-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathrun_glue.sh
69 lines (65 loc) · 1.55 KB
/
run_glue.sh
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
#!/bin/bash
TASK="RTE"
BERT_DIR="model/RTE/bert/"
DATA_DIR="data/RTE"
W_BERT="dynaw_bert/RTE/"
WD_BERT="dynawd_bert/RTE/"
echo $TASK
echo "dynaw training"
python run_glue.py \
--model_type bert \
--task_name $TASK \
--do_train \
--data_dir $DATA_DIR \
--model_dir $BERT_DIR \
--output_dir $W_BERT \
--max_seq_length 128 \
--learning_rate 2e-5 \
--per_gpu_train_batch_size 32 \
--per_gpu_eval_batch_size 32 \
--num_train_epochs 8 \
--width_mult_list 0.25,0.5,0.75,1.0 \
--depth_mult_list 1.0 \
--width_lambda1 1.0 \
--width_lambda2 0.1 \
--logging_steps 200 \
--training_phase dynabertw \
--graft_during_training
# --data_aug \
echo "dynawd training"
python run_glue.py \
--model_type bert \
--task_name $TASK \
--do_train \
--data_dir $DATA_DIR \
--model_dir $W_BERT \
--output_dir $WD_BERT \
--max_seq_length 128 \
--learning_rate 2e-5 \
--per_gpu_train_batch_size 32 \
--per_gpu_eval_batch_size 32 \
--num_train_epochs 8 \
--width_mult_list 0.25,0.5,0.75,1.0 \
--depth_mult_list 0.5,0.75,1.0 \
--depth_lambda1 1.0 \
--depth_lambda2 1.0 \
--logging_steps 200 \
--training_phase dynabert
# --data_aug \
echo "final fine-tuning"
python run_glue.py \
--model_type bert \
--task_name $TASK \
--do_train \
--data_dir $DATA_DIR \
--model_dir $WD_BERT \
--output_dir $WD_BERT \
--max_seq_length 128 \
--learning_rate 2e-5 \
--per_gpu_train_batch_size 32 \
--per_gpu_eval_batch_size 32 \
--num_train_epochs 3 \
--width_mult_list 0.25,0.5,0.75,1.0 \
--depth_mult_list 0.5,0.75,1.0 \
--logging_steps 20 \
--training_phase final_finetuning