We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
There was an error while loading. Please reload this page.
aa810fc
Added 77M (filtered) positions from Tibetan self-play
WDL linearly increases from 0.4 to 0.6 throughout training
Increase HL to 64
Lowered WDL from 0.75 to 0.4
Woo more, better datagen
The first net, surely there will be many more to come
Bump HL size to 32
More data
Increase HL size to 128
Double training superbatches, use cosine decay LR