| { | |
| "trainer": { | |
| "trainer_class": "BatchTopKTrainer", | |
| "dict_class": "BatchTopKSAE", | |
| "lr": 0.0001, | |
| "steps": 1221, | |
| "auxk_alpha": 0.03125, | |
| "warmup_steps": 500, | |
| "decay_start": null, | |
| "threshold_beta": 0.999, | |
| "threshold_start_step": 1000, | |
| "top_k_aux": 576, | |
| "seed": null, | |
| "activation_dim": 1152, | |
| "dict_size": 18432, | |
| "k": 100, | |
| "device": "cuda", | |
| "layer": 12, | |
| "lm_name": "stewy33/gemma-3-1b-it-0524_original_augmented_subtle_roman_concrete-a24a37e6-google/gemma-3-1b-it", | |
| "wandb_name": "SAEdiff_ftb-gemma3_1B-roman_concrete-L12-s1-t100-k100-lr1e-04-x16", | |
| "encoder_init_norm": 1.0 | |
| } | |
| } |