| { | |
| "block_size": 512, | |
| "vocab_size": 4096, | |
| "n_layer": 8, | |
| "n_embd": 32, | |
| "n_head": 4, | |
| "n_kv_head": 2, | |
| "n_hidden": 96, | |
| "dropout": 0.0, | |
| "use_rope": true, | |
| "norm_eps": 1e-5, | |
| "is_causal": true | |
| } | |
| { | |
| "block_size": 512, | |
| "vocab_size": 4096, | |
| "n_layer": 8, | |
| "n_embd": 32, | |
| "n_head": 4, | |
| "n_kv_head": 2, | |
| "n_hidden": 96, | |
| "dropout": 0.0, | |
| "use_rope": true, | |
| "norm_eps": 1e-5, | |
| "is_causal": true | |
| } | |