Update app.py

app.py

@@ -9,7 +9,7 @@ from datetime import datetime
 import numpy as np
 import os
 
-
+HF_TOKEN = os.environ['HF_TOKEN']
 # subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
 
 # models = {
@@ -33,7 +33,12 @@ def array_to_image_path(image_array):
     return full_path
     
 models = {
-    "Qwen/Qwen2-VL-72B-Instruct": Qwen2VLForConditionalGeneration.from_pretrained("Qwen/Qwen2-VL-72B-Instruct", trust_remote_code=True, torch_dtype="auto").cuda().eval()
+    "Qwen/Qwen2-VL-72B-Instruct": Qwen2VLForConditionalGeneration.from_pretrained(
+        "Qwen/Qwen2-VL-72B-Instruct", 
+        trust_remote_code=True, 
+        token=HF_TOKEN,
+        torch_dtype=torch.bfloat16
+    ).cuda().eval()
 
 }