Upload infer_qwen2_vl.py with huggingface_hub

infer_qwen2_vl.py

@@ -42,6 +42,7 @@ for img_path in glob.glob(f"{folder}/*.jpe"):
     images.append(img)
 
 num_image = len(images)
+print(f"beigin : {begin}, end : {end}, batch_size : {batch_size}"}")
 begin, end, batch_size= 0, 10, args.batch_size
 messages = [
     {
@@ -61,15 +62,18 @@ messages = [
         ],
     }
 ]
+
 from tqdm import tqdm
 # Preparation for inference
 ans = []
 counter = 0
 for batch_idx in tqdm(range(begin, end, batch_size)):
     batch = data[batch_idx: min(batch_idx + batch_size, end)]
+    print(f"data index range : {batch_idx} ~ {min(batch_idx + batch_size, end)}")
     image_inputs_batch, video_inputs_batch = [], []
     for idx,i in enumerate(batch):
         img = images[i]
+        print('gain image successfully !')
         messages[1]["content"][0]["image"] = img
         text = processor.apply_chat_template(
             messages, tokenize=False, add_generation_prompt=True
@@ -97,7 +101,7 @@ for batch_idx in tqdm(range(begin, end, batch_size)):
     #ans.append(output_text)
     save_path = "output.json"
     counter = counter + 1
-    if counter % 10 == 0:
+    if counter % 1 == 0:
         print(f"Saving data at iteration {idx + 1}")
         write_json(save_path, data)