Skip to main content
weblab_lecture
Projects
vip_20250224_bkup
Reports
gpt-2 large, iter600, batch_size8（小島さんとリハで検証してた設定です）
Log in
Sign up
Share
Comment
Star
gpt-2 large, iter600, batch_size8（小島さんとリハで検証してた設定です）
JEONG Seong Cheol
Created on January 21
|
Last edited on January 21
Comment
Section 1
val_loss
val_loss
100
200
300
400
500
600
Step
1.7
1.8
1.9
2
2.1
abci3-1625-gpt2-large-lr1e-05-iter600
train_loss
train_loss
100
200
300
400
500
Step
1.5
2
2.5
3
abci3-1625-gpt2-large-lr1e-05-iter600
test_loss
test_loss
0
200
400
600
800
1k
1.2k
Step
0
0.5
1
1.5
2
2.5
3
abci3-1625-gpt2-large-lr1e-05-iter600
learning_rate
learning_rate
0
200m
400m
600m
800m
1
Step
0
0.000005
0.00001
0.000015
0.00002
abci3-1625-gpt2-large-lr1e-05-iter600
Run set
1
Run set
1
Add a comment