Update app.py

app.py

@@ -1,64 +1,96 @@
 import gradio as gr
-from huggingface_hub import InferenceClient
+from transformers import AutoModelForCausalLM, AutoTokenizer
+import torch
+import re
 
-"""
-For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-"""
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
+# Model configuration
+MAX_REASONING_TOKENS = 4096
+MAX_RESPONSE_TOKENS = 2048
+MODEL_OPTIONS = ["beyoru/ThinkAgain1.2", "beyoru/ThinkAgain1.4", "beyoru/ThinkAgain1.5"]
 
+# Global variables for model and tokenizer
+model = None
+tokenizer = None
+messages = []
 
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-):
-    messages = [{"role": "system", "content": system_message}]
+# Function to extract text between <think> and </think> tags
+def extract_think_content(text):
+    match = re.search(r'<think>(.*?)</think>', text, re.DOTALL)
+    return match.group(1).strip() if match else text
 
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
+# Function to load model and tokenizer
+def load_model(model_name):
+    global model, tokenizer, messages
+    messages = []  # Reset conversation history
+    model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
+    tokenizer = AutoTokenizer.from_pretrained(model_name)
+    return f"Loaded model: {model_name}"
 
-    messages.append({"role": "user", "content": message})
+def generate_response(user_input, history=None, model_name=None):
+    global messages
+    
+    if history is None:
+        history = []
+    
+    # Append user message to history
+    messages.append({"role": "user", "content": user_input})
+    history.append((user_input, None))  # User message in chatbot
+    
+    # Generate reasoning
+    reasoning_template = tokenizer.apply_chat_template(messages, tokenize=False, add_reasoning_prompt=True)
+    reasoning_inputs = tokenizer(reasoning_template, return_tensors="pt").to(model.device)
+    reasoning_ids = model.generate(**reasoning_inputs, max_new_tokens=MAX_REASONING_TOKENS)
+    reasoning_output = tokenizer.decode(reasoning_ids[0, reasoning_inputs.input_ids.shape[1]:], skip_special_tokens=True)
+    
+    # Extract content from <think> tags
+    extracted_reasoning = extract_think_content(reasoning_output)
+    messages.append({"role": "reasoning", "content": extracted_reasoning})
+    
+    # Generate assistant response
+    response_template = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+    response_inputs = tokenizer(response_template, return_tensors="pt").to(model.device)
+    response_ids = model.generate(**response_inputs, max_new_tokens=MAX_RESPONSE_TOKENS)
+    response_output = tokenizer.decode(response_ids[0, response_inputs.input_ids.shape[1]:], skip_special_tokens=True)
+    
+    messages.append({"role": "assistant", "content": response_output})
+    
+    # Combine reasoning and assistant response in a single chatbot message with toggleable reasoning
+    combined_response = (
+        f"<details><summary>Reasoning</summary>{extracted_reasoning}</details>\n\n"
+        f"{response_output}"
+    )
+    history.append((None, combined_response))  # Assistant message with toggleable reasoning
+    
+    return history, ""
 
-    response = ""
+# Create Gradio interface
+with gr.Blocks() as demo:
+    gr.Markdown("# ThinkAgain Chatbot")
+    
+    # Model selection
+    model_dropdown = gr.Dropdown(choices=MODEL_OPTIONS, label="Select Model", value=MODEL_OPTIONS[-1])
+    model_load_status = gr.Textbox(label="Model Status", interactive=False)
+    
+    # Load model when dropdown changes
+    model_dropdown.change(
+        fn=load_model,
+        inputs=model_dropdown,
+        outputs=model_load_status
+    )
+    
+    chatbot = gr.Chatbot(label="Conversation")
+    user_input = gr.Textbox(label="Your Message", placeholder="Type your message here...")
+    submit = gr.Button("Send")
+    
+    # Clear input and update chatbot after submission
+    submit.click(
+        fn=generate_response,
+        inputs=[user_input, chatbot, model_dropdown],
+        outputs=[chatbot, user_input]
+    )
 
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
+# Load default model
+load_model(MODEL_OPTIONS[0])
 
-        response += token
-        yield response
-
-
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-demo = gr.ChatInterface(
-    respond,
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
-    ],
-)
-
-
-if __name__ == "__main__":
-    demo.launch()
+# Launch the app
+demo.launch()