Skip to content

upgrade: vllm==0.10.2 #246

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
junya-takayama merged 5 commits into from
Sep 22, 2025
Merged

upgrade: vllm==0.10.2 #246

Show file tree
Hide file tree
Changes from 4 commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
90d094d
Merge branch 'main' of github.com:sbintuitions/flexeval
junya-takayama Sep 8, 2025
2b2ebc8
upgrade: vllm==0.10.1.1
junya-takayama Sep 10, 2025
d4fcf64
use vllm==0.10.2
junya-takayama Sep 22, 2025
f9fe3ab
Adapting to interface changes in LLM.generate()
junya-takayama Sep 22, 2025
7e38325
Avoid passing as a keyword argument
junya-takayama Sep 22, 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
10 changes: 6 additions & 4 deletions flexeval/core/language_model/vllm_model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -187,15 +187,16 @@ def _batch_complete_text(
)

from vllm import RequestOutput, SamplingParams
from vllm.inputs import TokensPrompt

prompt_token_ids: list[list[int]] = model_inputs.input_ids
batch_prompt_token_ids: list[list[int]] = model_inputs.input_ids
sampling_params: list[SamplingParams] = []
skip_flag_list: list[bool] = []
for i, input_ids in enumerate(model_inputs.input_ids):
remaining = self.model_limit_tokens - len(input_ids)
instance_gen_kwargs = gen_kwargs.copy()
if remaining <= 0:
prompt_token_ids[i] = input_ids[:1]
batch_prompt_token_ids[i] = input_ids[:1]
instance_gen_kwargs["max_tokens"] = 1
msg = (
f"Received input that is longer than `model_limit_tokens = {self.model_limit_tokens}`. "
Expand All @@ -208,7 +209,7 @@ def _batch_complete_text(
skip_flag_list.append(remaining <= 0)

vllm_outputs: list[RequestOutput] = self.llm.generate(
prompt_token_ids=prompt_token_ids,
prompts=[TokensPrompt(prompt_token_ids=prompt_token_ids) for prompt_token_ids in batch_prompt_token_ids],
sampling_params=sampling_params,
use_tqdm=False,
)
Expand Down Expand Up @@ -309,6 +310,7 @@ def _batch_compute_log_probs(
sequence_length = max([len(input_ids) for input_ids in batch_input_ids])

from vllm import RequestOutput, SamplingParams
from vllm.inputs import TokensPrompt
from vllm.sequence import Logprob

sampling_params = SamplingParams(temperature=0.0, max_tokens=1, prompt_logprobs=1)
Expand All @@ -323,7 +325,7 @@ def _batch_compute_log_probs(
for chunk_input_ids in chunk_batch_input_ids
]
chunk_batch_outputs: list[RequestOutput] = self.llm.generate(
prompt_token_ids=chunk_batch_input_ids,
prompts=[TokensPrompt(prompt_token_ids=prompt_token_ids) for prompt_token_ids in chunk_batch_input_ids],

@junya-takayama junya-takayama Sep 22, 2025

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

prompt_token_ids in LLM.generate was removed.
vllm-project/vllm#18800

sampling_params=sampling_params,
use_tqdm=False,
)
Expand Down
Loading