[V0 deprecation] Remove VLLM_USE_V1 usage in most modules (#27955)

Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
This commit is contained in:
wangxiyuan
2025-11-04 20:51:16 -08:00
committed by GitHub
parent 878fd5a16f
commit 428bc7bf1c
19 changed files with 107 additions and 238 deletions
-13
View File
@@ -9,7 +9,6 @@ import numpy as np
import numpy.typing as npt
from PIL import Image
import vllm.envs as envs
from vllm.config.multimodal import (
AudioDummyOptions,
BaseDummyOptions,
@@ -306,18 +305,6 @@ class MultiModalProfiler(Generic[_I]):
if processor.pad_dummy_encoder_prompt:
num_tokens_to_pad = max(total_len, seq_len) - total_len
encoder_prompt_token_ids.extend([0] * num_tokens_to_pad)
# NOTE: Whisper allows total_len > seq_len.
elif total_len > seq_len and not envs.VLLM_USE_V1:
# `max_num_batched_tokens` is defined by `SchedulerConfig`
logger.warning_once(
"The encoder sequence length used for profiling (max_num_batched_tokens / max_num_seqs = %d) " # noqa: E501
"is too short to hold the multi-modal embeddings in the worst case (%d tokens in total, out of which %s are reserved for multi-modal embeddings). " # noqa: E501
"This may cause certain multi-modal inputs to fail during inference, even when the input text is short. " # noqa: E501
"To avoid this, you should increase `max_model_len`, reduce `max_num_seqs`, and/or reduce `mm_counts`.", # noqa: E501
seq_len,
total_len,
str(self._get_mm_num_tokens(mm_inputs)),
)
return DummyEncoderData(encoder_prompt_token_ids)