import gradio as gr
import spaces
import transformers
import torch

model_id = "GoToCompany/gemma2-9b-cpt-sahabatai-v1-instruct"

pipeline = transformers.pipeline(
    "text-generation",
    model=model_id,
    model_kwargs={"torch_dtype": torch.bfloat16},
    device_map="auto",
)

terminators = [
    pipeline.tokenizer.eos_token_id,
    pipeline.tokenizer.convert_tokens_to_ids("<|eot_id|>")
]

@spaces.GPU
def respond(
    message,
    history: list[tuple[str, str]],
    max_tokens,
    temperature,
    top_p,
):
    messages = []
    
    for val in history:
        if val[0]:
            messages.append({"role": "user", "content": val[0]})
        if val[1]:
            messages.append({"role": "assistant", "content": val[1]})

    messages.append({"role": "user", "content": message})

    outputs = pipeline(
        messages,
        max_new_tokens=max_tokens,
        do_sample = True,
        temperature=temperature,
        top_p=top_p,
        eos_token_id=terminators
    )
    
    yield outputs[0]["generated_text"][-1]["content"]

"""
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
"""
demo = gr.ChatInterface(
    respond,
    title = "🇮🇩 Sahabat AI (Gemma)",
    description = """This model is a fine-tuned version of SEA-LIONv3's Gemma model trained predominantly on Indonesian, Javanese, and Sundanese data.
    
    #### [Model page](https://huggingface.co/GoToCompany/gemma2-9b-cpt-sahabatai-v1-instruct)""",
    examples = [["Tolong carin resep sop buntut dong"], ["Sopo wae sing ana ing Punakawan?"], ["Kumaha caritana si Kabayan?"]],
    additional_inputs=[
        gr.Slider(minimum=1, maximum=2048, value=256, step=1, label="Max new tokens"),
        gr.Slider(minimum=0.1, maximum=1.0, 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()