update deprecated torch.cuda.amp

ammico/video_summary.py

@@ -8,7 +8,7 @@ from torchcodec.decoders import VideoDecoder
 from ammico.model import MultimodalSummaryModel
 from ammico.utils import AnalysisMethod
 
-from typing import List, Optional, Dict, Any, Generator, Tuple
+from typing import List, Dict, Any, Generator, Tuple
 from transformers import GenerationConfig
 
 
@@ -229,7 +229,7 @@ class VideoSummaryDetector(AnalysisMethod):
         with torch.inference_mode():
             try:
                 if self.summary_model.device == "cuda":
-                    with torch.cuda.amp.autocast(enabled=True):
+                    with torch.amp.autocast("cuda", enabled=True):
                         generated_ids = self.summary_model.model.generate(
                             **processor_inputs, generation_config=gen_conf
                         )