File size: 1,382 Bytes
2e969a8 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 | {
"model_type": "sentinel_r08_geodesic_v7_2a",
"architectures": [
"CascadeMeshModelV7_2"
],
"model_name": "Sentinel R-08 v7.2a DA ZERO",
"author": "Thomas Seri (Sentinel)",
"version": "7.2a",
"status": "Official Inference Release (Weights + Inference Runtime; Training Pipeline Proprietary)",
"d_model": 256,
"d_hidden": 256,
"frequency": 2,
"num_faces": 80,
"stages": 3,
"max_seq_len": 2048,
"max_new_tokens": 2048,
"max_output_tokens": 2048,
"n_attn_layers": 2,
"n_heads": 8,
"regional_hubs_per_stage": 12,
"pyramid_narrow": 128,
"pyramid_hidden": 512,
"causal_conv_kernel": 4,
"total_parameters": 51614115,
"pretrain_steps": 185000,
"pretrain_tokens": 740000000,
"pretrain_loss_final": 1.0469,
"midtrain_steps": 10000,
"midtrain_loss_final": 1.0428,
"benchmarks": {
"blimp_it_italian_accuracy": "82.51%",
"multiblimp_italian_accuracy": "76.71%",
"truthful_qa_facts_accuracy": "100.0%",
"truthful_qa_overall": "56.2%",
"kv_cache_memory_512_ctx": "0.8 MB",
"energy_consumption_per_tok": "2.40 mJ",
"repetition_loops": "Eliminated (0%)"
},
"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"
}
|