Upload infer_qwen2_vl.py with huggingface_hub

infer_qwen2_vl.py

@@ -42,7 +42,7 @@ file_names = os.listdir(folder)
 
 num_image = len(file_names)
 
-begin, end, batch_size= 0, num_image, 16
+begin, end, batch_size= 0, num_image, 12
 print(f"beigin : {begin}, end : {end}, batch_size : {batch_size}")
 messages = [
     {
@@ -68,8 +68,9 @@ from tqdm import tqdm
 ans = []
 counter = 0
 for batch_idx in tqdm(range(begin, end, batch_size)):
-    batch = file_names[batch_idx: min(batch_idx + batch_size, end)]
-    print(f"data index range : {batch_idx} ~ {min(batch_idx + batch_size, end)}")
+    up = min(batch_idx + batch_size, end)
+    batch = file_names[batch_idx: up]
+    print(f"data index range : {batch_idx} ~ {up}")
     image_inputs_batch, video_inputs_batch,text_batch = [], [], []
     for idx,i in enumerate(batch):
         #img = batch[i]
@@ -119,7 +120,7 @@ for batch_idx in tqdm(range(begin, end, batch_size)):
     ans.append(output_text)
     save_path = "output_final.json"
     counter = counter + 1
-    if counter % 1 == 0:
+    if counter % 10 == 0 or up + 10 >= end:
         print(f"Saving data at iteration {idx + 1}")
         write_json(save_path, ans)