mirror of
https://github.com/vllm-project/vllm-ascend.git
synced 2025-10-20 13:43:53 +08:00
2b726d8f90
1. vLLM commit45badd05d0changed the pooling check logic which broken vLLM Ascend. 2. vLLM commit3e04107d97requires higher version of transformers. The transformers version bug has been fixed bye936e401de. We can safe to remove the version limit now. 3. vLLM commit217937221badded a new input `enable_eplb` for FusedMoe Ops This PR fix the broken CI. - vLLM version: v0.9.2 - vLLM main:6a971ed692Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
28 lines
423 B
Plaintext
28 lines
423 B
Plaintext
# Should be mirrored in pyporject.toml
|
|
cmake>=3.26
|
|
decorator
|
|
einops
|
|
numpy<2.0.0
|
|
packaging
|
|
pip
|
|
pybind11
|
|
pyyaml
|
|
scipy
|
|
setuptools>=64
|
|
setuptools-scm>=8
|
|
torch>=2.5.1
|
|
torchvision<0.21.0
|
|
wheel
|
|
|
|
# requirements for disaggregated prefill
|
|
msgpack
|
|
quart
|
|
|
|
# Required for N-gram speculative decoding
|
|
numba
|
|
|
|
# Install torch_npu
|
|
--pre
|
|
--extra-index-url https://mirrors.huaweicloud.com/ascend/repos/pypi
|
|
torch-npu==2.5.1.post1.dev20250619
|