svwingerden's picture
Duplicate from georgeyw/ih-1k-large-batch
1037ff5 verified
raw
history blame contribute delete
231 Bytes
{
"d_model": 8,
"d_head": 8,
"n_layers": 2,
"n_ctx": 32,
"n_heads": 1,
"d_vocab": 10,
"normalization_type": "LN",
"attn_only": true,
"positional_embedding_type": "shortformer",
"seed": null
}