Update app.py

app.py

@@ -2,6 +2,11 @@ import gradio as gr
 from transformers import pipeline, AutoTokenizer, AutoModelForSeq2SeqLM
 from aksharamukha import transliterate
 import torch
+from dotenv import load_dotenv
+import os
+
+load_dotenv()
+access_token = os.getenv('ACCESS_TOKEN')
 
 # Set up device
 device = "cuda" if torch.cuda.is_available() else "cpu"
@@ -46,10 +51,17 @@ def transliterate_to_sinhala(text):
 # Placeholder for conversation model loading and pipeline setup
 # pipe1 = pipeline("text-generation", model="microsoft/Phi-3-mini-4k-instruct", trust_remote_code=True)
 
-interface = gr.Interface.load("huggingface/microsoft/Phi-3-mini-4k-instruct")
+# interface = gr.Interface.load("huggingface/microsoft/Phi-3-mini-4k-instruct")
+
+API_URL = "https://api-inference.huggingface.co/models/microsoft/Phi-3-mini-4k-instruct"
+headers = {"Authorization": f"Bearer {access_token}"}
 
-def conversation_predict(text):
-    return interface([text])[0]
+def query(payload):
+	response = requests.post(API_URL, headers=headers, json=payload)
+	return response.json()
+
+# def conversation_predict(text):
+#     return interface([text])[0]
 
 def ai_predicted(user_input):
     if user_input.lower() == 'exit':
@@ -58,8 +70,10 @@ def ai_predicted(user_input):
     user_input = translate_Singlish_to_sinhala(user_input)
     user_input = transliterate_to_sinhala(user_input)
     user_input = translate_sinhala_to_english(user_input)
-    # ai_response = pipe1([{"role": "user", "content": user_input}])
-    ai_response = conversation_predict(user_input)
+    ai_response = query({
+    	"inputs": user_input,
+    })
+    # ai_response = conversation_predict(user_input)
     ai_response_lines = ai_response.split("</s>")
 
     response = translate_english_to_sinhala(ai_response_lines[-1])
@@ -85,8 +99,6 @@ def respond(
 
     messages.append({"role": "user", "content": message})
 
-    
-
     response = ai_predicted(message)
 
     yield response