{"d_in": 2048, "dtype": "float32", "expansion_factor": 16, "l1_coefficient": 5.333333333333333e-05, "train_batch_size": 2048, "dead_feature_window": 1000, "use_ghost_grads": true, "d_sae": 32768} |
{"d_in": 2048, "dtype": "float32", "expansion_factor": 16, "l1_coefficient": 5.333333333333333e-05, "train_batch_size": 2048, "dead_feature_window": 1000, "use_ghost_grads": true, "d_sae": 32768} |