ufoufo1203x/llm-jp-3-13b-it-v4_lora

Uploaded model

• Developed by: ufoufo1203x
• License: apache-2.0
• Finetuned from model : llm-jp/llm-jp-3-13b

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

以下は、elyza-tasks-100-TV_0.jsonの回答のためのコードです。

from transformers import ( AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig, ) import torch from tqdm import tqdm import json

Replace with your actual Hugging Face access token

HF_TOKEN = "your_access_token" # Placeholder, replace with your token model_name = "ufoufo1203x/llm-jp-3-13b-it-v4_lora"

QLoRA configuration for 4-bit quantization (adjust as needed)

bnb_config = BitsAndBytesConfig( load_in_4bit=True, bnb_4bit_quant_type="nf4", bnb_4bit_compute_dtype=torch.bfloat16, bnb_4bit_use_double_quant=False, )

Load the model (ensure correct path and configuration)

try: model = AutoModelForCausalLM.from_pretrained( model_name, config=bnb_config if bnb_config else None, # Use QLoRA config if provided revision="main", # Specify revision if applicable (optional) use_auth_token=HF_TOKEN, # Use access token for private models ) except Exception as e: print(f"Error loading model: {e}") raise # Re-raise to signal failure

Provide further code for using the model with elyza-tasks-100-TV_0.json