{"d_in": 3072, "dtype": "float32", "expansion_factor": 32, "l1_coefficient": 3.9e-06, "train_batch_size": 8192, "dead_feature_window": 1000, "use_ghost_grads": true, "d_sae": 98304} |
{"d_in": 3072, "dtype": "float32", "expansion_factor": 32, "l1_coefficient": 3.9e-06, "train_batch_size": 8192, "dead_feature_window": 1000, "use_ghost_grads": true, "d_sae": 98304} |