Update README.md
Browse files
README.md
CHANGED
@@ -1 +1,27 @@
|
|
1 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# 설명
|
2 |
+
'\n'을 제거한 데이터셋으로 학습시킨 모델.
|
3 |
+
|
4 |
+
# 사용한 arguments
|
5 |
+
|
6 |
+
```bash
|
7 |
+
python train.py --output_dir ${MODEL_PATH} \
|
8 |
+
--wandb_name ${WANDB_NAME} \
|
9 |
+
--do_train --do_eval \
|
10 |
+
--model_name_or_path "klue/roberta-large" \
|
11 |
+
--dataset_name ../data/noNewline_train_dataset \
|
12 |
+
--save_strategy steps \
|
13 |
+
--save_steps 500 \
|
14 |
+
--save_total_limit 1 \
|
15 |
+
--logging_steps 100 \
|
16 |
+
--evaluation_strategy steps \
|
17 |
+
--eval_steps 500 \
|
18 |
+
--per_device_train_batch_size 8 \
|
19 |
+
--per_device_eval_batch_size 8 \
|
20 |
+
--learning_rate 5e-5 \
|
21 |
+
--weight_decay 0.0 \
|
22 |
+
--num_train_epochs 3.0 \
|
23 |
+
--lr_scheduler_type linear \
|
24 |
+
--warmup_ratio 0.0 \
|
25 |
+
--report_to wandb
|
26 |
+
|
27 |
+
```
|