| { | |
| "col_dim": 64, | |
| "col_blocks": 2, | |
| "col_heads": 4, | |
| "n_inducing": 64, | |
| "row_blocks": 3, | |
| "row_heads": 4, | |
| "n_cls": 4, | |
| "icl_blocks": 8, | |
| "icl_heads": 8, | |
| "icl_kv_heads_test": 1, | |
| "n_thinking": 16, | |
| "ff_factor": 2, | |
| "n_freq": 16, | |
| "n_ecdf_freq": 4, | |
| "group_offsets": [ | |
| 0, | |
| 1, | |
| 3 | |
| ], | |
| "label_slots": 16, | |
| "decoder_heads": 4, | |
| "rope_base": 10000.0, | |
| "cell_embed": "fourier", | |
| "ccmm": false, | |
| "ccmm_mask_fraction": 0.15, | |
| "row_mode": "summary", | |
| "row_refine": true, | |
| "row_refine_rounds": 3, | |
| "icl_drop_blocks": 1, | |
| "icl_ff_reallocation": 0 | |
| } | |