84nth08h/llm-jp-3-13b-it_lora_r128_max1024_loraAlpha128_dropout0.10_ichikaraAll

Uploaded model

• Developed by: 84nth08h
• 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.

Usage

Execute following code in Google Colab

max_seq_length = 1024 # Choose any! We auto support RoPE Scaling internally! 
dtype = None # None for auto detection. Float16 for Tesla T4, V100, Bfloat16 for Ampere+
load_in_4bit = True # Use 4bit quantization to reduce memory usage. Can be False.


# 必要なライブラリをインストール
%%capture
!pip install unsloth
!pip uninstall unsloth -y && pip install --upgrade --no-cache-dir "unsloth[colab-new] @ git+https://github.com/unslothai/unsloth.git"
!pip install -U torch
!pip install -U peft


# 必要なライブラリを読み込み
from unsloth import FastLanguageModel
from peft import PeftModel
import torch
import json
from tqdm import tqdm
import re

# ベースとなるモデルと学習したLoRAのアダプタ（Hugging FaceのIDを指定）。
model_id = "llm-jp/llm-jp-3-13b"
adapter_id = "84nth08h/llm-jp-3-13b-it_lora_r128_max1024_loraAlpha128_dropout0.10_ichikaraAll"


from google.colab import userdata
HF_W=userdata.get('HF_W')
HF_R=userdata.get('HF_R')

!huggingface-cli login --token $HF_R

%%time
dtype = None # Noneにしておけば自動で設定
load_in_4bit = True # 今回は13Bモデルを扱うためTrue

model, tokenizer = FastLanguageModel.from_pretrained(
    model_name=adapter_id,
    dtype=dtype,
    load_in_4bit=load_in_4bit,
    trust_remote_code=True,

    # use_gradient_checkpointing = "unsloth",
    # max_seq_length = max_seq_length,
)


# タスクとなるデータの読み込み。
# 事前にデータをアップロードしてください。
# ichikara_elyza_path

datasets = []
with open("./elyza-tasks-100-TV_0.jsonl", "r") as f:
# with open(ichikara_elyza_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 = ""

%%time
# モデルを用いてタスクの推論。

# 推論するためにモデルのモードを変更
FastLanguageModel.for_inference(model)

results = []
for dt in tqdm(datasets):
  input = dt["input"]

  prompt = f"""### 指示\n{input}\n### 回答\n"""

  inputs = tokenizer([prompt], return_tensors = "pt").to(model.device)
  outputs = model.generate(**inputs, max_new_tokens = 1024, use_cache = True, do_sample=True, repetition_penalty=1.2) 
  prediction = tokenizer.decode(outputs[0], skip_special_tokens=True).split('\n### 回答')[-1]
  print(prediction)
  results.append({"task_id": dt["task_id"], "input": input, "output": prediction})

  # 結果をjsonlで保存。

# ここではadapter_idを元にファイル名を決定しているが、ファイル名は任意で問題なし。
json_file_id = re.sub(".*/", "", adapter_id)
with open(f"/content/{json_file_id}_output.jsonl", 'w', encoding='utf-8') as f:
    for result in results:
        json.dump(result, f, ensure_ascii=False)
        f.write('\n')