| { | |
| "normalize_embed": true, | |
| "nth_layer": 13, | |
| "output_dim": 1, | |
| "precision": "32", | |
| "proj_act_fn": "gelu", | |
| "proj_dropout": 0.0, | |
| "proj_ln": true, | |
| "proj_num_layer": 5, | |
| "target_transform": { | |
| "CE": { | |
| "mean": 5.06865, | |
| "std": 1.93029 | |
| }, | |
| "CU": { | |
| "mean": 5.73633, | |
| "std": 1.75669 | |
| }, | |
| "PC": { | |
| "mean": 3.18591, | |
| "std": 1.86637 | |
| }, | |
| "PQ": { | |
| "mean": 6.57505, | |
| "std": 1.51466 | |
| } | |
| }, | |
| "use_weighted_layer_sum": true | |
| } |