-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathGUE_CNN.sh
158 lines (147 loc) · 5.72 KB
/
GUE_CNN.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
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
cd ..
cd ..
for dataset in H3_txt H3K4me1_txt H3K4me2_txt H3K4me3_txt H3K9ac_txt H3K14ac_txt H3K36me3_txt H3K79me3_txt H4_txt H4ac_txt; do
for lr in 1e-4; do
python -m train experiment=hg38/GUE \
model.d_model=256 \
model.layer._name_=CNN \
train.pretrained_model_path=weight/hyenadna/hyenadna-large-1m-seqlen \
dataset.dataset_name=$dataset \
optimizer.lr=$lr \
dataset=EMP \
dataset.max_length=128 \
wandb.mode=offline \
trainer.devices=4 \
dataset.batch_size=16 \
dataset.tokenizer_name=CNN \
train.global_batch_size=128 \
dataset.tokenizer_path=weight/hyenadna/hyenadna-large-1m-seqlen \
callbacks.early_stopping.monitor="val/accuracy" \
callbacks.model_checkpoint.monitor="val/accuracy" \
callbacks.model_checkpoint.filename="val/accuracy" \
callbacks.early_stopping.patience=5 \
wandb.id=GUE_EMP_${dataset}_CNN \
decoder=id
done
done
for dataset in 0_txt 1_txt 2_txt 3_txt 4_txt; do
for lr in 1e-4; do
python -m train experiment=hg38/GUE \
model.d_model=256 \
model.layer._name_=CNN \
train.pretrained_model_path=weight/hyenadna/hyenadna-large-1m-seqlen \
dataset.dataset_name=$dataset \
optimizer.lr=$lr \
dataset=mouse \
dataset.max_length=30 \
wandb.mode=offline \
trainer.devices=4 \
dataset.batch_size=28 \
dataset.tokenizer_name=CNN \
train.global_batch_size=128 \
callbacks.early_stopping.monitor="val/accuracy" \
callbacks.model_checkpoint.monitor="val/accuracy" \
callbacks.model_checkpoint.filename="val/accuracy" \
callbacks.early_stopping.patience=5 \
wandb.id=GUE_mouse_${dataset}_CNN \
decoder=id
done
done
for dataset in prom_300_all_txt prom_300_notata_txt prom_300_tata_txt prom_core_all_txt prom_core_notata_txt prom_core_tata_txt; do
for lr in 1e-4; do
python -m train experiment=hg38/GUE \
model.d_model=256 \
model.layer._name_=CNN \
train.pretrained_model_path=weight/hyenadna/hyenadna-large-1m-seqlen \
dataset.dataset_name=$dataset \
optimizer.lr=$lr \
dataset=prom \
dataset.max_length=70 \
wandb.mode=offline \
trainer.devices=4 \
dataset.batch_size=16 \
dataset.tokenizer_name=CNN \
train.global_batch_size=128 \
dataset.tokenizer_path=/weight/hyenadna/hyenadna-large-1m-seqlen \
callbacks.early_stopping.monitor="val/accuracy" \
callbacks.model_checkpoint.monitor="val/accuracy" \
callbacks.model_checkpoint.filename="val/accuracy" \
callbacks.early_stopping.patience=5 \
wandb.id=GUE_prom_${dataset}_CNN \
decoder=id
done
done
for dataset in reconstructed_txt; do
for lr in 1e-4; do
python -m train experiment=hg38/GUE \
model.d_model=256 \
model.layer._name_=CNN \
train.pretrained_model_path=/weight/hyenadna/hyenadna-large-1m-seqlen \
dataset.dataset_name=$dataset \
optimizer.lr=$lr \
dataset=splice \
dataset.max_length=80 \
wandb.mode=offline \
trainer.devices=4 \
dataset.batch_size=28 \
dataset.tokenizer_name=CNN \
dataset.tokenizer_path=/weight/hyenadna/hyenadna-large-1m-seqlen \
train.global_batch_size=128 \
trainer.gradient_clip_val=1.0 \
wandb.id=GUE_splice_${dataset}_CNN \
callbacks.early_stopping.monitor="val/accuracy" \
callbacks.model_checkpoint.monitor="val/accuracy" \
callbacks.model_checkpoint.filename="val/accuracy" \
callbacks.early_stopping.patience=3 \
wandb.id=GUE_splice_${dataset}_CNN \
decoder=id
done
done
for dataset in 0_txt 1_txt 2_txt 3_txt 4_txt; do
for lr in 1e-4; do
python -m train experiment=hg38/GUE \
model.d_model=256 \
model.layer._name_=CNN \
train.pretrained_model_path=/weight/hyenadna/hyenadna-large-1m-seqlen \
dataset.dataset_name=$dataset \
optimizer.lr=$lr \
dataset=tf \
dataset.max_length=30 \
wandb.mode=offline \
trainer.devices=4 \
dataset.batch_size=16 \
dataset.tokenizer_name=CNN \
train.global_batch_size=128 \
dataset.tokenizer_path=/weight/hyenadna/hyenadna-large-1m-seqlen \
callbacks.early_stopping.monitor="val/accuracy" \
callbacks.model_checkpoint.monitor="val/accuracy" \
callbacks.model_checkpoint.filename="val/accuracy" \
callbacks.early_stopping.patience=5 \
wandb.id=GUE_tf_${dataset}_CNN \
decoder=id
done
done
for dataset in covid_txt; do
for lr in 1e-4; do
python -m train experiment=hg38/GUE \
model.d_model=256 \
model.layer._name_=CNN \
train.pretrained_model_path=/weight/hyenadna/hyenadna-large-1m-seqlen \
dataset.dataset_name=$dataset \
optimizer.lr=$lr \
dataset=virus \
dataset.max_length=256 \
wandb.mode=offline \
trainer.devices=4 \
dataset.batch_size=16 \
dataset.tokenizer_name=CNN \
dataset.tokenizer_path=/weight/hyenadna/hyenadna-large-1m-seqlen \
train.global_batch_size=128 \
callbacks.early_stopping.monitor="val/accuracy" \
callbacks.model_checkpoint.monitor="val/accuracy" \
callbacks.model_checkpoint.filename="val/accuracy" \
callbacks.early_stopping.patience=5 \
wandb.id=GUE_virus_${dataset}_CNN \
decoder=id
done
done