OSE-8M-Research-Preview / architecture.json
MoniesTO's picture
Upload 19 files
6818ad5 verified
Raw History Blame Contribute Delete
888 Bytes
{
"family": "structured_residual_mlp",
"parameter_count": 7524640,
"state_shape": [
"B",
27
],
"candidates_shape": [
"B",
"N",
16
],
"output_shape": [
"B",
"N+1"
],
"NOUL_index": "N",
"genome": {
"schema_id": "dgr-yuzou.ose-dm-genome.v0.3",
"cap_millions": 8,
"family": "residual_mlp",
"width": 512,
"logical_depth": 3,
"physical_layers": 3,
"heads": 1,
"ffn_ratio": 4,
"bottleneck_size": 128,
"sharing_mode": "residual_mlp",
"parameter_count": 7524640,
"genome_id_sha256": "3c0f871dd890f301e074c0b43285abcbffb22fe03e6a994daae18f335ab41b3a"
},
"pooling": "Mean of admitted candidate representations; forbidden candidates must be removed first.",
"tokenizer": null,
"auxiliary_heads": "Weights retained; this inference returns decision logits only."
}