deepspeedai / deepspeedai/DeepSpeedExamples

Llama2 as actor using zero_stage3

Open
#814 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
6.8k
Forks
1.1k
Avg merge
2d 16h
Merged PRs (30d)
1

Description

Hello! Did anyone meet the following bug when using zero_stage3 for Lllama2?
step3_rlhf_finetuning/rlhf_engine.py:61 in init
│ │
│ 58 │ │ self.num_total_iters = num_total_iters │
│ 59 │ │ self.tokenizer = tokenizer │
│ 60 │ │ │
│ ❱ 61 │ │ self.actor = self._init_actor(actor_model_name_or_path=actor_model_name_or_path)

AttributeError: 'LlamaAttention' object has no attribute 'rope_theta'.

Note that OPT works, and using zero_stage2 also works.

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 step3_rlhf_finetuning/rlhf_engine.py at the actor initialization on line 61, then reproduce the Llama2 failure with zero_stage3. Compare that path with the working OPT and zero_stage2 cases, and inspect how LlamaAttention is accessed. Done means Llama2 actor initialization completes without the rope_theta AttributeError.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
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.