| { | |
| "vocab_size": 50304, | |
| "n_layer": 2, | |
| "n_head": 4, | |
| "n_embd": 64, | |
| "squared_mlp": false, | |
| "bilinear": false, | |
| "expansion_factor": 4, | |
| "gated": false, | |
| "squared_attn": false, | |
| "step": 500 | |
| } |
| { | |
| "vocab_size": 50304, | |
| "n_layer": 2, | |
| "n_head": 4, | |
| "n_embd": 64, | |
| "squared_mlp": false, | |
| "bilinear": false, | |
| "expansion_factor": 4, | |
| "gated": false, | |
| "squared_attn": false, | |
| "step": 500 | |
| } |