training_logs/mlc__seed2.json

{
  "loss": [
    108364.0234375,
    4026.425048828125,
    3392.5234375,
    3163.514892578125,
    3035.2373046875,
    2914.15185546875,
    2829.30126953125,
    2792.417724609375,
    2742.762451171875,
    2776.29541015625,
    2653.72314453125,
    2669.077880859375,
    2629.986572265625,
    2664.68994140625,
    2601.24609375,
    2580.51806640625,
    2639.58642578125,
    2563.909912109375,
    2561.493408203125,
    2594.917724609375,
    2565.618408203125
  ],
  "l0": [
    500.0,
    492.274658203125,
    494.14306640625,
    492.90283203125,
    493.268310546875,
    492.36572265625,
    493.102294921875,
    493.1220703125,
    491.611083984375,
    493.525146484375,
    490.8408203125,
    490.330810546875,
    488.69091796875,
    491.362548828125,
    491.734375,
    490.404052734375,
    489.929443359375,
    489.762451171875,
    491.277099609375,
    489.604736328125,
    489.1552734375
  ],
  "steps_logged": [
    0,
    200,
    400,
    600,
    800,
    1000,
    1200,
    1400,
    1600,
    1800,
    2000,
    2200,
    2400,
    2600,
    2800,
    3000,
    3200,
    3400,
    3600,
    3800,
    4000
  ],
  "final_step": 4000,
  "converged": true,
  "plateau_last": 0.016735184303764403,
  "elapsed_s": 768.920223236084,
  "row": 4,
  "arch_id": "mlc",
  "arch": "mlc",
  "group": 1,
  "src_class": "MultiLayerCrosscoder",
  "src_module": "src.architectures.mlc",
  "T": null,
  "T_max": null,
  "t_sample": null,
  "n_layers": 5,
  "k_win": 500,
  "k_pos": 100,
  "shifts": null,
  "alpha": null,
  "gamma": null,
  "n_scales": null,
  "seed": 2,
  "d_in": 2304,
  "d_sae": 18432,
  "subject_model": "google/gemma-2-2b",
  "anchor_layer": 12,
  "mlc_layers": [
    10,
    11,
    12,
    13,
    14
  ],
  "phase": "phase7_unification",
  "run_id": "mlc__seed2"
}