| { |
| "model_name": "R08-v7.2", |
| "architectures": [ |
| "R08Model" |
| ], |
| "model_type": "geodesic_mesh_pyramid_v7_2", |
| "architecture_name": "cascade_stages_attn_mesh_v7_2", |
| "wiring": "serial_emb_mesh_x3_head_pezzi_causal_conv_regional_hubs", |
| "d_model": 256, |
| "max_seq_len": 512, |
| "seq_len": 256, |
| "vocab_pezzi": 1251, |
| "vocab_parole": 17908, |
| "vocab_unioni": 20006, |
| "n_attn_layers": 2, |
| "n_heads": 8, |
| "pyramid_center": true, |
| "regional_pyramid": { |
| "num_regions": 12, |
| "icosahedron_poles": 12, |
| "faces_partitioned": 80, |
| "d_stretto": 128, |
| "d_centro": 512, |
| "stadi": 3, |
| "parametri_hubs_totali": 7266816 |
| }, |
| "causal_convolutions_1d": { |
| "enabled": true, |
| "stages": [ |
| "conv_12", |
| "conv_23" |
| ], |
| "kernel_size": 4, |
| "d_model": 256, |
| "zero_init_residual": true |
| }, |
| "mesh_topology": { |
| "geometry": "geodesic_sphere_icosahedron", |
| "frequency": 2, |
| "total_faces": 80, |
| "total_vertices": 42, |
| "neighbors_per_face": 3 |
| }, |
| "routing": { |
| "hash_routing": true, |
| "router_v6": true, |
| "no_walk": true, |
| "compass_spiral": true |
| }, |
| "total_parameters": 51614115, |
| "benchmarks": { |
| "blimp_it": 0.8202, |
| "multiblimp_ita": 0.7346, |
| "truthfulqa_zero_shot": 0.562, |
| "blimp_en": 0.55 |
| }, |
| "torch_dtype": "float32", |
| "transformers_version": "4.45.0", |
| "author": "Thomas Seri (Sentinel)", |
| "license": "R08-Non-Commercial-Ethical-License-v2.0", |
| "jurisdiction": "Court of Milan (Sezione Specializzata in Materia di Impresa), Italy", |
| "prohibited_uses": "Commercial, Military, Defense, Clean-Room Reverse Engineering, Model Distillation" |
| } |
|
|
