ken0x0a/llm-jp-3-13b-finetune

出力方法

以下の1~3を順に実行する。
(omnicampusの開発環境での実行を想定)

1. "elyza-tasks-100-TV_0.jsonl" を current directory に配置する。

2. ライブラリをインストール（アップデート）する。

pip install -U bitsandbytes transformers accelerate datasets peft

3. 以下の内容の python script を作り実行する。

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

model_id = "models/models--llm-jp--llm-jp-3-13b/snapshots/cd3823f4c1fcbb0ad2e2af46036ab1b0ca13192a"
# model_id = "llm-jp/llm-jp-3-13b" 
adapter_id = "ken0x0a/llm-jp-3-13b-finetune"
out_filename = "./llm-jp-3-13b-outputs.jsonl"

# 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",
)

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

model = PeftModel.from_pretrained(model, adapter_id)

# 実行
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)
    attention_mask = torch.ones_like(tokenized_input)
    with torch.no_grad():
        outputs = model.generate(
            tokenized_input,
            attention_mask=attention_mask,
            max_new_tokens=100,
            do_sample=False,
            repetition_penalty=1.2,
            pad_token_id=tokenizer.eos_token_id
        )[0]
    output = tokenizer.decode(outputs[tokenized_input.size(1):], skip_special_tokens=True)
    results.append({"task_id": data["task_id"], "output": output})

with open(out_filename, 'w', encoding='utf-8') as f:
    for result in results:
        json.dump(result, f, ensure_ascii=False)
        f.write('\n')