[Bugfix] Add missing dynamic_arg_dims for Qwen3-ASR torch.compile (#35869)

Signed-off-by: Nathan Price <nathan@abridge.com>
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Co-authored-by: Isotr0py <mozf@mail2.sysu.edu.cn>

vllm/model_executor/models/qwen3_asr_realtime.py

@@ -22,7 +22,6 @@ from collections.abc import AsyncGenerator, Mapping
 import numpy as np
 import torch
 
-from vllm.compilation.decorators import support_torch_compile
 from vllm.config import ModelConfig, SpeechToTextConfig, VllmConfig
 from vllm.inputs.data import PromptType, TokensPrompt
 from vllm.logger import init_logger
@@ -177,7 +176,6 @@ class Qwen3ASRRealtimeMultiModalProcessor(Qwen3ASRMultiModalProcessor):
     info=Qwen3ASRProcessingInfo,
     dummy_inputs=Qwen3ASRDummyInputsBuilder,
 )
-@support_torch_compile
 class Qwen3ASRRealtimeGeneration(Qwen3ASRForConditionalGeneration, SupportsRealtime):
     realtime_max_tokens = 64