mlcommons / mlcommons/inference
Is beam search strategy mandatory in gptj-6b benchmark?
Open
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 1.6k
- Forks
- 650
- Avg merge
- 1d 22h
- Merged PRs (30d)
- 6
Description
hi, I found that vllm will drop beam search feature in future releases due to complicated implementation and beam search is not so popular in real use.
So could I know if gptj-6b benchmark test will drop beam search strategy too?
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
No file, test, or entry point is named. Start by locating the GPT-J-6B benchmark definition and its beam-search configuration, then compare that behavior with the applicable MLPerf inference requirements. Done would require a resolved decision on whether beam search remains mandatory and, if a change is approved, an explicitly scoped implementation and validation plan.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- machine-learning
- Domain
- machine-learning, testing-qa
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 20/100