Create app.py

app.py

@@ -48,21 +48,21 @@ def prepare_data(temp_text, audio_prompt):
     example_embeddings = torch.tensor(example["speaker_embeddings"]).unsqueeze(0)
     return example_embeddings
 
-def generate_gpt4_response(user_text, print_output=False):
-    """
-    Query OpenAI GPT-4 for the specific key and get back a response
-    :type user_text: str the user's text to query for
-    :type print_output: boolean whether or not to print the raw output JSON
-    """
-    message=[{"role": "user", "content": user_text+'in just 2 very small sentences'}]
-    completions = ai.ChatCompletion.create(
-         model="gpt-4",
-         messages=message,
-         max_tokens=250
-     )
+# def generate_gpt4_response(user_text, print_output=False):
+#     """
+#     Query OpenAI GPT-4 for the specific key and get back a response
+#     :type user_text: str the user's text to query for
+#     :type print_output: boolean whether or not to print the raw output JSON
+#     """
+#     message=[{"role": "user", "content": user_text+'in just 2 very small sentences'}]
+#     completions = ai.ChatCompletion.create(
+#          model="gpt-4",
+#          messages=message,
+#          max_tokens=250
+#      )
 
-    # Return the first choice's text
-    return completions['choices'][0]['message']['content']
+#     # Return the first choice's text
+#     return completions['choices'][0]['message']['content']
 
 
 def predict(temp_text, temp_audio, record_audio_prompt, prompt_text):
@@ -71,7 +71,8 @@ def predict(temp_text, temp_audio, record_audio_prompt, prompt_text):
     else:
       audio_prompt = record_audio_prompt
     
-    text = generate_gpt4_response(prompt_text)
+    # text = generate_gpt4_response(prompt_text)
+    text=prompt_text
     embeddings=prepare_data(temp_text, audio_prompt)
     inputs = processor(text=text, return_tensors="pt")
     spectrogram = model.generate_speech(inputs["input_ids"], embeddings)