| { | |
| "hidden_dim": 16, | |
| "n_layers": 2, | |
| "n_heads": 2, | |
| "seq_len": 16, | |
| "vocab_size": 16, | |
| "post_embed_norm": false, | |
| "weight_tying": false | |
| } | |
| { | |
| "hidden_dim": 16, | |
| "n_layers": 2, | |
| "n_heads": 2, | |
| "seq_len": 16, | |
| "vocab_size": 16, | |
| "post_embed_norm": false, | |
| "weight_tying": false | |
| } | |