Sakalti's picture
Update app.py
f6418af verified
raw
history blame
2.01 kB
import gradio as gr
from modelscope.pipelines import pipeline
from modelscope.utils.constant import Tasks
import swift.llm # swift.llm パッケージを直接インポート
# 使用可能なモデルのリスト
models = ["Qwen/Qwen2.5-7B-Instruct", "Qwen/Qwen2.5-14B-Instruct", "Qwen/Qwen2.5-32B-Instruct"]
def respond(
message,
history: list[tuple[str, str]],
system_message,
max_tokens,
temperature,
top_p,
selected_model
):
# 型変換: selected_modelを文字列に変換
selected_model = str(selected_model)
# 選択したモデルに基づいてPipelineを初期化
pipe = pipeline(task=Tasks.text_generation, model=selected_model)
messages = [{"role": "system", "content": system_message}]
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})
response = ""
# モデルの推論
for message in pipe(messages, max_tokens=max_tokens, temperature=temperature, top_p=top_p, stream=True):
token = message.get("choices", [{}])[0].get("delta", {}).get("content", "")
response += token
return response
# インターフェース
demo = gr.ChatInterface(
respond,
additional_inputs=[
gr.Textbox(value="あなたはフレンドリーなチャットボットです。", label="システムメッセージ"),
gr.Slider(minimum=1, maximum=2048, value=768, step=1, label="新規トークン最大"),
gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="温度"),
gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05,Label="Top-p (核 sampling)"),
gr.Dropdown(choices=models, value=models[0], label="モデル"),
],
concurrency_limit=30 # 例: 同時に30つのリクエストを処理
)
if __name__ == "__main__":
demo.launch(share=True)