Update app.py

app.py

@@ -24,7 +24,7 @@ from models.unet import UNet3DConditionModel
 from models.controlnet import ControlNetModel3D
 from models.RIFE.IFNet_HDv3 import IFNet
 
-
+hf_token = os.environ['HF_TOKEN']
 device = "cuda"
 
 sd_path = "checkpoints/stable-diffusion-v1-5"
@@ -85,13 +85,13 @@ def infer(prompt, video_path, condition, video_length, is_long_video):
     else:
         annotator = controlnet_parser_dict[condition]()
 
-    tokenizer = CLIPTokenizer.from_pretrained(sd_path, subfolder="tokenizer")
-    text_encoder = CLIPTextModel.from_pretrained(sd_path, subfolder="text_encoder").to(dtype=torch.float16)
-    vae = AutoencoderKL.from_pretrained(sd_path, subfolder="vae").to(dtype=torch.float16)
-    unet = UNet3DConditionModel.from_pretrained_2d(sd_path, subfolder="unet").to(dtype=torch.float16)
+    tokenizer = CLIPTokenizer.from_pretrained(sd_path, subfolder="tokenizer", use_auth_token=hf_token)
+    text_encoder = CLIPTextModel.from_pretrained(sd_path, subfolder="text_encoder", use_auth_token=hf_token).to(dtype=torch.float16)
+    vae = AutoencoderKL.from_pretrained(sd_path, subfolder="vae", use_auth_token=hf_token).to(dtype=torch.float16)
+    unet = UNet3DConditionModel.from_pretrained_2d(sd_path, subfolder="unet", use_auth_token=hf_token).to(dtype=torch.float16)
     controlnet = ControlNetModel3D.from_pretrained_2d(controlnet_dict[condition]).to(dtype=torch.float16)
     interpolater = IFNet(ckpt_path=inter_path).to(dtype=torch.float16)
-    scheduler=DDIMScheduler.from_pretrained(sd_path, subfolder="scheduler")
+    scheduler=DDIMScheduler.from_pretrained(sd_path, subfolder="scheduler", use_auth_token=hf_token)
 
     pipe = ControlVideoPipeline(
             vae=vae, text_encoder=text_encoder, tokenizer=tokenizer, unet=unet,