import torch
from transformers import AutoTokenizer, AutoModelForCausalLM
base_model = 'bigdefence/Llama-3.1-8B-Ko-bigdefence'
device = 'cuda' if torch.cuda.is_available() else 'cpu'

tokenizer = AutoTokenizer.from_pretrained(base_model)
model = AutoModelForCausalLM.from_pretrained(base_model, torch_dtype=torch.float16, device_map="auto")
model.eval()
def generate_response(prompt, model, tokenizer, text_streamer,max_new_tokens=256):
    inputs = tokenizer(prompt, return_tensors="pt", add_special_tokens=True)
    inputs = inputs.to(model.device)

    with torch.no_grad():
        outputs = model.generate(
            **inputs,
            streamer=text_streamer,
            max_new_tokens=max_new_tokens,
            do_sample=True,
            pad_token_id=tokenizer.eos_token_id
        )

    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
    return response.replace(prompt, '').strip()
key = "안녕?"
prompt = f"""Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.

### Instruction:
{key}

### Response:
"""
text_streamer = TextStreamer(tokenizer)
response = generate_response(prompt, model, tokenizer,text_streamer)
print(response)

Uploaded model

• Developed by: Bigdefence
• License: apache-2.0
• Finetuned from model : meta-llama/Meta-Llama-3.1-8B
• Dataset : MarkrAI/KoCommercial-Dataset

Thanks

• 한국어 LLM 오픈생태계에 많은 공헌을 해주신, Beomi 님과 maywell 님, MarkrAI님 감사의 인사 드립니다.

This llama model was trained 2x faster with Unsloth and Huggingface's TRL library.