3LM-MLX / config.json
GeneLab's picture
3LM-MLX: 35.66M / 367.6M tokens pretrained + SFT
0fb999f verified
Raw
History Blame Contribute Delete
198 Bytes
{
"vocab_size": 32000,
"block_size": 512,
"n_layer": 6,
"n_head": 8,
"n_embd": 512,
"dropout": 0.05,
"arch": "3lm",
"mlp_hidden": 1408,
"rope_theta": 10000.0,
"norm_eps": 1e-05
}