app.py

import os
from threading import Thread
from typing import Iterator

import gradio as gr
import spaces
import torch
from transformers import AutoModelForCausalLM, Qwen2TokenizerFast, TextIteratorStreamer

# based on gemma demo code

DESCRIPTION = """\
# SeaLLMs Med

SeaLLMs Med is a medical version of SeaLLMs
"""

MAX_MAX_NEW_TOKENS = 2048
DEFAULT_MAX_NEW_TOKENS = 1024
MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")

#model_id = "SeaLLMs/SeaLLMs-v3-7B-Chat"
model_id = "SeaLLMs/debug_6"
tokenizer = Qwen2TokenizerFast.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    device_map="auto",
    torch_dtype=torch.bfloat16,
)
model.config.sliding_window = 4096
model.eval()


system_prompt = "You are a helpful assistant."


@spaces.GPU(duration=90)
def generate(
    message: str,
    chat_history: list[tuple[str, str]],
    max_new_tokens: int = 1024,
    temperature: float = 0.6,
    top_p: float = 0.9,
    top_k: int = 50,
    repetition_penalty: float = 1.2,
) -> Iterator[str]:
    conversation = [{"role": "system", "content": system_prompt}]
    for user, assistant in chat_history:
        conversation.extend(
            [
                {"role": "user", "content": user},
                {"role": "assistant", "content": assistant},
            ]
        )
    conversation.append({"role": "user", "content": message})

    input_ids = tokenizer.apply_chat_template(conversation, add_generation_prompt=True, return_tensors="pt")
    if input_ids.shape[1] > MAX_INPUT_TOKEN_LENGTH:
        input_ids = input_ids[:, -MAX_INPUT_TOKEN_LENGTH:]
        gr.Warning(f"Trimmed input from conversation as it was longer than {MAX_INPUT_TOKEN_LENGTH} tokens.")
    input_ids = input_ids.to(model.device)

    streamer = TextIteratorStreamer(tokenizer, timeout=60.0, skip_prompt=True, skip_special_tokens=True)
    generate_kwargs = dict(
        {"input_ids": input_ids},
        streamer=streamer,
        max_new_tokens=max_new_tokens,
        do_sample=True,
        top_p=top_p,
        top_k=top_k,
        temperature=temperature,
        num_beams=1,
        repetition_penalty=repetition_penalty,
        eos_token_id=tokenizer.eos_token_id,
    )
    t = Thread(target=model.generate, kwargs=generate_kwargs)
    t.start()

    outputs = []
    for text in streamer:
        outputs.append(text)
        yield "".join(outputs)


chat_interface = gr.ChatInterface(
    fn=generate,
    additional_inputs=[],
    stop_btn=None,
    examples=[
        ["Explain the Symptoms of Covid-19"],
    ]
    cache_examples=False,
)

with gr.Blocks(css="style.css", fill_height=True) as demo:
    gr.Markdown(DESCRIPTION)
    gr.DuplicateButton(value="Duplicate Space for private use", elem_id="duplicate-button")
    chat_interface.render()

if __name__ == "__main__":
    demo.queue(max_size=20).launch()