app.py

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


tokenizer = AutoTokenizer.from_pretrained("IlyaGusev/saiga_llama3_8b")
model = AutoModelForCausalLM.from_pretrained("IlyaGusev/saiga_llama3_8b", torch_dtype=torch.bfloat16)
model = model  #.to('cuda')


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


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

    # messages = "".join(["".join(["<|start_header_id|>user<|end_header_id|>\n"+item[0], 
    #                              "<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n"+item[1]])
    #             for item in history_transformer_format])
    messages = [{"role": item[0], "user": item[1]} for item in history_transformer_format]
    print(messages)

    # model_inputs = tokenizer([messages], return_tensors="pt") # .to("cuda")
    model_inputs = tokenizer.apply_chat_template(messages, return_tensors="pt")
    streamer = TextIteratorStreamer(tokenizer, timeout=10., skip_prompt=True, skip_special_tokens=True)
    generate_kwargs = dict(
        model_inputs,
        streamer=streamer,
        max_new_tokens=1024,
        do_sample=True,
        top_p=0.95,
        top_k=1000,
        temperature=1.0,
        num_beams=1,
        stopping_criteria=StoppingCriteriaList([stop])
        )
    t = Thread(target=model.generate, kwargs=generate_kwargs)
    t.start()

    partial_message = ""
    for new_token in streamer:
        if new_token != '<':
            partial_message += new_token
            yield partial_message

gr.ChatInterface(predict).launch(share=True)