Skip to content

PTQ-Level Efficiency, QAT-Level Accuracy | 训练后量化的效率,量化感知训练的精度

Remove cuda UT dependencies version restrictions (#651)

Sun, Xuehao35cf154f22d4
diff --git a/test/test_cuda/requirements_vlm.txt b/test/test_cuda/requirements_vlm.txt
index cfe352536..0bfb0c39f 100644
--- a/test/test_cuda/requirements_vlm.txt
+++ b/test/test_cuda/requirements_vlm.txt
@@ -3,21 +3,21 @@
# pip install -v git+https://github.com/casper-hansen/[email protected] --no-build-isolation
accelerate
# autoawq
-bitsandbytes==0.44.0
+bitsandbytes
datasets
einops
-flash-attn==2.5.8
+flash-attn
intel-extension-for-transformers
lm-eval>=0.4.2,<0.5
numpy < 2.0
optimum
pandas
-protobuf==3.20.2
+protobuf
pillow
py-cpuinfo
-torch==2.3.0
+torch
torchvision
-triton==2.3.0
+triton
tqdm
-transformers==4.45.0
+transformers
xformers