DeepSpeed output as the checkpoint, particularly the point after both ranks report trainable parameters. Inspect where distributed initialization and training should proceed; done means the run advances past that point or the cause of the hang is documented with a reproducible result.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
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.