kittokito commited on
Commit
d592964
1 Parent(s): eac24d9

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +89 -0
README.md CHANGED
@@ -9,6 +9,9 @@ tags:
9
  license: apache-2.0
10
  language:
11
  - en
 
 
 
12
  ---
13
 
14
  # Uploaded model
@@ -20,3 +23,89 @@ language:
20
  This llama model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.
21
 
22
  [<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9
  license: apache-2.0
10
  language:
11
  - en
12
+ - ja
13
+ datasets:
14
+ - elyza/ELYZA-tasks-100
15
  ---
16
 
17
  # Uploaded model
 
23
  This llama model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.
24
 
25
  [<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)
26
+
27
+ # Useage
28
+ 以下のコードを Google Colab で実行してください。
29
+
30
+ ``` python
31
+ # 必要なライブラリをインストール
32
+ !pip install unsloth
33
+ !pip uninstall unsloth -y && pip install --upgrade --no-cache-dir "unsloth[colab-new] @ git+https://github.com/unslothai/unsloth.git"
34
+ !pip install -U torch
35
+ !pip install -U peft
36
+
37
+ # 必要なライブラリを読み込み
38
+ from unsloth import FastLanguageModel
39
+ from peft import PeftModel
40
+ import torch
41
+ import json
42
+ from tqdm import tqdm
43
+
44
+ # 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)。
45
+ model_id = "llm-jp/llm-jp-3-13b"
46
+ adapter_id = "kittokito/llm-jp-3-13b-it-202412170007"
47
+
48
+ from google.colab import userdata
49
+ HF_TOKEN=userdata.get('HF_TOKEN')
50
+
51
+ # unslothのFastLanguageModelで元のモデルをロード。
52
+ dtype = None # Noneにしておけば自動で設定
53
+ load_in_4bit = True # 4bit量子化でモデルのパラメーターをダウンロード
54
+
55
+ model, tokenizer = FastLanguageModel.from_pretrained(
56
+ model_name=model_id,
57
+ dtype=dtype,
58
+ load_in_4bit=load_in_4bit,
59
+ trust_remote_code=True,
60
+ )
61
+
62
+ # 元のモデルにLoRAのアダプタを統合。
63
+ model = PeftModel.from_pretrained(model, adapter_id, token = HF_TOKEN)
64
+
65
+ # タスクとなるデータの読み込み。
66
+ # 事前にデータをアップロードしてください。
67
+ datasets = []
68
+ with open("./elyza-tasks-100-TV_0.jsonl", "r") as f:
69
+ item = ""
70
+ for line in f:
71
+ line = line.strip()
72
+ item += line
73
+ if item.endswith("}"):
74
+ datasets.append(json.loads(item))
75
+ item = ""
76
+
77
+ # モデルを用いてタスクの推論。
78
+
79
+ # 推論するためにモデルのモードを変更
80
+ FastLanguageModel.for_inference(model)
81
+
82
+ results = []
83
+ for dt in tqdm(datasets):
84
+ input = dt["input"]
85
+
86
+ prompt = f"""### 指示\n{input}\n### 回答\n"""
87
+
88
+ inputs = tokenizer([prompt], return_tensors = "pt").to(model.device)
89
+
90
+ outputs = model.generate(**inputs, max_new_tokens = 512, use_cache = True, do_sample=False, repetition_penalty=1.2)
91
+ prediction = tokenizer.decode(outputs[0], skip_special_tokens=True).split('\n### 回答')[-1]
92
+
93
+ results.append({"task_id": dt["task_id"], "input": input, "output": prediction})
94
+
95
+ # 結果をjsonlで保存。
96
+ with open(f"/content/{json_file_id}_output.jsonl", 'w', encoding='utf-8') as f:
97
+ for result in results:
98
+ json.dump(result, f, ensure_ascii=False)
99
+ f.write('\n')
100
+
101
+ ```
102
+
103
+ # Datasets
104
+
105
+ ## Instruction Tuning
106
+ The models have been fine-tuned on the following datasets.
107
+
108
+ | **Language** | **Dataset** | **Description** |
109
+ |--------------|----------------------------------------------|-------------------------------------------------------------------------------|
110
+ | Japanese | [ichikara-instruction-003-001-1.json](http://liat-aip.sakura.ne.jp/wp/llmのための日本語インストラクションデータ作成/llmのための日本語インストラクションデータ-公開/) | A manually constructed instruction dataset. |
111
+ | | Synthesized data from [Elyza-tasks-100](https://huggingface.co/datasets/elyza/ELYZA-tasks-100) | Synthesized data from [Elyza-tasks-100](https://huggingface.co/datasets/elyza/ELYZA-tasks-100) by using LLM (Mixtral-8x22B). |