app.py

import os
import gradio as gr
import spaces
from transformers import AutoTokenizer, AutoModelForCausalLM, TextGenerationPipeline

model_id = "google/gemma-3-12b-it"
hf_token = os.environ.get("HUGGINGFACE_TOKEN")

# 包含模型加载 + 推理
@spaces.GPU
def generate(prompt):
    tokenizer = AutoTokenizer.from_pretrained(model_id, token=hf_token)
    model = AutoModelForCausalLM.from_pretrained(model_id, token=hf_token, device_map="auto")
    pipeline = TextGenerationPipeline(model=model, tokenizer=tokenizer)
    output = pipeline(prompt, max_new_tokens=100, do_sample=True, temperature=0.7)
    return output[0]["generated_text"]

# 构建界面
gr.Interface(
    fn=generate,
    inputs=gr.Text(label="Enter your prompt"),
    outputs=gr.Textbox(label="Generated Text"),
    title="Gemma-3-27B Inference (ZeroGPU)"
).launch()