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.pyindex 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