import os from threading import Thread from typing import Iterator import gradio as gr import spaces import torch from transformers import AutoModelForCausalLM, GemmaTokenizerFast, TextIteratorStreamer 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 = "google/gemma-2-9b-it" tokenizer = GemmaTokenizerFast.from_pretrained(model_id) model = AutoModelForCausalLM.from_pretrained( model_id, device_map="auto", torch_dtype=torch.bfloat16, ) model.config.sliding_window = 4096 model.eval() @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 = [] 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=20.0, 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, ) 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, chatbot=gr.Chatbot(height=500, label = "日本語アシスタント", show_label=True), textbox=gr.Textbox(placeholder="メッセージを入力してください", container=False, scale=7), additional_inputs=[ gr.Slider( label="テキスト作成時の最大単語数", minimum=1, maximum=MAX_MAX_NEW_TOKENS, step=1, value=DEFAULT_MAX_NEW_TOKENS, ), gr.Slider( label="創造", minimum=0.1, maximum=4.0, step=0.1, value=0.2, ), gr.Slider( label="最も確率の高い単語のグループ", minimum=0.05, maximum=1.0, step=0.05, value=0.9, ), gr.Slider( label="上位の単語の確率が最も高い(top-k)", minimum=1, maximum=1000, step=1, value=50, ), gr.Slider( label="懲罰を繰り返す", minimum=1.0, maximum=2.0, step=0.05, value=1.1, ), ], theme="soft", stop_btn=None, examples = [ ["寿司の作り方"], ["美しい着物ドレスの選び方"], ["地震が起きたらどうするか"], ["どうすれば幸せに生きられるか"], ["魚を食べることの利点"], ["グループで効果的に作業する方法"] ], cache_examples=False, title = "日本語アシスタント", clear_btn="🗑️ 消す", undo_btn="↩️ 元に戻す", submit_btn="🚀 送信", retry_btn="🔄 リトライ", additional_inputs_accordion="高度なカスタマイズ", ) if __name__ == "__main__": chat_interface.queue(max_size=20).launch()