6 ms·
Achieving the improvements from K2.5 -> composer 2.5 with just post-training is actually more impressive. Though I believe their next model is trained from scra
by shoeb00m 3mo ago
Achieving the improvements from K2.5 -> composer 2.5 with just post-training is actually more impressive. Though I believe their next model is trained from scratch.