Update demo/t2v.py

demo/t2v.py

@@ -57,7 +57,6 @@ class Text2Video():
         torch.cuda.empty_cache()
         self.agent.to('cuda')
         self.clip.to('cuda')
-        self.tokenizer.to('cuda')
 
         print('start:', prompt, time.strftime('%Y-%m-%d %H:%M:%S',time.localtime(time.time())))
         start = time.time()
@@ -77,7 +76,7 @@ class Text2Video():
             for text in labels_list:
                 with torch.no_grad():
                     text_feat.append(self.clip.get_txt_feat(text,))
-            text_feat = torch.stack(text_feat, dim=0).to(self.clip.device)
+            text_feat = torch.stack(text_feat, dim=0).to('cuda')
 
             video_embed = text_feat