Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
akoboldfrying
3 months ago
|
parent
|
context
|
favorite
| on:
Gram Newton-Schulz: A Fast, Hardware-Aware Newton-...
Only read the first section but this sounds really impressive -- up to 50% of up to 17% of training time when using the Muon optimiser, so up to around 7% of basically pure improvement with no downside.
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search: