Upload app.py

app.py

@@ -58,11 +58,11 @@ def generate(message, chat_history, max_new_tokens):
     final_prompt += "User: " + message + "\n"
     final_prompt += "Output:"
 
-    if (
-        len(tokenizer.tokenize(final_prompt))
-        >= tokenizer.model_max_length - max_new_tokens
-    ):
-        final_prompt = "Instruction: Say 'Input exceeded context size, please clear the chat history and retry!' Output:"
+    # if (
+    #     len(tokenizer.tokenize(final_prompt))
+    #     >= tokenizer.model_max_length - max_new_tokens
+    # ):
+    #     final_prompt = "Instruction: Say 'Input exceeded context size, please clear the chat history and retry!' Output:"
 
     # Streamer
     streamer = TextIteratorStreamer(
@@ -99,7 +99,7 @@ with gr.Blocks() as demo:
   # Phi-2 Scientific Question Chatbot
   This chatbot was created using Microsoft's 2.7 billion parameter [phi-2](https://huggingface.co/microsoft/phi-2) Transformer model.
 
-  Phi-2 model was fine-tuned with questions including physics chemistry biology QA using SciQ dataset. In order to reduce the response time on this hardware, `max_new_tokens` has been set to `21` in the text generation pipeline. With this default configuration, it takes approximately `60 seconds` for the response to start being generated, and streamed one word at a time. Use the slider below to increase or decrease the length of the generated text.
+  Phi-2 model was fine-tuned with questions including highschool level physics chemistry biology QA using SciQ dataset. In order to reduce the response time on this hardware, `max_new_tokens` has been set to `21` in the text generation pipeline. With this default configuration, it takes approximately `60 seconds` for the response to start being generated, and streamed one word at a time. Use the slider below to increase or decrease the length of the generated text.
 
   For the safetensor: huggingface.co/mudogruer
   """