2 ms·
This is probably less likely with this model, as it’s almost certainly a further RL training continuation of 3.5 27b. The bugs with this architecture were worke
by jjcm 6mo ago
This is probably less likely with this model, as it’s almost certainly a further RL training continuation of 3.5 27b. The bugs with this architecture were worked out when that dropped.
- originalvichy 6mo agoValuable note!