README.md

metadata

library_name: transformers
tags:
  - llama
  - trl
datasets:
  - elyza/ELYZA-tasks-100
language:
  - ja
base_model:
  - llm-jp/llm-jp-3-13b

Model Card for Model ID

Model Details

Model Description

東大松尾研LLM講座2024の最終課題向けのelyza-tasks-100-TV_0.jsonlの出力用にFinetuningしたモデルです。 モデルの利用については、提供いただいたOmmniCampusの環境およびサンプルコードに沿ったものとなっております。

• Developed by: maktag
• Language(s) (NLP): Japanese
• Finetuned from model [optional]: llm-jp/llm-jp-3-13b

How to Get Started with the Model

from transformers import AutoTokenizer, AutoModelForCausalLM

# Load the fine-tuned model and tokenizer
base_model_id = "llm-jp/llm-jp-3-13b" 
adapter_id = "maktag/llm-jp-3-13b-finetune8"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id)

# QLoRA config
bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype=torch.bfloat16,
)

# Load model
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    quantization_config=bnb_config,
    device_map="auto",
    token = HF_TOKEN
)

# Load tokenizer
tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True, token = HF_TOKEN)

# 元のモデルにLoRAのアダプタを統合。
model = PeftModel.from_pretrained(model, adapter_id, token = HF_TOKEN)

[More Information Needed]

Training Details

• Fine-Tuning Framework: LoRA-based PEFT (Parameter-Efficient Fine-Tuning).
• Dataset: Proprietary Japanese instruction-following dataset.
• Sequence Length: 512 tokens.
• Hyperparameters:
  • Batch size: 32
  • Learning rate: 1e-5
  • Epochs: 3

Training Data

• elyza/ELYZA-tasks-100
• Ichikara Instruction