Merge daf6cc0c2e into 83e52f6a27

eval_llm.py

@@ -22,8 +22,7 @@ def init_model(args):
         ckp = f'./{args.save_dir}/{args.weight}_{args.hidden_size}{moe_suffix}.pth'
         model.load_state_dict(torch.load(ckp, map_location=args.device), strict=True)
         if args.lora_weight != 'None':
-            apply_lora(model)
-            load_lora(model, f'./{args.save_dir}/lora/{args.lora_weight}_{args.hidden_size}.pth')
+            merge_lora(model, f'./{args.save_dir}/lora/{args.lora_weight}_{args.hidden_size}.pth')
     else:
         model = AutoModelForCausalLM.from_pretrained(args.load_from, trust_remote_code=True)
     get_model_params(model, model.config)

model/model_lora.py

@@ -41,6 +41,23 @@ def load_lora(model, path):
             lora_state = {k.replace(f'{name}.lora.', ''): v for k, v in state_dict.items() if f'{name}.lora.' in k}
             module.lora.load_state_dict(lora_state)
 
+def merge_lora(model, path):
+    state_dict = torch.load(path, map_location=model.device)
+    # 移除可能的module前缀，确保key与模型层级名称一致
+    state_dict = {(k[7:] if k.startswith('module.') else k): v for k, v in state_dict.items()}
+
+    for name, module in model.named_modules():
+        if isinstance(module, nn.Linear) and module.weight.shape[0] == module.weight.shape[1]:
+            key_A = f"{name}.lora.A.weight"
+            key_B = f"{name}.lora.B.weight"
+            if key_A in state_dict and key_B in state_dict:
+                # 直接合并权重: W_new = W_old + B @ A
+                device = module.weight.device
+                A_weight = state_dict[key_A].to(device)
+                B_weight = state_dict[key_B].to(device)
+                module.weight.data += B_weight @ A_weight
+                
+    return model
 
 def save_lora(model, path):
     raw_model = getattr(model, '_orig_mod', model)