Quazim0t0 commited on
Commit
f74dfb4
·
verified ·
1 Parent(s): 8deab7a

Upload model.py with huggingface_hub

Browse files
Files changed (1) hide show
  1. model.py +4 -0
model.py CHANGED
@@ -718,6 +718,10 @@ class QuazimotoLM(nn.Module):
718
  "use_ring_specialists", "ring_n_specialists", "ring_spec_key_dim",
719
  "ring_spec_slot_dim", "ring_spec_top_k", "ring_spec_write_lr",
720
  "use_fractal_phase_seed",
 
 
 
 
721
  # Hydra tip mixer geometry -- so checkpoints rebuild at the exact colony
722
  "wdw_modes", "wdw_steps", "wdw_dt", "wdw_constraint_weight")}
723
  n = sum(p.numel() for p in self.parameters())
 
718
  "use_ring_specialists", "ring_n_specialists", "ring_spec_key_dim",
719
  "ring_spec_slot_dim", "ring_spec_top_k", "ring_spec_write_lr",
720
  "use_fractal_phase_seed",
721
+ # morphable hybrid + TriAttention -- so a hybrid checkpoint is self-describing
722
+ # and rebuilds with the linear twins / gates / compressor in place.
723
+ "use_morphable", "morph_alpha_init", "morph_k_full", "morph_reg_lambda",
724
+ "use_triattention", "triattn_budget", "triattn_window", "triattn_max_offset",
725
  # Hydra tip mixer geometry -- so checkpoints rebuild at the exact colony
726
  "wdw_modes", "wdw_steps", "wdw_dt", "wdw_constraint_weight")}
727
  n = sum(p.numel() for p in self.parameters())