Upload infer_3.py with huggingface_hub

infer_3.py

@@ -19,11 +19,7 @@ def write_json(file_path, data):
 # default: Load the model on the available device(s)
 print(torch.cuda.device_count())
 model_path = "/home/zbz5349/WorkSpace/aigeeks/Qwen2.5-VL/LLaMA-Factory/output/Qwen2.5-VL-3B_all"
-# model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
-#     model_path, torch_dtype="auto", device_map="auto"
-# )
 
-# We recommend enabling flash_attention_2 for better acceleration and memory saving, especially in multi-image and video scenarios.
 model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
     model_path,
     torch_dtype=torch.bfloat16,
@@ -50,8 +46,8 @@ for batch_idx in tqdm(range(begin, end, batch_size)):
     image_list = []
     input_text_list = []
     data_list = []
-    save_data = []
-    
+    save_list = []
+    sd_ans = []
     # while True:
     for idx, i in enumerate(batch):
         save_ =  {
@@ -66,7 +62,8 @@ for batch_idx in tqdm(range(begin, end, batch_size)):
             {"type": "image", "image": "file:///path/to/image2.jpg"},
             {"type": "text", "text": "Describe this video."},
         ],
-        "answer":""
+        "answer":"None",
+        "result":"None",
     }
         messages =  {
         "role": "user",
@@ -95,8 +92,9 @@ for batch_idx in tqdm(range(begin, end, batch_size)):
         save_['content'][1]['image'] = image_path
         save_['content'][2]['text'] = question
         save_['answer'] = answer
+        sd_ans.append(answer)
         data_list.append(messages)
-        save_data.append(save_)
+        save_list.append(save_)
 
     text = processor.apply_chat_template(data_list, tokenize=False, add_generation_prompt=True)
     image_inputs, video_inputs, video_kwargs = process_vision_info(data_list, return_video_kwargs=True)
@@ -119,11 +117,13 @@ for batch_idx in tqdm(range(begin, end, batch_size)):
     output_text = processor.batch_decode(
         generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
     )
-    save_["answer"] = output_text
-    if output_text == answer:
-        correct_num = correct_num + 1
-    save_data.append(save_)
+    for idx,x in enumerate(output_text):
+        save_list[idx]['result'] = x
+        save_data.append(save_list[idx])
+        
 print("correct_num", correct_num)
 write_json("infer_answer_finetune.json",save_data)
 
+        
+