import gradio as gr
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, StoppingCriteria, StoppingCriteriaList, TextIteratorStreamer
from threading import Thread
import spaces

finetuned_model = "CONCREE/adia-llm"

# Charge le modele
model = AutoModelForCausalLM.from_pretrained(
        finetuned_model,
        device_map="auto",
        trust_remote_code=True,
)

# Load tokenizer
tokenizer = AutoTokenizer.from_pretrained(finetuned_model,
                                         trust_remote_code=True,
                                         padding=True,
                                         truncation=True)

class StopOnTokens(StoppingCriteria):
    def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor, **kwargs) -> bool:
        stop_ids = [29, 0]
        for stop_id in stop_ids:
            if input_ids[0][-1] == stop_id:
                return True
        return False

@spaces.GPU
def predict(message, history):
    history_transformer_format = history + [[message, ""]]
    stop = StopOnTokens()

    messages = "".join(["".join(["\n[INST]:"+item[0], "\n[/INST]:"+item[1]]) for item in history_transformer_format])

    model_inputs = tokenizer([messages], return_tensors="pt").to("cuda")
    streamer = TextIteratorStreamer(tokenizer, timeout=10., skip_prompt=True, skip_special_tokens=True)
    generate_kwargs = dict(
        model_inputs,
        streamer=streamer,
        max_new_tokens=1024,
        num_beams=1,
        stopping_criteria=StoppingCriteriaList([stop])
    )
    t = Thread(target=model.generate, kwargs=generate_kwargs)
    t.start()
    partial_message = ""
    start_flag = True  # Flag to ignore initial newline

    for new_token in streamer:
        if start_flag and new_token == '\n':
            continue
        start_flag = False
        partial_message += new_token
        yield partial_message


demo = gr.ChatInterface(predict).launch()


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