Voltline's picture
Release VimeML V2.1 step40000 FP32 and Core ML INT8 (GPL-2.0)
29f25be verified
Raw History Blame Contribute Delete
281 Bytes
{
"vocab_size": 16384,
"context_length": 128,
"d_model": 320,
"n_heads": 5,
"n_layers": 6,
"d_ff": 832,
"dropout": 0.0,
"norm": "rmsnorm",
"norm_eps": 1e-05,
"activation": "swiglu",
"bias": false,
"tie_embeddings": true,
"position_embedding": "learned"
}