from transformers import (
    AutoModelForCausalLM,
    AutoTokenizer,
    BitsAndBytesConfig,
    TrainingArguments,
    logging,
)
from peft import (
    LoraConfig,
    PeftModel,
    get_peft_model,
)

import os, torch, gc, json
from tqdm import tqdm
from datasets import load_dataset
import bitsandbytes as bnb
from trl import SFTTrainer
from google.colab import userdata

# Hugging Face Token
os.environ["HF_TOKEN"] = userdata.get("HF_TOKEN")

# 推論データ準備
datasets = []

inference_data_path = '/content/drive/MyDrive/your_path'
with open(f"{inference_data_path}/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 = ""

# モデルとトークナイザー準備
new_model_id = "yottan-wywy/llm-jp-3-13b-instruct-finetune_1217"

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

model = AutoModelForCausalLM.from_pretrained(
    new_model_id,
    quantization_config=bnb_config,
    device_map="auto"
)

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

# 推論実行
results = []
system_text = "以下は、タスクを説明する指示です。要求を適切に満たす回答を**簡潔に**書きなさい。"
for data in tqdm(datasets):

  input_text = data["input"]

  prompt = f"""
  {system_text}
  ### 指示
  {input_text}
  ### 応答
  """

  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"], "input": input_text, "output": output})

Model Details

Model type: Transformer-based Language Model

Datasets

Instruction tuning

Language	Dataset	description
Japanese	elyza/ELYZA-tasks-100	A manually constructed instruction dataset

License

Apache License, Version 2.0