mlcommons / mlcommons/inference
Performance improvement- GPT-J and BERT Offline scenario
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 1.6k
- Forks
- 650
- Avg merge
- 1d 22h
- Merged PRs (30d)
- 6
Description
The current implementation of GPT-J and BERT carries out the prediction in sequential manner. Could the performance of GPT-J and BERT be improved by implementing parallel processing through threads rather than sequential processing?
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start with language/gpt-j/backend.py at line 72 and language/bert/pytorch_SUT.py at line 68 to trace the sequential prediction path in the Offline scenario. Compare how each path handles requests before deciding how to measure parallel execution. Done means both GPT-J and BERT use the agreed threaded approach and show improved offline performance.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python, pytorch
- Domain
- machine-learning, performance
- Issue type
- Feature
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100