| { | |
| "d_sae": 16384, | |
| "actual_l0": 159.6283643, | |
| "effective_l0": 64, | |
| "cross_entropy": 414.545231640625, | |
| "sae_type": "saebench", | |
| "sae_release": "canrager/saebench_gemma-2-2b_width-2pow14_date-0107", | |
| "num_tokens": 10000000, | |
| "num_epochs": 5, | |
| "lr": 0.0005, | |
| "loaded_dag": null, | |
| "allow_negative_influence": false, | |
| "script_runtime_seconds": 5197.718630075455, | |
| "script_runtime_minutes": 86.62864383459092, | |
| "history": { | |
| "train_loss": [ | |
| 732.1907091413226, | |
| 560.2760546003069, | |
| 514.9291493007115, | |
| 491.22300894601005, | |
| 476.2275621686663, | |
| 465.3508274623326, | |
| 457.40275312151226, | |
| 451.09822518484935, | |
| 446.30048413957866, | |
| 441.738911655971, | |
| 441.61235983712334, | |
| 438.730587550572, | |
| 435.8709974016462, | |
| 433.6602375575474, | |
| 431.13352857317244, | |
| 429.5512477329799, | |
| 427.9140635899135, | |
| 426.9687508719308, | |
| 425.8828778948103, | |
| 424.7388517107282, | |
| 423.9662421090262, | |
| 423.96841408865794, | |
| 422.5081399100167, | |
| 421.8477118355887, | |
| 421.43183070591516, | |
| 420.86571916852677, | |
| 420.3802429199219, | |
| 419.57361580984934, | |
| 419.46634172712055, | |
| 418.8761511666434, | |
| 418.72140110560827, | |
| 417.92284327915735, | |
| 417.9393042428153, | |
| 417.46479753766744, | |
| 417.235406930106, | |
| 417.29322139195034, | |
| 416.66914716448105, | |
| 416.730729675293, | |
| 416.1867268153599, | |
| 416.17149549211774, | |
| 415.9424224853516, | |
| 415.75340663364955, | |
| 415.61898738316125, | |
| 415.6563940865653, | |
| 415.4041767665318, | |
| 415.323065839495, | |
| 415.1418197631836, | |
| 415.0046354021345, | |
| 414.66681213378905, | |
| 414.5555570329939, | |
| 414.46331852504187, | |
| 414.56583448137553, | |
| 414.94642486572263, | |
| 414.2539657592773, | |
| 414.27054879324777 | |
| ], | |
| "train_steps": [ | |
| 140, | |
| 280, | |
| 420, | |
| 560, | |
| 700, | |
| 840, | |
| 980, | |
| 1120, | |
| 1260, | |
| 1400, | |
| 1407, | |
| 1547, | |
| 1687, | |
| 1827, | |
| 1967, | |
| 2107, | |
| 2247, | |
| 2387, | |
| 2527, | |
| 2667, | |
| 2807, | |
| 2814, | |
| 2954, | |
| 3094, | |
| 3234, | |
| 3374, | |
| 3514, | |
| 3654, | |
| 3794, | |
| 3934, | |
| 4074, | |
| 4214, | |
| 4221, | |
| 4361, | |
| 4501, | |
| 4641, | |
| 4781, | |
| 4921, | |
| 5061, | |
| 5201, | |
| 5341, | |
| 5481, | |
| 5621, | |
| 5628, | |
| 5768, | |
| 5908, | |
| 6048, | |
| 6188, | |
| 6328, | |
| 6468, | |
| 6608, | |
| 6748, | |
| 6888, | |
| 7028, | |
| 7035 | |
| ], | |
| "val_loss": [ | |
| 599.6144170898438, | |
| 531.20299453125, | |
| 500.5198718261719, | |
| 482.2030783691406, | |
| 469.7318541015625, | |
| 460.6147657714844, | |
| 453.6471915527344, | |
| 448.1693904296875, | |
| 443.7288943359375, | |
| 440.08709370117185, | |
| 440.0859723144531, | |
| 436.948605859375, | |
| 434.4268278808594, | |
| 432.26872094726565, | |
| 430.40091801757814, | |
| 428.79407412109373, | |
| 427.3747754882813, | |
| 426.132505859375, | |
| 425.0184524414062, | |
| 424.043707421875, | |
| 423.15604125976563, | |
| 423.2965830078125, | |
| 422.35915874023436, | |
| 421.66575834960935, | |
| 421.0220482910156, | |
| 420.4452380859375, | |
| 419.91448291015627, | |
| 419.4309670410156, | |
| 418.988144921875, | |
| 418.58148408203124, | |
| 418.18817978515625, | |
| 417.8346670898438, | |
| 417.99942548828125, | |
| 417.5140052734375, | |
| 417.2248073730469, | |
| 416.9462147949219, | |
| 416.6832835449219, | |
| 416.4367142578125, | |
| 416.230258203125, | |
| 416.02604897460935, | |
| 415.8633370117187, | |
| 415.68912651367185, | |
| 415.53023618164065, | |
| 415.7061401367188, | |
| 415.4003775390625, | |
| 415.27109868164064, | |
| 415.1384886230469, | |
| 415.018615234375, | |
| 414.89894208984373, | |
| 414.7949107421875, | |
| 414.68661064453124, | |
| 414.6003978515625, | |
| 414.48824331054686, | |
| 414.40605263671875, | |
| 414.545231640625 | |
| ], | |
| "val_cross_entropy": [ | |
| 599.6144170898438, | |
| 531.20299453125, | |
| 500.5198718261719, | |
| 482.2030783691406, | |
| 469.7318541015625, | |
| 460.6147657714844, | |
| 453.6471915527344, | |
| 448.1693904296875, | |
| 443.7288943359375, | |
| 440.08709370117185, | |
| 440.0859723144531, | |
| 436.948605859375, | |
| 434.4268278808594, | |
| 432.26872094726565, | |
| 430.40091801757814, | |
| 428.79407412109373, | |
| 427.3747754882813, | |
| 426.132505859375, | |
| 425.0184524414062, | |
| 424.043707421875, | |
| 423.15604125976563, | |
| 423.2965830078125, | |
| 422.35915874023436, | |
| 421.66575834960935, | |
| 421.0220482910156, | |
| 420.4452380859375, | |
| 419.91448291015627, | |
| 419.4309670410156, | |
| 418.988144921875, | |
| 418.58148408203124, | |
| 418.18817978515625, | |
| 417.8346670898438, | |
| 417.99942548828125, | |
| 417.5140052734375, | |
| 417.2248073730469, | |
| 416.9462147949219, | |
| 416.6832835449219, | |
| 416.4367142578125, | |
| 416.230258203125, | |
| 416.02604897460935, | |
| 415.8633370117187, | |
| 415.68912651367185, | |
| 415.53023618164065, | |
| 415.7061401367188, | |
| 415.4003775390625, | |
| 415.27109868164064, | |
| 415.1384886230469, | |
| 415.018615234375, | |
| 414.89894208984373, | |
| 414.7949107421875, | |
| 414.68661064453124, | |
| 414.6003978515625, | |
| 414.48824331054686, | |
| 414.40605263671875, | |
| 414.545231640625 | |
| ], | |
| "effective_l0": [ | |
| 99, | |
| 86, | |
| 80, | |
| 77, | |
| 74, | |
| 73, | |
| 71, | |
| 70, | |
| 70, | |
| 69, | |
| 69, | |
| 68, | |
| 68, | |
| 67, | |
| 67, | |
| 67, | |
| 67, | |
| 66, | |
| 66, | |
| 66, | |
| 66, | |
| 66, | |
| 66, | |
| 66, | |
| 65, | |
| 65, | |
| 65, | |
| 65, | |
| 65, | |
| 65, | |
| 65, | |
| 65, | |
| 65, | |
| 65, | |
| 65, | |
| 65, | |
| 65, | |
| 65, | |
| 65, | |
| 65, | |
| 64, | |
| 64, | |
| 64, | |
| 64, | |
| 64, | |
| 64, | |
| 64, | |
| 64, | |
| 64, | |
| 64, | |
| 64, | |
| 64, | |
| 64, | |
| 64, | |
| 64 | |
| ], | |
| "val_steps": [ | |
| 140, | |
| 280, | |
| 420, | |
| 560, | |
| 700, | |
| 840, | |
| 980, | |
| 1120, | |
| 1260, | |
| 1400, | |
| 1407, | |
| 1547, | |
| 1687, | |
| 1827, | |
| 1967, | |
| 2107, | |
| 2247, | |
| 2387, | |
| 2527, | |
| 2667, | |
| 2807, | |
| 2814, | |
| 2954, | |
| 3094, | |
| 3234, | |
| 3374, | |
| 3514, | |
| 3654, | |
| 3794, | |
| 3934, | |
| 4074, | |
| 4214, | |
| 4221, | |
| 4361, | |
| 4501, | |
| 4641, | |
| 4781, | |
| 4921, | |
| 5061, | |
| 5201, | |
| 5341, | |
| 5481, | |
| 5621, | |
| 5628, | |
| 5768, | |
| 5908, | |
| 6048, | |
| 6188, | |
| 6328, | |
| 6468, | |
| 6608, | |
| 6748, | |
| 6888, | |
| 7028, | |
| 7035 | |
| ] | |
| }, | |
| "sae_id": "gemma-2-2b_matryoshka_batch_top_k_width-2pow14_date-0107/resid_post_layer_12/trainer_3" | |
| } |