Update app.py

app.py

@@ -32,9 +32,15 @@ def process_transcript(model, processor, language, audio_path):
         return "Please provide some input audio: either upload an audio file or use the microphone."
     else:
         id_language = dict_languages[language]
-        inputs = processor.apply_transcrition_request(language=id_language, audio=audio_path, model_id=model_name)
+        raw_inputs = processor.apply_transcrition_request(language=id_language, audio=audio_path, model_id=model_name)
 #        inputs = inputs.to(device, dtype=torch.bfloat16)
-        inputs = {k: v.to(device, dtype=torch.bfloat16) for k, v in inputs.items()}
+#        inputs = {k: v.to(device, dtype=torch.bfloat16) for k, v in inputs.items()}
+        inputs = {}
+        for k, v in raw_inputs.items():
+            if k == "input_ids":
+                inputs[k] = v.to(device=device, dtype=torch.long)
+            else:
+                inputs[k] = v.to(device=device, dtype=torch.bfloat16)
         outputs = model.generate(**inputs, max_new_tokens=MAX_TOKENS)
         decoded_outputs = processor.batch_decode(outputs[:, inputs.input_ids.shape[1]:], skip_special_tokens=True)