[Bugfix][Intel] Fix XPU Dockerfile Build

.dockerignore

@@ -1,4 +1,7 @@
-vllm/*.so
+/.github/
 /.venv
 /build
+/docs
 dist
+Dockerfile*
+vllm/*.so

Dockerfile.xpu

@@ -1,4 +1,4 @@
-FROM intel/oneapi-basekit:2024.1.0-devel-ubuntu20.04
+FROM intel/oneapi-basekit:2024.1.0-devel-ubuntu22.04 AS vllm-base
 
 RUN wget -O- https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB | gpg --dearmor | tee /usr/share/keyrings/intel-oneapi-archive-keyring.gpg > /dev/null && \
     echo "deb [signed-by=/usr/share/keyrings/intel-oneapi-archive-keyring.gpg] https://apt.repos.intel.com/oneapi all main " | tee /etc/apt/sources.list.d/oneAPI.list && \
@@ -8,15 +8,37 @@ RUN wget -O- https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRO
     echo "deb [arch=amd64,i386 signed-by=/usr/share/keyrings/intel-graphics.gpg] https://repositories.intel.com/graphics/ubuntu jammy arc" | tee /etc/apt/sources.list.d/intel.gpu.jammy.list && \
     chmod 644 /usr/share/keyrings/intel-graphics.gpg
 
-RUN apt-get update  -y \
-&& apt-get install -y curl libicu70 lsb-release git wget vim numactl python3 python3-pip
+RUN apt-get update -y && \
+    apt-get install -y --no-install-recommends --fix-missing \
+    curl \
+    git \
+    libsndfile1 \
+    lsb-release \
+    numactl \
+    python3 \
+    python3-dev \
+    python3-pip \
+    # vim \
+    wget
+
+WORKDIR /workspace/vllm
+COPY requirements-xpu.txt /workspace/vllm/requirements-xpu.txt
+COPY requirements-common.txt /workspace/vllm/requirements-common.txt
+
+RUN pip install --no-cache-dir -r requirements-xpu.txt
 
 COPY ./ /workspace/vllm
 
-WORKDIR /workspace/vllm
+ENV VLLM_TARGET_DEVICE=xpu
+RUN python3 setup.py install
+
+FROM vllm-base AS vllm-openai
 
-RUN pip install -v -r requirements-xpu.txt
+# install additional dependencies for openai api server
+RUN --mount=type=cache,target=/root/.cache/pip \
+    pip install accelerate hf_transfer 'modelscope!=1.15.0'
 
-RUN VLLM_TARGET_DEVICE=xpu python3 setup.py install
+ENV VLLM_USAGE_SOURCE production-docker-image \
+    TRITON_XPU_PROFILE 1
 
-CMD ["/bin/bash"]
+ENTRYPOINT ["python3", "-m", "vllm.entrypoints.openai.api_server"]

requirements-common.txt

@@ -1,11 +1,11 @@
 psutil
 sentencepiece  # Required for LLaMA tokenizer.
 numpy < 2.0.0
-requests
+requests >= 2.26.0
 tqdm
 py-cpuinfo
 transformers >= 4.43.2  # Required for Chameleon and Llama 3.1 hotfox.
-tokenizers >= 0.19.1  # Required for Llama 3.
+tokenizers == 0.19.1  # Required for Llama 3.
 protobuf # Required by LlamaTokenizer.
 fastapi
 aiohttp

requirements-xpu.txt

@@ -1,11 +1,13 @@
 # Common dependencies
 -r requirements-common.txt
 
-setuptools < 70.0.0 # IPEX's torch have some dependency. to be removed.
+setuptools == 69.5.1 # IPEX XPU dependency
 
-torch @ https://intel-extension-for-pytorch.s3.amazonaws.com/ipex_dev/xpu/torch-2.1.0.post1%2Bcxx11.abi-cp310-cp310-linux_x86_64.whl
-intel_extension_for_pytorch @ https://intel-extension-for-pytorch.s3.amazonaws.com/ipex_dev/xpu/intel_extension_for_pytorch-2.1.30a0-cp310-cp310-linux_x86_64.whl
-oneccl_bind_pt @ https://intel-extension-for-pytorch.s3.amazonaws.com/ipex_stable/xpu/oneccl_bind_pt-2.1.200%2Bxpu-cp310-cp310-linux_x86_64.whl
-
-triton @ https://github.com/intel/intel-xpu-backend-for-triton/releases/download/v2.1.0/triton-2.1.0-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
+torch == 2.1.0.post3+cxx11.abi
+torchvision == 0.16.0.post3+cxx11.abi
+torchaudio == 2.1.0.post3+cxx11.abi
+intel_extension_for_pytorch == 2.1.40+xpu
+oneccl_bind_pt == 2.1.400+xpu
 
+triton-xpu @ https://github.com/intel/intel-xpu-backend-for-triton/releases/download/v3.0.0b2/triton_xpu-3.0.0b2-cp310-cp310-linux_x86_64.whl
+--extra-index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/us/

setup.py

@@ -413,6 +413,8 @@ def _read_requirements(filename: str) -> List[str]:
         for line in requirements:
             if line.startswith("-r "):
                 resolved_requirements += _read_requirements(line.split()[1])
+            elif line.startswith("--"):
+                continue
             else:
                 resolved_requirements.append(line)
         return resolved_requirements

vllm/platforms/__init__.py

@@ -42,6 +42,15 @@
 except Exception:
     pass
 
+is_xpu = False
+
+try:
+    import torch
+    if hasattr(torch, 'xpu') and torch.xpu.is_available():
+        is_xpu = True
+except AttributeError:
+    pass
+
 if is_tpu:
     # people might install pytorch built with cuda but run on tpu
     # so we need to check tpu first
@@ -53,6 +62,9 @@
 elif is_rocm:
     from .rocm import RocmPlatform
     current_platform = RocmPlatform()
+elif is_xpu:
+    from .xpu import XPUPlatform
+    current_platform = XPUPlatform()
 else:
     current_platform = UnspecifiedPlatform()
 

vllm/platforms/interface.py

@@ -8,6 +8,7 @@ class PlatformEnum(enum.Enum):
     CUDA = enum.auto()
     ROCM = enum.auto()
     TPU = enum.auto()
+    XPU = enum.auto()
     UNSPECIFIED = enum.auto()
 
 
@@ -23,6 +24,9 @@ def is_rocm(self) -> bool:
     def is_tpu(self) -> bool:
         return self._enum == PlatformEnum.TPU
 
+    def is_xpu(self) -> bool:
+        return self._enum == PlatformEnum.XPU
+
     @staticmethod
     def get_device_capability(device_id: int = 0) -> Tuple[int, int]:
         raise NotImplementedError

vllm/platforms/xpu.py

@@ -0,0 +1,17 @@
+from typing import Tuple
+
+import torch
+
+from .interface import Platform, PlatformEnum
+
+
+class XPUPlatform(Platform):
+    _enum = PlatformEnum.XPU
+
+    @staticmethod
+    def get_device_capability(device_id: int = 0) -> Tuple[int, int]:
+        return torch.xpu.get_device_capability(device_id)
+
+    @staticmethod
+    def get_device_name(device_id: int = 0) -> str:
+        return torch.xpu.get_device_name(device_id)