Skip to main content
homebrewnlp
Projects
gpt
Reports
Loss/Median64 vs Steps (22/12/31 09:17:31)
Log in
Sign up
Share
Comment
Star
Share
Comment
Star
Loss/Median64 vs Steps (22/12/31 09:17:31)
Lucas Nestler
Created on December 31
|
Last edited on December 31
Comment
Loss/Median64 vs Steps
Loss/Median64 vs Steps
3k
4k
5k
6k
7k
8k
9k
10k
20k
30k
40k
50k
60k
70k
80k
90k
Step
0.7
0.8
0.9
1
group: param64
group: embedding-gradient-shrink
group: no-truegrad
group: fp64_loss-higher_eps
group: correct-decay
group: fp32-v12
group: remove-dead-code
group: tied-moe-modulo
Run set
60
Add a comment