Skip to main content
vit-prisma
Projects
Imagenet
Reports
Imagenet Training report
Log in
Sign up
Share
Comment
Star
Share
Comment
Star
Imagenet Training report
Yash Vadi
Created on November 30
|
Last edited on November 30
Comment
Training Metrics for the two models
Notations:
happy-wildflower-150: 4 layers Attention Only
hearty-paper-146: 4 layers Attention+MLP
train-Average Loss
train-Average Loss
0
20k
40k
60k
80k
Step
50
100
150
200
250
300
Val Loss
Val Loss
20k
40k
60k
80k
Step
1400
1600
1800
2000
2200
2400
2600
Run set
8
Add a comment