Update training.ipynb
Browse files- training.ipynb +477 -233
training.ipynb
CHANGED
@@ -1,278 +1,522 @@
|
|
1 |
{
|
2 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
{
|
4 |
-
|
5 |
-
"
|
6 |
-
|
7 |
-
"source": [
|
8 |
-
"The primary codes below are based on [akpe12/JP-KR-ocr-translator-for-travel](https://github.com/akpe12/JP-KR-ocr-translator-for-travel)."
|
9 |
]
|
10 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
11 |
{
|
12 |
-
|
13 |
-
"
|
14 |
-
|
15 |
-
},
|
16 |
-
"source": [
|
17 |
-
"## Import"
|
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 |
-
"from transformers.models.encoder_decoder.modeling_encoder_decoder import EncoderDecoderModel\n",
|
46 |
-
"\n",
|
47 |
-
"# encoder_model_name = \"xlm-roberta-base\"\n",
|
48 |
-
"encoder_model_name = \"cl-tohoku/bert-base-japanese-v2\"\n",
|
49 |
-
"decoder_model_name = \"skt/kogpt2-base-v2\""
|
50 |
-
]
|
51 |
-
},
|
52 |
{
|
53 |
-
|
54 |
-
"
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
59 |
-
"source": [
|
60 |
-
"device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")\n",
|
61 |
-
"# device = torch.device(\"cpu\")\n",
|
62 |
-
"device, torch.cuda.device_count()"
|
63 |
]
|
|
|
|
|
|
|
64 |
},
|
65 |
{
|
66 |
-
|
67 |
-
"
|
68 |
-
|
69 |
-
|
|
|
70 |
},
|
71 |
-
"
|
72 |
-
|
73 |
-
"class GPT2Tokenizer(PreTrainedTokenizerFast):\n",
|
74 |
-
" def build_inputs_with_special_tokens(self, token_ids: List[int]) -> List[int]:\n",
|
75 |
-
" return token_ids + [self.eos_token_id] \n",
|
76 |
-
"\n",
|
77 |
-
"src_tokenizer = BertJapaneseTokenizer.from_pretrained(encoder_model_name)\n",
|
78 |
-
"trg_tokenizer = GPT2Tokenizer.from_pretrained(decoder_model_name, bos_token='</s>', eos_token='</s>', unk_token='<unk>',\n",
|
79 |
-
" pad_token='<pad>', mask_token='<mask>')"
|
80 |
]
|
|
|
|
|
|
|
81 |
},
|
82 |
{
|
83 |
-
|
84 |
-
"
|
85 |
-
|
86 |
-
|
87 |
-
"
|
88 |
-
|
89 |
]
|
|
|
|
|
|
|
90 |
},
|
91 |
{
|
92 |
-
|
93 |
-
"
|
94 |
-
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
"source": [
|
99 |
-
"class PairedDataset:\n",
|
100 |
-
" def __init__(self, \n",
|
101 |
-
" src_tokenizer: PreTrainedTokenizerFast, tgt_tokenizer: PreTrainedTokenizerFast,\n",
|
102 |
-
" file_path: str\n",
|
103 |
-
" ):\n",
|
104 |
-
" self.src_tokenizer = src_tokenizer\n",
|
105 |
-
" self.trg_tokenizer = tgt_tokenizer\n",
|
106 |
-
" with open(file_path, 'r') as fd:\n",
|
107 |
-
" reader = csv.reader(fd)\n",
|
108 |
-
" next(reader)\n",
|
109 |
-
" self.data = [row for row in reader]\n",
|
110 |
-
"\n",
|
111 |
-
" def __getitem__(self, index: int) -> Dict[str, torch.Tensor]:\n",
|
112 |
-
"# with open('train_log.txt', 'a+') as log_file:\n",
|
113 |
-
"# log_file.write(f'reading data[{index}] {self.data[index]}\\n')\n",
|
114 |
-
" src, trg = self.data[index]\n",
|
115 |
-
" embeddings = self.src_tokenizer(src, return_attention_mask=False, return_token_type_ids=False)\n",
|
116 |
-
" embeddings['labels'] = self.trg_tokenizer.build_inputs_with_special_tokens(self.trg_tokenizer(trg, return_attention_mask=False)['input_ids'])\n",
|
117 |
-
"\n",
|
118 |
-
" return embeddings\n",
|
119 |
-
"\n",
|
120 |
-
" def __len__(self):\n",
|
121 |
-
" return len(self.data)\n",
|
122 |
-
" \n",
|
123 |
-
"DATA_ROOT = './output'\n",
|
124 |
-
"FILE_FFAC_FULL = 'ffac_full.csv'\n",
|
125 |
-
"FILE_FFAC_TEST = 'ffac_test.csv'\n",
|
126 |
-
"FILE_JA_KO_TRAIN = 'ja_ko_train.csv'\n",
|
127 |
-
"FILE_JA_KO_TEST = 'ja_ko_test.csv'\n",
|
128 |
-
"\n",
|
129 |
-
"# train_dataset = PairedDataset(src_tokenizer, trg_tokenizer, f'{DATA_ROOT}/{FILE_FFAC_FULL}')\n",
|
130 |
-
"# eval_dataset = PairedDataset(src_tokenizer, trg_tokenizer, f'{DATA_ROOT}/{FILE_FFAC_TEST}') \n",
|
131 |
-
"train_dataset = PairedDataset(src_tokenizer, trg_tokenizer, f'{DATA_ROOT}/{FILE_JA_KO_TRAIN}')\n",
|
132 |
-
"eval_dataset = PairedDataset(src_tokenizer, trg_tokenizer, f'{DATA_ROOT}/{FILE_JA_KO_TEST}') "
|
133 |
]
|
|
|
|
|
|
|
134 |
},
|
135 |
{
|
136 |
-
|
137 |
-
"
|
138 |
-
|
139 |
-
|
140 |
-
"
|
141 |
-
|
142 |
-
"# at the `src, trg = self.data[index]`\n",
|
143 |
-
"# The `cat ffac_full.csv tteb_train.csv > ja_ko_train.csv` command may be the reason.\n",
|
144 |
-
"# the last row of first csv and first row of second csv is merged and that's why 3rd column is created (which arouse ValueError)\n",
|
145 |
-
"# debug_data = train_dataset.data\n"
|
146 |
]
|
|
|
|
|
|
|
147 |
},
|
148 |
{
|
149 |
-
|
150 |
-
"
|
151 |
-
|
|
|
|
|
152 |
},
|
153 |
-
"
|
154 |
-
|
155 |
]
|
|
|
|
|
|
|
156 |
},
|
157 |
{
|
158 |
-
|
159 |
-
"
|
160 |
-
|
161 |
-
|
162 |
-
|
163 |
-
|
164 |
-
"source": [
|
165 |
-
"model = EncoderDecoderModel.from_encoder_decoder_pretrained(\n",
|
166 |
-
" encoder_model_name,\n",
|
167 |
-
" decoder_model_name,\n",
|
168 |
-
" pad_token_id=trg_tokenizer.bos_token_id,\n",
|
169 |
-
")\n",
|
170 |
-
"model.config.decoder_start_token_id = trg_tokenizer.bos_token_id"
|
171 |
]
|
|
|
|
|
|
|
172 |
},
|
173 |
{
|
174 |
-
|
175 |
-
"
|
176 |
-
|
177 |
-
|
178 |
-
|
179 |
-
|
180 |
-
"source": [
|
181 |
-
"# for Trainer\n",
|
182 |
-
"import wandb\n",
|
183 |
-
"\n",
|
184 |
-
"collate_fn = DataCollatorForSeq2Seq(src_tokenizer, model)\n",
|
185 |
-
"wandb.init(project=\"fftr-poc1\", name='jbert+kogpt2')\n",
|
186 |
-
"\n",
|
187 |
-
"arguments = Seq2SeqTrainingArguments(\n",
|
188 |
-
" output_dir='dump',\n",
|
189 |
-
" do_train=True,\n",
|
190 |
-
" do_eval=True,\n",
|
191 |
-
" evaluation_strategy=\"epoch\",\n",
|
192 |
-
" save_strategy=\"epoch\",\n",
|
193 |
-
" num_train_epochs=3,\n",
|
194 |
-
" # num_train_epochs=25,\n",
|
195 |
-
" per_device_train_batch_size=30,\n",
|
196 |
-
" # per_device_train_batch_size=64,\n",
|
197 |
-
" per_device_eval_batch_size=30,\n",
|
198 |
-
" # per_device_eval_batch_size=64,\n",
|
199 |
-
" warmup_ratio=0.1,\n",
|
200 |
-
" gradient_accumulation_steps=4,\n",
|
201 |
-
" save_total_limit=5,\n",
|
202 |
-
" dataloader_num_workers=1,\n",
|
203 |
-
" fp16=True,\n",
|
204 |
-
" load_best_model_at_end=True,\n",
|
205 |
-
" report_to='wandb'\n",
|
206 |
-
")\n",
|
207 |
-
"\n",
|
208 |
-
"trainer = Trainer(\n",
|
209 |
-
" model,\n",
|
210 |
-
" arguments,\n",
|
211 |
-
" data_collator=collate_fn,\n",
|
212 |
-
" train_dataset=train_dataset,\n",
|
213 |
-
" eval_dataset=eval_dataset\n",
|
214 |
-
")"
|
215 |
]
|
|
|
|
|
|
|
216 |
},
|
217 |
{
|
218 |
-
|
219 |
-
"
|
220 |
-
|
221 |
-
|
222 |
-
"
|
223 |
-
|
224 |
]
|
|
|
|
|
|
|
225 |
},
|
226 |
{
|
227 |
-
|
228 |
-
"
|
229 |
-
|
230 |
-
|
231 |
-
|
232 |
-
|
233 |
-
"source": [
|
234 |
-
"# model = EncoderDecoderModel.from_encoder_decoder_pretrained(\"xlm-roberta-base\", \"skt/kogpt2-base-v2\")"
|
235 |
]
|
|
|
|
|
|
|
236 |
},
|
237 |
{
|
238 |
-
|
239 |
-
"
|
240 |
-
|
241 |
-
|
242 |
-
|
243 |
-
|
244 |
-
"source": [
|
245 |
-
"trainer.train()\n",
|
246 |
-
"\n",
|
247 |
-
"model.save_pretrained(\"dump/best_model\")\n",
|
248 |
-
"src_tokenizer.save_pretrained(\"dump/best_model/src_tokenizer\")\n",
|
249 |
-
"trg_tokenizer.save_pretrained(\"dump/best_model/trg_tokenizer\")"
|
250 |
]
|
|
|
|
|
|
|
251 |
}
|
252 |
-
|
253 |
-
|
254 |
-
"
|
255 |
-
|
256 |
-
|
257 |
-
|
258 |
-
"
|
259 |
-
"
|
260 |
-
|
261 |
-
|
262 |
-
|
263 |
-
"
|
264 |
-
|
265 |
-
|
266 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
267 |
},
|
268 |
-
"
|
269 |
-
|
270 |
-
|
271 |
-
|
272 |
-
|
273 |
-
|
274 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
275 |
},
|
276 |
-
"
|
277 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
278 |
}
|
|
|
1 |
{
|
2 |
+
"cells": [
|
3 |
+
{
|
4 |
+
"attachments": {},
|
5 |
+
"cell_type": "markdown",
|
6 |
+
"metadata": {},
|
7 |
+
"source": [
|
8 |
+
"The primary codes below are based on [akpe12/JP-KR-ocr-translator-for-travel](https://github.com/akpe12/JP-KR-ocr-translator-for-travel)."
|
9 |
+
]
|
10 |
+
},
|
11 |
+
{
|
12 |
+
"cell_type": "markdown",
|
13 |
+
"metadata": {
|
14 |
+
"id": "TrHlPFqwFAgj"
|
15 |
+
},
|
16 |
+
"source": [
|
17 |
+
"## Import"
|
18 |
+
]
|
19 |
+
},
|
20 |
+
{
|
21 |
+
"cell_type": "code",
|
22 |
+
"execution_count": 1,
|
23 |
+
"metadata": {
|
24 |
+
"id": "t-jXeSJKE1WM"
|
25 |
+
},
|
26 |
+
"outputs": [],
|
27 |
+
"source": [
|
28 |
+
"from typing import Dict, List\n",
|
29 |
+
"import csv\n",
|
30 |
+
"\n",
|
31 |
+
"import datasets\n",
|
32 |
+
"import torch\n",
|
33 |
+
"from transformers import (\n",
|
34 |
+
" PreTrainedTokenizerFast,\n",
|
35 |
+
" DataCollatorForSeq2Seq,\n",
|
36 |
+
" Seq2SeqTrainingArguments,\n",
|
37 |
+
" BertJapaneseTokenizer,\n",
|
38 |
+
" Trainer\n",
|
39 |
+
")\n",
|
40 |
+
"from transformers.models.encoder_decoder.modeling_encoder_decoder import EncoderDecoderModel\n",
|
41 |
+
"\n",
|
42 |
+
"from datasets import load_dataset\n",
|
43 |
+
"\n",
|
44 |
+
"# encoder_model_name = \"xlm-roberta-base\"\n",
|
45 |
+
"encoder_model_name = \"cl-tohoku/bert-base-japanese-v2\"\n",
|
46 |
+
"decoder_model_name = \"skt/kogpt2-base-v2\""
|
47 |
+
]
|
48 |
+
},
|
49 |
+
{
|
50 |
+
"cell_type": "code",
|
51 |
+
"execution_count": 2,
|
52 |
+
"metadata": {
|
53 |
+
"id": "nEW5trBtbykK"
|
54 |
+
},
|
55 |
+
"outputs": [
|
56 |
{
|
57 |
+
"data": {
|
58 |
+
"text/plain": [
|
59 |
+
"(device(type='cpu'), 0)"
|
|
|
|
|
60 |
]
|
61 |
+
},
|
62 |
+
"execution_count": 2,
|
63 |
+
"metadata": {},
|
64 |
+
"output_type": "execute_result"
|
65 |
+
}
|
66 |
+
],
|
67 |
+
"source": [
|
68 |
+
"device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")\n",
|
69 |
+
"# device = torch.device(\"cpu\")\n",
|
70 |
+
"device, torch.cuda.device_count()"
|
71 |
+
]
|
72 |
+
},
|
73 |
+
{
|
74 |
+
"cell_type": "code",
|
75 |
+
"execution_count": 3,
|
76 |
+
"metadata": {
|
77 |
+
"id": "5ic7pUUBFU_v"
|
78 |
+
},
|
79 |
+
"outputs": [],
|
80 |
+
"source": [
|
81 |
+
"class GPT2Tokenizer(PreTrainedTokenizerFast):\n",
|
82 |
+
" def build_inputs_with_special_tokens(self, token_ids: List[int]) -> List[int]:\n",
|
83 |
+
" return token_ids + [self.eos_token_id] \n",
|
84 |
+
"\n",
|
85 |
+
"src_tokenizer = BertJapaneseTokenizer.from_pretrained(encoder_model_name)\n",
|
86 |
+
"trg_tokenizer = GPT2Tokenizer.from_pretrained(decoder_model_name, bos_token='</s>', eos_token='</s>', unk_token='<unk>',\n",
|
87 |
+
" pad_token='<pad>', mask_token='<mask>')"
|
88 |
+
]
|
89 |
+
},
|
90 |
+
{
|
91 |
+
"cell_type": "markdown",
|
92 |
+
"metadata": {
|
93 |
+
"id": "DTf4U1fmFQFh"
|
94 |
+
},
|
95 |
+
"source": [
|
96 |
+
"## Data"
|
97 |
+
]
|
98 |
+
},
|
99 |
+
{
|
100 |
+
"cell_type": "code",
|
101 |
+
"execution_count": 4,
|
102 |
+
"metadata": {
|
103 |
+
"collapsed": false
|
104 |
+
},
|
105 |
+
"outputs": [],
|
106 |
+
"source": [
|
107 |
+
"dataset = load_dataset(\"sappho192/Tatoeba-Challenge-jpn-kor\")\n",
|
108 |
+
"# dataset = load_dataset(\"D:\\\\REPO\\\\Tatoeba-Challenge-jpn-kor\")\n",
|
109 |
+
"\n",
|
110 |
+
"train_dataset = dataset['train']\n",
|
111 |
+
"test_dataset = dataset['test']\n",
|
112 |
+
"\n",
|
113 |
+
"train_first_row = train_dataset[0]\n",
|
114 |
+
"test_first_row = test_dataset[0]"
|
115 |
+
]
|
116 |
+
},
|
117 |
+
{
|
118 |
+
"cell_type": "code",
|
119 |
+
"execution_count": 5,
|
120 |
+
"metadata": {
|
121 |
+
"id": "65L4O1c5FLKt"
|
122 |
+
},
|
123 |
+
"outputs": [],
|
124 |
+
"source": [
|
125 |
+
"class PairedDataset:\n",
|
126 |
+
" def __init__(self, \n",
|
127 |
+
" source_tokenizer: PreTrainedTokenizerFast, target_tokenizer: PreTrainedTokenizerFast,\n",
|
128 |
+
" file_path: str = None,\n",
|
129 |
+
" dataset_raw: datasets.Dataset = None\n",
|
130 |
+
" ):\n",
|
131 |
+
" self.src_tokenizer = source_tokenizer\n",
|
132 |
+
" self.trg_tokenizer = target_tokenizer\n",
|
133 |
+
" \n",
|
134 |
+
" if file_path is not None:\n",
|
135 |
+
" with open(file_path, 'r') as fd:\n",
|
136 |
+
" reader = csv.reader(fd)\n",
|
137 |
+
" next(reader)\n",
|
138 |
+
" self.data = [row for row in reader]\n",
|
139 |
+
" elif dataset_raw is not None:\n",
|
140 |
+
" self.data = dataset_raw\n",
|
141 |
+
" else:\n",
|
142 |
+
" raise ValueError('file_path or dataset_raw must be specified')\n",
|
143 |
+
"\n",
|
144 |
+
" def __getitem__(self, index: int) -> Dict[str, torch.Tensor]:\n",
|
145 |
+
"# with open('train_log.txt', 'a+') as log_file:\n",
|
146 |
+
"# log_file.write(f'reading data[{index}] {self.data[index]}\\n')\n",
|
147 |
+
" if isinstance(self.data, datasets.Dataset):\n",
|
148 |
+
" src, trg = self.data[index]['sourceString'], self.data[index]['targetString']\n",
|
149 |
+
" else:\n",
|
150 |
+
" src, trg = self.data[index]\n",
|
151 |
+
" embeddings = self.src_tokenizer(src, return_attention_mask=False, return_token_type_ids=False)\n",
|
152 |
+
" embeddings['labels'] = self.trg_tokenizer.build_inputs_with_special_tokens(self.trg_tokenizer(trg, return_attention_mask=False)['input_ids'])\n",
|
153 |
+
"\n",
|
154 |
+
" return embeddings\n",
|
155 |
+
"\n",
|
156 |
+
" def __len__(self):\n",
|
157 |
+
" return len(self.data)"
|
158 |
+
]
|
159 |
+
},
|
160 |
+
{
|
161 |
+
"cell_type": "code",
|
162 |
+
"execution_count": 6,
|
163 |
+
"metadata": {
|
164 |
+
"collapsed": false
|
165 |
+
},
|
166 |
+
"outputs": [],
|
167 |
+
"source": [
|
168 |
+
"DATA_ROOT = './output'\n",
|
169 |
+
"FILE_FFAC_FULL = 'ffac_full.csv'\n",
|
170 |
+
"FILE_FFAC_TEST = 'ffac_test.csv'\n",
|
171 |
+
"FILE_JA_KO_TRAIN = 'ja_ko_train.csv'\n",
|
172 |
+
"FILE_JA_KO_TEST = 'ja_ko_test.csv'\n",
|
173 |
+
"\n",
|
174 |
+
"# train_dataset = PairedDataset(src_tokenizer, trg_tokenizer, file_path=f'{DATA_ROOT}/{FILE_FFAC_FULL}')\n",
|
175 |
+
"# eval_dataset = PairedDataset(src_tokenizer, trg_tokenizer, file_path=f'{DATA_ROOT}/{FILE_FFAC_TEST}') \n",
|
176 |
+
"\n",
|
177 |
+
"# train_dataset = PairedDataset(src_tokenizer, trg_tokenizer, file_path=f'{DATA_ROOT}/{FILE_JA_KO_TRAIN}')\n",
|
178 |
+
"# eval_dataset = PairedDataset(src_tokenizer, trg_tokenizer, file_path=f'{DATA_ROOT}/{FILE_JA_KO_TEST}')"
|
179 |
+
]
|
180 |
+
},
|
181 |
+
{
|
182 |
+
"cell_type": "code",
|
183 |
+
"execution_count": 7,
|
184 |
+
"metadata": {
|
185 |
+
"collapsed": false
|
186 |
+
},
|
187 |
+
"outputs": [
|
188 |
{
|
189 |
+
"data": {
|
190 |
+
"text/plain": [
|
191 |
+
"{'input_ids': [2, 33, 2181, 1402, 893, 15200, 893, 13507, 881, 933, 882, 829, 3], 'labels': [9085, 10936, 10993, 23363, 9134, 18368, 8006, 389, 1]}"
|
|
|
|
|
|
|
192 |
]
|
193 |
+
},
|
194 |
+
"execution_count": 7,
|
195 |
+
"metadata": {},
|
196 |
+
"output_type": "execute_result"
|
197 |
+
}
|
198 |
+
],
|
199 |
+
"source": [
|
200 |
+
"train_dataset = PairedDataset(src_tokenizer, trg_tokenizer, dataset_raw=train_dataset)\n",
|
201 |
+
"eval_dataset = PairedDataset(src_tokenizer, trg_tokenizer, dataset_raw=test_dataset)\n",
|
202 |
+
"eval_dataset[0]"
|
203 |
+
]
|
204 |
+
},
|
205 |
+
{
|
206 |
+
"cell_type": "code",
|
207 |
+
"execution_count": 8,
|
208 |
+
"metadata": {},
|
209 |
+
"outputs": [],
|
210 |
+
"source": [
|
211 |
+
"# be sure to check the column count of each dataset if you encounter \"ValueError: too many values to unpack (expected 2)\"\n",
|
212 |
+
"# at the `src, trg = self.data[index]`\n",
|
213 |
+
"# The `cat ffac_full.csv tteb_train.csv > ja_ko_train.csv` command may be the reason.\n",
|
214 |
+
"# the last row of first csv and first row of second csv is merged and that's why 3rd column is created (which arouse ValueError)\n",
|
215 |
+
"# debug_data = train_dataset.data\n"
|
216 |
+
]
|
217 |
+
},
|
218 |
+
{
|
219 |
+
"cell_type": "markdown",
|
220 |
+
"metadata": {
|
221 |
+
"id": "uCBiLouSFiZY"
|
222 |
+
},
|
223 |
+
"source": [
|
224 |
+
"## Model"
|
225 |
+
]
|
226 |
+
},
|
227 |
+
{
|
228 |
+
"cell_type": "code",
|
229 |
+
"execution_count": 9,
|
230 |
+
"metadata": {
|
231 |
+
"id": "I7uFbFYJFje8"
|
232 |
+
},
|
233 |
+
"outputs": [
|
234 |
{
|
235 |
+
"name": "stderr",
|
236 |
+
"output_type": "stream",
|
237 |
+
"text": [
|
238 |
+
"Some weights of GPT2LMHeadModel were not initialized from the model checkpoint at skt/kogpt2-base-v2 and are newly initialized: ['transformer.h.0.crossattention.c_attn.bias', 'transformer.h.0.crossattention.c_attn.weight', 'transformer.h.0.crossattention.c_proj.bias', 'transformer.h.0.crossattention.c_proj.weight', 'transformer.h.0.crossattention.q_attn.bias', 'transformer.h.0.crossattention.q_attn.weight', 'transformer.h.0.ln_cross_attn.bias', 'transformer.h.0.ln_cross_attn.weight', 'transformer.h.1.crossattention.c_attn.bias', 'transformer.h.1.crossattention.c_attn.weight', 'transformer.h.1.crossattention.c_proj.bias', 'transformer.h.1.crossattention.c_proj.weight', 'transformer.h.1.crossattention.q_attn.bias', 'transformer.h.1.crossattention.q_attn.weight', 'transformer.h.1.ln_cross_attn.bias', 'transformer.h.1.ln_cross_attn.weight', 'transformer.h.10.crossattention.c_attn.bias', 'transformer.h.10.crossattention.c_attn.weight', 'transformer.h.10.crossattention.c_proj.bias', 'transformer.h.10.crossattention.c_proj.weight', 'transformer.h.10.crossattention.q_attn.bias', 'transformer.h.10.crossattention.q_attn.weight', 'transformer.h.10.ln_cross_attn.bias', 'transformer.h.10.ln_cross_attn.weight', 'transformer.h.11.crossattention.c_attn.bias', 'transformer.h.11.crossattention.c_attn.weight', 'transformer.h.11.crossattention.c_proj.bias', 'transformer.h.11.crossattention.c_proj.weight', 'transformer.h.11.crossattention.q_attn.bias', 'transformer.h.11.crossattention.q_attn.weight', 'transformer.h.11.ln_cross_attn.bias', 'transformer.h.11.ln_cross_attn.weight', 'transformer.h.2.crossattention.c_attn.bias', 'transformer.h.2.crossattention.c_attn.weight', 'transformer.h.2.crossattention.c_proj.bias', 'transformer.h.2.crossattention.c_proj.weight', 'transformer.h.2.crossattention.q_attn.bias', 'transformer.h.2.crossattention.q_attn.weight', 'transformer.h.2.ln_cross_attn.bias', 'transformer.h.2.ln_cross_attn.weight', 'transformer.h.3.crossattention.c_attn.bias', 'transformer.h.3.crossattention.c_attn.weight', 'transformer.h.3.crossattention.c_proj.bias', 'transformer.h.3.crossattention.c_proj.weight', 'transformer.h.3.crossattention.q_attn.bias', 'transformer.h.3.crossattention.q_attn.weight', 'transformer.h.3.ln_cross_attn.bias', 'transformer.h.3.ln_cross_attn.weight', 'transformer.h.4.crossattention.c_attn.bias', 'transformer.h.4.crossattention.c_attn.weight', 'transformer.h.4.crossattention.c_proj.bias', 'transformer.h.4.crossattention.c_proj.weight', 'transformer.h.4.crossattention.q_attn.bias', 'transformer.h.4.crossattention.q_attn.weight', 'transformer.h.4.ln_cross_attn.bias', 'transformer.h.4.ln_cross_attn.weight', 'transformer.h.5.crossattention.c_attn.bias', 'transformer.h.5.crossattention.c_attn.weight', 'transformer.h.5.crossattention.c_proj.bias', 'transformer.h.5.crossattention.c_proj.weight', 'transformer.h.5.crossattention.q_attn.bias', 'transformer.h.5.crossattention.q_attn.weight', 'transformer.h.5.ln_cross_attn.bias', 'transformer.h.5.ln_cross_attn.weight', 'transformer.h.6.crossattention.c_attn.bias', 'transformer.h.6.crossattention.c_attn.weight', 'transformer.h.6.crossattention.c_proj.bias', 'transformer.h.6.crossattention.c_proj.weight', 'transformer.h.6.crossattention.q_attn.bias', 'transformer.h.6.crossattention.q_attn.weight', 'transformer.h.6.ln_cross_attn.bias', 'transformer.h.6.ln_cross_attn.weight', 'transformer.h.7.crossattention.c_attn.bias', 'transformer.h.7.crossattention.c_attn.weight', 'transformer.h.7.crossattention.c_proj.bias', 'transformer.h.7.crossattention.c_proj.weight', 'transformer.h.7.crossattention.q_attn.bias', 'transformer.h.7.crossattention.q_attn.weight', 'transformer.h.7.ln_cross_attn.bias', 'transformer.h.7.ln_cross_attn.weight', 'transformer.h.8.crossattention.c_attn.bias', 'transformer.h.8.crossattention.c_attn.weight', 'transformer.h.8.crossattention.c_proj.bias', 'transformer.h.8.crossattention.c_proj.weight', 'transformer.h.8.crossattention.q_attn.bias', 'transformer.h.8.crossattention.q_attn.weight', 'transformer.h.8.ln_cross_attn.bias', 'transformer.h.8.ln_cross_attn.weight', 'transformer.h.9.crossattention.c_attn.bias', 'transformer.h.9.crossattention.c_attn.weight', 'transformer.h.9.crossattention.c_proj.bias', 'transformer.h.9.crossattention.c_proj.weight', 'transformer.h.9.crossattention.q_attn.bias', 'transformer.h.9.crossattention.q_attn.weight', 'transformer.h.9.ln_cross_attn.bias', 'transformer.h.9.ln_cross_attn.weight']\n",
|
239 |
+
"You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n"
|
240 |
+
]
|
241 |
+
}
|
242 |
+
],
|
243 |
+
"source": [
|
244 |
+
"model = EncoderDecoderModel.from_encoder_decoder_pretrained(\n",
|
245 |
+
" encoder_model_name,\n",
|
246 |
+
" decoder_model_name,\n",
|
247 |
+
" pad_token_id=trg_tokenizer.bos_token_id,\n",
|
248 |
+
")\n",
|
249 |
+
"model.config.decoder_start_token_id = trg_tokenizer.bos_token_id"
|
250 |
+
]
|
251 |
+
},
|
252 |
+
{
|
253 |
+
"cell_type": "code",
|
254 |
+
"execution_count": 11,
|
255 |
+
"metadata": {
|
256 |
+
"id": "YFq2GyOAUV0W"
|
257 |
+
},
|
258 |
+
"outputs": [
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
259 |
{
|
260 |
+
"data": {
|
261 |
+
"text/html": [
|
262 |
+
"Finishing last run (ID:1vwqqxps) before initializing another..."
|
263 |
+
],
|
264 |
+
"text/plain": [
|
265 |
+
"<IPython.core.display.HTML object>"
|
|
|
|
|
|
|
|
|
266 |
]
|
267 |
+
},
|
268 |
+
"metadata": {},
|
269 |
+
"output_type": "display_data"
|
270 |
},
|
271 |
{
|
272 |
+
"data": {
|
273 |
+
"application/vnd.jupyter.widget-view+json": {
|
274 |
+
"model_id": "a82aa19a250b43f28d7ecc72eeebc88d",
|
275 |
+
"version_major": 2,
|
276 |
+
"version_minor": 0
|
277 |
},
|
278 |
+
"text/plain": [
|
279 |
+
"VBox(children=(Label(value='0.001 MB of 0.010 MB uploaded\\r'), FloatProgress(value=0.10972568578553615, max=1.…"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
280 |
]
|
281 |
+
},
|
282 |
+
"metadata": {},
|
283 |
+
"output_type": "display_data"
|
284 |
},
|
285 |
{
|
286 |
+
"data": {
|
287 |
+
"text/html": [
|
288 |
+
" View run <strong style=\"color:#cdcd00\">jbert+kogpt2</strong> at: <a href='https://wandb.ai/sappho192/fftr-poc1/runs/1vwqqxps' target=\"_blank\">https://wandb.ai/sappho192/fftr-poc1/runs/1vwqqxps</a><br/>Synced 5 W&B file(s), 0 media file(s), 0 artifact file(s) and 0 other file(s)"
|
289 |
+
],
|
290 |
+
"text/plain": [
|
291 |
+
"<IPython.core.display.HTML object>"
|
292 |
]
|
293 |
+
},
|
294 |
+
"metadata": {},
|
295 |
+
"output_type": "display_data"
|
296 |
},
|
297 |
{
|
298 |
+
"data": {
|
299 |
+
"text/html": [
|
300 |
+
"Find logs at: <code>.\\wandb\\run-20240131_135356-1vwqqxps\\logs</code>"
|
301 |
+
],
|
302 |
+
"text/plain": [
|
303 |
+
"<IPython.core.display.HTML object>"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
304 |
]
|
305 |
+
},
|
306 |
+
"metadata": {},
|
307 |
+
"output_type": "display_data"
|
308 |
},
|
309 |
{
|
310 |
+
"data": {
|
311 |
+
"text/html": [
|
312 |
+
"Successfully finished last run (ID:1vwqqxps). Initializing new run:<br/>"
|
313 |
+
],
|
314 |
+
"text/plain": [
|
315 |
+
"<IPython.core.display.HTML object>"
|
|
|
|
|
|
|
|
|
316 |
]
|
317 |
+
},
|
318 |
+
"metadata": {},
|
319 |
+
"output_type": "display_data"
|
320 |
},
|
321 |
{
|
322 |
+
"data": {
|
323 |
+
"application/vnd.jupyter.widget-view+json": {
|
324 |
+
"model_id": "c2cd7f6fb5b1428b98b80a3cc82ec303",
|
325 |
+
"version_major": 2,
|
326 |
+
"version_minor": 0
|
327 |
},
|
328 |
+
"text/plain": [
|
329 |
+
"VBox(children=(Label(value='Waiting for wandb.init()...\\r'), FloatProgress(value=0.011288888888884685, max=1.0…"
|
330 |
]
|
331 |
+
},
|
332 |
+
"metadata": {},
|
333 |
+
"output_type": "display_data"
|
334 |
},
|
335 |
{
|
336 |
+
"data": {
|
337 |
+
"text/html": [
|
338 |
+
"Tracking run with wandb version 0.16.2"
|
339 |
+
],
|
340 |
+
"text/plain": [
|
341 |
+
"<IPython.core.display.HTML object>"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
342 |
]
|
343 |
+
},
|
344 |
+
"metadata": {},
|
345 |
+
"output_type": "display_data"
|
346 |
},
|
347 |
{
|
348 |
+
"data": {
|
349 |
+
"text/html": [
|
350 |
+
"Run data is saved locally in <code>d:\\REPO\\ffxiv-ja-ko-translator\\wandb\\run-20240131_135421-etxsdxw2</code>"
|
351 |
+
],
|
352 |
+
"text/plain": [
|
353 |
+
"<IPython.core.display.HTML object>"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
354 |
]
|
355 |
+
},
|
356 |
+
"metadata": {},
|
357 |
+
"output_type": "display_data"
|
358 |
},
|
359 |
{
|
360 |
+
"data": {
|
361 |
+
"text/html": [
|
362 |
+
"Syncing run <strong><a href='https://wandb.ai/sappho192/fftr-poc1/runs/etxsdxw2' target=\"_blank\">jbert+kogpt2</a></strong> to <a href='https://wandb.ai/sappho192/fftr-poc1' target=\"_blank\">Weights & Biases</a> (<a href='https://wandb.me/run' target=\"_blank\">docs</a>)<br/>"
|
363 |
+
],
|
364 |
+
"text/plain": [
|
365 |
+
"<IPython.core.display.HTML object>"
|
366 |
]
|
367 |
+
},
|
368 |
+
"metadata": {},
|
369 |
+
"output_type": "display_data"
|
370 |
},
|
371 |
{
|
372 |
+
"data": {
|
373 |
+
"text/html": [
|
374 |
+
" View project at <a href='https://wandb.ai/sappho192/fftr-poc1' target=\"_blank\">https://wandb.ai/sappho192/fftr-poc1</a>"
|
375 |
+
],
|
376 |
+
"text/plain": [
|
377 |
+
"<IPython.core.display.HTML object>"
|
|
|
|
|
378 |
]
|
379 |
+
},
|
380 |
+
"metadata": {},
|
381 |
+
"output_type": "display_data"
|
382 |
},
|
383 |
{
|
384 |
+
"data": {
|
385 |
+
"text/html": [
|
386 |
+
" View run at <a href='https://wandb.ai/sappho192/fftr-poc1/runs/etxsdxw2' target=\"_blank\">https://wandb.ai/sappho192/fftr-poc1/runs/etxsdxw2</a>"
|
387 |
+
],
|
388 |
+
"text/plain": [
|
389 |
+
"<IPython.core.display.HTML object>"
|
|
|
|
|
|
|
|
|
|
|
|
|
390 |
]
|
391 |
+
},
|
392 |
+
"metadata": {},
|
393 |
+
"output_type": "display_data"
|
394 |
}
|
395 |
+
],
|
396 |
+
"source": [
|
397 |
+
"# for Trainer\n",
|
398 |
+
"import wandb\n",
|
399 |
+
"\n",
|
400 |
+
"collate_fn = DataCollatorForSeq2Seq(src_tokenizer, model)\n",
|
401 |
+
"wandb.init(project=\"fftr-poc1\", name='jbert+kogpt2')\n",
|
402 |
+
"\n",
|
403 |
+
"arguments = Seq2SeqTrainingArguments(\n",
|
404 |
+
" output_dir='dump',\n",
|
405 |
+
" do_train=True,\n",
|
406 |
+
" do_eval=True,\n",
|
407 |
+
" evaluation_strategy=\"epoch\",\n",
|
408 |
+
" save_strategy=\"epoch\",\n",
|
409 |
+
" num_train_epochs=3,\n",
|
410 |
+
" # num_train_epochs=25,\n",
|
411 |
+
" per_device_train_batch_size=1,\n",
|
412 |
+
" # per_device_train_batch_size=30, # takes 40GB\n",
|
413 |
+
" # per_device_train_batch_size=64,\n",
|
414 |
+
" per_device_eval_batch_size=1,\n",
|
415 |
+
" # per_device_eval_batch_size=30,\n",
|
416 |
+
" # per_device_eval_batch_size=64,\n",
|
417 |
+
" warmup_ratio=0.1,\n",
|
418 |
+
" gradient_accumulation_steps=4,\n",
|
419 |
+
" save_total_limit=5,\n",
|
420 |
+
" dataloader_num_workers=1,\n",
|
421 |
+
" # fp16=True, # ENABLE if CUDA is enabled\n",
|
422 |
+
" load_best_model_at_end=True,\n",
|
423 |
+
" report_to='wandb'\n",
|
424 |
+
")\n",
|
425 |
+
"\n",
|
426 |
+
"trainer = Trainer(\n",
|
427 |
+
" model,\n",
|
428 |
+
" arguments,\n",
|
429 |
+
" data_collator=collate_fn,\n",
|
430 |
+
" train_dataset=train_dataset,\n",
|
431 |
+
" eval_dataset=eval_dataset\n",
|
432 |
+
")"
|
433 |
+
]
|
434 |
+
},
|
435 |
+
{
|
436 |
+
"cell_type": "markdown",
|
437 |
+
"metadata": {
|
438 |
+
"id": "pPsjDHO5Vc3y"
|
439 |
+
},
|
440 |
+
"source": [
|
441 |
+
"## Training"
|
442 |
+
]
|
443 |
+
},
|
444 |
+
{
|
445 |
+
"cell_type": "code",
|
446 |
+
"execution_count": null,
|
447 |
+
"metadata": {
|
448 |
+
"id": "_T4P4XunmK-C"
|
449 |
+
},
|
450 |
+
"outputs": [],
|
451 |
+
"source": [
|
452 |
+
"# model = EncoderDecoderModel.from_encoder_decoder_pretrained(\"xlm-roberta-base\", \"skt/kogpt2-base-v2\")"
|
453 |
+
]
|
454 |
+
},
|
455 |
+
{
|
456 |
+
"cell_type": "code",
|
457 |
+
"execution_count": 12,
|
458 |
+
"metadata": {
|
459 |
+
"id": "7vTqAgW6Ve3J"
|
460 |
+
},
|
461 |
+
"outputs": [
|
462 |
+
{
|
463 |
+
"data": {
|
464 |
+
"application/vnd.jupyter.widget-view+json": {
|
465 |
+
"model_id": "0afe460e9f614d9a90379cf99fcf8af3",
|
466 |
+
"version_major": 2,
|
467 |
+
"version_minor": 0
|
468 |
},
|
469 |
+
"text/plain": [
|
470 |
+
" 0%| | 0/9671328 [00:00<?, ?it/s]"
|
471 |
+
]
|
472 |
+
},
|
473 |
+
"metadata": {},
|
474 |
+
"output_type": "display_data"
|
475 |
}
|
476 |
+
],
|
477 |
+
"source": [
|
478 |
+
"trainer.train()\n",
|
479 |
+
"\n",
|
480 |
+
"model.save_pretrained(\"dump/best_model\")\n",
|
481 |
+
"src_tokenizer.save_pretrained(\"dump/best_model/src_tokenizer\")\n",
|
482 |
+
"trg_tokenizer.save_pretrained(\"dump/best_model/trg_tokenizer\")"
|
483 |
+
]
|
484 |
+
},
|
485 |
+
{
|
486 |
+
"cell_type": "code",
|
487 |
+
"execution_count": 2,
|
488 |
+
"metadata": {},
|
489 |
+
"outputs": [],
|
490 |
+
"source": [
|
491 |
+
"# import wandb\n",
|
492 |
+
"# wandb.finish()"
|
493 |
+
]
|
494 |
+
}
|
495 |
+
],
|
496 |
+
"metadata": {
|
497 |
+
"colab": {
|
498 |
+
"machine_shape": "hm",
|
499 |
+
"provenance": []
|
500 |
+
},
|
501 |
+
"gpuClass": "premium",
|
502 |
+
"kernelspec": {
|
503 |
+
"display_name": "Python 3 (ipykernel)",
|
504 |
+
"language": "python",
|
505 |
+
"name": "python3"
|
506 |
},
|
507 |
+
"language_info": {
|
508 |
+
"codemirror_mode": {
|
509 |
+
"name": "ipython",
|
510 |
+
"version": 3
|
511 |
+
},
|
512 |
+
"file_extension": ".py",
|
513 |
+
"mimetype": "text/x-python",
|
514 |
+
"name": "python",
|
515 |
+
"nbconvert_exporter": "python",
|
516 |
+
"pygments_lexer": "ipython3",
|
517 |
+
"version": "3.10.13"
|
518 |
+
}
|
519 |
+
},
|
520 |
+
"nbformat": 4,
|
521 |
+
"nbformat_minor": 0
|
522 |
}
|