| { | |
| "d_sae": 4096, | |
| "actual_l0": 165.8764343, | |
| "effective_l0": 100, | |
| "cross_entropy": 464.4294169433594, | |
| "sae_type": "saebench", | |
| "sae_release": "canrager/saebench_gemma-2-2b_width-2pow12_date-0107", | |
| "num_tokens": 10000000, | |
| "num_epochs": 1, | |
| "lr": 0.0005, | |
| "loaded_dag": null, | |
| "allow_negative_influence": false, | |
| "history": { | |
| "train_loss": [ | |
| 756.2381443568638, | |
| 571.8783499581473, | |
| 527.0960534232004, | |
| 504.57795976911274, | |
| 491.4717749459403, | |
| 482.7741740635463, | |
| 477.00267617361885, | |
| 472.19471849714006, | |
| 468.3589043753488, | |
| 464.8287867954799, | |
| 464.81527361188614 | |
| ], | |
| "train_steps": [ | |
| 140, | |
| 280, | |
| 420, | |
| 560, | |
| 700, | |
| 840, | |
| 980, | |
| 1120, | |
| 1260, | |
| 1400, | |
| 1407 | |
| ], | |
| "val_loss": [ | |
| 613.0383723632813, | |
| 543.270251953125, | |
| 513.7303801757813, | |
| 497.2721671386719, | |
| 486.76615244140623, | |
| 479.4650443359375, | |
| 474.1357983886719, | |
| 470.0854748535156, | |
| 466.90237919921873, | |
| 464.3813434570313, | |
| 464.4294169433594 | |
| ], | |
| "val_cross_entropy": [ | |
| 613.0383723632813, | |
| 543.270251953125, | |
| 513.7303801757813, | |
| 497.2721671386719, | |
| 486.76615244140623, | |
| 479.4650443359375, | |
| 474.1357983886719, | |
| 470.0854748535156, | |
| 466.90237919921873, | |
| 464.3813434570313, | |
| 464.4294169433594 | |
| ], | |
| "effective_l0": [ | |
| 142, | |
| 121, | |
| 113, | |
| 108, | |
| 106, | |
| 104, | |
| 102, | |
| 101, | |
| 100, | |
| 100, | |
| 100 | |
| ], | |
| "val_steps": [ | |
| 140, | |
| 280, | |
| 420, | |
| 560, | |
| 700, | |
| 840, | |
| 980, | |
| 1120, | |
| 1260, | |
| 1400, | |
| 1407 | |
| ] | |
| }, | |
| "sae_id": "gemma-2-2b_jump_relu_width-2pow12_date-0107/resid_post_layer_12/trainer_3" | |
| } |