Skip to content

[Bugfix] Fixes prefix-repetition benchmark script #26828

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
ruisearch42 merged 1 commit into from
Oct 15, 2025
Merged

[Bugfix] Fixes prefix-repetition benchmark script #26828

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 4 additions & 3 deletions vllm/benchmarks/datasets.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2850,13 +2850,14 @@ def _generate_exact_length_tokens(target_length: int) -> list[int]:
requests = []
token_mismatch_total = 0
for _ in range(num_prefixes):
prefix_tokens = _generate_exact_length_tokens(prefix_len)
prefix_tokens, prefix_mismatch = _generate_exact_length_tokens(prefix_len)
token_mismatch_total += prefix_mismatch

for _ in range(prompts_per_prefix):
suffix_tokens, token_mistmatch = _generate_exact_length_tokens(
suffix_tokens, suffix_mismatch = _generate_exact_length_tokens(
suffix_len
)
token_mismatch_total += token_mistmatch
token_mismatch_total += suffix_mismatch
combined_tokens = prefix_tokens + suffix_tokens
prompt = tokenizer.decode(combined_tokens)
prompt_len = len(combined_tokens)
Expand Down
Loading