Update README.md
Browse files
README.md
CHANGED
@@ -1,6 +1,9 @@
|
|
1 |
---
|
2 |
library_name: transformers
|
3 |
-
|
|
|
|
|
|
|
4 |
---
|
5 |
|
6 |
# Model Card for Model ID
|
@@ -14,6 +17,13 @@ tags: []
|
|
14 |
### Model Description
|
15 |
|
16 |
<!-- Provide a longer summary of what this model is. -->
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
17 |
|
18 |
This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
|
19 |
|
@@ -37,6 +47,66 @@ This is the model card of a 🤗 transformers model that has been pushed on the
|
|
37 |
|
38 |
<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
|
39 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
40 |
### Direct Use
|
41 |
|
42 |
<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
|
|
|
1 |
---
|
2 |
library_name: transformers
|
3 |
+
datasets:
|
4 |
+
- llm-jp/magpie-sft-v1.0
|
5 |
+
base_model:
|
6 |
+
- google/gemma-2-9b
|
7 |
---
|
8 |
|
9 |
# Model Card for Model ID
|
|
|
17 |
### Model Description
|
18 |
|
19 |
<!-- Provide a longer summary of what this model is. -->
|
20 |
+
gemma-2-9bを4bit量子化しQloraでllm-jp/magpie-sft-v0.1を用いInstruction Turnnigしたモデルです。
|
21 |
+
|
22 |
+
以下のチャットテンプレートを定義しています。
|
23 |
+
<bos>{%- for message in messages %}
|
24 |
+
<start_of_turn>{{ message.role }}: {{ message.content }}<end_of_turn>
|
25 |
+
{%- endfor %}{% if add_generation_prompt %}
|
26 |
+
<start_of_turn>assistant: {% endif %}<eos>
|
27 |
|
28 |
This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
|
29 |
|
|
|
47 |
|
48 |
<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
|
49 |
|
50 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
|
51 |
+
import torch
|
52 |
+
from peft import PeftModel, PeftConfig
|
53 |
+
|
54 |
+
model_name = "mssfj/gemma-2-9b-bnb-4bit-chat-template"
|
55 |
+
lora_weight = "mssfj/gemma-2-9b-4bit-magpie"
|
56 |
+
|
57 |
+
# 量子化設定
|
58 |
+
quantization_config = BitsAndBytesConfig(
|
59 |
+
load_in_4bit=False,
|
60 |
+
bnb_4bit_compute_dtype=torch.bfloat16,
|
61 |
+
bnb_4bit_quant_type="nf4",
|
62 |
+
bnb_4bit_use_double_quant=False
|
63 |
+
)
|
64 |
+
|
65 |
+
# ベースモデルのロード
|
66 |
+
base_model = AutoModelForCausalLM.from_pretrained(
|
67 |
+
model_name,
|
68 |
+
quantization_config=quantization_config,
|
69 |
+
device_map="auto"
|
70 |
+
)
|
71 |
+
|
72 |
+
# QLoRA済みモデルの適用
|
73 |
+
model = PeftModel.from_pretrained(base_model, lora_weight)
|
74 |
+
|
75 |
+
# トークナイザのロード
|
76 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
77 |
+
|
78 |
+
input="""「図書館で本を読んだ。」という文は「どこで本を読んだ?」という疑問文に直すことができます。
|
79 |
+
このとき、「図書館」は「どこ」の疑問詞タグを持ちます。
|
80 |
+
|
81 |
+
それでは、「本」という単語はどのような疑問詞タグを持つでしょうか? 全て選んでください。対応するものがない場合は「なし」と答えてください。
|
82 |
+
"""
|
83 |
+
|
84 |
+
messages = [
|
85 |
+
{"role": "system", "content": """日本で一番高い山は?
|
86 |
+
"""},
|
87 |
+
{"role": "user", "content": input},
|
88 |
+
]
|
89 |
+
|
90 |
+
# チャットテンプレートを適用
|
91 |
+
input_ids = tokenizer.apply_chat_template(
|
92 |
+
messages,
|
93 |
+
tokenize=True,
|
94 |
+
add_generation_prompt=True,
|
95 |
+
return_tensors="pt"
|
96 |
+
).to(model.device)
|
97 |
+
|
98 |
+
outputs = model.generate(
|
99 |
+
input_ids,
|
100 |
+
max_new_tokens=256,
|
101 |
+
temperature=0.2,
|
102 |
+
do_sample=True,
|
103 |
+
eos_token_id=tokenizer.eos_token_id,
|
104 |
+
pad_token_id=tokenizer.pad_token_id,
|
105 |
+
early_stopping=True,
|
106 |
+
)
|
107 |
+
|
108 |
+
response = tokenizer.decode(outputs[0][input_ids.shape[1]:], skip_special_tokens=True)
|
109 |
+
|
110 |
### Direct Use
|
111 |
|
112 |
<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
|