mlcommons / mlcommons/inference
[Small LLM] Max tokens fixed at 128?
Open
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 1.6k
- Forks
- 650
- Avg merge
- 1d 22h
- Merged PRs (30d)
- 6
Description
The reference implementation appears to fix the maximum number of output tokens to 128.
Despite this, the reference scores:
{
'rouge1': 38.7792,
'rouge2': 15.9075,
'rougeL': 24.4957,
'rougeLsum': 35.793,
'gen_len': 8167644,
'gen_num': 13368,
}
imply that the average number of output tokens is ~611 (= gen_len / gen_num).
What's going on?
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/llama3.1-8b/SUT_VLLM.py around the referenced line 75, then compare its token limit with the accuracy-target scores and generation statistics linked in the issue. Trace how gen_len and gen_num are produced, and document or correct the explanation for the apparent mismatch so the reported behavior is consistent.
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
- 30/100