Skip to main content
marin-community
Projects
marin
Reports
Qwen 3 speedruns (QK Norm, Muon)
Log in
Sign up
Share
Comment
Star
Qwen 3 speedruns (QK Norm, Muon)
Calvin Xu
Created on September 12
|
Last edited on September 12
Comment
Model configs match the previous Llama Muon speedruns:
https://github.com/marin-community/marin/pull/1405
Section 1
train/loss
train/loss
0
5k
10k
15k
20k
Step
4
6
8
10
12
qwen3_300m_muon_4096-ee4f99
qwen3_130m_muon_4096-04770b
qwen3_520m_muon_4096-361875
qwen3_1_2b_muon_4096-d117d2
Run set
4
Run set
4
Run set
4
Add a comment