| { | |
| "model_type": "tabum", | |
| "model_config": { | |
| "d_stage1": 128, | |
| "n_inducing": 48, | |
| "stage1_blocks": 2, | |
| "stage1_heads": 4, | |
| "n_fourier_freqs": 6, | |
| "max_classes": 100, | |
| "n_cls_tokens": 4, | |
| "stage2_blocks": 2, | |
| "stage2_heads": 4, | |
| "d_model": 320, | |
| "stage3_blocks": 10, | |
| "stage3_heads": 8, | |
| "n_reg_bins": 100, | |
| "reg_support": 5.0, | |
| "head_dim": 128 | |
| }, | |
| "trained_steps": 20000, | |
| "n_parameters": 14306258 | |
| } | |