import torch
from PIL import Image
import gradio as gr
import spaces
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
import os
from threading import Thread


HF_TOKEN = os.environ.get("HF_TOKEN", None)
MODEL_ID = "THUDM/glm-4-9b-chat"
MODEL_ID2 = "THUDM/glm-4-9b-chat-1m"
MODELS = os.environ.get("MODELS")
MODEL_NAME = MODELS.split("/")[-1]

TITLE = "<h1><center>GLM-4-9B</center></h1>"

DESCRIPTION = f'<h3><center>MODEL: <a href="https://hf.co/{MODELS}">{MODEL_NAME}</a></center></h3>'

CSS = """
.duplicate-button {
  margin: auto !important;
  color: white !important;
  background: black !important;
  border-radius: 100vh !important;
}
"""

model = AutoModelForCausalLM.from_pretrained(
        MODELS,
        torch_dtype=torch.bfloat16,
        low_cpu_mem_usage=True,
        trust_remote_code=True,
        ).to(0).eval()

tokenizer = AutoTokenizer.from_pretrained(MODELS,trust_remote_code=True)

@spaces.GPU
def stream_chat(message: str, history: list, temperature: float, max_new_tokens: int):
    print(f'message is - {message}')
    print(f'history is - {history}')
    conversation = []
    for prompt, answer in history:
        conversation.extend([{"role": "user", "content": prompt}, {"role": "assistant", "content": answer}])
    conversation.append({"role": "user", "content": message})

    print(f"Conversation is -\n{conversation}")
    
    input_ids = tokenizer.apply_chat_template(conversation, tokenize=True, add_generation_prompt=True, return_tensors="pt", return_dict=True).to(model.device)
    
    streamer = TextIteratorStreamer(tokenizer, **{"skip_special_tokens": True, "skip_prompt": True, 'clean_up_tokenization_spaces':False,})

    generate_kwargs = dict(
        input_ids=input_ids, 
        streamer=streamer,
        max_new_tokens=max_new_tokens, 
        do_sample=True, 
        temperature=temperature,
        repetition_penalty=1.2,
        eos_token_id=model.config.eos_token_id,
    )
    
    thread = Thread(target=model.generate, kwargs=generate_kwargs)
    thread.start()

    buffer = ""
    for new_text in streamer:
        buffer += new_text
        yield buffer



chatbot = gr.Chatbot(height=450)

with gr.Blocks(css=CSS) as demo:
    gr.HTML(TITLE)
    gr.HTML(DESCRIPTION)
    gr.DuplicateButton(value="Duplicate Space for private use", elem_classes="duplicate-button")
    gr.ChatInterface(
        fn=stream_chat,
        chatbot=chatbot,
        fill_height=True,
        additional_inputs_accordion=gr.Accordion(label="⚙️ Parameters", open=False, render=False),
        additional_inputs=[
            gr.Slider(
                minimum=0,
                maximum=1,
                step=0.1,
                value=0.8,
                label="Temperature",
                render=False,
            ),
            gr.Slider(
                minimum=128,
                maximum=4096,
                step=1,
                value=1024,
                label="Max new tokens",
                render=False,
            ),
        ],
        examples=[
            ["Help me study vocabulary: write a sentence for me to fill in the blank, and I'll try to pick the correct option."],
            ["What are 5 creative things I could do with my kids' art? I don't want to throw them away, but it's also so much clutter."],
            ["Tell me a random fun fact about the Roman Empire."],
            ["Show me a code snippet of a website's sticky header in CSS and JavaScript."],
        ],
        cache_examples=False,
    )


if __name__ == "__main__":
    demo.launch()