File size: 6,736 Bytes
5a7326f d414f05 5a7326f |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 |
{
"cells": [
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [],
"source": [
"!pip install -Uqq git+https://github.com/huggingface/peft.git\n",
"!pip install -Uqq transformers datasets accelerate bitsandbytes\n",
"!pip install sentencepiece\n",
"!pip install einops"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
]
}
],
"source": [
"import torch\n",
"from peft import PeftModel\n",
"from transformers import AutoModelForCausalLM, LlamaTokenizer\n",
"\n",
"MODEL_ID = \"stabilityai/japanese-stablelm-base-alpha-7b\"\n",
"LORA_MODEL_ID = \"tsukemono/japanese-stablelm-base-alpha-7b-qlora-marisa\""
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "VOmiOziuEr6N",
"outputId": "678b317e-f235-43f2-e443-1df05bd20253"
},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"The model weights are not tied. Please use the `tie_weights` method before using the `infer_auto_device` function.\n",
"Loading checkpoint shards: 100%|██████████| 3/3 [01:14<00:00, 24.88s/it]\n"
]
}
],
"source": [
"# model設定\n",
"model = AutoModelForCausalLM.from_pretrained(MODEL_ID, device_map='auto', load_in_8bit=True, torch_dtype=torch.float16, trust_remote_code=True)\n",
"model.eval()\n",
"model = PeftModel.from_pretrained(model, LORA_MODEL_ID, device_map='auto')"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"You are using the legacy behaviour of the <class 'transformers.models.llama.tokenization_llama.LlamaTokenizer'>. This means that tokens that come after special tokens will not be properly handled. We recommend you to read the related pull request available at https://github.com/huggingface/transformers/pull/24565\n"
]
}
],
"source": [
"# tokenizer設定\n",
"tokenizer = LlamaTokenizer.from_pretrained(MODEL_ID, useFast=False)\n",
"ret_token = tokenizer(\"\\n\", truncation=True, add_special_tokens=False)['input_ids'][-1]"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {},
"outputs": [],
"source": [
"# テキスト生成関数の定義\n",
"def generate(text,input=None,maxTokens=512):\n",
" prompt = f\"ユーザー: {text}\\n魔理沙: \"\n",
" input_ids = tokenizer(prompt, \n",
" return_tensors=\"pt\", \n",
" truncation=True, \n",
" add_special_tokens=False\n",
" ).input_ids.cuda()\n",
" with torch.no_grad():\n",
" outputs = model.generate(\n",
" input_ids = input_ids,\n",
" max_length=maxTokens,\n",
" # max_new_tokens=50,\n",
" do_sample=True,\n",
" temperature=0.1,\n",
" top_p=0.9, \n",
" top_k=20,\n",
" no_repeat_ngram_size=2,\n",
" repetition_penalty=1.15,\n",
" pad_token_id=tokenizer.pad_token_id,\n",
" # bad_words_ids=[[186]], # 改行記号\n",
" eos_token_id = [tokenizer.eos_token_id,ret_token]\n",
" )\n",
" outputs = tokenizer.decode(outputs.tolist()[0][input_ids.size(1):],skip_special_tokens=True)\n",
" return outputs.replace(\"\\n\",\"\")"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"'そうだな。今は、この異変を解決する為に動いているからな！その過程で鍛えられたんだと思うぜ。'"
]
},
"execution_count": 5,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"generate(\"強さの秘訣はなんですか?\")"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"'仏教は、人間の煩悩を否定している。しかし、私は人間だ！だから、私の欲望も肯定するべきなんだぜ！'"
]
},
"execution_count": 6,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"generate(\"ブッダの思想についてどう思う?\")"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"'富士山だ。'"
]
},
"execution_count": 7,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"generate(\"日本で一番高い山は?\")"
]
},
{
"cell_type": "code",
"execution_count": 10,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"'そうだな。まず、妖怪がたくさんいるな！あと人間も結構多いぜ。'"
]
},
"execution_count": 10,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"generate(\"幻想郷ってどんな場所?\")"
]
}
],
"metadata": {
"accelerator": "GPU",
"colab": {
"gpuType": "T4",
"provenance": []
},
"gpuClass": "standard",
"kernelspec": {
"display_name": "Python 3",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.9.5"
}
},
"nbformat": 4,
"nbformat_minor": 0
}
|