Update app.py

app.py

@@ -1,17 +1,21 @@
 import gradio as gr
-from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor, GPT2LMHeadModel, GPT2Tokenizer, VitsProcessor, VitsForConditionalGeneration
+from transformers import AutoTokenizer, AutoModelForCausalLM, Speech2TextProcessor, Speech2TextForConditionalGeneration, VitsProcessor, VitsForConditionalGeneration
 
 # Load the ASR model and processor
-asr_processor = Wav2Vec2Processor.from_pretrained("/path/to/canary/processor")
-asr_model = Wav2Vec2ForCTC.from_pretrained("/path/to/canary/model")
+asr_processor = Speech2TextProcessor.from_pretrained("/path/to/canary/processor")
+asr_model = Speech2TextForConditionalGeneration.from_pretrained("/path/to/canary/model")
 
 # Load the text processing model and tokenizer
-proc_tokenizer = GPT2Tokenizer.from_pretrained("/path/to/phi3/tokenizer")
-proc_model = GPT2LMHeadModel.from_pretrained("/path/to/phi3/model")
+proc_tokenizer = AutoTokenizer.from_pretrained("microsoft/Phi-3-mini-128k-instruct")
+proc_model = AutoModelForCausalLM.from_pretrained(
+    "microsoft/Phi-3-mini-128k-instruct", 
+    device_map="cuda", 
+    torch_dtype="auto", 
+    trust_remote_code=True, 
 
 # Load the TTS model and processor
-tts_processor = VitsProcessor.from_pretrained("facebook/vits-base")
-tts_model = VitsForConditionalGeneration.from_pretrained("facebook/vits-base")
+tts_processor = VitsProcessor.from_pretrained("/path/to/vits/processor")
+tts_model = VitsForConditionalGeneration.from_pretrained("/path/to/vits/model")
 
 def process_speech(speech):
     # Convert the speech to text