PTQ-Level Efficiency, QAT-Level Accuracy | 训练后量化的效率,量化感知训练的精度
set disable_opt_rtn as default for int8 RTN (#1201)
Weiwei56ad9d8292ff
diff --git a/auto_round/compressors/base.py b/auto_round/compressors/base.pyindex d9f1e6054..c44c752ee 100644--- a/auto_round/compressors/base.py+++ b/auto_round/compressors/base.py@@ -354,6 +354,9 @@ class BaseCompressor(object):logger.warning("for bits <= 2, it is recommended to enable `auto-round-best` " "and turn on `--enable_alg_ext` ")+ if self.bits == 8 and self.iters == 0 and self.data_type == "int" and not disable_opt_rtn:+ logger.warning("for INT8 RTN quantization, set `--disable_opt_rtn` as default.")+ disable_opt_rtn = Trueself.minmax_lr = minmax_lr or self.lrself.enable_alg_ext = enable_alg_ext