llm-jp-3-13b-ry-ft1

Uploaded model

• Developed by: ryomac
• Finetuned from model : llm-jp/llm-jp-3-13b

Sample Use

以下は、elyza-tasks-100-TV_0.jsonlの回答のための推論用のコードです。

!pip install -U pip==24.3.1
!pip install -U transformers==4.46.3
!pip install -U bitsandbytes==0.45.0
!pip install -U accelerate==1.2.1
!pip install -U datasets==3.2.0
!pip install -U peft==0.14.0
!pip install -U trl==0.12.2

from transformers import (
    AutoModelForCausalLM,
    AutoTokenizer,
    BitsAndBytesConfig,
    TrainingArguments,
    logging,
)
from peft import (
    LoraConfig,
    PeftModel,
    get_peft_model,
)
import os, torch, gc, re
from datasets import load_dataset
import bitsandbytes as bnb
from trl import SFTTrainer

# 各自HugginFaceのトークンを取得してください
HF_TOKEN = "your-token"

model_id = "llm-jp/llm-jp-3-13b"
adapter_id =  "ryomac/llm-jp-3-13b-ry-ft1"


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


# モデル読み込み
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    quantization_config=bnb_config,
    device_map="auto",
    token=HF_TOKEN
)

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

# Peftモデルを適用
model = PeftModel.from_pretrained(model, adapter_id, token=HF_TOKEN)


# データセットの読み込み。
datasets = []
with open("./elyza-tasks-100-TV_0.jsonl", "r") as f:
    item = ""
    for line in f:
      line = line.strip()
      item += line
      if item.endswith("}"):
        datasets.append(json.loads(item))
        item = ""
results = []
for data in tqdm(datasets):
  input = data["input"]

  prompt = f"""### 指示
  {input}
  ### 回答：
  """

  tokenized_input = tokenizer.encode(prompt, add_special_tokens=False, return_tensors="pt").to(model.device)
  with torch.no_grad():
      outputs = model.generate(
          tokenized_input,
          max_new_tokens=300,
          do_sample=False,
          repetition_penalty=1.2
      )[0]
  output = tokenizer.decode(outputs[tokenized_input.size(1):], skip_special_tokens=True)
  results.append({"task_id": data["task_id"], "input": input, "output": output})

model_name = re.sub(".*/", "", model_name)
with open(f"./{model_name}-my-original-outputs.jsonl", 'w', encoding='utf-8') as f:
    for result in results:
        json.dump(result, f, ensure_ascii=False)  # ensure_ascii=False for handling non-ASCII characters
        f.write('\n')