Upload mmalaya_arch.py

mmalaya_arch.py

@@ -303,15 +303,15 @@ class MMAlayaMetaForCausalLM(ABC):
         stopping_criteria = KeywordsStoppingCriteria(
             [conv.sep2], 
             tokenizer, 
-            torch.tensor(input_ids, dtype=torch.long),
+            torch.tensor(input_ids, dtype=torch.long).unsqueeze(0),
             )
         # 加载图像
         image_processor = model.get_vision_tower().image_processor
-        image_tensor = image_processor(image, return_tensors='pt')['pixel_values'].half().cuda()
+        image_tensor = image_processor(image, return_tensors='pt')['pixel_values'].half()
 
         if return_tensors is not None:
             if return_tensors == 'pt':
-                return torch.tensor(input_ids, dtype=torch.long), image_tensor, stopping_criteria
+                return torch.tensor(input_ids, dtype=torch.long).unsqueeze(0), image_tensor, stopping_criteria
             raise ValueError(f'Unsupported tensor type: {return_tensors}')
         
         return input_ids, image_tensor, stopping_criteria