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 some issues,still have some issues

wenhuach215ba94a178117
diff --git a/auto_round/autoround.py b/auto_round/autoround.py
index dd9e37b81..db736d1b7 100644
--- a/auto_round/autoround.py
+++ b/auto_round/autoround.py
@@ -1043,7 +1043,7 @@ class AutoRound(object):
Returns:
The specified optimizer.
"""
- from sign_sgd import SGD
+ from auto_round.sign_sgd import SGD
return SGD
diff --git a/examples/main.py b/examples/main.py
index a1a0bb766..7e461d6b1 100644
--- a/examples/main.py
+++ b/examples/main.py
@@ -45,7 +45,7 @@ if __name__ == '__main__':
parser.add_argument("--sym", action='store_true',
help=" sym quantization")
- parser.add_argument("--iters", default=400, type=int,
+ parser.add_argument("--iters", default=200, type=int,
help=" iters")
parser.add_argument("--use_quant_input", action='store_true',