phamff/vietnamese-legal-lora-adapter

from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel
import torch

# Load base model
base_model = AutoModelForCausalLM.from_pretrained(
    "1TuanPham/T-VisStar-7B-v0.1",
    torch_dtype=torch.bfloat16,
    device_map="auto"
)

# Load LoRA adapter
model = PeftModel.from_pretrained(base_model, "phamff/vietnamese-legal-lora-adapter")
tokenizer = AutoTokenizer.from_pretrained("1TuanPham/T-VisStar-7B-v0.1")

# Generate
question = "Quyền và nghĩa vụ của công dân là gì?"
prompt = f"<|user|>\n{question}\n<|assistant|>\n"
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_length=512, temperature=0.7)
answer = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(answer.split("<|assistant|>\n")[-1])