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

fix critic disable_opt_rtn regression (#963)

Wenhua Cheng282aab668f3d
diff --git a/auto_round/compressors/base.py b/auto_round/compressors/base.py
index 869fc74de..c56e10750 100644
--- a/auto_round/compressors/base.py
+++ b/auto_round/compressors/base.py
@@ -141,7 +141,7 @@ class BaseCompressor(object):
device_map: Union[str, torch.device, int, dict] = 0,
enable_torch_compile: bool = False,
enable_alg_ext: bool = False,
- disable_opt_rtn: bool = True,
+ disable_opt_rtn: bool = False,
seed: int = 42,
**kwargs,
):