Update app_template.py

app_template.py

@@ -5,39 +5,12 @@ import requests
 
 tulu = "https://tonic1-tulu.hf.space/--replicas/t5vxm/"
 
-HF_TOKEN = os.getenv("HF_TOKEN")
-HEADERS = {"Authorization": f"Bearer {HF_TOKEN}"}
-
-def build_input_prompt(message, chatbot, system_prompt):
-    """
-    Constructs the input prompt string from the chatbot interactions and the current message.
-    """
-    input_prompt = "<|system|>\n" + system_prompt + "</s>\n<|user|>\n"
-    for interaction in chatbot:
-        input_prompt = input_prompt + str(interaction[0]) + "</s>\n<|assistant|>\n" + str(interaction[1]) + "\n</s>\n<|user|>\n"
-
-    input_prompt = input_prompt + str(message) + "</s>\n<|assistant|>"
-    return input_prompt
-
-
-def post_request_beta(payload):
-    """
-    Sends a POST request to the predefined Tulu URL and returns the JSON response.
-    """
-    response = requests.post(tulu, headers=HEADERS, json=payload)
-    response.raise_for_status()  # Will raise an HTTPError if the HTTP request returned an unsuccessful status code
-    return response.json()
-
 
 def predict_beta(message, chatbot=[], system_prompt=""):
-    client = Client(tulu)  # Assuming Client is properly defined and tulu is a valid argument
-
-    # Build the input prompt
-    input_prompt = build_input_prompt(message, chatbot, system_prompt)  # Ensure this function is defined
+    client = Client(tulu)
 
     try:
-        # Adjust these parameters as needed
-        max_new_tokens = 1200
+        max_new_tokens = 800
         temperature = 0.4
         top_p = 0.9
         repetition_penalty = 0.9
@@ -45,7 +18,8 @@ def predict_beta(message, chatbot=[], system_prompt=""):
 
         # Making the prediction
         result = client.predict(
-            input_prompt,  # Using the built input prompt
+            message,
+            system_prompt
             max_new_tokens,
             temperature,
             top_p,
@@ -54,9 +28,8 @@ def predict_beta(message, chatbot=[], system_prompt=""):
             fn_index=0
         )
 
-        # Extracting the response
         if result is not None and len(result) > 0:
-            bot_message = result[0]  # Assuming the response is in the first element
+            bot_message = result[0] 
             return bot_message
         else:
             raise gr.Error("No response received from the model.")
@@ -67,8 +40,6 @@ def predict_beta(message, chatbot=[], system_prompt=""):
 
 def test_preview_chatbot(message, history):
     response = predict_beta(message, history, SYSTEM_PROMPT)
-    text_start = response.rfind("<|assistant|>", ) + len("<|assistant|>")
-    response = response[text_start:]
     return response