Skip to content

[Frontend] Add optional token-level progress bar to LLM.beam_search #19301

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
merged 7 commits into from
Jun 19, 2025
+14 −1
Merged

[Frontend] Add optional token-level progress bar to LLM.beam_search #19301

Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
649c6ef
Add progress bar for `LLM.beam_search`
NekoMimiUnagi Jun 6, 2025
3c3c4ac
Update the comment for `use_tqdm`
NekoMimiUnagi Jun 6, 2025
cf90871
Update llm.py based on Code Review
NekoMimiUnagi Jun 6, 2025
99d7ab5
Fix use_tqdm bug and reformat the file
NekoMimiUnagi Jun 7, 2025
418f086
Update vllm/entrypoints/llm.py
NekoMimiUnagi Jun 9, 2025
00f12b4
update the log based on the suggestion from @aarnphm
NekoMimiUnagi Jun 17, 2025
3e27f64
update code format
NekoMimiUnagi Jun 17, 2025
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
15 changes: 14 additions & 1 deletion vllm/entrypoints/llm.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -531,6 +531,7 @@ def beam_search(
prompts: list[Union[TokensPrompt, TextPrompt]],
params: BeamSearchParams,
lora_request: Optional[Union[list[LoRARequest], LoRARequest]] = None,
use_tqdm: bool = False,
) -> list[BeamSearchOutput]:
"""
Generate sequences using beam search.
Expand All @@ -540,6 +541,7 @@ def beam_search(
of token IDs.
params: The beam search parameters.
lora_request: LoRA request to use for generation, if any.
use_tqdm: Whether to use tqdm to display the progress bar.
"""
# TODO: how does beam search work together with length penalty,
# frequency, penalty, and stopping criteria, etc.?
Expand Down Expand Up @@ -602,7 +604,18 @@ def create_tokens_prompt_from_beam(
**mm_kwargs,
), )

for _ in range(max_tokens):
token_iter = range(max_tokens)
if use_tqdm:
token_iter = tqdm(token_iter,
desc="Beam search",
unit="token",
unit_scale=False)
logger.warning(
"The progress bar shows the upper bound on token steps and "
"may finish early due to stopping conditions. It does not "
"reflect instance-level progress.")

for _ in token_iter:
all_beams: list[BeamSearchSequence] = list(
sum((instance.beams for instance in instances), []))
pos = [0] + list(
Expand Down