Bump vllm minimum version to 0.17.1 (#165)

Some latest enhancements we want to grab

Signed-off-by: Eric Curtin <eric.curtin@docker.com>

Author: ericcurtin

install.sh

@@ -123,7 +123,7 @@ main() {
 
   ensure_venv "$venv"
 
-  local vllm_v="0.14.1"
+  local vllm_v="0.17.1"
   local url_base="https://github.com/vllm-project/vllm/releases/download"
   local filename="vllm-$vllm_v.tar.gz"
   curl -OL $url_base/v$vllm_v/$filename

pyproject.toml

@@ -43,7 +43,7 @@ dependencies = [
 ]
 
 [project.optional-dependencies]
-vllm = ["vllm>=0.14.0"]
+vllm = ["vllm>=0.17.1"]
 stt = [
     # Speech-to-text audio processing (Whisper models)
     "librosa>=0.10.2",

vllm_metal/stt/hf_config.py

@@ -196,11 +196,11 @@ def _make_stub_class():
     )
     from vllm.multimodal.parse import MultiModalDataItems
     from vllm.multimodal.processing import (
+        BaseDummyInputsBuilder,
         BaseMultiModalProcessor,
         BaseProcessingInfo,
         PromptUpdate,
     )
-    from vllm.multimodal.profiling import BaseDummyInputsBuilder
     from vllm.multimodal.registry import _ProcessorFactories
     from vllm.tokenizers import get_tokenizer
 
@@ -228,6 +228,8 @@ def get_tokenizer(self):
             return self.ctx.get_tokenizer()
 
     class _DummyInputs(BaseDummyInputsBuilder[_Info]):
+        """Minimal dummy inputs builder for Qwen3-ASR audio."""
+
         def get_dummy_text(self, mm_counts: Mapping[str, int]) -> str:
             return "<|audio_pad|>"