[ROCm][Bugfix] Patch for the `Multi-Modal Processor Test` group (#29702)

Signed-off-by: Andreas Karatzas <akaratza@amd.com>

docker/Dockerfile.rocm

@@ -65,6 +65,8 @@ COPY --from=build_vllm ${COMMON_WORKDIR}/vllm/tests /tests
 COPY --from=build_vllm ${COMMON_WORKDIR}/vllm/examples /examples
 COPY --from=build_vllm ${COMMON_WORKDIR}/vllm/docker/Dockerfile.rocm /docker/
 COPY --from=build_vllm ${COMMON_WORKDIR}/vllm/.buildkite /.buildkite
+# Centralized v1 package - copied to both test and final stages
+COPY --from=build_vllm ${COMMON_WORKDIR}/vllm/vllm/v1 /vllm_v1
 
 # -----------------------
 # Test vLLM image
@@ -88,10 +90,22 @@ COPY --from=build_vllm ${COMMON_WORKDIR}/vllm /vllm-workspace
 
 # install development dependencies (for testing)
 RUN cd /vllm-workspace \
-    && rm -rf vllm \
     && python3 -m pip install -e tests/vllm_test_utils \
     && python3 -m pip install pytest-shard
 
+# enable fast downloads from hf (for testing)
+RUN --mount=type=cache,target=/root/.cache/uv \
+    uv pip install --system hf_transfer
+ENV HF_HUB_ENABLE_HF_TRANSFER=1
+
+# Copy in the v1 package
+COPY --from=export_vllm /vllm_v1 /usr/local/lib/python${PYTHON_VERSION}/dist-packages/vllm/v1
+
+# Source code is used in the `python_only_compile.sh` test
+# We hide it inside `src/` so that this source code
+# will not be imported by other tests
+RUN mkdir src && mv vllm src/vllm
+
 # -----------------------
 # Final vLLM image
 FROM base AS final
@@ -116,6 +130,9 @@ RUN --mount=type=bind,from=export_vllm,src=/,target=/install \
     && pip uninstall -y vllm \
     && uv pip install --system *.whl
 
+# Copy in the v1 package
+COPY --from=export_vllm /vllm_v1 /usr/local/lib/python${PYTHON_VERSION}/dist-packages/vllm/v1
+
 ARG COMMON_WORKDIR
 
 # Copy over the benchmark scripts as well

docker/Dockerfile.rocm_base

@@ -5,6 +5,8 @@ ARG PYTORCH_BRANCH="1c57644d"
 ARG PYTORCH_VISION_BRANCH="v0.23.0"
 ARG PYTORCH_REPO="https://github.com/ROCm/pytorch.git"
 ARG PYTORCH_VISION_REPO="https://github.com/pytorch/vision.git"
+ARG PYTORCH_AUDIO_BRANCH="v2.9.0"
+ARG PYTORCH_AUDIO_REPO="https://github.com/pytorch/audio.git"
 ARG FA_BRANCH="0e60e394"
 ARG FA_REPO="https://github.com/Dao-AILab/flash-attention.git"
 ARG AITER_BRANCH="59bd8ff2"
@@ -23,6 +25,7 @@ ENV AITER_ROCM_ARCH=gfx942;gfx950
 ENV HSA_NO_SCRATCH_RECLAIM=1
 
 ARG PYTHON_VERSION=3.12
+ENV PYTHON_VERSION=${PYTHON_VERSION}
 
 RUN mkdir -p /app
 WORKDIR /app
@@ -45,6 +48,7 @@ RUN apt-get update -y \
     && python3 --version && python3 -m pip --version
 
 RUN pip install -U packaging 'cmake<4' ninja wheel 'setuptools<80' pybind11 Cython
+RUN apt-get update && apt-get install -y libjpeg-dev libsox-dev libsox-fmt-all sox && rm -rf /var/lib/apt/lists/*
 
 FROM base AS build_triton
 ARG TRITON_BRANCH
@@ -66,11 +70,14 @@ RUN mkdir -p /app/install && cp /opt/rocm/share/amd_smi/dist/*.whl /app/install
 FROM base AS build_pytorch
 ARG PYTORCH_BRANCH
 ARG PYTORCH_VISION_BRANCH
+ARG PYTORCH_AUDIO_BRANCH
 ARG PYTORCH_REPO
 ARG PYTORCH_VISION_REPO
+ARG PYTORCH_AUDIO_REPO
+
 RUN git clone ${PYTORCH_REPO} pytorch
-RUN cd pytorch && git checkout ${PYTORCH_BRANCH} && \
-    pip install -r requirements.txt && git submodule update --init --recursive \
+RUN cd pytorch && git checkout ${PYTORCH_BRANCH} \
+    && pip install -r requirements.txt && git submodule update --init --recursive \
     && python3 tools/amd_build/build_amd.py \
     && CMAKE_PREFIX_PATH=$(python3 -c 'import sys; print(sys.prefix)') python3 setup.py bdist_wheel --dist-dir=dist \
     && pip install dist/*.whl
@@ -78,8 +85,15 @@ RUN git clone ${PYTORCH_VISION_REPO} vision
 RUN cd vision && git checkout ${PYTORCH_VISION_BRANCH} \
     && python3 setup.py bdist_wheel --dist-dir=dist \
     && pip install dist/*.whl
+RUN git clone ${PYTORCH_AUDIO_REPO} audio
+RUN cd audio && git checkout ${PYTORCH_AUDIO_BRANCH} \
+    && git submodule update --init --recursive \
+    && pip install -r requirements.txt \
+    && python3 setup.py bdist_wheel --dist-dir=dist \
+    && pip install dist/*.whl
 RUN mkdir -p /app/install && cp /app/pytorch/dist/*.whl /app/install \
-    && cp /app/vision/dist/*.whl /app/install
+    && cp /app/vision/dist/*.whl /app/install \
+    && cp /app/audio/dist/*.whl /app/install
 
 FROM base AS build_fa
 ARG FA_BRANCH
@@ -130,6 +144,8 @@ ARG PYTORCH_BRANCH
 ARG PYTORCH_VISION_BRANCH
 ARG PYTORCH_REPO
 ARG PYTORCH_VISION_REPO
+ARG PYTORCH_AUDIO_BRANCH
+ARG PYTORCH_AUDIO_REPO
 ARG FA_BRANCH
 ARG FA_REPO
 ARG AITER_BRANCH
@@ -141,7 +157,9 @@ RUN echo "BASE_IMAGE: ${BASE_IMAGE}" > /app/versions.txt \
     && echo "PYTORCH_VISION_BRANCH: ${PYTORCH_VISION_BRANCH}" >> /app/versions.txt \
     && echo "PYTORCH_REPO: ${PYTORCH_REPO}" >> /app/versions.txt \
     && echo "PYTORCH_VISION_REPO: ${PYTORCH_VISION_REPO}" >> /app/versions.txt \
+    && echo "PYTORCH_AUDIO_BRANCH: ${PYTORCH_AUDIO_BRANCH}" >> /app/versions.txt \
+    && echo "PYTORCH_AUDIO_REPO: ${PYTORCH_AUDIO_REPO}" >> /app/versions.txt \
     && echo "FA_BRANCH: ${FA_BRANCH}" >> /app/versions.txt \
     && echo "FA_REPO: ${FA_REPO}" >> /app/versions.txt \
     && echo "AITER_BRANCH: ${AITER_BRANCH}" >> /app/versions.txt \
-    && echo "AITER_REPO: ${AITER_REPO}" >> /app/versions.txt
+    && echo "AITER_REPO: ${AITER_REPO}" >> /app/versions.txt
\ No newline at end of file

requirements/rocm-test.txt

 # Common dependencies
 -r common.txt
+
+# Test infrastructure
 tblib==3.1.0
-bm25s==0.2.13
-pystemmer==3.0.0
+pytest==8.3.5
+pytest-asyncio==0.24.0
+pytest-timeout==2.3.1
+pytest-cov==6.3.0
+pytest-forked==1.6.0
+pytest-rerunfailures==14.0
+pytest-shard==0.1.2
+
+# Async/HTTP dependencies
+anyio==4.6.2.post1
+    # via httpx, starlette
+aiohttp==3.13.0
+    # via gpt-oss
+httpx==0.27.2
+    # HTTP testing
 
-# Entrypoints test
-# librosa==0.10.2.post1 # required by audio tests in entrypoints/openai
+# Audio processing dependencies
 audioread==3.0.1
+    # via librosa
 cffi==1.17.1
+    # via soundfile
 decorator==5.2.1
+    # via librosa
 lazy-loader==0.4
+    # via librosa
 platformdirs==4.3.6
+    # via pooch
 pooch==1.8.2
-#pycparse==2.22
+    # via librosa
 soundfile==0.13.1
+    # via librosa
 soxr==0.5.0.post1
+    # via librosa
 librosa==0.10.2.post1
 
-# Entrypoints test
-#vllm[video] # required by entrypoints/openai/test_video.py
-decord==0.6.0
-
-# Entrypoints test
-#sentence-transformers # required by entrypoints/openai/test_score.py
-sentence-transformers==3.4.1
-
-# Basic Models Test
-matplotlib==3.10.3
+# Retrieval and search
+bm25s==0.2.13
+    # via mteb
+pystemmer==3.0.0
+    # via mteb
 
-# Multi-Modal Models Test (Extended) 3
+# Multi-modal processing
 blobfile==3.0.0
+    # Multi-Modal Models Test
+decord==0.6.0
+    # video processing, required by entrypoints/openai/test_video.py
 
-# Required for openai schema test.
+# OpenAI compatibility and testing
+gpt-oss==0.0.8
+    # OpenAI compatibility tests
 schemathesis==3.39.15
+    # OpenAI schema test
 
-# Required for mteb test
-mteb[bm25s]>=1.38.11, <2
-
-# Required for eval tests
+# Evaluation and benchmarking
 lm-eval[api] @ git+https://github.com/EleutherAI/lm-evaluation-harness.git@206b7722158f58c35b7ffcd53b035fdbdda5126d
 
-# Required for multiprocessed tests that use spawn method
+# Required for multiprocessed tests that use spawn method, Datasets and Evaluate Test
 multiprocess==0.70.16
 
 # Plugins test
 terratorch @ git+https://github.com/IBM/terratorch.git@07184fcf91a1324f831ff521dd238d97fe350e3e
 torchgeo==0.7.0
+    # via terratorch
+# MTEB Benchmark Test
+mteb==2.1.2
+
+# Data processing
+xgrammar @ git+https://github.com/mlc-ai/xgrammar.git@eafd4db51b78acc64b3f0764ef27dfd206c28628
+    # Test async scheduling
+
+# Utilities
+num2words==0.5.14
+    # via lm-eval
+pqdm==0.2.0
+    # via lm-eval
 
 # Required for suffix decoding test
 arctic-inference == 0.1.1
+# Required for Nemotron test
+open-clip-torch==2.32.0

tests/models/multimodal/processing/test_tensor_schema.py

@@ -30,6 +30,7 @@ from vllm.model_executor.models.interfaces import (
 from vllm.multimodal import MULTIMODAL_REGISTRY, BatchedTensorInputs
 from vllm.multimodal.processing import BaseMultiModalProcessor, InputProcessingContext
 from vllm.multimodal.utils import group_mm_kwargs_by_modality
+from vllm.platforms import current_platform
 from vllm.transformers_utils.tokenizer import cached_tokenizer_from_config
 from vllm.utils.collection_utils import is_list_of
 from vllm.utils.torch_utils import set_default_torch_dtype
@@ -176,6 +177,12 @@ def test_model_tensor_schema(model_id: str):
         exist_overrides=model_info.hf_overrides,
     )
 
+    # ROCm: Detect if model uses AWQ quantization and set appropriate dtype
+    if "awq" in model_id.lower() and current_platform.is_rocm():
+        dtype = "float16"
+    else:
+        dtype = model_info.dtype
+
     model_config = ModelConfig(
         model_id,
         tokenizer=model_info.tokenizer or model_id,
@@ -187,7 +194,7 @@ def test_model_tensor_schema(model_id: str):
         enable_prompt_embeds=model_info.require_embed_inputs,
         enable_mm_embeds=model_info.require_embed_inputs,
         enforce_eager=model_info.enforce_eager,
-        dtype=model_info.dtype,
+        dtype=dtype,
     )
 
     model_cls = MULTIMODAL_REGISTRY._get_model_cls(model_config)