Skip to content

fix passing vllm model config's max_len to _generate_token_mode #3

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
wangshangsam merged 1 commit into from
Jan 29, 2026
Merged

fix passing vllm model config's max_len to _generate_token_mode #3

Show file tree
Hide file tree
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
9 changes: 4 additions & 5 deletions components/src/dynamo/vllm/handlers.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -164,11 +164,10 @@ def build_sampling_params(
provided_max_tokens = request.get("stop_conditions", {}).get("max_tokens", None)
token_ids = request.get("token_ids", [])
input_length = len(token_ids)
if model_max_len is not None and (provided_max_tokens is None):
# Ensure at least 1 token generation by default when possible
dynamic_default = max(1, model_max_len - input_length)
sampling_params.max_tokens = dynamic_default

dynamic_default = max(1, model_max_len - input_length)
model_config_max_tokens = default_sampling_params.get("max_tokens")
sampling_params.max_tokens = min(filter(lambda x: x is not None,
[provided_max_tokens, dynamic_default, model_config_max_tokens]))
return sampling_params


Expand Down
2 changes: 2 additions & 0 deletions examples/multimodal/utils/chat_processor.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -91,6 +91,8 @@ async def _parse_raw_request(
preprocess_result.engine_prompt["prompt_token_ids"]
)

default_max_tokens = min(default_max_tokens, self.model_config.max_model_len)

sampling_params = request.to_sampling_params(
default_max_tokens,
self.model_config.logits_processor_pattern,
Expand Down
Loading