app.py

# import gradio as gr
# import os
# os.environ["UNSLOTH_DEVICE"] = "cuda"
# from unsloth import FastLanguageModel
# import torch
# HF_TOKEN = os.environ["HF_TOKEN"]
# # -------------------- Load Model --------------------
# model, tokenizer = FastLanguageModel.from_pretrained(
#     model_name="ak0601/gpt-oss-20b-persona-chat",  # your trained model
#     max_seq_length=1024,
#     dtype=None,
#     load_in_4bit=True,
#     device_map="auto",
#     token=HF_TOKEN
# )

# # -------------------- Conversation Formatter --------------------
# def format_conversation(conversation):
#     text = ""
#     for turn in conversation:
#         if turn["role"] == "system":
#             text += f"[SYSTEM] {turn['content']}\n"
#         elif turn["role"] == "user":
#             text += f"[USER] {turn['content']}\n"
#         elif turn["role"] == "assistant":
#             text += f"[ASSISTANT] {turn['content']}\n"
#     text += "[ASSISTANT]"
#     return text

# def generate_reply(conversation):
#     inputs = tokenizer(
#         format_conversation(conversation),
#         return_tensors="pt"
#     ).to(model.device)

#     output_ids = model.generate(
#         **inputs,
#         max_new_tokens=256,
#         temperature=0.7,
#         top_p=0.9,
#         repetition_penalty=1.1,
#         eos_token_id=tokenizer.eos_token_id,
#     )

#     response = tokenizer.decode(output_ids[0], skip_special_tokens=True)
#     response = response.split("[ASSISTANT]")[-1].strip()
#     return response

# # -------------------- Gradio Functions --------------------
# def start_chat(persona):
#     conversation = [
#     {"role": "system", "content": f"""You are a digital twin.
# ONLY respond based on persona and user input.
# \nPersona: {persona}"""},
# ]
#     return conversation, [(None, "How can I help you?")]

# def chat(user_message, history, conversation):
#     conversation.append({"role": "user", "content": user_message})
#     reply = generate_reply(conversation)
#     conversation.append({"role": "assistant", "content": reply})
#     history.append((user_message, reply))
#     return history, conversation

# # -------------------- Gradio UI --------------------
# with gr.Blocks() as demo:
#     gr.Markdown("## 🤖 Digital Twin Chat")

#     persona_box = gr.Textbox(label="Enter your persona",
#                              value="I am male. I am unsociable. I have a weakness for sweets. I am a jack of all, master of none.")
#     start_btn = gr.Button("Start Chat")

#     chatbot = gr.Chatbot()
#     msg = gr.Textbox(label="Your message")

#     state_conversation = gr.State([])
#     state_history = gr.State([])

#     start_btn.click(start_chat, inputs=persona_box, outputs=[state_conversation, chatbot])
#     msg.submit(chat, inputs=[msg, chatbot, state_conversation], outputs=[chatbot, state_conversation])

# demo.launch()



import gradio as gr
import torch
import os
from transformers import AutoModelForCausalLM, AutoTokenizer

HF_TOKEN = os.environ["HF_TOKEN"]

# -------------------- Load Model --------------------
model_name = "ak0601/gpt-oss-20b-persona-chat"

tokenizer = AutoTokenizer.from_pretrained(model_name, token=HF_TOKEN)

model = AutoModelForCausalLM.from_pretrained(
    model_name,
    device_map="auto",         # automatically places model on GPU
    torch_dtype=torch.float16, # efficient for H200
    load_in_4bit=True,         # quantization if available
    token=HF_TOKEN
)

# -------------------- Conversation Formatter --------------------
def format_conversation(conversation):
    text = ""
    for turn in conversation:
        if turn["role"] == "system":
            text += f"[SYSTEM] {turn['content']}\n"
        elif turn["role"] == "user":
            text += f"[USER] {turn['content']}\n"
        elif turn["role"] == "assistant":
            text += f"[ASSISTANT] {turn['content']}\n"
    text += "[ASSISTANT]"
    return text

def generate_reply(conversation):
    inputs = tokenizer(
        format_conversation(conversation),
        return_tensors="pt"
    ).to(model.device)

    output_ids = model.generate(
        **inputs,
        max_new_tokens=256,
        temperature=0.7,
        top_p=0.9,
        repetition_penalty=1.1,
        eos_token_id=tokenizer.eos_token_id,
    )

    response = tokenizer.decode(output_ids[0], skip_special_tokens=True)
    response = response.split("[ASSISTANT]")[-1].strip()
    return response

# -------------------- Gradio Functions --------------------
def start_chat(persona):
    conversation = [
        {"role": "system", "content": f"""You are a digital twin.
ONLY respond based on persona and user input.
\nPersona: {persona}"""},
    ]
    return conversation, [(None, "How can I help you?")]

def chat(user_message, history, conversation):
    conversation.append({"role": "user", "content": user_message})
    reply = generate_reply(conversation)
    conversation.append({"role": "assistant", "content": reply})
    history.append((user_message, reply))
    return history, conversation

# -------------------- Gradio UI --------------------
with gr.Blocks() as demo:
    gr.Markdown("## 🤖 Digital Twin Chat")

    persona_box = gr.Textbox(label="Enter your persona",
                             value="I am male. I am unsociable. I have a weakness for sweets. I am a jack of all, master of none.")
    start_btn = gr.Button("Start Chat")

    chatbot = gr.Chatbot()
    msg = gr.Textbox(label="Your message")

    state_conversation = gr.State([])
    state_history = gr.State([])

    start_btn.click(start_chat, inputs=persona_box, outputs=[state_conversation, chatbot])
    msg.submit(chat, inputs=[msg, chatbot, state_conversation], outputs=[chatbot, state_conversation])

demo.launch()