Update libra/model/builder.py

libra/model/builder.py

@@ -77,7 +77,7 @@ def load_pretrained_model(model_path, model_base, model_name, device="cpu"):
             model = LibraLlamaForCausalLM.from_pretrained(model_base, low_cpu_mem_usage=True, config=cfg_pretrained, **kwargs)
 
             mm_projector_weights = torch.load(os.path.join(model_path, 'mm_projector.bin'), map_location='cpu')
-            mm_projector_weights = {k: v.to(torch.float16) for k, v in mm_projector_weights.items()}
+            mm_projector_weights = {k: v.to(torch.bfloat16) for k, v in mm_projector_weights.items()}
             model.load_state_dict(mm_projector_weights, strict=False)
         else:
             tokenizer = AutoTokenizer.from_pretrained(model_path, use_fast=False)
@@ -94,11 +94,11 @@ def load_pretrained_model(model_path, model_base, model_name, device="cpu"):
             print(f"Merging weights")
             model = model.merge_and_unload()
             print('Convert to FP16...')
-            model.to(torch.float16)
+            model.to(torch.bfloat16)
         else:
             use_fast = False
             tokenizer = AutoTokenizer.from_pretrained(model_path, use_fast=False)
-            model = AutoModelForCausalLM.from_pretrained(model_path, low_cpu_mem_usage=True,torch_dtype=torch.bfloat16, **kwargs)
+            model = AutoModelForCausalLM.from_pretrained(model_path, low_cpu_mem_usage=True, **kwargs)
 
     image_processor = None