| { | |
| "vocab_size": 24000, | |
| "n_layers": 12, | |
| "d_model": 768, | |
| "n_heads": 12, | |
| "n_kv_heads": 4, | |
| "ffn_hidden": 2304, | |
| "max_seq_len": 2048, | |
| "rope_theta": 10000.0, | |
| "norm_eps": 1e-05, | |
| "pos_emb": "rope" | |
| } |
| { | |
| "vocab_size": 24000, | |
| "n_layers": 12, | |
| "d_model": 768, | |
| "n_heads": 12, | |
| "n_kv_heads": 4, | |
| "ffn_hidden": 2304, | |
| "max_seq_len": 2048, | |
| "rope_theta": 10000.0, | |
| "norm_eps": 1e-05, | |
| "pos_emb": "rope" | |
| } |