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

Freeze compressed-tensors & llmc version temporary (#1562)

chen, suyue4cc0cce4b5d0
diff --git a/test/test_cpu/requirements.txt b/test/test_cpu/requirements.txt
index 367f0ffd8..80f981849 100644
--- a/test/test_cpu/requirements.txt
+++ b/test/test_cpu/requirements.txt
@@ -5,7 +5,7 @@ sentencepiece
torchvision
pillow
numba
-llmcompressor @ git+https://github.com/vllm-project/llm-compressor.git@main
+llmcompressor @ git+https://github.com/vllm-project/llm-compressor.git@370c04c34191c0e5fb88cf889e16578bf8c7fe20
compressed-tensors==0.14.1a20260313 # temporary pin for llmcompressor
lm_eval >= 0.4.10 # for transformers >= 5.0.0
diffusers
diff --git a/test/test_cuda/requirements_llmc.txt b/test/test_cuda/requirements_llmc.txt
index c82328a23..d03556a6b 100644
--- a/test/test_cuda/requirements_llmc.txt
+++ b/test/test_cuda/requirements_llmc.txt
@@ -1,3 +1,3 @@
-llmcompressor @ git+https://github.com/vllm-project/llm-compressor.git@main
+llmcompressor @ git+https://github.com/vllm-project/llm-compressor.git@370c04c34191c0e5fb88cf889e16578bf8c7fe20
compressed-tensors==0.14.1a20260313 # temporary pin for llmcompressor
parameterized