@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / 80cbe10c59
Fork 0

A high-throughput and memory-efficient inference and serving engine for LLMs

Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git

Synced Nov 16, 19:30 UTC

[OpenVINO] migrate to latest dependencies versions (#7251)

Ilya Lavrenov committed Aug 7, 2024 at 20:49 UTC 80cbe10c59a3354d0fc7c841fdc6422fc64899aa
4 files changed +6 -33
Dockerfile.openvino
+1 -1
@@ -21,7 +21,7 @@ COPY setup.py /workspace/vllm/
21 # install build requirements
22 RUN PIP_EXTRA_INDEX_URL="https://download.pytorch.org/whl/cpu" python3 -m pip install -r /workspace/vllm/requirements-build.txt
23 # build vLLM with OpenVINO backend
24 -RUN PIP_EXTRA_INDEX_URL="https://download.pytorch.org/whl/cpu https://storage.openvinotoolkit.org/simple/wheels/pre-release" VLLM_TARGET_DEVICE="openvino" python3 -m pip install /workspace/vllm/
24 +RUN PIP_EXTRA_INDEX_URL="https://download.pytorch.org/whl/cpu" VLLM_TARGET_DEVICE="openvino" python3 -m pip install /workspace/vllm/
25
26 COPY examples/ /workspace/vllm/examples
27 COPY benchmarks/ /workspace/vllm/benchmarks
docs/source/getting_started/openvino-installation.rst
+1 -1
@@ -57,7 +57,7 @@ Install from source
57
58 .. code-block:: console
59
60 - $ PIP_EXTRA_INDEX_URL="https://download.pytorch.org/whl/cpu https://storage.openvinotoolkit.org/simple/wheels/pre-release" VLLM_TARGET_DEVICE=openvino python -m pip install -v .
60 + $ PIP_EXTRA_INDEX_URL="https://download.pytorch.org/whl/cpu" VLLM_TARGET_DEVICE=openvino python -m pip install -v .
61
62 .. _openvino_backend_performance_tips:
63
requirements-openvino.txt
+3 -30
@@ -1,34 +1,7 @@
1 # Common dependencies
2 -# -r requirements-common.txt
3 -# TODO: remove temporary copy of all common dependencies once Optimum Intel will support Transformers >= 4.43.2
4 -cmake >= 3.21
5 -ninja # For faster builds.
6 -psutil
7 -sentencepiece # Required for LLaMA tokenizer.
8 -numpy < 2.0.0
9 -requests
10 -tqdm
11 -py-cpuinfo
12 -transformers < 4.43
13 -tokenizers >= 0.19.1 # Required for Llama 3.
14 -fastapi
15 -aiohttp
16 -openai
17 -uvicorn[standard]
18 -pydantic >= 2.0 # Required for OpenAI server.
19 -pillow # Required for image processing
20 -prometheus_client >= 0.18.0
21 -prometheus-fastapi-instrumentator >= 7.0.0
22 -tiktoken >= 0.6.0 # Required for DBRX tokenizer
23 -lm-format-enforcer == 0.10.3
24 -outlines >= 0.0.43, < 0.1 # Requires torch >= 2.1.0
25 -typing_extensions
26 -filelock >= 3.10.4 # filelock starts to support `mode` argument from 3.10.4
27 -pyzmq
28 -gguf == 0.9.1
2 +-r requirements-common.txt
3
4 # OpenVINO dependencies
5 torch >= 2.1.2
32 -openvino ~= 2024.3.0.dev
33 -openvino-tokenizers[transformers] ~= 2024.3.0.0.dev
34 -optimum-intel[openvino] >= 1.18.1
6 +openvino ~= 2024.3.0
7 +optimum-intel[openvino] >= 1.18.2
setup.py
+1 -1
@@ -272,7 +272,7 @@ def _build_custom_ops() -> bool:
272
273
274 def _build_core_ext() -> bool:
275 - return not _is_neuron() and not _is_tpu()
275 + return not _is_neuron() and not _is_tpu() and not _is_openvino()
276
277
278 def get_hipcc_rocm_version():