clef-FP8 / clef_head /joint_head_config.json
aleksblago's picture
Add clef-FP8
122e4f1 verified
Raw History Blame
119 Bytes
{
"hidden_size": 5120,
"width": 1024,
"routing_layers": 2,
"layers": 4,
"heads": 16,
"feedforward": 4096
}