deepspeedai / deepspeedai/DeepSpeed

[BUG] run_zero_quant.sh seems not working

Open
#3,334 2 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

bug compression
Dominant language
Python
Stars
43.1k
Forks
5k
Avg merge
4d 15h
Merged PRs (30d)
112

Description

Describe the bug
I am trying run_zero_quant for GPT-J model, but I find the output model size is not compressed, the model file size is the same as origin model, also after I load the model, the GPU memory is the same as origin model,

To Reproduce
Steps to reproduce the behavior:

  1. Go to DeepSpeedExamples/model_compression/bert
  2. Run:
    pip install -r requirements.txt
    bash bash_script/run_zero_quant.sh
    If I don't do any change to sh, it will use origin gpt-2 model, also I tried gpt-j model. Neither output model file size are compressed

Is there a bug? Could you please check if this issue can be reproduced?

Contributor guide

Open the contributing guide

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 in DeepSpeedExamples/model_compression/bert and inspect bash_script/run_zero_quant.sh, first reproducing the GPT-2 and GPT-J runs described. Compare the saved model size and GPU memory after loading with the original model. Done means determining whether the script produces a quantized model and documenting or fixing the discrepancy with a focused check.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning, performance
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.