{"d_in": 2048, "dtype": "float32", "expansion_factor": 32, "l1_coefficient": 3.555555555555555e-05, "train_batch_size": 2048, "dead_feature_window": 1000, "use_ghost_grads": true, "d_sae": 65536} |
{"d_in": 2048, "dtype": "float32", "expansion_factor": 32, "l1_coefficient": 3.555555555555555e-05, "train_batch_size": 2048, "dead_feature_window": 1000, "use_ghost_grads": true, "d_sae": 65536} |