from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained(
    "bh1scw/qwen3-0.6b-notifai",
    device_map="auto",
    trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("bh1scw/qwen3-0.6b-notifai")

# 推理
messages = [{"role": "user", "content": "App: Slack\nTitle: #critical\nBody: PROD DOWN"}]
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=50)
print(tokenizer.decode(outputs[0]))

模型架构

Base Model: Qwen/Qwen3-0.6B-Base
Fine-tuning: LoRA
Quantization: 4-bit (bitsandbytes)
Model Size: ~529 MB

任务

该模型用于通知分类：

文件夹分类: Work, Personal, Alerts, Promotions
优先级预测: 1-5 (5 = 最紧急)

许可证

Apache 2.0