| { |
| "auxiliary_active_at_checkpoint": false, |
| "auxiliary_coefficient_at_checkpoint": 0.0, |
| "calibration_minibatch": 0, |
| "checkpoint_words": 30000000, |
| "condition": "shuffled", |
| "factor_scales": { |
| "associative": 0.6297903656959534, |
| "hierarchical": 6.804664611816406, |
| "order": 0.7903814315795898 |
| }, |
| "initial_aux_gradient_ratio": 0.25, |
| "input_output_embeddings_tied": false, |
| "lambda0": 0.16559839248657227, |
| "projectors": { |
| "concatenated_condition_number": 1.0, |
| "cross_block_overlap_frobenius": [ |
| 0.0, |
| 0.0, |
| 0.0 |
| ], |
| "mode": "shuffled", |
| "ranks": [ |
| 128, |
| 128, |
| 128 |
| ], |
| "within_block_orthonormal_error": [ |
| 0.0, |
| 0.0, |
| 0.0 |
| ] |
| }, |
| "schema_version": 1, |
| "variant": "dial_shuffled_s0", |
| "withdrawal_words": 10000000 |
| } |
|
|