Update transcription_diarization.py

transcription_diarization.py

@@ -33,7 +33,7 @@ class LazyTranscriptionPipeline:
     @spaces.GPU(duration=100)
     def get_pipeline(self, language):
         if self.pipe is None:
-            model_id = "openai/whisper-medium"
+            model_id = "openai/whisper-large-v3"
             self.model = AutoModelForSpeechSeq2Seq.from_pretrained(
                 model_id, torch_dtype=torch.float16, low_cpu_mem_usage=True, use_safetensors=True
             )
@@ -44,11 +44,11 @@ class LazyTranscriptionPipeline:
                 model=self.model,
                 tokenizer=self.processor.tokenizer,
                 feature_extractor=self.processor.feature_extractor,
-                max_new_tokens=128,
-                chunk_length_s=30,
+                max_new_tokens=64,
+                chunk_length_s=15,
                 batch_size=1,
                 return_timestamps=True,
-                torch_dtype=torch.float16,
+                torch_dtype=torch.bfloat16,
                 device=torch.device("cuda"),
                 generate_kwargs={"language": language}
             )