SPB Git

spb/wp31_uqo

Public
458 B · 24 lines yaml
1# PRISME reference configuration: prisme_1180m (see docs/architecture_spec.md)2model:3  name: prisme_1180m4  vocab_size: 327685  d_model: 20486  n_layers: 247  n_heads: 328  n_kv_heads: 89  ffn_width: 563210  max_seq_len: 409611  rope_theta: 10000.012  norm_eps: 1.0e-513  tie_embeddings: true14  use_workspace: true15  w: 76816  ffn_w: 204817  n_slots: 1618  n_editor_blocks: 219  n_editor_heads: 1220  r_max: 821  eta_max: 0.2522  clip_c: 1.023  readout_zero_init: true24
Files · main
Menu

Theme