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 awq exporting (#358)

Wenhua Chengca9e655ef8f5
diff --git a/auto_round/export/export_to_awq/export.py b/auto_round/export/export_to_awq/export.py
index bba2423a1..09f454837 100644
--- a/auto_round/export/export_to_awq/export.py
+++ b/auto_round/export/export_to_awq/export.py
@@ -123,7 +123,7 @@ def save_quantized_as_autoawq(output_dir, inplace=True, **kwargs):
with ThreadPoolExecutor(max_workers=2) as executor:
with tqdm(total=len(names), leave=True) as pbar:
def wrapper(name):
- pack_layer(name, model, layer_config, backend, pbar)
+ pack_layer(name, compressed_model, layer_config, backend, pbar)
for _ in executor.map(wrapper, names):
pass