Supermix v97 (v97_scf_connectome.selected.pt)

Standalone experimental release for v97_scf_connectome.selected.pt from kai9987kai/Supermix β€” the first SCF Cognitive Fabric connectome continuation introducing fast Hebbian trace plasticity (cns_hebbian_trace=True) and normalized latent-entropy early halting (thinking_entropy_halt=True).

Key Additions & Measured Receipts

  • Warm-started from v96_physio_neurogenesis.selected.pt on datasets/v97/v97_evolution_mix.jsonl (9,415-token vocabulary, 6 layers, 832-slot bilateral male-CNS connectome).
  • Adds bounded zero-init fast Hebbian gain modulation (cns_hebbian_trace=True) and per-cycle entropy-stability halting (thinking_entropy_halt=True).
Evaluation Result Notes
Selected Dev Loss 0.059501 Step 12 of 36 (extra.dev_loss_at_write)
SCF Hybrid (python -m scf bench --seed 82) 41 / 41 (1.0000) Deterministic verification + connectome atlas

Per Rule 9 promotion gating, v96_physio_neurogenesis.selected.pt remains #1 Promoted Default; v97_scf_connectome.selected.pt is released as an experimental checkpoint.

Quick Start

pip install torch numpy flask huggingface_hub
huggingface-cli download Kai9987kai/supermix-v97 --local-dir .
python example_usage.py
Downloads last month
-
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support

Model tree for Kai9987kai/supermix-v97

Finetuned
(1)
this model