| { |
| "best_global_step": null, |
| "best_metric": null, |
| "best_model_checkpoint": null, |
| "epoch": 6.0, |
| "eval_steps": 500, |
| "global_step": 3750, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 0.008, |
| "grad_norm": 0.8981367352393554, |
| "learning_rate": 4.266666666666667e-07, |
| "loss": 0.4888, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.47944098711013794, |
| "step": 5, |
| "valid_targets_mean": 4582.5, |
| "valid_targets_min": 993 |
| }, |
| { |
| "epoch": 0.016, |
| "grad_norm": 0.8928800645664441, |
| "learning_rate": 9.600000000000001e-07, |
| "loss": 0.4944, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.482805073261261, |
| "step": 10, |
| "valid_targets_mean": 4469.9, |
| "valid_targets_min": 2036 |
| }, |
| { |
| "epoch": 0.024, |
| "grad_norm": 0.8429262427173801, |
| "learning_rate": 1.4933333333333336e-06, |
| "loss": 0.4868, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4876168966293335, |
| "step": 15, |
| "valid_targets_mean": 4845.4, |
| "valid_targets_min": 1993 |
| }, |
| { |
| "epoch": 0.032, |
| "grad_norm": 0.6219287564170936, |
| "learning_rate": 2.0266666666666666e-06, |
| "loss": 0.4796, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4330214858055115, |
| "step": 20, |
| "valid_targets_mean": 6251.3, |
| "valid_targets_min": 4098 |
| }, |
| { |
| "epoch": 0.04, |
| "grad_norm": 0.646321971592786, |
| "learning_rate": 2.56e-06, |
| "loss": 0.4672, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4708998501300812, |
| "step": 25, |
| "valid_targets_mean": 5169.9, |
| "valid_targets_min": 2296 |
| }, |
| { |
| "epoch": 0.048, |
| "grad_norm": 0.6126810504508798, |
| "learning_rate": 3.093333333333334e-06, |
| "loss": 0.4883, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46361398696899414, |
| "step": 30, |
| "valid_targets_mean": 5248.1, |
| "valid_targets_min": 2025 |
| }, |
| { |
| "epoch": 0.056, |
| "grad_norm": 0.6433339746949742, |
| "learning_rate": 3.6266666666666674e-06, |
| "loss": 0.472, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4749748110771179, |
| "step": 35, |
| "valid_targets_mean": 4852.9, |
| "valid_targets_min": 3273 |
| }, |
| { |
| "epoch": 0.064, |
| "grad_norm": 0.5586685257232964, |
| "learning_rate": 4.16e-06, |
| "loss": 0.4768, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46658235788345337, |
| "step": 40, |
| "valid_targets_mean": 5065.5, |
| "valid_targets_min": 3320 |
| }, |
| { |
| "epoch": 0.072, |
| "grad_norm": 0.5467911493384551, |
| "learning_rate": 4.693333333333334e-06, |
| "loss": 0.4751, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4425419867038727, |
| "step": 45, |
| "valid_targets_mean": 5127.7, |
| "valid_targets_min": 3395 |
| }, |
| { |
| "epoch": 0.08, |
| "grad_norm": 0.48340368890311297, |
| "learning_rate": 5.226666666666667e-06, |
| "loss": 0.4717, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46598565578460693, |
| "step": 50, |
| "valid_targets_mean": 5332.1, |
| "valid_targets_min": 3277 |
| }, |
| { |
| "epoch": 0.088, |
| "grad_norm": 0.5183342546094541, |
| "learning_rate": 5.76e-06, |
| "loss": 0.4745, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4834442734718323, |
| "step": 55, |
| "valid_targets_mean": 5207.1, |
| "valid_targets_min": 2983 |
| }, |
| { |
| "epoch": 0.096, |
| "grad_norm": 0.5336611077189376, |
| "learning_rate": 6.293333333333334e-06, |
| "loss": 0.4586, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.47208839654922485, |
| "step": 60, |
| "valid_targets_mean": 4723.8, |
| "valid_targets_min": 1847 |
| }, |
| { |
| "epoch": 0.104, |
| "grad_norm": 0.5503452018179987, |
| "learning_rate": 6.826666666666667e-06, |
| "loss": 0.4601, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46985557675361633, |
| "step": 65, |
| "valid_targets_mean": 4936.9, |
| "valid_targets_min": 3173 |
| }, |
| { |
| "epoch": 0.112, |
| "grad_norm": 0.5503447086350417, |
| "learning_rate": 7.360000000000001e-06, |
| "loss": 0.4577, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4540197253227234, |
| "step": 70, |
| "valid_targets_mean": 4607.8, |
| "valid_targets_min": 2625 |
| }, |
| { |
| "epoch": 0.12, |
| "grad_norm": 0.5505781724923542, |
| "learning_rate": 7.893333333333335e-06, |
| "loss": 0.4686, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.470470130443573, |
| "step": 75, |
| "valid_targets_mean": 4661.5, |
| "valid_targets_min": 2744 |
| }, |
| { |
| "epoch": 0.128, |
| "grad_norm": 0.4836336821293739, |
| "learning_rate": 8.426666666666667e-06, |
| "loss": 0.4591, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4442409873008728, |
| "step": 80, |
| "valid_targets_mean": 5319.5, |
| "valid_targets_min": 3392 |
| }, |
| { |
| "epoch": 0.136, |
| "grad_norm": 0.46392512933107694, |
| "learning_rate": 8.96e-06, |
| "loss": 0.4458, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44882267713546753, |
| "step": 85, |
| "valid_targets_mean": 5385.0, |
| "valid_targets_min": 2173 |
| }, |
| { |
| "epoch": 0.144, |
| "grad_norm": 0.49894048280220865, |
| "learning_rate": 9.493333333333334e-06, |
| "loss": 0.4462, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4637455344200134, |
| "step": 90, |
| "valid_targets_mean": 5103.9, |
| "valid_targets_min": 2723 |
| }, |
| { |
| "epoch": 0.152, |
| "grad_norm": 0.5759424094197596, |
| "learning_rate": 1.0026666666666667e-05, |
| "loss": 0.4522, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4512324333190918, |
| "step": 95, |
| "valid_targets_mean": 4100.9, |
| "valid_targets_min": 1487 |
| }, |
| { |
| "epoch": 0.16, |
| "grad_norm": 0.5141467900138702, |
| "learning_rate": 1.056e-05, |
| "loss": 0.4483, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4794415235519409, |
| "step": 100, |
| "valid_targets_mean": 5131.4, |
| "valid_targets_min": 1477 |
| }, |
| { |
| "epoch": 0.168, |
| "grad_norm": 0.4832362926255411, |
| "learning_rate": 1.1093333333333334e-05, |
| "loss": 0.4372, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4561840891838074, |
| "step": 105, |
| "valid_targets_mean": 5314.9, |
| "valid_targets_min": 3103 |
| }, |
| { |
| "epoch": 0.176, |
| "grad_norm": 0.4387442376620969, |
| "learning_rate": 1.1626666666666668e-05, |
| "loss": 0.4368, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40504753589630127, |
| "step": 110, |
| "valid_targets_mean": 5826.3, |
| "valid_targets_min": 3311 |
| }, |
| { |
| "epoch": 0.184, |
| "grad_norm": 0.4988011448745755, |
| "learning_rate": 1.216e-05, |
| "loss": 0.4356, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4684132933616638, |
| "step": 115, |
| "valid_targets_mean": 5494.0, |
| "valid_targets_min": 1580 |
| }, |
| { |
| "epoch": 0.192, |
| "grad_norm": 0.5199813050326368, |
| "learning_rate": 1.2693333333333336e-05, |
| "loss": 0.4552, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46420836448669434, |
| "step": 120, |
| "valid_targets_mean": 4813.0, |
| "valid_targets_min": 2362 |
| }, |
| { |
| "epoch": 0.2, |
| "grad_norm": 0.4954878865603237, |
| "learning_rate": 1.3226666666666668e-05, |
| "loss": 0.4289, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40875816345214844, |
| "step": 125, |
| "valid_targets_mean": 5424.5, |
| "valid_targets_min": 2813 |
| }, |
| { |
| "epoch": 0.208, |
| "grad_norm": 0.4452330550923181, |
| "learning_rate": 1.376e-05, |
| "loss": 0.4344, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41843193769454956, |
| "step": 130, |
| "valid_targets_mean": 5507.5, |
| "valid_targets_min": 2958 |
| }, |
| { |
| "epoch": 0.216, |
| "grad_norm": 0.4748202511158756, |
| "learning_rate": 1.4293333333333334e-05, |
| "loss": 0.4406, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4301196336746216, |
| "step": 135, |
| "valid_targets_mean": 5481.1, |
| "valid_targets_min": 3238 |
| }, |
| { |
| "epoch": 0.224, |
| "grad_norm": 0.44786763570245125, |
| "learning_rate": 1.4826666666666666e-05, |
| "loss": 0.4526, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4539117217063904, |
| "step": 140, |
| "valid_targets_mean": 5623.2, |
| "valid_targets_min": 2022 |
| }, |
| { |
| "epoch": 0.232, |
| "grad_norm": 0.4617246062077554, |
| "learning_rate": 1.5360000000000002e-05, |
| "loss": 0.4319, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4338736832141876, |
| "step": 145, |
| "valid_targets_mean": 5401.1, |
| "valid_targets_min": 2397 |
| }, |
| { |
| "epoch": 0.24, |
| "grad_norm": 0.4760513710425182, |
| "learning_rate": 1.5893333333333333e-05, |
| "loss": 0.4396, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42178189754486084, |
| "step": 150, |
| "valid_targets_mean": 5432.9, |
| "valid_targets_min": 2134 |
| }, |
| { |
| "epoch": 0.248, |
| "grad_norm": 0.4436846015572444, |
| "learning_rate": 1.642666666666667e-05, |
| "loss": 0.439, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4323717951774597, |
| "step": 155, |
| "valid_targets_mean": 5910.3, |
| "valid_targets_min": 4008 |
| }, |
| { |
| "epoch": 0.256, |
| "grad_norm": 0.4825167197308148, |
| "learning_rate": 1.696e-05, |
| "loss": 0.432, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4239029288291931, |
| "step": 160, |
| "valid_targets_mean": 5440.1, |
| "valid_targets_min": 2784 |
| }, |
| { |
| "epoch": 0.264, |
| "grad_norm": 0.5019443261993475, |
| "learning_rate": 1.7493333333333334e-05, |
| "loss": 0.4323, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42029061913490295, |
| "step": 165, |
| "valid_targets_mean": 5427.6, |
| "valid_targets_min": 2802 |
| }, |
| { |
| "epoch": 0.272, |
| "grad_norm": 0.5133271968210673, |
| "learning_rate": 1.8026666666666668e-05, |
| "loss": 0.4286, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41252821683883667, |
| "step": 170, |
| "valid_targets_mean": 4884.2, |
| "valid_targets_min": 2893 |
| }, |
| { |
| "epoch": 0.28, |
| "grad_norm": 0.5326517913856371, |
| "learning_rate": 1.8560000000000002e-05, |
| "loss": 0.4315, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44338303804397583, |
| "step": 175, |
| "valid_targets_mean": 5285.8, |
| "valid_targets_min": 2915 |
| }, |
| { |
| "epoch": 0.288, |
| "grad_norm": 0.6155212273227975, |
| "learning_rate": 1.9093333333333336e-05, |
| "loss": 0.4313, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4674159288406372, |
| "step": 180, |
| "valid_targets_mean": 4593.1, |
| "valid_targets_min": 1775 |
| }, |
| { |
| "epoch": 0.296, |
| "grad_norm": 0.48135487173733127, |
| "learning_rate": 1.9626666666666666e-05, |
| "loss": 0.4295, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4108573794364929, |
| "step": 185, |
| "valid_targets_mean": 5198.2, |
| "valid_targets_min": 1828 |
| }, |
| { |
| "epoch": 0.304, |
| "grad_norm": 0.466594444336494, |
| "learning_rate": 2.016e-05, |
| "loss": 0.4301, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4188600182533264, |
| "step": 190, |
| "valid_targets_mean": 5310.8, |
| "valid_targets_min": 3286 |
| }, |
| { |
| "epoch": 0.312, |
| "grad_norm": 0.5123632031907194, |
| "learning_rate": 2.0693333333333334e-05, |
| "loss": 0.4349, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41983145475387573, |
| "step": 195, |
| "valid_targets_mean": 4830.8, |
| "valid_targets_min": 1700 |
| }, |
| { |
| "epoch": 0.32, |
| "grad_norm": 0.5363639175944347, |
| "learning_rate": 2.1226666666666668e-05, |
| "loss": 0.4176, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4081733822822571, |
| "step": 200, |
| "valid_targets_mean": 4285.8, |
| "valid_targets_min": 1231 |
| }, |
| { |
| "epoch": 0.328, |
| "grad_norm": 0.4650182407586379, |
| "learning_rate": 2.1760000000000002e-05, |
| "loss": 0.4227, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40866294503211975, |
| "step": 205, |
| "valid_targets_mean": 5516.6, |
| "valid_targets_min": 1704 |
| }, |
| { |
| "epoch": 0.336, |
| "grad_norm": 0.4846136798368386, |
| "learning_rate": 2.2293333333333336e-05, |
| "loss": 0.4226, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4383664131164551, |
| "step": 210, |
| "valid_targets_mean": 5284.0, |
| "valid_targets_min": 2303 |
| }, |
| { |
| "epoch": 0.344, |
| "grad_norm": 0.4901923802225922, |
| "learning_rate": 2.282666666666667e-05, |
| "loss": 0.4238, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4387369751930237, |
| "step": 215, |
| "valid_targets_mean": 5364.0, |
| "valid_targets_min": 2255 |
| }, |
| { |
| "epoch": 0.352, |
| "grad_norm": 0.4525182770711486, |
| "learning_rate": 2.336e-05, |
| "loss": 0.4139, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42604705691337585, |
| "step": 220, |
| "valid_targets_mean": 5948.7, |
| "valid_targets_min": 4283 |
| }, |
| { |
| "epoch": 0.36, |
| "grad_norm": 0.5447522003513544, |
| "learning_rate": 2.3893333333333337e-05, |
| "loss": 0.4229, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42427945137023926, |
| "step": 225, |
| "valid_targets_mean": 4549.4, |
| "valid_targets_min": 1931 |
| }, |
| { |
| "epoch": 0.368, |
| "grad_norm": 0.4766641176989221, |
| "learning_rate": 2.442666666666667e-05, |
| "loss": 0.4267, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4418550133705139, |
| "step": 230, |
| "valid_targets_mean": 5490.9, |
| "valid_targets_min": 2740 |
| }, |
| { |
| "epoch": 0.376, |
| "grad_norm": 0.48675005290579404, |
| "learning_rate": 2.496e-05, |
| "loss": 0.4314, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4508809447288513, |
| "step": 235, |
| "valid_targets_mean": 5645.6, |
| "valid_targets_min": 3469 |
| }, |
| { |
| "epoch": 0.384, |
| "grad_norm": 0.4648012694534012, |
| "learning_rate": 2.5493333333333335e-05, |
| "loss": 0.435, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43246546387672424, |
| "step": 240, |
| "valid_targets_mean": 5152.1, |
| "valid_targets_min": 3420 |
| }, |
| { |
| "epoch": 0.392, |
| "grad_norm": 0.6822657669786886, |
| "learning_rate": 2.6026666666666666e-05, |
| "loss": 0.4265, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43735986948013306, |
| "step": 245, |
| "valid_targets_mean": 5612.2, |
| "valid_targets_min": 3068 |
| }, |
| { |
| "epoch": 0.4, |
| "grad_norm": 0.4915802928551818, |
| "learning_rate": 2.6560000000000003e-05, |
| "loss": 0.424, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4244071841239929, |
| "step": 250, |
| "valid_targets_mean": 5370.2, |
| "valid_targets_min": 1593 |
| }, |
| { |
| "epoch": 0.408, |
| "grad_norm": 0.4717635624202091, |
| "learning_rate": 2.7093333333333337e-05, |
| "loss": 0.419, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39868617057800293, |
| "step": 255, |
| "valid_targets_mean": 5211.9, |
| "valid_targets_min": 2201 |
| }, |
| { |
| "epoch": 0.416, |
| "grad_norm": 0.4834575277536717, |
| "learning_rate": 2.7626666666666668e-05, |
| "loss": 0.4236, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41014114022254944, |
| "step": 260, |
| "valid_targets_mean": 4866.6, |
| "valid_targets_min": 1999 |
| }, |
| { |
| "epoch": 0.424, |
| "grad_norm": 0.4718501005994946, |
| "learning_rate": 2.816e-05, |
| "loss": 0.4257, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40851637721061707, |
| "step": 265, |
| "valid_targets_mean": 5743.9, |
| "valid_targets_min": 3042 |
| }, |
| { |
| "epoch": 0.432, |
| "grad_norm": 0.5195934705554233, |
| "learning_rate": 2.869333333333334e-05, |
| "loss": 0.4317, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4213217496871948, |
| "step": 270, |
| "valid_targets_mean": 5105.2, |
| "valid_targets_min": 2092 |
| }, |
| { |
| "epoch": 0.44, |
| "grad_norm": 0.517032230430082, |
| "learning_rate": 2.922666666666667e-05, |
| "loss": 0.4251, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4187217950820923, |
| "step": 275, |
| "valid_targets_mean": 4924.5, |
| "valid_targets_min": 2085 |
| }, |
| { |
| "epoch": 0.448, |
| "grad_norm": 0.4444674110850732, |
| "learning_rate": 2.9760000000000003e-05, |
| "loss": 0.4073, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4063284993171692, |
| "step": 280, |
| "valid_targets_mean": 5549.7, |
| "valid_targets_min": 3180 |
| }, |
| { |
| "epoch": 0.456, |
| "grad_norm": 0.5092069033343127, |
| "learning_rate": 3.0293333333333334e-05, |
| "loss": 0.4242, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43576300144195557, |
| "step": 285, |
| "valid_targets_mean": 5190.6, |
| "valid_targets_min": 2082 |
| }, |
| { |
| "epoch": 0.464, |
| "grad_norm": 0.5014606635124516, |
| "learning_rate": 3.0826666666666674e-05, |
| "loss": 0.433, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44080421328544617, |
| "step": 290, |
| "valid_targets_mean": 4883.8, |
| "valid_targets_min": 2618 |
| }, |
| { |
| "epoch": 0.472, |
| "grad_norm": 0.5000718880535248, |
| "learning_rate": 3.1360000000000005e-05, |
| "loss": 0.418, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4310905933380127, |
| "step": 295, |
| "valid_targets_mean": 5341.3, |
| "valid_targets_min": 2633 |
| }, |
| { |
| "epoch": 0.48, |
| "grad_norm": 0.49357888550178053, |
| "learning_rate": 3.1893333333333335e-05, |
| "loss": 0.4258, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4273624122142792, |
| "step": 300, |
| "valid_targets_mean": 5813.1, |
| "valid_targets_min": 2239 |
| }, |
| { |
| "epoch": 0.488, |
| "grad_norm": 0.5145638706813439, |
| "learning_rate": 3.2426666666666666e-05, |
| "loss": 0.4245, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42333951592445374, |
| "step": 305, |
| "valid_targets_mean": 4995.4, |
| "valid_targets_min": 2135 |
| }, |
| { |
| "epoch": 0.496, |
| "grad_norm": 0.4942735878387689, |
| "learning_rate": 3.296e-05, |
| "loss": 0.4072, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4325196444988251, |
| "step": 310, |
| "valid_targets_mean": 4867.6, |
| "valid_targets_min": 1483 |
| }, |
| { |
| "epoch": 0.504, |
| "grad_norm": 0.436855962833709, |
| "learning_rate": 3.349333333333334e-05, |
| "loss": 0.417, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4261610507965088, |
| "step": 315, |
| "valid_targets_mean": 6261.0, |
| "valid_targets_min": 1394 |
| }, |
| { |
| "epoch": 0.512, |
| "grad_norm": 0.4594886268841487, |
| "learning_rate": 3.402666666666667e-05, |
| "loss": 0.4076, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4086250364780426, |
| "step": 320, |
| "valid_targets_mean": 5693.5, |
| "valid_targets_min": 1794 |
| }, |
| { |
| "epoch": 0.52, |
| "grad_norm": 0.4879088558081727, |
| "learning_rate": 3.456e-05, |
| "loss": 0.4025, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4034465253353119, |
| "step": 325, |
| "valid_targets_mean": 4798.7, |
| "valid_targets_min": 1519 |
| }, |
| { |
| "epoch": 0.528, |
| "grad_norm": 0.5182535180179841, |
| "learning_rate": 3.509333333333333e-05, |
| "loss": 0.4092, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42085325717926025, |
| "step": 330, |
| "valid_targets_mean": 4988.7, |
| "valid_targets_min": 3132 |
| }, |
| { |
| "epoch": 0.536, |
| "grad_norm": 0.4722584950170572, |
| "learning_rate": 3.562666666666667e-05, |
| "loss": 0.3953, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3744209110736847, |
| "step": 335, |
| "valid_targets_mean": 6070.1, |
| "valid_targets_min": 3547 |
| }, |
| { |
| "epoch": 0.544, |
| "grad_norm": 0.4871456702546813, |
| "learning_rate": 3.6160000000000006e-05, |
| "loss": 0.4177, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41021811962127686, |
| "step": 340, |
| "valid_targets_mean": 4361.2, |
| "valid_targets_min": 2180 |
| }, |
| { |
| "epoch": 0.552, |
| "grad_norm": 0.4985836713557574, |
| "learning_rate": 3.669333333333334e-05, |
| "loss": 0.4072, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.415152907371521, |
| "step": 345, |
| "valid_targets_mean": 5523.1, |
| "valid_targets_min": 2892 |
| }, |
| { |
| "epoch": 0.56, |
| "grad_norm": 0.44107792470345925, |
| "learning_rate": 3.722666666666667e-05, |
| "loss": 0.4197, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.406636118888855, |
| "step": 350, |
| "valid_targets_mean": 5612.4, |
| "valid_targets_min": 3296 |
| }, |
| { |
| "epoch": 0.568, |
| "grad_norm": 0.49588015008123604, |
| "learning_rate": 3.7760000000000004e-05, |
| "loss": 0.4167, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4159216582775116, |
| "step": 355, |
| "valid_targets_mean": 5005.8, |
| "valid_targets_min": 1793 |
| }, |
| { |
| "epoch": 0.576, |
| "grad_norm": 0.5198631524949012, |
| "learning_rate": 3.8293333333333335e-05, |
| "loss": 0.4052, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4168175458908081, |
| "step": 360, |
| "valid_targets_mean": 4999.4, |
| "valid_targets_min": 1858 |
| }, |
| { |
| "epoch": 0.584, |
| "grad_norm": 0.45756417733746413, |
| "learning_rate": 3.882666666666667e-05, |
| "loss": 0.3922, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3764269948005676, |
| "step": 365, |
| "valid_targets_mean": 5129.8, |
| "valid_targets_min": 1737 |
| }, |
| { |
| "epoch": 0.592, |
| "grad_norm": 0.47545556762209873, |
| "learning_rate": 3.936e-05, |
| "loss": 0.4188, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4313249886035919, |
| "step": 370, |
| "valid_targets_mean": 5349.0, |
| "valid_targets_min": 1851 |
| }, |
| { |
| "epoch": 0.6, |
| "grad_norm": 0.472636669300859, |
| "learning_rate": 3.989333333333333e-05, |
| "loss": 0.4129, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38715893030166626, |
| "step": 375, |
| "valid_targets_mean": 5179.4, |
| "valid_targets_min": 2021 |
| }, |
| { |
| "epoch": 0.608, |
| "grad_norm": 0.5116851198377632, |
| "learning_rate": 3.9999861365387784e-05, |
| "loss": 0.4168, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4201098084449768, |
| "step": 380, |
| "valid_targets_mean": 5033.2, |
| "valid_targets_min": 3420 |
| }, |
| { |
| "epoch": 0.616, |
| "grad_norm": 0.5199362462098336, |
| "learning_rate": 3.9999298165569614e-05, |
| "loss": 0.4151, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.413302481174469, |
| "step": 385, |
| "valid_targets_mean": 4725.6, |
| "valid_targets_min": 2527 |
| }, |
| { |
| "epoch": 0.624, |
| "grad_norm": 0.4979804160336678, |
| "learning_rate": 3.999830174807269e-05, |
| "loss": 0.4007, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39216095209121704, |
| "step": 390, |
| "valid_targets_mean": 5273.8, |
| "valid_targets_min": 2823 |
| }, |
| { |
| "epoch": 0.632, |
| "grad_norm": 0.4403695246628613, |
| "learning_rate": 3.9996872134481036e-05, |
| "loss": 0.3946, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3717777132987976, |
| "step": 395, |
| "valid_targets_mean": 5351.7, |
| "valid_targets_min": 3408 |
| }, |
| { |
| "epoch": 0.64, |
| "grad_norm": 0.48911707796868703, |
| "learning_rate": 3.999500935576245e-05, |
| "loss": 0.4159, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.429704487323761, |
| "step": 400, |
| "valid_targets_mean": 5340.6, |
| "valid_targets_min": 3747 |
| }, |
| { |
| "epoch": 0.648, |
| "grad_norm": 0.4681019703407759, |
| "learning_rate": 3.999271345226776e-05, |
| "loss": 0.4098, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41974306106567383, |
| "step": 405, |
| "valid_targets_mean": 5763.4, |
| "valid_targets_min": 4144 |
| }, |
| { |
| "epoch": 0.656, |
| "grad_norm": 0.5102854487026843, |
| "learning_rate": 3.9989984473730035e-05, |
| "loss": 0.4146, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4262502193450928, |
| "step": 410, |
| "valid_targets_mean": 5314.6, |
| "valid_targets_min": 2325 |
| }, |
| { |
| "epoch": 0.664, |
| "grad_norm": 0.496751192405806, |
| "learning_rate": 3.998682247926343e-05, |
| "loss": 0.4136, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.45322519540786743, |
| "step": 415, |
| "valid_targets_mean": 5338.8, |
| "valid_targets_min": 3976 |
| }, |
| { |
| "epoch": 0.672, |
| "grad_norm": 0.43167646005396765, |
| "learning_rate": 3.998322753736193e-05, |
| "loss": 0.3954, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36269742250442505, |
| "step": 420, |
| "valid_targets_mean": 5213.8, |
| "valid_targets_min": 2677 |
| }, |
| { |
| "epoch": 0.68, |
| "grad_norm": 0.44609008933518335, |
| "learning_rate": 3.99791997258979e-05, |
| "loss": 0.4088, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3873952031135559, |
| "step": 425, |
| "valid_targets_mean": 5529.1, |
| "valid_targets_min": 2602 |
| }, |
| { |
| "epoch": 0.688, |
| "grad_norm": 0.4542608446447402, |
| "learning_rate": 3.997473913212036e-05, |
| "loss": 0.4004, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3982234001159668, |
| "step": 430, |
| "valid_targets_mean": 5639.7, |
| "valid_targets_min": 2569 |
| }, |
| { |
| "epoch": 0.696, |
| "grad_norm": 0.5032832670291768, |
| "learning_rate": 3.9969845852653087e-05, |
| "loss": 0.3998, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38907137513160706, |
| "step": 435, |
| "valid_targets_mean": 4954.6, |
| "valid_targets_min": 2602 |
| }, |
| { |
| "epoch": 0.704, |
| "grad_norm": 0.4696848718898221, |
| "learning_rate": 3.996451999349258e-05, |
| "loss": 0.3972, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40265825390815735, |
| "step": 440, |
| "valid_targets_mean": 4970.3, |
| "valid_targets_min": 2452 |
| }, |
| { |
| "epoch": 0.712, |
| "grad_norm": 0.46045090559352714, |
| "learning_rate": 3.995876167000569e-05, |
| "loss": 0.3935, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4230922758579254, |
| "step": 445, |
| "valid_targets_mean": 4800.4, |
| "valid_targets_min": 3248 |
| }, |
| { |
| "epoch": 0.72, |
| "grad_norm": 0.4891482904717523, |
| "learning_rate": 3.9952571006927186e-05, |
| "loss": 0.3987, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4035327434539795, |
| "step": 450, |
| "valid_targets_mean": 4973.6, |
| "valid_targets_min": 3934 |
| }, |
| { |
| "epoch": 0.728, |
| "grad_norm": 0.49539778824595054, |
| "learning_rate": 3.9945948138356995e-05, |
| "loss": 0.4193, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42373305559158325, |
| "step": 455, |
| "valid_targets_mean": 4801.6, |
| "valid_targets_min": 2645 |
| }, |
| { |
| "epoch": 0.736, |
| "grad_norm": 0.4252026603501911, |
| "learning_rate": 3.993889320775735e-05, |
| "loss": 0.3925, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37491941452026367, |
| "step": 460, |
| "valid_targets_mean": 5669.7, |
| "valid_targets_min": 2359 |
| }, |
| { |
| "epoch": 0.744, |
| "grad_norm": 0.44075639148681955, |
| "learning_rate": 3.9931406367949627e-05, |
| "loss": 0.4115, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3941989243030548, |
| "step": 465, |
| "valid_targets_mean": 4925.9, |
| "valid_targets_min": 2254 |
| }, |
| { |
| "epoch": 0.752, |
| "grad_norm": 0.4544184671048398, |
| "learning_rate": 3.9923487781111106e-05, |
| "loss": 0.4033, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3993281126022339, |
| "step": 470, |
| "valid_targets_mean": 5728.8, |
| "valid_targets_min": 1549 |
| }, |
| { |
| "epoch": 0.76, |
| "grad_norm": 0.4301508746553947, |
| "learning_rate": 3.9915137618771386e-05, |
| "loss": 0.3996, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40640246868133545, |
| "step": 475, |
| "valid_targets_mean": 5621.4, |
| "valid_targets_min": 3501 |
| }, |
| { |
| "epoch": 0.768, |
| "grad_norm": 0.4926350477068029, |
| "learning_rate": 3.9906356061808713e-05, |
| "loss": 0.4004, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43261945247650146, |
| "step": 480, |
| "valid_targets_mean": 5281.9, |
| "valid_targets_min": 3595 |
| }, |
| { |
| "epoch": 0.776, |
| "grad_norm": 0.4333840111854536, |
| "learning_rate": 3.9897143300446055e-05, |
| "loss": 0.4011, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.392102986574173, |
| "step": 485, |
| "valid_targets_mean": 5049.9, |
| "valid_targets_min": 2764 |
| }, |
| { |
| "epoch": 0.784, |
| "grad_norm": 0.8363055244819876, |
| "learning_rate": 3.988749953424696e-05, |
| "loss": 0.4143, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39420437812805176, |
| "step": 490, |
| "valid_targets_mean": 5572.1, |
| "valid_targets_min": 1556 |
| }, |
| { |
| "epoch": 0.792, |
| "grad_norm": 0.5978874990580179, |
| "learning_rate": 3.9877424972111264e-05, |
| "loss": 0.3906, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41103875637054443, |
| "step": 495, |
| "valid_targets_mean": 4687.1, |
| "valid_targets_min": 3192 |
| }, |
| { |
| "epoch": 0.8, |
| "grad_norm": 0.41911173088649056, |
| "learning_rate": 3.9866919832270554e-05, |
| "loss": 0.4018, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3836507797241211, |
| "step": 500, |
| "valid_targets_mean": 5841.7, |
| "valid_targets_min": 3554 |
| }, |
| { |
| "epoch": 0.808, |
| "grad_norm": 0.47310284509318157, |
| "learning_rate": 3.9855984342283414e-05, |
| "loss": 0.4192, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4343651533126831, |
| "step": 505, |
| "valid_targets_mean": 5008.1, |
| "valid_targets_min": 3127 |
| }, |
| { |
| "epoch": 0.816, |
| "grad_norm": 0.4622380887479421, |
| "learning_rate": 3.9844618739030545e-05, |
| "loss": 0.4163, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3995232582092285, |
| "step": 510, |
| "valid_targets_mean": 5015.6, |
| "valid_targets_min": 2736 |
| }, |
| { |
| "epoch": 0.824, |
| "grad_norm": 0.43922212490128154, |
| "learning_rate": 3.98328232687096e-05, |
| "loss": 0.407, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3930421471595764, |
| "step": 515, |
| "valid_targets_mean": 5085.0, |
| "valid_targets_min": 3059 |
| }, |
| { |
| "epoch": 0.832, |
| "grad_norm": 0.47784197846119747, |
| "learning_rate": 3.982059818682986e-05, |
| "loss": 0.4284, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4223286807537079, |
| "step": 520, |
| "valid_targets_mean": 4909.1, |
| "valid_targets_min": 3166 |
| }, |
| { |
| "epoch": 0.84, |
| "grad_norm": 0.4558204453099838, |
| "learning_rate": 3.980794375820669e-05, |
| "loss": 0.404, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4094545841217041, |
| "step": 525, |
| "valid_targets_mean": 5132.7, |
| "valid_targets_min": 2419 |
| }, |
| { |
| "epoch": 0.848, |
| "grad_norm": 0.46704626735444055, |
| "learning_rate": 3.9794860256955825e-05, |
| "loss": 0.4106, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3970775008201599, |
| "step": 530, |
| "valid_targets_mean": 4925.8, |
| "valid_targets_min": 2110 |
| }, |
| { |
| "epoch": 0.856, |
| "grad_norm": 0.4650313639088827, |
| "learning_rate": 3.9781347966487415e-05, |
| "loss": 0.3963, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3976203203201294, |
| "step": 535, |
| "valid_targets_mean": 5517.6, |
| "valid_targets_min": 3413 |
| }, |
| { |
| "epoch": 0.864, |
| "grad_norm": 0.4351063387412498, |
| "learning_rate": 3.9767407179499875e-05, |
| "loss": 0.3989, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3733922839164734, |
| "step": 540, |
| "valid_targets_mean": 5028.8, |
| "valid_targets_min": 2367 |
| }, |
| { |
| "epoch": 0.872, |
| "grad_norm": 0.43593401113352803, |
| "learning_rate": 3.975303819797358e-05, |
| "loss": 0.3918, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3817777633666992, |
| "step": 545, |
| "valid_targets_mean": 5394.4, |
| "valid_targets_min": 3362 |
| }, |
| { |
| "epoch": 0.88, |
| "grad_norm": 0.4521458862271579, |
| "learning_rate": 3.973824133316431e-05, |
| "loss": 0.411, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39881613850593567, |
| "step": 550, |
| "valid_targets_mean": 4746.9, |
| "valid_targets_min": 1841 |
| }, |
| { |
| "epoch": 0.888, |
| "grad_norm": 0.4598437426940062, |
| "learning_rate": 3.972301690559645e-05, |
| "loss": 0.3963, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41156241297721863, |
| "step": 555, |
| "valid_targets_mean": 5072.2, |
| "valid_targets_min": 2351 |
| }, |
| { |
| "epoch": 0.896, |
| "grad_norm": 0.45285867156543996, |
| "learning_rate": 3.970736524505615e-05, |
| "loss": 0.3912, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3909609913825989, |
| "step": 560, |
| "valid_targets_mean": 4792.6, |
| "valid_targets_min": 2658 |
| }, |
| { |
| "epoch": 0.904, |
| "grad_norm": 0.43784154625587784, |
| "learning_rate": 3.969128669058411e-05, |
| "loss": 0.393, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3765137195587158, |
| "step": 565, |
| "valid_targets_mean": 5121.5, |
| "valid_targets_min": 1949 |
| }, |
| { |
| "epoch": 0.912, |
| "grad_norm": 0.418306220764273, |
| "learning_rate": 3.9674781590468256e-05, |
| "loss": 0.3956, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4005352854728699, |
| "step": 570, |
| "valid_targets_mean": 5744.1, |
| "valid_targets_min": 2584 |
| }, |
| { |
| "epoch": 0.92, |
| "grad_norm": 0.4657930394698072, |
| "learning_rate": 3.9657850302236184e-05, |
| "loss": 0.3925, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4179115891456604, |
| "step": 575, |
| "valid_targets_mean": 5052.2, |
| "valid_targets_min": 3259 |
| }, |
| { |
| "epoch": 0.928, |
| "grad_norm": 0.4049052068455232, |
| "learning_rate": 3.964049319264744e-05, |
| "loss": 0.4059, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38576745986938477, |
| "step": 580, |
| "valid_targets_mean": 5334.1, |
| "valid_targets_min": 2420 |
| }, |
| { |
| "epoch": 0.936, |
| "grad_norm": 0.5094071278854767, |
| "learning_rate": 3.962271063768555e-05, |
| "loss": 0.4037, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40250417590141296, |
| "step": 585, |
| "valid_targets_mean": 4488.8, |
| "valid_targets_min": 2558 |
| }, |
| { |
| "epoch": 0.944, |
| "grad_norm": 0.45363433070212433, |
| "learning_rate": 3.960450302254989e-05, |
| "loss": 0.4073, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4112010598182678, |
| "step": 590, |
| "valid_targets_mean": 5494.6, |
| "valid_targets_min": 2142 |
| }, |
| { |
| "epoch": 0.952, |
| "grad_norm": 0.48015860579066305, |
| "learning_rate": 3.958587074164735e-05, |
| "loss": 0.4107, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42106136679649353, |
| "step": 595, |
| "valid_targets_mean": 5029.9, |
| "valid_targets_min": 1599 |
| }, |
| { |
| "epoch": 0.96, |
| "grad_norm": 0.4726418125981115, |
| "learning_rate": 3.956681419858376e-05, |
| "loss": 0.4129, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44462063908576965, |
| "step": 600, |
| "valid_targets_mean": 4962.4, |
| "valid_targets_min": 3008 |
| }, |
| { |
| "epoch": 0.968, |
| "grad_norm": 0.4714956138332078, |
| "learning_rate": 3.954733380615516e-05, |
| "loss": 0.4084, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.421886146068573, |
| "step": 605, |
| "valid_targets_mean": 5476.6, |
| "valid_targets_min": 2883 |
| }, |
| { |
| "epoch": 0.976, |
| "grad_norm": 0.4453487361232741, |
| "learning_rate": 3.95274299863389e-05, |
| "loss": 0.405, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4126083552837372, |
| "step": 610, |
| "valid_targets_mean": 5330.6, |
| "valid_targets_min": 2021 |
| }, |
| { |
| "epoch": 0.984, |
| "grad_norm": 0.4367895422493744, |
| "learning_rate": 3.950710317028443e-05, |
| "loss": 0.4052, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38854730129241943, |
| "step": 615, |
| "valid_targets_mean": 5959.2, |
| "valid_targets_min": 3182 |
| }, |
| { |
| "epoch": 0.992, |
| "grad_norm": 0.44765703609963614, |
| "learning_rate": 3.9486353798303996e-05, |
| "loss": 0.4033, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4119337201118469, |
| "step": 620, |
| "valid_targets_mean": 5644.2, |
| "valid_targets_min": 2831 |
| }, |
| { |
| "epoch": 1.0, |
| "grad_norm": 0.43974684229797334, |
| "learning_rate": 3.946518231986313e-05, |
| "loss": 0.4028, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4035848379135132, |
| "step": 625, |
| "valid_targets_mean": 5190.4, |
| "valid_targets_min": 3827 |
| }, |
| { |
| "epoch": 1.008, |
| "grad_norm": 0.48855337491728335, |
| "learning_rate": 3.9443589193570847e-05, |
| "loss": 0.3957, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4080842435359955, |
| "step": 630, |
| "valid_targets_mean": 4767.4, |
| "valid_targets_min": 2321 |
| }, |
| { |
| "epoch": 1.016, |
| "grad_norm": 0.4907773997756432, |
| "learning_rate": 3.942157488716976e-05, |
| "loss": 0.3958, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3828607201576233, |
| "step": 635, |
| "valid_targets_mean": 4518.8, |
| "valid_targets_min": 3041 |
| }, |
| { |
| "epoch": 1.024, |
| "grad_norm": 0.5085197835637683, |
| "learning_rate": 3.939913987752595e-05, |
| "loss": 0.3973, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3916488289833069, |
| "step": 640, |
| "valid_targets_mean": 4417.9, |
| "valid_targets_min": 1737 |
| }, |
| { |
| "epoch": 1.032, |
| "grad_norm": 0.4433613038481427, |
| "learning_rate": 3.9376284650618605e-05, |
| "loss": 0.4011, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3774226903915405, |
| "step": 645, |
| "valid_targets_mean": 5779.4, |
| "valid_targets_min": 2021 |
| }, |
| { |
| "epoch": 1.04, |
| "grad_norm": 0.42357973309466423, |
| "learning_rate": 3.935300970152952e-05, |
| "loss": 0.3856, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3880053162574768, |
| "step": 650, |
| "valid_targets_mean": 5261.9, |
| "valid_targets_min": 3166 |
| }, |
| { |
| "epoch": 1.048, |
| "grad_norm": 0.5134718214392691, |
| "learning_rate": 3.932931553443235e-05, |
| "loss": 0.3935, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40429461002349854, |
| "step": 655, |
| "valid_targets_mean": 5344.0, |
| "valid_targets_min": 2752 |
| }, |
| { |
| "epoch": 1.056, |
| "grad_norm": 0.4703039395300262, |
| "learning_rate": 3.930520266258173e-05, |
| "loss": 0.3811, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3813926577568054, |
| "step": 660, |
| "valid_targets_mean": 4612.1, |
| "valid_targets_min": 1858 |
| }, |
| { |
| "epoch": 1.064, |
| "grad_norm": 0.44274665847983213, |
| "learning_rate": 3.928067160830208e-05, |
| "loss": 0.3862, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3840048313140869, |
| "step": 665, |
| "valid_targets_mean": 5517.1, |
| "valid_targets_min": 2024 |
| }, |
| { |
| "epoch": 1.072, |
| "grad_norm": 0.45775020925899956, |
| "learning_rate": 3.925572290297638e-05, |
| "loss": 0.398, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4003162384033203, |
| "step": 670, |
| "valid_targets_mean": 4584.3, |
| "valid_targets_min": 2766 |
| }, |
| { |
| "epoch": 1.08, |
| "grad_norm": 0.4584692707936968, |
| "learning_rate": 3.9230357087034606e-05, |
| "loss": 0.3908, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40750008821487427, |
| "step": 675, |
| "valid_targets_mean": 5665.4, |
| "valid_targets_min": 1957 |
| }, |
| { |
| "epoch": 1.088, |
| "grad_norm": 0.5369024824104618, |
| "learning_rate": 3.9204574709942036e-05, |
| "loss": 0.3958, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4147145450115204, |
| "step": 680, |
| "valid_targets_mean": 5024.9, |
| "valid_targets_min": 3214 |
| }, |
| { |
| "epoch": 1.096, |
| "grad_norm": 0.48799889720184964, |
| "learning_rate": 3.917837633018734e-05, |
| "loss": 0.391, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3953361511230469, |
| "step": 685, |
| "valid_targets_mean": 4575.1, |
| "valid_targets_min": 2150 |
| }, |
| { |
| "epoch": 1.104, |
| "grad_norm": 0.45972476296051695, |
| "learning_rate": 3.915176251527051e-05, |
| "loss": 0.3857, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37637484073638916, |
| "step": 690, |
| "valid_targets_mean": 4653.4, |
| "valid_targets_min": 1847 |
| }, |
| { |
| "epoch": 1.112, |
| "grad_norm": 0.5133988937565749, |
| "learning_rate": 3.912473384169051e-05, |
| "loss": 0.3871, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3966331481933594, |
| "step": 695, |
| "valid_targets_mean": 4091.7, |
| "valid_targets_min": 1848 |
| }, |
| { |
| "epoch": 1.12, |
| "grad_norm": 0.4375799130461317, |
| "learning_rate": 3.9097290894932866e-05, |
| "loss": 0.3785, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38700032234191895, |
| "step": 700, |
| "valid_targets_mean": 5802.2, |
| "valid_targets_min": 4130 |
| }, |
| { |
| "epoch": 1.1280000000000001, |
| "grad_norm": 0.4341977270934447, |
| "learning_rate": 3.906943426945691e-05, |
| "loss": 0.3898, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3827499449253082, |
| "step": 705, |
| "valid_targets_mean": 5576.7, |
| "valid_targets_min": 3034 |
| }, |
| { |
| "epoch": 1.1360000000000001, |
| "grad_norm": 0.41678110923933936, |
| "learning_rate": 3.9041164568682955e-05, |
| "loss": 0.384, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3722837567329407, |
| "step": 710, |
| "valid_targets_mean": 5462.0, |
| "valid_targets_min": 2135 |
| }, |
| { |
| "epoch": 1.144, |
| "grad_norm": 0.4028190267926392, |
| "learning_rate": 3.90124824049792e-05, |
| "loss": 0.3839, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3803715705871582, |
| "step": 715, |
| "valid_targets_mean": 5823.0, |
| "valid_targets_min": 3934 |
| }, |
| { |
| "epoch": 1.152, |
| "grad_norm": 0.41756133225911835, |
| "learning_rate": 3.8983388399648465e-05, |
| "loss": 0.3868, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37416183948516846, |
| "step": 720, |
| "valid_targets_mean": 5653.7, |
| "valid_targets_min": 1393 |
| }, |
| { |
| "epoch": 1.16, |
| "grad_norm": 0.4954710439019623, |
| "learning_rate": 3.895388318291474e-05, |
| "loss": 0.3974, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4164339303970337, |
| "step": 725, |
| "valid_targets_mean": 4792.9, |
| "valid_targets_min": 2921 |
| }, |
| { |
| "epoch": 1.168, |
| "grad_norm": 0.45821028932979313, |
| "learning_rate": 3.892396739390952e-05, |
| "loss": 0.3923, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.398240327835083, |
| "step": 730, |
| "valid_targets_mean": 4954.4, |
| "valid_targets_min": 1895 |
| }, |
| { |
| "epoch": 1.176, |
| "grad_norm": 0.43628958347005947, |
| "learning_rate": 3.8893641680657986e-05, |
| "loss": 0.3933, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3736591339111328, |
| "step": 735, |
| "valid_targets_mean": 5429.6, |
| "valid_targets_min": 3109 |
| }, |
| { |
| "epoch": 1.184, |
| "grad_norm": 0.44462335380637497, |
| "learning_rate": 3.886290670006495e-05, |
| "loss": 0.3858, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37904173135757446, |
| "step": 740, |
| "valid_targets_mean": 4757.2, |
| "valid_targets_min": 2954 |
| }, |
| { |
| "epoch": 1.192, |
| "grad_norm": 0.44905859122506986, |
| "learning_rate": 3.8831763117900605e-05, |
| "loss": 0.3909, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40427857637405396, |
| "step": 745, |
| "valid_targets_mean": 5420.8, |
| "valid_targets_min": 2351 |
| }, |
| { |
| "epoch": 1.2, |
| "grad_norm": 0.5152020198894951, |
| "learning_rate": 3.8800211608786166e-05, |
| "loss": 0.3835, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3733412027359009, |
| "step": 750, |
| "valid_targets_mean": 4586.8, |
| "valid_targets_min": 1665 |
| }, |
| { |
| "epoch": 1.208, |
| "grad_norm": 0.4172979085097833, |
| "learning_rate": 3.876825285617918e-05, |
| "loss": 0.3928, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3518708646297455, |
| "step": 755, |
| "valid_targets_mean": 5490.5, |
| "valid_targets_min": 3223 |
| }, |
| { |
| "epoch": 1.216, |
| "grad_norm": 0.48039928277142646, |
| "learning_rate": 3.873588755235876e-05, |
| "loss": 0.3977, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41430026292800903, |
| "step": 760, |
| "valid_targets_mean": 4947.5, |
| "valid_targets_min": 2813 |
| }, |
| { |
| "epoch": 1.224, |
| "grad_norm": 0.4547738953925324, |
| "learning_rate": 3.870311639841062e-05, |
| "loss": 0.3872, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3734089434146881, |
| "step": 765, |
| "valid_targets_mean": 4527.4, |
| "valid_targets_min": 3284 |
| }, |
| { |
| "epoch": 1.232, |
| "grad_norm": 0.4560915305401523, |
| "learning_rate": 3.866994010421182e-05, |
| "loss": 0.3793, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3907153010368347, |
| "step": 770, |
| "valid_targets_mean": 5002.2, |
| "valid_targets_min": 2989 |
| }, |
| { |
| "epoch": 1.24, |
| "grad_norm": 0.43808178968104833, |
| "learning_rate": 3.863635938841545e-05, |
| "loss": 0.3896, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3927520215511322, |
| "step": 775, |
| "valid_targets_mean": 5138.7, |
| "valid_targets_min": 2969 |
| }, |
| { |
| "epoch": 1.248, |
| "grad_norm": 0.5731993222857884, |
| "learning_rate": 3.8602374978435015e-05, |
| "loss": 0.3921, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38371872901916504, |
| "step": 780, |
| "valid_targets_mean": 4440.9, |
| "valid_targets_min": 2082 |
| }, |
| { |
| "epoch": 1.256, |
| "grad_norm": 0.4717052818348134, |
| "learning_rate": 3.8567987610428705e-05, |
| "loss": 0.387, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40044158697128296, |
| "step": 785, |
| "valid_targets_mean": 5114.4, |
| "valid_targets_min": 2573 |
| }, |
| { |
| "epoch": 1.264, |
| "grad_norm": 0.48156686109239966, |
| "learning_rate": 3.853319802928345e-05, |
| "loss": 0.3943, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3929145336151123, |
| "step": 790, |
| "valid_targets_mean": 4854.8, |
| "valid_targets_min": 3001 |
| }, |
| { |
| "epoch": 1.272, |
| "grad_norm": 0.42737780873914105, |
| "learning_rate": 3.849800698859877e-05, |
| "loss": 0.3833, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3718585968017578, |
| "step": 795, |
| "valid_targets_mean": 5414.2, |
| "valid_targets_min": 3182 |
| }, |
| { |
| "epoch": 1.28, |
| "grad_norm": 0.505661228019235, |
| "learning_rate": 3.846241525067047e-05, |
| "loss": 0.3875, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39211171865463257, |
| "step": 800, |
| "valid_targets_mean": 5127.4, |
| "valid_targets_min": 2633 |
| }, |
| { |
| "epoch": 1.288, |
| "grad_norm": 0.4332272857601429, |
| "learning_rate": 3.842642358647411e-05, |
| "loss": 0.3909, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40590980648994446, |
| "step": 805, |
| "valid_targets_mean": 5665.1, |
| "valid_targets_min": 3288 |
| }, |
| { |
| "epoch": 1.296, |
| "grad_norm": 0.5304002248780334, |
| "learning_rate": 3.839003277564831e-05, |
| "loss": 0.3866, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39189791679382324, |
| "step": 810, |
| "valid_targets_mean": 5237.4, |
| "valid_targets_min": 2540 |
| }, |
| { |
| "epoch": 1.304, |
| "grad_norm": 0.40568382951306065, |
| "learning_rate": 3.835324360647785e-05, |
| "loss": 0.3775, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36877959966659546, |
| "step": 815, |
| "valid_targets_mean": 5780.4, |
| "valid_targets_min": 2490 |
| }, |
| { |
| "epoch": 1.312, |
| "grad_norm": 0.48693087500901144, |
| "learning_rate": 3.831605687587663e-05, |
| "loss": 0.3772, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3799806833267212, |
| "step": 820, |
| "valid_targets_mean": 5663.4, |
| "valid_targets_min": 1780 |
| }, |
| { |
| "epoch": 1.32, |
| "grad_norm": 0.4586016595355896, |
| "learning_rate": 3.827847338937037e-05, |
| "loss": 0.3855, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3993930220603943, |
| "step": 825, |
| "valid_targets_mean": 4757.1, |
| "valid_targets_min": 2664 |
| }, |
| { |
| "epoch": 1.328, |
| "grad_norm": 0.4204839258732377, |
| "learning_rate": 3.824049396107918e-05, |
| "loss": 0.3852, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37551194429397583, |
| "step": 830, |
| "valid_targets_mean": 5558.9, |
| "valid_targets_min": 3316 |
| }, |
| { |
| "epoch": 1.336, |
| "grad_norm": 0.4847839806564338, |
| "learning_rate": 3.8202119413699914e-05, |
| "loss": 0.3779, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3712822198867798, |
| "step": 835, |
| "valid_targets_mean": 4409.2, |
| "valid_targets_min": 2397 |
| }, |
| { |
| "epoch": 1.3439999999999999, |
| "grad_norm": 0.4688095526412658, |
| "learning_rate": 3.8163350578488366e-05, |
| "loss": 0.3882, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3980256915092468, |
| "step": 840, |
| "valid_targets_mean": 4826.0, |
| "valid_targets_min": 2713 |
| }, |
| { |
| "epoch": 1.3519999999999999, |
| "grad_norm": 0.42823903795415735, |
| "learning_rate": 3.812418829524124e-05, |
| "loss": 0.3781, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.367824524641037, |
| "step": 845, |
| "valid_targets_mean": 5280.9, |
| "valid_targets_min": 3041 |
| }, |
| { |
| "epoch": 1.3599999999999999, |
| "grad_norm": 0.43097329227868486, |
| "learning_rate": 3.8084633412277974e-05, |
| "loss": 0.3748, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36401140689849854, |
| "step": 850, |
| "valid_targets_mean": 5287.2, |
| "valid_targets_min": 2617 |
| }, |
| { |
| "epoch": 1.3679999999999999, |
| "grad_norm": 0.47489003592515217, |
| "learning_rate": 3.804468678642238e-05, |
| "loss": 0.3772, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38549166917800903, |
| "step": 855, |
| "valid_targets_mean": 4679.8, |
| "valid_targets_min": 2089 |
| }, |
| { |
| "epoch": 1.376, |
| "grad_norm": 0.44326773611975706, |
| "learning_rate": 3.800434928298403e-05, |
| "loss": 0.3926, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36632058024406433, |
| "step": 860, |
| "valid_targets_mean": 4626.1, |
| "valid_targets_min": 3101 |
| }, |
| { |
| "epoch": 1.384, |
| "grad_norm": 0.4684954024856344, |
| "learning_rate": 3.796362177573957e-05, |
| "loss": 0.3844, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3859478235244751, |
| "step": 865, |
| "valid_targets_mean": 4761.1, |
| "valid_targets_min": 2658 |
| }, |
| { |
| "epoch": 1.392, |
| "grad_norm": 0.7067972177367762, |
| "learning_rate": 3.792250514691378e-05, |
| "loss": 0.378, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37382084131240845, |
| "step": 870, |
| "valid_targets_mean": 5656.4, |
| "valid_targets_min": 1740 |
| }, |
| { |
| "epoch": 1.4, |
| "grad_norm": 0.44998762757060823, |
| "learning_rate": 3.788100028716043e-05, |
| "loss": 0.3825, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38379499316215515, |
| "step": 875, |
| "valid_targets_mean": 5077.1, |
| "valid_targets_min": 2134 |
| }, |
| { |
| "epoch": 1.408, |
| "grad_norm": 0.4425491939748669, |
| "learning_rate": 3.7839108095543016e-05, |
| "loss": 0.3908, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4144558608531952, |
| "step": 880, |
| "valid_targets_mean": 4891.1, |
| "valid_targets_min": 2359 |
| }, |
| { |
| "epoch": 1.416, |
| "grad_norm": 0.44872903982587553, |
| "learning_rate": 3.7796829479515295e-05, |
| "loss": 0.404, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4046860635280609, |
| "step": 885, |
| "valid_targets_mean": 5185.1, |
| "valid_targets_min": 3102 |
| }, |
| { |
| "epoch": 1.424, |
| "grad_norm": 0.44652067385500843, |
| "learning_rate": 3.775416535490159e-05, |
| "loss": 0.3876, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40474092960357666, |
| "step": 890, |
| "valid_targets_mean": 6059.2, |
| "valid_targets_min": 2870 |
| }, |
| { |
| "epoch": 1.432, |
| "grad_norm": 0.46151481662731947, |
| "learning_rate": 3.7711116645876984e-05, |
| "loss": 0.3834, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40247803926467896, |
| "step": 895, |
| "valid_targets_mean": 4519.4, |
| "valid_targets_min": 2420 |
| }, |
| { |
| "epoch": 1.44, |
| "grad_norm": 0.41798293534167597, |
| "learning_rate": 3.7667684284947286e-05, |
| "loss": 0.3801, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37212830781936646, |
| "step": 900, |
| "valid_targets_mean": 5178.4, |
| "valid_targets_min": 1612 |
| }, |
| { |
| "epoch": 1.448, |
| "grad_norm": 0.44789313580261725, |
| "learning_rate": 3.762386921292885e-05, |
| "loss": 0.3808, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4068279266357422, |
| "step": 905, |
| "valid_targets_mean": 5443.9, |
| "valid_targets_min": 3051 |
| }, |
| { |
| "epoch": 1.456, |
| "grad_norm": 0.4463145091315363, |
| "learning_rate": 3.757967237892818e-05, |
| "loss": 0.384, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39448195695877075, |
| "step": 910, |
| "valid_targets_mean": 4667.4, |
| "valid_targets_min": 3121 |
| }, |
| { |
| "epoch": 1.464, |
| "grad_norm": 0.3960629663939336, |
| "learning_rate": 3.7535094740321334e-05, |
| "loss": 0.3931, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37656766176223755, |
| "step": 915, |
| "valid_targets_mean": 6040.2, |
| "valid_targets_min": 3172 |
| }, |
| { |
| "epoch": 1.472, |
| "grad_norm": 0.39965812607505785, |
| "learning_rate": 3.749013726273328e-05, |
| "loss": 0.3757, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36183857917785645, |
| "step": 920, |
| "valid_targets_mean": 5720.5, |
| "valid_targets_min": 3462 |
| }, |
| { |
| "epoch": 1.48, |
| "grad_norm": 0.42101522641403266, |
| "learning_rate": 3.7444800920016875e-05, |
| "loss": 0.3966, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3867553770542145, |
| "step": 925, |
| "valid_targets_mean": 5015.6, |
| "valid_targets_min": 3374 |
| }, |
| { |
| "epoch": 1.488, |
| "grad_norm": 0.38589699903465097, |
| "learning_rate": 3.7399086694231864e-05, |
| "loss": 0.3882, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3751162886619568, |
| "step": 930, |
| "valid_targets_mean": 5944.3, |
| "valid_targets_min": 3089 |
| }, |
| { |
| "epoch": 1.496, |
| "grad_norm": 0.3843724703047966, |
| "learning_rate": 3.735299557562352e-05, |
| "loss": 0.3631, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35290801525115967, |
| "step": 935, |
| "valid_targets_mean": 6191.9, |
| "valid_targets_min": 3515 |
| }, |
| { |
| "epoch": 1.504, |
| "grad_norm": 0.45800864124740515, |
| "learning_rate": 3.7306528562601245e-05, |
| "loss": 0.3895, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4088478684425354, |
| "step": 940, |
| "valid_targets_mean": 4674.3, |
| "valid_targets_min": 1993 |
| }, |
| { |
| "epoch": 1.512, |
| "grad_norm": 0.4226144961241278, |
| "learning_rate": 3.7259686661716945e-05, |
| "loss": 0.3689, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3802897334098816, |
| "step": 945, |
| "valid_targets_mean": 4855.2, |
| "valid_targets_min": 2936 |
| }, |
| { |
| "epoch": 1.52, |
| "grad_norm": 0.39803148199970795, |
| "learning_rate": 3.7212470887643204e-05, |
| "loss": 0.3742, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37292391061782837, |
| "step": 950, |
| "valid_targets_mean": 5653.2, |
| "valid_targets_min": 3089 |
| }, |
| { |
| "epoch": 1.528, |
| "grad_norm": 0.38056071926163415, |
| "learning_rate": 3.7164882263151315e-05, |
| "loss": 0.3726, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3680902123451233, |
| "step": 955, |
| "valid_targets_mean": 5920.5, |
| "valid_targets_min": 3506 |
| }, |
| { |
| "epoch": 1.536, |
| "grad_norm": 0.41331189136225627, |
| "learning_rate": 3.711692181908913e-05, |
| "loss": 0.3869, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3711453080177307, |
| "step": 960, |
| "valid_targets_mean": 5234.6, |
| "valid_targets_min": 2221 |
| }, |
| { |
| "epoch": 1.544, |
| "grad_norm": 0.7383242086636586, |
| "learning_rate": 3.706859059435871e-05, |
| "loss": 0.3825, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40064120292663574, |
| "step": 965, |
| "valid_targets_mean": 5673.0, |
| "valid_targets_min": 3180 |
| }, |
| { |
| "epoch": 1.552, |
| "grad_norm": 0.4338073690757923, |
| "learning_rate": 3.701988963589384e-05, |
| "loss": 0.3891, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4085204601287842, |
| "step": 970, |
| "valid_targets_mean": 4975.8, |
| "valid_targets_min": 2134 |
| }, |
| { |
| "epoch": 1.56, |
| "grad_norm": 0.41899503144858735, |
| "learning_rate": 3.697081999863736e-05, |
| "loss": 0.3712, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37785467505455017, |
| "step": 975, |
| "valid_targets_mean": 5241.7, |
| "valid_targets_min": 1960 |
| }, |
| { |
| "epoch": 1.568, |
| "grad_norm": 0.4215046894151013, |
| "learning_rate": 3.692138274551828e-05, |
| "loss": 0.3824, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3735452890396118, |
| "step": 980, |
| "valid_targets_mean": 5824.8, |
| "valid_targets_min": 2889 |
| }, |
| { |
| "epoch": 1.576, |
| "grad_norm": 0.45298127897609947, |
| "learning_rate": 3.687157894742878e-05, |
| "loss": 0.3857, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4039228558540344, |
| "step": 985, |
| "valid_targets_mean": 4839.1, |
| "valid_targets_min": 2152 |
| }, |
| { |
| "epoch": 1.584, |
| "grad_norm": 0.4638740398775489, |
| "learning_rate": 3.682140968320101e-05, |
| "loss": 0.3749, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3689330518245697, |
| "step": 990, |
| "valid_targets_mean": 5185.7, |
| "valid_targets_min": 2303 |
| }, |
| { |
| "epoch": 1.592, |
| "grad_norm": 0.42428253337113275, |
| "learning_rate": 3.6770876039583725e-05, |
| "loss": 0.3808, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36522096395492554, |
| "step": 995, |
| "valid_targets_mean": 5266.6, |
| "valid_targets_min": 3739 |
| }, |
| { |
| "epoch": 1.6, |
| "grad_norm": 0.5241631533901762, |
| "learning_rate": 3.671997911121871e-05, |
| "loss": 0.3824, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4003037214279175, |
| "step": 1000, |
| "valid_targets_mean": 5543.6, |
| "valid_targets_min": 2715 |
| }, |
| { |
| "epoch": 1.608, |
| "grad_norm": 0.45414191158699585, |
| "learning_rate": 3.6668720000617126e-05, |
| "loss": 0.3887, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4138500690460205, |
| "step": 1005, |
| "valid_targets_mean": 4949.1, |
| "valid_targets_min": 2059 |
| }, |
| { |
| "epoch": 1.616, |
| "grad_norm": 0.4442828380024753, |
| "learning_rate": 3.661709981813558e-05, |
| "loss": 0.3785, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4071786403656006, |
| "step": 1010, |
| "valid_targets_mean": 5250.0, |
| "valid_targets_min": 2771 |
| }, |
| { |
| "epoch": 1.624, |
| "grad_norm": 0.44627869250200586, |
| "learning_rate": 3.6565119681952086e-05, |
| "loss": 0.3873, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41742074489593506, |
| "step": 1015, |
| "valid_targets_mean": 4925.9, |
| "valid_targets_min": 2742 |
| }, |
| { |
| "epoch": 1.6320000000000001, |
| "grad_norm": 0.469256122789057, |
| "learning_rate": 3.651278071804186e-05, |
| "loss": 0.3869, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4260239601135254, |
| "step": 1020, |
| "valid_targets_mean": 5018.3, |
| "valid_targets_min": 3424 |
| }, |
| { |
| "epoch": 1.6400000000000001, |
| "grad_norm": 0.43986136175087603, |
| "learning_rate": 3.646008406015291e-05, |
| "loss": 0.3831, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38492852449417114, |
| "step": 1025, |
| "valid_targets_mean": 5456.1, |
| "valid_targets_min": 1461 |
| }, |
| { |
| "epoch": 1.6480000000000001, |
| "grad_norm": 0.3939131634877155, |
| "learning_rate": 3.6407030849781475e-05, |
| "loss": 0.3743, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3802257478237152, |
| "step": 1030, |
| "valid_targets_mean": 5886.1, |
| "valid_targets_min": 4150 |
| }, |
| { |
| "epoch": 1.6560000000000001, |
| "grad_norm": 0.4287998513054192, |
| "learning_rate": 3.635362223614733e-05, |
| "loss": 0.3733, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37610316276550293, |
| "step": 1035, |
| "valid_targets_mean": 5026.8, |
| "valid_targets_min": 2254 |
| }, |
| { |
| "epoch": 1.6640000000000001, |
| "grad_norm": 0.44506795404948923, |
| "learning_rate": 3.629985937616884e-05, |
| "loss": 0.3942, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41211822628974915, |
| "step": 1040, |
| "valid_targets_mean": 4870.8, |
| "valid_targets_min": 3079 |
| }, |
| { |
| "epoch": 1.6720000000000002, |
| "grad_norm": 0.43123203753239964, |
| "learning_rate": 3.624574343443794e-05, |
| "loss": 0.3699, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3657289743423462, |
| "step": 1045, |
| "valid_targets_mean": 5300.8, |
| "valid_targets_min": 3194 |
| }, |
| { |
| "epoch": 1.6800000000000002, |
| "grad_norm": 0.7361413745068606, |
| "learning_rate": 3.619127558319492e-05, |
| "loss": 0.3768, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3692248463630676, |
| "step": 1050, |
| "valid_targets_mean": 5129.8, |
| "valid_targets_min": 2976 |
| }, |
| { |
| "epoch": 1.688, |
| "grad_norm": 0.4341745369454553, |
| "learning_rate": 3.613645700230298e-05, |
| "loss": 0.3839, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36595362424850464, |
| "step": 1055, |
| "valid_targets_mean": 5220.9, |
| "valid_targets_min": 1879 |
| }, |
| { |
| "epoch": 1.696, |
| "grad_norm": 0.4868617194303009, |
| "learning_rate": 3.6081288879222696e-05, |
| "loss": 0.383, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39306819438934326, |
| "step": 1060, |
| "valid_targets_mean": 5158.3, |
| "valid_targets_min": 2253 |
| }, |
| { |
| "epoch": 1.704, |
| "grad_norm": 0.3969693801378097, |
| "learning_rate": 3.602577240898633e-05, |
| "loss": 0.3778, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3763543963432312, |
| "step": 1065, |
| "valid_targets_mean": 6068.9, |
| "valid_targets_min": 3251 |
| }, |
| { |
| "epoch": 1.712, |
| "grad_norm": 0.44603055994792007, |
| "learning_rate": 3.596990879417188e-05, |
| "loss": 0.383, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41039323806762695, |
| "step": 1070, |
| "valid_targets_mean": 5020.9, |
| "valid_targets_min": 2578 |
| }, |
| { |
| "epoch": 1.72, |
| "grad_norm": 0.4003812844721183, |
| "learning_rate": 3.591369924487711e-05, |
| "loss": 0.3852, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38508373498916626, |
| "step": 1075, |
| "valid_targets_mean": 5231.2, |
| "valid_targets_min": 2404 |
| }, |
| { |
| "epoch": 1.728, |
| "grad_norm": 0.3986461945731081, |
| "learning_rate": 3.585714497869326e-05, |
| "loss": 0.3693, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3564768433570862, |
| "step": 1080, |
| "valid_targets_mean": 5398.2, |
| "valid_targets_min": 2189 |
| }, |
| { |
| "epoch": 1.736, |
| "grad_norm": 0.47541635078479144, |
| "learning_rate": 3.580024722067872e-05, |
| "loss": 0.3967, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41275617480278015, |
| "step": 1085, |
| "valid_targets_mean": 4953.2, |
| "valid_targets_min": 3378 |
| }, |
| { |
| "epoch": 1.744, |
| "grad_norm": 0.4510022618524394, |
| "learning_rate": 3.574300720333247e-05, |
| "loss": 0.3711, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3906727433204651, |
| "step": 1090, |
| "valid_targets_mean": 5614.8, |
| "valid_targets_min": 3518 |
| }, |
| { |
| "epoch": 1.752, |
| "grad_norm": 0.446466329084181, |
| "learning_rate": 3.568542616656739e-05, |
| "loss": 0.3974, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4059438407421112, |
| "step": 1095, |
| "valid_targets_mean": 5190.3, |
| "valid_targets_min": 3449 |
| }, |
| { |
| "epoch": 1.76, |
| "grad_norm": 0.4516579341292395, |
| "learning_rate": 3.5627505357683404e-05, |
| "loss": 0.3741, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37381842732429504, |
| "step": 1100, |
| "valid_targets_mean": 5287.5, |
| "valid_targets_min": 2893 |
| }, |
| { |
| "epoch": 1.768, |
| "grad_norm": 0.47680958011694785, |
| "learning_rate": 3.5569246031340474e-05, |
| "loss": 0.3716, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40289732813835144, |
| "step": 1105, |
| "valid_targets_mean": 4751.4, |
| "valid_targets_min": 1979 |
| }, |
| { |
| "epoch": 1.776, |
| "grad_norm": 0.4304497982436774, |
| "learning_rate": 3.5510649449531375e-05, |
| "loss": 0.3855, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3671915531158447, |
| "step": 1110, |
| "valid_targets_mean": 5969.1, |
| "valid_targets_min": 2748 |
| }, |
| { |
| "epoch": 1.784, |
| "grad_norm": 0.4553411443656608, |
| "learning_rate": 3.545171688155441e-05, |
| "loss": 0.3843, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39203453063964844, |
| "step": 1115, |
| "valid_targets_mean": 5388.7, |
| "valid_targets_min": 3187 |
| }, |
| { |
| "epoch": 1.792, |
| "grad_norm": 0.41842466080908763, |
| "learning_rate": 3.5392449603985894e-05, |
| "loss": 0.3799, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3623499870300293, |
| "step": 1120, |
| "valid_targets_mean": 5962.8, |
| "valid_targets_min": 3785 |
| }, |
| { |
| "epoch": 1.8, |
| "grad_norm": 0.4290247094853605, |
| "learning_rate": 3.53328489006525e-05, |
| "loss": 0.3807, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3769965171813965, |
| "step": 1125, |
| "valid_targets_mean": 5425.0, |
| "valid_targets_min": 3813 |
| }, |
| { |
| "epoch": 1.808, |
| "grad_norm": 0.42191366505701916, |
| "learning_rate": 3.527291606260345e-05, |
| "loss": 0.3679, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3785749077796936, |
| "step": 1130, |
| "valid_targets_mean": 5525.8, |
| "valid_targets_min": 2135 |
| }, |
| { |
| "epoch": 1.8159999999999998, |
| "grad_norm": 0.4149505889318829, |
| "learning_rate": 3.521265238808255e-05, |
| "loss": 0.3809, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37426960468292236, |
| "step": 1135, |
| "valid_targets_mean": 5296.0, |
| "valid_targets_min": 1813 |
| }, |
| { |
| "epoch": 1.8239999999999998, |
| "grad_norm": 0.4134306493825934, |
| "learning_rate": 3.515205918250007e-05, |
| "loss": 0.3904, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3777231276035309, |
| "step": 1140, |
| "valid_targets_mean": 5227.3, |
| "valid_targets_min": 1935 |
| }, |
| { |
| "epoch": 1.8319999999999999, |
| "grad_norm": 0.4167841510646067, |
| "learning_rate": 3.5091137758404456e-05, |
| "loss": 0.3861, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3775944411754608, |
| "step": 1145, |
| "valid_targets_mean": 5361.8, |
| "valid_targets_min": 3991 |
| }, |
| { |
| "epoch": 1.8399999999999999, |
| "grad_norm": 0.4086620515264784, |
| "learning_rate": 3.5029889435453924e-05, |
| "loss": 0.378, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3771355152130127, |
| "step": 1150, |
| "valid_targets_mean": 5253.3, |
| "valid_targets_min": 2110 |
| }, |
| { |
| "epoch": 1.8479999999999999, |
| "grad_norm": 0.4013192088363942, |
| "learning_rate": 3.496831554038784e-05, |
| "loss": 0.371, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37708157300949097, |
| "step": 1155, |
| "valid_targets_mean": 5485.9, |
| "valid_targets_min": 2232 |
| }, |
| { |
| "epoch": 1.8559999999999999, |
| "grad_norm": 0.4217553187319389, |
| "learning_rate": 3.490641740699801e-05, |
| "loss": 0.3775, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38679879903793335, |
| "step": 1160, |
| "valid_targets_mean": 5816.8, |
| "valid_targets_min": 2999 |
| }, |
| { |
| "epoch": 1.8639999999999999, |
| "grad_norm": 0.4456023349690564, |
| "learning_rate": 3.484419637609977e-05, |
| "loss": 0.3645, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36329349875450134, |
| "step": 1165, |
| "valid_targets_mean": 4806.6, |
| "valid_targets_min": 1480 |
| }, |
| { |
| "epoch": 1.8719999999999999, |
| "grad_norm": 0.4093899518312936, |
| "learning_rate": 3.478165379550292e-05, |
| "loss": 0.387, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37503308057785034, |
| "step": 1170, |
| "valid_targets_mean": 5244.5, |
| "valid_targets_min": 3192 |
| }, |
| { |
| "epoch": 1.88, |
| "grad_norm": 0.4463869131952069, |
| "learning_rate": 3.471879101998262e-05, |
| "loss": 0.3755, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40678396821022034, |
| "step": 1175, |
| "valid_targets_mean": 4832.0, |
| "valid_targets_min": 1743 |
| }, |
| { |
| "epoch": 1.888, |
| "grad_norm": 0.43274693527655883, |
| "learning_rate": 3.465560941124992e-05, |
| "loss": 0.3936, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4071776270866394, |
| "step": 1180, |
| "valid_targets_mean": 4963.8, |
| "valid_targets_min": 2659 |
| }, |
| { |
| "epoch": 1.896, |
| "grad_norm": 0.4132096224970363, |
| "learning_rate": 3.459211033792233e-05, |
| "loss": 0.382, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3747752904891968, |
| "step": 1185, |
| "valid_targets_mean": 5074.2, |
| "valid_targets_min": 3205 |
| }, |
| { |
| "epoch": 1.904, |
| "grad_norm": 0.4287060109784466, |
| "learning_rate": 3.4528295175494194e-05, |
| "loss": 0.3765, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3766595423221588, |
| "step": 1190, |
| "valid_targets_mean": 5658.9, |
| "valid_targets_min": 3545 |
| }, |
| { |
| "epoch": 1.912, |
| "grad_norm": 0.4268405730014428, |
| "learning_rate": 3.4464165306306845e-05, |
| "loss": 0.3741, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36618363857269287, |
| "step": 1195, |
| "valid_targets_mean": 4413.0, |
| "valid_targets_min": 2111 |
| }, |
| { |
| "epoch": 1.92, |
| "grad_norm": 0.469374544591111, |
| "learning_rate": 3.4399722119518675e-05, |
| "loss": 0.3823, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38940536975860596, |
| "step": 1200, |
| "valid_targets_mean": 4680.4, |
| "valid_targets_min": 2461 |
| }, |
| { |
| "epoch": 1.928, |
| "grad_norm": 0.535795483968215, |
| "learning_rate": 3.433496701107506e-05, |
| "loss": 0.3786, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3481036424636841, |
| "step": 1205, |
| "valid_targets_mean": 4684.6, |
| "valid_targets_min": 1836 |
| }, |
| { |
| "epoch": 1.936, |
| "grad_norm": 0.4550337466460218, |
| "learning_rate": 3.426990138367813e-05, |
| "loss": 0.3854, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40872815251350403, |
| "step": 1210, |
| "valid_targets_mean": 4863.5, |
| "valid_targets_min": 2161 |
| }, |
| { |
| "epoch": 1.944, |
| "grad_norm": 0.4498032082793084, |
| "learning_rate": 3.420452664675633e-05, |
| "loss": 0.3836, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4176926016807556, |
| "step": 1215, |
| "valid_targets_mean": 4703.0, |
| "valid_targets_min": 2786 |
| }, |
| { |
| "epoch": 1.952, |
| "grad_norm": 0.4043181581349687, |
| "learning_rate": 3.4138844216433946e-05, |
| "loss": 0.3872, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39830559492111206, |
| "step": 1220, |
| "valid_targets_mean": 5234.8, |
| "valid_targets_min": 3487 |
| }, |
| { |
| "epoch": 1.96, |
| "grad_norm": 0.43672221627069385, |
| "learning_rate": 3.407285551550041e-05, |
| "loss": 0.3817, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.341196209192276, |
| "step": 1225, |
| "valid_targets_mean": 5404.8, |
| "valid_targets_min": 3205 |
| }, |
| { |
| "epoch": 1.968, |
| "grad_norm": 0.4387643826831486, |
| "learning_rate": 3.4006561973379466e-05, |
| "loss": 0.3814, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36825552582740784, |
| "step": 1230, |
| "valid_targets_mean": 4670.6, |
| "valid_targets_min": 1700 |
| }, |
| { |
| "epoch": 1.976, |
| "grad_norm": 0.44378332080147553, |
| "learning_rate": 3.3939965026098245e-05, |
| "loss": 0.3739, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38368356227874756, |
| "step": 1235, |
| "valid_targets_mean": 4895.3, |
| "valid_targets_min": 1954 |
| }, |
| { |
| "epoch": 1.984, |
| "grad_norm": 0.4544182188550218, |
| "learning_rate": 3.38730661162561e-05, |
| "loss": 0.3843, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40588533878326416, |
| "step": 1240, |
| "valid_targets_mean": 4783.8, |
| "valid_targets_min": 1487 |
| }, |
| { |
| "epoch": 1.992, |
| "grad_norm": 0.4462569799261606, |
| "learning_rate": 3.3805866692993414e-05, |
| "loss": 0.3722, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3766114115715027, |
| "step": 1245, |
| "valid_targets_mean": 4928.6, |
| "valid_targets_min": 1280 |
| }, |
| { |
| "epoch": 2.0, |
| "grad_norm": 0.41884441928566246, |
| "learning_rate": 3.373836821196018e-05, |
| "loss": 0.3638, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.343788743019104, |
| "step": 1250, |
| "valid_targets_mean": 5560.9, |
| "valid_targets_min": 1955 |
| }, |
| { |
| "epoch": 2.008, |
| "grad_norm": 0.45758483790830873, |
| "learning_rate": 3.3670572135284456e-05, |
| "loss": 0.3734, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37212687730789185, |
| "step": 1255, |
| "valid_targets_mean": 4559.3, |
| "valid_targets_min": 2254 |
| }, |
| { |
| "epoch": 2.016, |
| "grad_norm": 0.4146250246536376, |
| "learning_rate": 3.360247993154073e-05, |
| "loss": 0.3572, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34013664722442627, |
| "step": 1260, |
| "valid_targets_mean": 4974.4, |
| "valid_targets_min": 1665 |
| }, |
| { |
| "epoch": 2.024, |
| "grad_norm": 0.4735563264830304, |
| "learning_rate": 3.35340930757181e-05, |
| "loss": 0.3703, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35172104835510254, |
| "step": 1265, |
| "valid_targets_mean": 4493.6, |
| "valid_targets_min": 1828 |
| }, |
| { |
| "epoch": 2.032, |
| "grad_norm": 0.4530635233659161, |
| "learning_rate": 3.3465413049188276e-05, |
| "loss": 0.3702, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3541629910469055, |
| "step": 1270, |
| "valid_targets_mean": 5138.2, |
| "valid_targets_min": 1716 |
| }, |
| { |
| "epoch": 2.04, |
| "grad_norm": 0.4327890729586313, |
| "learning_rate": 3.3396441339673564e-05, |
| "loss": 0.3706, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37398430705070496, |
| "step": 1275, |
| "valid_targets_mean": 5744.1, |
| "valid_targets_min": 2447 |
| }, |
| { |
| "epoch": 2.048, |
| "grad_norm": 0.432678115955485, |
| "learning_rate": 3.3327179441214574e-05, |
| "loss": 0.3643, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37051498889923096, |
| "step": 1280, |
| "valid_targets_mean": 4579.2, |
| "valid_targets_min": 1799 |
| }, |
| { |
| "epoch": 2.056, |
| "grad_norm": 0.39637524913173877, |
| "learning_rate": 3.325762885413791e-05, |
| "loss": 0.3689, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3396611511707306, |
| "step": 1285, |
| "valid_targets_mean": 5177.8, |
| "valid_targets_min": 1955 |
| }, |
| { |
| "epoch": 2.064, |
| "grad_norm": 0.4411502284822464, |
| "learning_rate": 3.318779108502362e-05, |
| "loss": 0.3583, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35129696130752563, |
| "step": 1290, |
| "valid_targets_mean": 5425.0, |
| "valid_targets_min": 3122 |
| }, |
| { |
| "epoch": 2.072, |
| "grad_norm": 0.4572350445926835, |
| "learning_rate": 3.3117667646672616e-05, |
| "loss": 0.3658, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3516935706138611, |
| "step": 1295, |
| "valid_targets_mean": 4652.6, |
| "valid_targets_min": 1461 |
| }, |
| { |
| "epoch": 2.08, |
| "grad_norm": 0.3958070011442368, |
| "learning_rate": 3.304726005807386e-05, |
| "loss": 0.3599, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3486027121543884, |
| "step": 1300, |
| "valid_targets_mean": 5498.5, |
| "valid_targets_min": 3917 |
| }, |
| { |
| "epoch": 2.088, |
| "grad_norm": 0.425189458556415, |
| "learning_rate": 3.297656984437148e-05, |
| "loss": 0.3706, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36821645498275757, |
| "step": 1305, |
| "valid_targets_mean": 5508.2, |
| "valid_targets_min": 3300 |
| }, |
| { |
| "epoch": 2.096, |
| "grad_norm": 0.41140267030024497, |
| "learning_rate": 3.2905598536831715e-05, |
| "loss": 0.365, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3791000247001648, |
| "step": 1310, |
| "valid_targets_mean": 5259.4, |
| "valid_targets_min": 2908 |
| }, |
| { |
| "epoch": 2.104, |
| "grad_norm": 0.393915211029325, |
| "learning_rate": 3.2834347672809776e-05, |
| "loss": 0.3668, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3591312766075134, |
| "step": 1315, |
| "valid_targets_mean": 5570.9, |
| "valid_targets_min": 3683 |
| }, |
| { |
| "epoch": 2.112, |
| "grad_norm": 0.4122697360983724, |
| "learning_rate": 3.276281879571651e-05, |
| "loss": 0.3639, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3598926067352295, |
| "step": 1320, |
| "valid_targets_mean": 5165.6, |
| "valid_targets_min": 1960 |
| }, |
| { |
| "epoch": 2.12, |
| "grad_norm": 0.37850087516654707, |
| "learning_rate": 3.2691013454985006e-05, |
| "loss": 0.3656, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3287518322467804, |
| "step": 1325, |
| "valid_targets_mean": 5557.1, |
| "valid_targets_min": 1528 |
| }, |
| { |
| "epoch": 2.128, |
| "grad_norm": 0.4027815519497277, |
| "learning_rate": 3.2618933206036994e-05, |
| "loss": 0.3587, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3566942811012268, |
| "step": 1330, |
| "valid_targets_mean": 5634.6, |
| "valid_targets_min": 2278 |
| }, |
| { |
| "epoch": 2.136, |
| "grad_norm": 0.41190104024524776, |
| "learning_rate": 3.2546579610249177e-05, |
| "loss": 0.3649, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3564939498901367, |
| "step": 1335, |
| "valid_targets_mean": 5933.2, |
| "valid_targets_min": 3399 |
| }, |
| { |
| "epoch": 2.144, |
| "grad_norm": 0.4504816850058378, |
| "learning_rate": 3.2473954234919386e-05, |
| "loss": 0.3762, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3878824710845947, |
| "step": 1340, |
| "valid_targets_mean": 5479.0, |
| "valid_targets_min": 3139 |
| }, |
| { |
| "epoch": 2.152, |
| "grad_norm": 0.40295612651291646, |
| "learning_rate": 3.240105865323266e-05, |
| "loss": 0.3613, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33742016553878784, |
| "step": 1345, |
| "valid_targets_mean": 5556.9, |
| "valid_targets_min": 3431 |
| }, |
| { |
| "epoch": 2.16, |
| "grad_norm": 0.4566123576068589, |
| "learning_rate": 3.232789444422714e-05, |
| "loss": 0.3515, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3618641495704651, |
| "step": 1350, |
| "valid_targets_mean": 4496.8, |
| "valid_targets_min": 1826 |
| }, |
| { |
| "epoch": 2.168, |
| "grad_norm": 0.4350606966987809, |
| "learning_rate": 3.225446319275988e-05, |
| "loss": 0.3674, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3733912706375122, |
| "step": 1355, |
| "valid_targets_mean": 5226.0, |
| "valid_targets_min": 2978 |
| }, |
| { |
| "epoch": 2.176, |
| "grad_norm": 0.43998798572028774, |
| "learning_rate": 3.218076648947251e-05, |
| "loss": 0.3722, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3630021810531616, |
| "step": 1360, |
| "valid_targets_mean": 5146.9, |
| "valid_targets_min": 2461 |
| }, |
| { |
| "epoch": 2.184, |
| "grad_norm": 0.4257274437045283, |
| "learning_rate": 3.2106805930756804e-05, |
| "loss": 0.37, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3703768253326416, |
| "step": 1365, |
| "valid_targets_mean": 5433.6, |
| "valid_targets_min": 2622 |
| }, |
| { |
| "epoch": 2.192, |
| "grad_norm": 0.42990688273597316, |
| "learning_rate": 3.2032583118720045e-05, |
| "loss": 0.3627, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3537941575050354, |
| "step": 1370, |
| "valid_targets_mean": 4474.9, |
| "valid_targets_min": 1839 |
| }, |
| { |
| "epoch": 2.2, |
| "grad_norm": 0.41580843580366766, |
| "learning_rate": 3.195809966115038e-05, |
| "loss": 0.37, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36605304479599, |
| "step": 1375, |
| "valid_targets_mean": 5248.1, |
| "valid_targets_min": 2124 |
| }, |
| { |
| "epoch": 2.208, |
| "grad_norm": 0.40633646622793795, |
| "learning_rate": 3.188335717148195e-05, |
| "loss": 0.3694, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3577083945274353, |
| "step": 1380, |
| "valid_targets_mean": 5359.4, |
| "valid_targets_min": 2519 |
| }, |
| { |
| "epoch": 2.216, |
| "grad_norm": 0.41902422314686816, |
| "learning_rate": 3.1808357268759964e-05, |
| "loss": 0.3654, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3516870141029358, |
| "step": 1385, |
| "valid_targets_mean": 5334.6, |
| "valid_targets_min": 2893 |
| }, |
| { |
| "epoch": 2.224, |
| "grad_norm": 0.49088087001730035, |
| "learning_rate": 3.173310157760563e-05, |
| "loss": 0.3654, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3908011317253113, |
| "step": 1390, |
| "valid_targets_mean": 4560.0, |
| "valid_targets_min": 3054 |
| }, |
| { |
| "epoch": 2.232, |
| "grad_norm": 0.4056739625240814, |
| "learning_rate": 3.165759172818093e-05, |
| "loss": 0.346, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33683454990386963, |
| "step": 1395, |
| "valid_targets_mean": 5666.7, |
| "valid_targets_min": 2651 |
| }, |
| { |
| "epoch": 2.24, |
| "grad_norm": 0.3875675112631321, |
| "learning_rate": 3.158182935615336e-05, |
| "loss": 0.3585, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34061503410339355, |
| "step": 1400, |
| "valid_targets_mean": 5319.4, |
| "valid_targets_min": 3354 |
| }, |
| { |
| "epoch": 2.248, |
| "grad_norm": 0.4407351320062524, |
| "learning_rate": 3.150581610266046e-05, |
| "loss": 0.3605, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3685963749885559, |
| "step": 1405, |
| "valid_targets_mean": 5144.6, |
| "valid_targets_min": 2960 |
| }, |
| { |
| "epoch": 2.2560000000000002, |
| "grad_norm": 0.46446135164691993, |
| "learning_rate": 3.1429553614274256e-05, |
| "loss": 0.3713, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38366982340812683, |
| "step": 1410, |
| "valid_targets_mean": 5096.9, |
| "valid_targets_min": 2278 |
| }, |
| { |
| "epoch": 2.2640000000000002, |
| "grad_norm": 0.40570746499830124, |
| "learning_rate": 3.1353043542965636e-05, |
| "loss": 0.3585, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34492629766464233, |
| "step": 1415, |
| "valid_targets_mean": 5507.0, |
| "valid_targets_min": 4010 |
| }, |
| { |
| "epoch": 2.2720000000000002, |
| "grad_norm": 0.41254742312273185, |
| "learning_rate": 3.1276287546068536e-05, |
| "loss": 0.3758, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37195152044296265, |
| "step": 1420, |
| "valid_targets_mean": 5251.9, |
| "valid_targets_min": 2689 |
| }, |
| { |
| "epoch": 2.2800000000000002, |
| "grad_norm": 0.42950613372726965, |
| "learning_rate": 3.1199287286244047e-05, |
| "loss": 0.3637, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3722546100616455, |
| "step": 1425, |
| "valid_targets_mean": 5077.6, |
| "valid_targets_min": 2110 |
| }, |
| { |
| "epoch": 2.288, |
| "grad_norm": 0.4265459839739803, |
| "learning_rate": 3.112204443144438e-05, |
| "loss": 0.3735, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3684571385383606, |
| "step": 1430, |
| "valid_targets_mean": 4924.7, |
| "valid_targets_min": 2992 |
| }, |
| { |
| "epoch": 2.296, |
| "grad_norm": 0.4465717320574725, |
| "learning_rate": 3.1044560654876775e-05, |
| "loss": 0.3649, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36094534397125244, |
| "step": 1435, |
| "valid_targets_mean": 5348.5, |
| "valid_targets_min": 2173 |
| }, |
| { |
| "epoch": 2.304, |
| "grad_norm": 0.4254056526983632, |
| "learning_rate": 3.0966837634967215e-05, |
| "loss": 0.3702, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.380215048789978, |
| "step": 1440, |
| "valid_targets_mean": 5071.0, |
| "valid_targets_min": 3342 |
| }, |
| { |
| "epoch": 2.312, |
| "grad_norm": 0.4201825670688606, |
| "learning_rate": 3.088887705532409e-05, |
| "loss": 0.3625, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3729965090751648, |
| "step": 1445, |
| "valid_targets_mean": 5385.3, |
| "valid_targets_min": 3069 |
| }, |
| { |
| "epoch": 2.32, |
| "grad_norm": 0.4215304817349782, |
| "learning_rate": 3.081068060470174e-05, |
| "loss": 0.3686, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37321168184280396, |
| "step": 1450, |
| "valid_targets_mean": 5369.3, |
| "valid_targets_min": 2480 |
| }, |
| { |
| "epoch": 2.328, |
| "grad_norm": 0.43121898683951576, |
| "learning_rate": 3.073224997696385e-05, |
| "loss": 0.365, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37154775857925415, |
| "step": 1455, |
| "valid_targets_mean": 5396.8, |
| "valid_targets_min": 2656 |
| }, |
| { |
| "epoch": 2.336, |
| "grad_norm": 0.429703927833946, |
| "learning_rate": 3.065358687104675e-05, |
| "loss": 0.3586, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3900029957294464, |
| "step": 1460, |
| "valid_targets_mean": 5136.2, |
| "valid_targets_min": 2737 |
| }, |
| { |
| "epoch": 2.344, |
| "grad_norm": 0.4154076196755879, |
| "learning_rate": 3.057469299092264e-05, |
| "loss": 0.3675, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3497335910797119, |
| "step": 1465, |
| "valid_targets_mean": 5225.2, |
| "valid_targets_min": 1599 |
| }, |
| { |
| "epoch": 2.352, |
| "grad_norm": 0.435768634602551, |
| "learning_rate": 3.0495570045562686e-05, |
| "loss": 0.3712, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39625436067581177, |
| "step": 1470, |
| "valid_targets_mean": 4737.6, |
| "valid_targets_min": 2981 |
| }, |
| { |
| "epoch": 2.36, |
| "grad_norm": 0.37640679021873186, |
| "learning_rate": 3.041621974889996e-05, |
| "loss": 0.3615, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3373177647590637, |
| "step": 1475, |
| "valid_targets_mean": 5284.9, |
| "valid_targets_min": 1879 |
| }, |
| { |
| "epoch": 2.368, |
| "grad_norm": 0.4903548424846744, |
| "learning_rate": 3.0336643819792342e-05, |
| "loss": 0.3658, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37226593494415283, |
| "step": 1480, |
| "valid_targets_mean": 5414.7, |
| "valid_targets_min": 2378 |
| }, |
| { |
| "epoch": 2.376, |
| "grad_norm": 0.4528321418446191, |
| "learning_rate": 3.0256843981985295e-05, |
| "loss": 0.3692, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3730866611003876, |
| "step": 1485, |
| "valid_targets_mean": 4879.6, |
| "valid_targets_min": 2368 |
| }, |
| { |
| "epoch": 2.384, |
| "grad_norm": 0.4185724718104171, |
| "learning_rate": 3.0176821964074503e-05, |
| "loss": 0.3683, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35924726724624634, |
| "step": 1490, |
| "valid_targets_mean": 4952.4, |
| "valid_targets_min": 2567 |
| }, |
| { |
| "epoch": 2.392, |
| "grad_norm": 0.42965895192742065, |
| "learning_rate": 3.009657949946844e-05, |
| "loss": 0.3749, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3830062747001648, |
| "step": 1495, |
| "valid_targets_mean": 5606.8, |
| "valid_targets_min": 3378 |
| }, |
| { |
| "epoch": 2.4, |
| "grad_norm": 0.3802459422946915, |
| "learning_rate": 3.00161183263508e-05, |
| "loss": 0.3587, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3240050673484802, |
| "step": 1500, |
| "valid_targets_mean": 5342.9, |
| "valid_targets_min": 2748 |
| }, |
| { |
| "epoch": 2.408, |
| "grad_norm": 0.6407155519650329, |
| "learning_rate": 2.993544018764289e-05, |
| "loss": 0.3587, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3805801272392273, |
| "step": 1505, |
| "valid_targets_mean": 5248.1, |
| "valid_targets_min": 3093 |
| }, |
| { |
| "epoch": 2.416, |
| "grad_norm": 0.40301729661651065, |
| "learning_rate": 2.9854546830965833e-05, |
| "loss": 0.3558, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3492342233657837, |
| "step": 1510, |
| "valid_targets_mean": 5280.9, |
| "valid_targets_min": 3374 |
| }, |
| { |
| "epoch": 2.424, |
| "grad_norm": 0.4022590473663137, |
| "learning_rate": 2.9773440008602736e-05, |
| "loss": 0.3784, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3651634156703949, |
| "step": 1515, |
| "valid_targets_mean": 5080.6, |
| "valid_targets_min": 1695 |
| }, |
| { |
| "epoch": 2.432, |
| "grad_norm": 0.5291739029975226, |
| "learning_rate": 2.96921214774607e-05, |
| "loss": 0.3601, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36488720774650574, |
| "step": 1520, |
| "valid_targets_mean": 6123.9, |
| "valid_targets_min": 4685 |
| }, |
| { |
| "epoch": 2.44, |
| "grad_norm": 0.4146868610693962, |
| "learning_rate": 2.9610592999032815e-05, |
| "loss": 0.3754, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3789938688278198, |
| "step": 1525, |
| "valid_targets_mean": 5279.3, |
| "valid_targets_min": 2611 |
| }, |
| { |
| "epoch": 2.448, |
| "grad_norm": 0.40020425255047276, |
| "learning_rate": 2.9528856339359973e-05, |
| "loss": 0.3521, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3602893054485321, |
| "step": 1530, |
| "valid_targets_mean": 5018.1, |
| "valid_targets_min": 1841 |
| }, |
| { |
| "epoch": 2.456, |
| "grad_norm": 0.4059901423762991, |
| "learning_rate": 2.9446913268992588e-05, |
| "loss": 0.3664, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37402212619781494, |
| "step": 1535, |
| "valid_targets_mean": 5850.1, |
| "valid_targets_min": 2704 |
| }, |
| { |
| "epoch": 2.464, |
| "grad_norm": 0.41248635350696844, |
| "learning_rate": 2.936476556295229e-05, |
| "loss": 0.3672, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3938930630683899, |
| "step": 1540, |
| "valid_targets_mean": 5445.2, |
| "valid_targets_min": 3938 |
| }, |
| { |
| "epoch": 2.472, |
| "grad_norm": 0.4044493543396456, |
| "learning_rate": 2.928241500069346e-05, |
| "loss": 0.3663, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36537793278694153, |
| "step": 1545, |
| "valid_targets_mean": 5683.9, |
| "valid_targets_min": 2931 |
| }, |
| { |
| "epoch": 2.48, |
| "grad_norm": 0.4233510289414247, |
| "learning_rate": 2.9199863366064655e-05, |
| "loss": 0.3627, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3810930550098419, |
| "step": 1550, |
| "valid_targets_mean": 5118.5, |
| "valid_targets_min": 3142 |
| }, |
| { |
| "epoch": 2.488, |
| "grad_norm": 0.4679709309754674, |
| "learning_rate": 2.9117112447270007e-05, |
| "loss": 0.3708, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33751729130744934, |
| "step": 1555, |
| "valid_targets_mean": 5638.9, |
| "valid_targets_min": 2578 |
| }, |
| { |
| "epoch": 2.496, |
| "grad_norm": 0.4937100798126159, |
| "learning_rate": 2.9034164036830462e-05, |
| "loss": 0.3585, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39013129472732544, |
| "step": 1560, |
| "valid_targets_mean": 4878.2, |
| "valid_targets_min": 3109 |
| }, |
| { |
| "epoch": 2.504, |
| "grad_norm": 0.42930461587710605, |
| "learning_rate": 2.8951019931544975e-05, |
| "loss": 0.3761, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38630640506744385, |
| "step": 1565, |
| "valid_targets_mean": 4970.0, |
| "valid_targets_min": 2923 |
| }, |
| { |
| "epoch": 2.512, |
| "grad_norm": 0.4105397778729547, |
| "learning_rate": 2.8867681932451544e-05, |
| "loss": 0.3616, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3673209846019745, |
| "step": 1570, |
| "valid_targets_mean": 5441.9, |
| "valid_targets_min": 1621 |
| }, |
| { |
| "epoch": 2.52, |
| "grad_norm": 0.45779135009056754, |
| "learning_rate": 2.8784151844788267e-05, |
| "loss": 0.367, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3952563405036926, |
| "step": 1575, |
| "valid_targets_mean": 4770.4, |
| "valid_targets_min": 2691 |
| }, |
| { |
| "epoch": 2.528, |
| "grad_norm": 0.3825595852437832, |
| "learning_rate": 2.8700431477954155e-05, |
| "loss": 0.3594, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34684157371520996, |
| "step": 1580, |
| "valid_targets_mean": 6202.9, |
| "valid_targets_min": 3764 |
| }, |
| { |
| "epoch": 2.536, |
| "grad_norm": 0.45585952033378213, |
| "learning_rate": 2.8616522645470012e-05, |
| "loss": 0.3733, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4025506377220154, |
| "step": 1585, |
| "valid_targets_mean": 5011.0, |
| "valid_targets_min": 1993 |
| }, |
| { |
| "epoch": 2.544, |
| "grad_norm": 0.40357030619469253, |
| "learning_rate": 2.8532427164939086e-05, |
| "loss": 0.3588, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34323740005493164, |
| "step": 1590, |
| "valid_targets_mean": 5032.5, |
| "valid_targets_min": 3101 |
| }, |
| { |
| "epoch": 2.552, |
| "grad_norm": 0.415330626676679, |
| "learning_rate": 2.844814685800776e-05, |
| "loss": 0.3597, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3511185050010681, |
| "step": 1595, |
| "valid_targets_mean": 5447.2, |
| "valid_targets_min": 3202 |
| }, |
| { |
| "epoch": 2.56, |
| "grad_norm": 0.4519020463484335, |
| "learning_rate": 2.8363683550326028e-05, |
| "loss": 0.3686, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37612107396125793, |
| "step": 1600, |
| "valid_targets_mean": 5222.8, |
| "valid_targets_min": 1862 |
| }, |
| { |
| "epoch": 2.568, |
| "grad_norm": 0.44220348830564793, |
| "learning_rate": 2.8279039071508024e-05, |
| "loss": 0.353, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35742953419685364, |
| "step": 1605, |
| "valid_targets_mean": 4658.1, |
| "valid_targets_min": 1813 |
| }, |
| { |
| "epoch": 2.576, |
| "grad_norm": 0.40146559114127295, |
| "learning_rate": 2.81942152550923e-05, |
| "loss": 0.3584, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34263306856155396, |
| "step": 1610, |
| "valid_targets_mean": 5279.0, |
| "valid_targets_min": 2362 |
| }, |
| { |
| "epoch": 2.584, |
| "grad_norm": 0.39271944727239555, |
| "learning_rate": 2.810921393850219e-05, |
| "loss": 0.35, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32437071204185486, |
| "step": 1615, |
| "valid_targets_mean": 5344.3, |
| "valid_targets_min": 1687 |
| }, |
| { |
| "epoch": 2.592, |
| "grad_norm": 0.42643929145552506, |
| "learning_rate": 2.802403696300595e-05, |
| "loss": 0.3597, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37210166454315186, |
| "step": 1620, |
| "valid_targets_mean": 5013.1, |
| "valid_targets_min": 2911 |
| }, |
| { |
| "epoch": 2.6, |
| "grad_norm": 0.41469146403208523, |
| "learning_rate": 2.7938686173676915e-05, |
| "loss": 0.3653, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3780370056629181, |
| "step": 1625, |
| "valid_targets_mean": 4959.9, |
| "valid_targets_min": 2159 |
| }, |
| { |
| "epoch": 2.608, |
| "grad_norm": 0.41375741859837173, |
| "learning_rate": 2.7853163419353505e-05, |
| "loss": 0.353, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3571500778198242, |
| "step": 1630, |
| "valid_targets_mean": 4978.8, |
| "valid_targets_min": 1854 |
| }, |
| { |
| "epoch": 2.616, |
| "grad_norm": 0.41387845742512414, |
| "learning_rate": 2.776747055259918e-05, |
| "loss": 0.3599, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36002427339553833, |
| "step": 1635, |
| "valid_targets_mean": 5488.3, |
| "valid_targets_min": 2766 |
| }, |
| { |
| "epoch": 2.624, |
| "grad_norm": 0.40194650921564135, |
| "learning_rate": 2.768160942966233e-05, |
| "loss": 0.3546, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3453819453716278, |
| "step": 1640, |
| "valid_targets_mean": 5861.9, |
| "valid_targets_min": 2739 |
| }, |
| { |
| "epoch": 2.632, |
| "grad_norm": 0.6163428202363138, |
| "learning_rate": 2.759558191043603e-05, |
| "loss": 0.3711, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38815101981163025, |
| "step": 1645, |
| "valid_targets_mean": 5131.2, |
| "valid_targets_min": 3011 |
| }, |
| { |
| "epoch": 2.64, |
| "grad_norm": 0.43515423042205953, |
| "learning_rate": 2.7509389858417783e-05, |
| "loss": 0.3683, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37336793541908264, |
| "step": 1650, |
| "valid_targets_mean": 5366.9, |
| "valid_targets_min": 3841 |
| }, |
| { |
| "epoch": 2.648, |
| "grad_norm": 0.37763678955379976, |
| "learning_rate": 2.7423035140669147e-05, |
| "loss": 0.3529, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33463242650032043, |
| "step": 1655, |
| "valid_targets_mean": 5664.6, |
| "valid_targets_min": 1874 |
| }, |
| { |
| "epoch": 2.656, |
| "grad_norm": 0.41867644698915196, |
| "learning_rate": 2.7336519627775288e-05, |
| "loss": 0.3648, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34850186109542847, |
| "step": 1660, |
| "valid_targets_mean": 4854.1, |
| "valid_targets_min": 1998 |
| }, |
| { |
| "epoch": 2.664, |
| "grad_norm": 0.3937534197938621, |
| "learning_rate": 2.724984519380444e-05, |
| "loss": 0.3651, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33096402883529663, |
| "step": 1665, |
| "valid_targets_mean": 5396.0, |
| "valid_targets_min": 1930 |
| }, |
| { |
| "epoch": 2.672, |
| "grad_norm": 0.4320372464178654, |
| "learning_rate": 2.7163013716267353e-05, |
| "loss": 0.3771, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37169164419174194, |
| "step": 1670, |
| "valid_targets_mean": 4900.2, |
| "valid_targets_min": 1997 |
| }, |
| { |
| "epoch": 2.68, |
| "grad_norm": 0.43988576961386305, |
| "learning_rate": 2.707602707607659e-05, |
| "loss": 0.3583, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.374626100063324, |
| "step": 1675, |
| "valid_targets_mean": 4813.2, |
| "valid_targets_min": 2771 |
| }, |
| { |
| "epoch": 2.6879999999999997, |
| "grad_norm": 0.4852947653577413, |
| "learning_rate": 2.6988887157505786e-05, |
| "loss": 0.3665, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3685493469238281, |
| "step": 1680, |
| "valid_targets_mean": 5290.1, |
| "valid_targets_min": 2777 |
| }, |
| { |
| "epoch": 2.6959999999999997, |
| "grad_norm": 0.5017200522515853, |
| "learning_rate": 2.6901595848148842e-05, |
| "loss": 0.3592, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3635026514530182, |
| "step": 1685, |
| "valid_targets_mean": 4304.8, |
| "valid_targets_min": 2284 |
| }, |
| { |
| "epoch": 2.7039999999999997, |
| "grad_norm": 0.446802045763841, |
| "learning_rate": 2.681415503887904e-05, |
| "loss": 0.3661, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3783327639102936, |
| "step": 1690, |
| "valid_targets_mean": 4806.3, |
| "valid_targets_min": 1342 |
| }, |
| { |
| "epoch": 2.7119999999999997, |
| "grad_norm": 0.42879352594913495, |
| "learning_rate": 2.672656662380805e-05, |
| "loss": 0.3648, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35883185267448425, |
| "step": 1695, |
| "valid_targets_mean": 4599.1, |
| "valid_targets_min": 2676 |
| }, |
| { |
| "epoch": 2.7199999999999998, |
| "grad_norm": 0.4073296260357778, |
| "learning_rate": 2.6638832500244967e-05, |
| "loss": 0.3637, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3522590398788452, |
| "step": 1700, |
| "valid_targets_mean": 4777.3, |
| "valid_targets_min": 1485 |
| }, |
| { |
| "epoch": 2.7279999999999998, |
| "grad_norm": 0.4103216921880463, |
| "learning_rate": 2.655095456865514e-05, |
| "loss": 0.3614, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3730509877204895, |
| "step": 1705, |
| "valid_targets_mean": 5273.6, |
| "valid_targets_min": 3391 |
| }, |
| { |
| "epoch": 2.7359999999999998, |
| "grad_norm": 0.4375692949795071, |
| "learning_rate": 2.6462934732619047e-05, |
| "loss": 0.372, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37239909172058105, |
| "step": 1710, |
| "valid_targets_mean": 4954.7, |
| "valid_targets_min": 2578 |
| }, |
| { |
| "epoch": 2.7439999999999998, |
| "grad_norm": 0.4273338717206872, |
| "learning_rate": 2.6374774898791047e-05, |
| "loss": 0.3745, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3949906826019287, |
| "step": 1715, |
| "valid_targets_mean": 5317.1, |
| "valid_targets_min": 2890 |
| }, |
| { |
| "epoch": 2.752, |
| "grad_norm": 0.4630342842695557, |
| "learning_rate": 2.6286476976858084e-05, |
| "loss": 0.3614, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3776775002479553, |
| "step": 1720, |
| "valid_targets_mean": 4465.7, |
| "valid_targets_min": 1477 |
| }, |
| { |
| "epoch": 2.76, |
| "grad_norm": 0.4381992370831232, |
| "learning_rate": 2.619804287949831e-05, |
| "loss": 0.371, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3866676688194275, |
| "step": 1725, |
| "valid_targets_mean": 4869.6, |
| "valid_targets_min": 2292 |
| }, |
| { |
| "epoch": 2.768, |
| "grad_norm": 0.41227489226020997, |
| "learning_rate": 2.6109474522339676e-05, |
| "loss": 0.3648, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3621484041213989, |
| "step": 1730, |
| "valid_targets_mean": 5246.6, |
| "valid_targets_min": 3141 |
| }, |
| { |
| "epoch": 2.776, |
| "grad_norm": 0.4030593082017024, |
| "learning_rate": 2.6020773823918414e-05, |
| "loss": 0.3686, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3880328834056854, |
| "step": 1735, |
| "valid_targets_mean": 5665.9, |
| "valid_targets_min": 2531 |
| }, |
| { |
| "epoch": 2.784, |
| "grad_norm": 0.4331140416021422, |
| "learning_rate": 2.5931942705637473e-05, |
| "loss": 0.3561, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39987117052078247, |
| "step": 1740, |
| "valid_targets_mean": 5368.9, |
| "valid_targets_min": 1986 |
| }, |
| { |
| "epoch": 2.792, |
| "grad_norm": 0.4829613862636031, |
| "learning_rate": 2.5842983091724923e-05, |
| "loss": 0.3714, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3804318606853485, |
| "step": 1745, |
| "valid_targets_mean": 4986.2, |
| "valid_targets_min": 2559 |
| }, |
| { |
| "epoch": 2.8, |
| "grad_norm": 0.412528215308002, |
| "learning_rate": 2.575389690919226e-05, |
| "loss": 0.3638, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3335714340209961, |
| "step": 1750, |
| "valid_targets_mean": 5083.1, |
| "valid_targets_min": 2029 |
| }, |
| { |
| "epoch": 2.808, |
| "grad_norm": 0.41666700357741726, |
| "learning_rate": 2.5664686087792658e-05, |
| "loss": 0.3636, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36872315406799316, |
| "step": 1755, |
| "valid_targets_mean": 5352.8, |
| "valid_targets_min": 1549 |
| }, |
| { |
| "epoch": 2.816, |
| "grad_norm": 0.4478892842078491, |
| "learning_rate": 2.5575352559979188e-05, |
| "loss": 0.3632, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3751943111419678, |
| "step": 1760, |
| "valid_targets_mean": 5114.4, |
| "valid_targets_min": 3244 |
| }, |
| { |
| "epoch": 2.824, |
| "grad_norm": 0.4078550371702614, |
| "learning_rate": 2.5485898260862936e-05, |
| "loss": 0.3603, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3371888995170593, |
| "step": 1765, |
| "valid_targets_mean": 5394.6, |
| "valid_targets_min": 3537 |
| }, |
| { |
| "epoch": 2.832, |
| "grad_norm": 0.41361661877878947, |
| "learning_rate": 2.5396325128171072e-05, |
| "loss": 0.3643, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37865155935287476, |
| "step": 1770, |
| "valid_targets_mean": 5159.8, |
| "valid_targets_min": 3559 |
| }, |
| { |
| "epoch": 2.84, |
| "grad_norm": 0.40957836632126954, |
| "learning_rate": 2.5306635102204942e-05, |
| "loss": 0.3532, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36653077602386475, |
| "step": 1775, |
| "valid_targets_mean": 5626.7, |
| "valid_targets_min": 2279 |
| }, |
| { |
| "epoch": 2.848, |
| "grad_norm": 0.4117360240447823, |
| "learning_rate": 2.5216830125797943e-05, |
| "loss": 0.365, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35501328110694885, |
| "step": 1780, |
| "valid_targets_mean": 4968.1, |
| "valid_targets_min": 1313 |
| }, |
| { |
| "epoch": 2.856, |
| "grad_norm": 0.39139739211634533, |
| "learning_rate": 2.5126912144273517e-05, |
| "loss": 0.3553, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3554091155529022, |
| "step": 1785, |
| "valid_targets_mean": 5330.7, |
| "valid_targets_min": 2456 |
| }, |
| { |
| "epoch": 2.864, |
| "grad_norm": 0.3995353415245557, |
| "learning_rate": 2.5036883105402985e-05, |
| "loss": 0.3681, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3680855631828308, |
| "step": 1790, |
| "valid_targets_mean": 5626.8, |
| "valid_targets_min": 3205 |
| }, |
| { |
| "epoch": 2.872, |
| "grad_norm": 0.41857296112037734, |
| "learning_rate": 2.4946744959363343e-05, |
| "loss": 0.3596, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.369193971157074, |
| "step": 1795, |
| "valid_targets_mean": 5265.1, |
| "valid_targets_min": 2849 |
| }, |
| { |
| "epoch": 2.88, |
| "grad_norm": 0.4072725749604882, |
| "learning_rate": 2.4856499658695018e-05, |
| "loss": 0.3654, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35496389865875244, |
| "step": 1800, |
| "valid_targets_mean": 5219.5, |
| "valid_targets_min": 1763 |
| }, |
| { |
| "epoch": 2.888, |
| "grad_norm": 0.4134570618193474, |
| "learning_rate": 2.4766149158259603e-05, |
| "loss": 0.3709, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36438286304473877, |
| "step": 1805, |
| "valid_targets_mean": 5537.2, |
| "valid_targets_min": 2762 |
| }, |
| { |
| "epoch": 2.896, |
| "grad_norm": 0.45496611973474876, |
| "learning_rate": 2.4675695415197476e-05, |
| "loss": 0.367, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35853463411331177, |
| "step": 1810, |
| "valid_targets_mean": 5137.8, |
| "valid_targets_min": 2022 |
| }, |
| { |
| "epoch": 2.904, |
| "grad_norm": 0.4044790746858865, |
| "learning_rate": 2.458514038888543e-05, |
| "loss": 0.3592, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3444103002548218, |
| "step": 1815, |
| "valid_targets_mean": 5248.6, |
| "valid_targets_min": 2875 |
| }, |
| { |
| "epoch": 2.912, |
| "grad_norm": 0.41497694373040794, |
| "learning_rate": 2.4494486040894208e-05, |
| "loss": 0.3647, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3692508935928345, |
| "step": 1820, |
| "valid_targets_mean": 5677.1, |
| "valid_targets_min": 3785 |
| }, |
| { |
| "epoch": 2.92, |
| "grad_norm": 0.40205583087492847, |
| "learning_rate": 2.440373433494603e-05, |
| "loss": 0.3541, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3686904013156891, |
| "step": 1825, |
| "valid_targets_mean": 5355.6, |
| "valid_targets_min": 3219 |
| }, |
| { |
| "epoch": 2.928, |
| "grad_norm": 0.4342133437384953, |
| "learning_rate": 2.4312887236872066e-05, |
| "loss": 0.3675, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38688552379608154, |
| "step": 1830, |
| "valid_targets_mean": 4944.9, |
| "valid_targets_min": 2883 |
| }, |
| { |
| "epoch": 2.936, |
| "grad_norm": 0.44085569762386745, |
| "learning_rate": 2.4221946714569803e-05, |
| "loss": 0.3653, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37348800897598267, |
| "step": 1835, |
| "valid_targets_mean": 4917.7, |
| "valid_targets_min": 2132 |
| }, |
| { |
| "epoch": 2.944, |
| "grad_norm": 0.4010003283793989, |
| "learning_rate": 2.4130914737960472e-05, |
| "loss": 0.3612, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3380041718482971, |
| "step": 1840, |
| "valid_targets_mean": 5156.8, |
| "valid_targets_min": 1960 |
| }, |
| { |
| "epoch": 2.952, |
| "grad_norm": 0.40187249175981854, |
| "learning_rate": 2.4039793278946358e-05, |
| "loss": 0.3618, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3490532636642456, |
| "step": 1845, |
| "valid_targets_mean": 5288.4, |
| "valid_targets_min": 2303 |
| }, |
| { |
| "epoch": 2.96, |
| "grad_norm": 0.4136531163452904, |
| "learning_rate": 2.394858431136806e-05, |
| "loss": 0.3625, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33788761496543884, |
| "step": 1850, |
| "valid_targets_mean": 4856.2, |
| "valid_targets_min": 3121 |
| }, |
| { |
| "epoch": 2.968, |
| "grad_norm": 0.4121724833901122, |
| "learning_rate": 2.385728981096178e-05, |
| "loss": 0.37, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35700806975364685, |
| "step": 1855, |
| "valid_targets_mean": 4855.5, |
| "valid_targets_min": 2715 |
| }, |
| { |
| "epoch": 2.976, |
| "grad_norm": 0.41954397876497573, |
| "learning_rate": 2.3765911755316503e-05, |
| "loss": 0.3616, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36838629841804504, |
| "step": 1860, |
| "valid_targets_mean": 5086.6, |
| "valid_targets_min": 1855 |
| }, |
| { |
| "epoch": 2.984, |
| "grad_norm": 0.3936183308773185, |
| "learning_rate": 2.3674452123831125e-05, |
| "loss": 0.3606, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35266947746276855, |
| "step": 1865, |
| "valid_targets_mean": 5625.2, |
| "valid_targets_min": 1768 |
| }, |
| { |
| "epoch": 2.992, |
| "grad_norm": 0.3710906031622613, |
| "learning_rate": 2.358291289767165e-05, |
| "loss": 0.3627, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33334052562713623, |
| "step": 1870, |
| "valid_targets_mean": 5494.1, |
| "valid_targets_min": 3664 |
| }, |
| { |
| "epoch": 3.0, |
| "grad_norm": 0.39469556021566654, |
| "learning_rate": 2.3491296059728202e-05, |
| "loss": 0.3582, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3548370599746704, |
| "step": 1875, |
| "valid_targets_mean": 5301.9, |
| "valid_targets_min": 1778 |
| }, |
| { |
| "epoch": 3.008, |
| "grad_norm": 0.409771374224646, |
| "learning_rate": 2.339960359457212e-05, |
| "loss": 0.359, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3643002510070801, |
| "step": 1880, |
| "valid_targets_mean": 4959.6, |
| "valid_targets_min": 2645 |
| }, |
| { |
| "epoch": 3.016, |
| "grad_norm": 0.4380701674659553, |
| "learning_rate": 2.3307837488412955e-05, |
| "loss": 0.3515, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3382589817047119, |
| "step": 1885, |
| "valid_targets_mean": 5437.6, |
| "valid_targets_min": 3537 |
| }, |
| { |
| "epoch": 3.024, |
| "grad_norm": 0.39049135319375, |
| "learning_rate": 2.3215999729055437e-05, |
| "loss": 0.3375, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32954859733581543, |
| "step": 1890, |
| "valid_targets_mean": 5263.8, |
| "valid_targets_min": 3341 |
| }, |
| { |
| "epoch": 3.032, |
| "grad_norm": 0.4403802859371559, |
| "learning_rate": 2.312409230585641e-05, |
| "loss": 0.3564, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36996719241142273, |
| "step": 1895, |
| "valid_targets_mean": 5193.2, |
| "valid_targets_min": 2255 |
| }, |
| { |
| "epoch": 3.04, |
| "grad_norm": 0.41714580390836, |
| "learning_rate": 2.3032117209681782e-05, |
| "loss": 0.3509, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35327672958374023, |
| "step": 1900, |
| "valid_targets_mean": 5415.9, |
| "valid_targets_min": 2303 |
| }, |
| { |
| "epoch": 3.048, |
| "grad_norm": 0.3798579904046935, |
| "learning_rate": 2.2940076432863335e-05, |
| "loss": 0.3491, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3159680962562561, |
| "step": 1905, |
| "valid_targets_mean": 5538.6, |
| "valid_targets_min": 1884 |
| }, |
| { |
| "epoch": 3.056, |
| "grad_norm": 0.4442775508701002, |
| "learning_rate": 2.2847971969155626e-05, |
| "loss": 0.3413, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3351253867149353, |
| "step": 1910, |
| "valid_targets_mean": 4157.2, |
| "valid_targets_min": 1793 |
| }, |
| { |
| "epoch": 3.064, |
| "grad_norm": 0.41237447343015426, |
| "learning_rate": 2.275580581369276e-05, |
| "loss": 0.3552, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34021955728530884, |
| "step": 1915, |
| "valid_targets_mean": 5412.8, |
| "valid_targets_min": 3395 |
| }, |
| { |
| "epoch": 3.072, |
| "grad_norm": 0.3977316593518725, |
| "learning_rate": 2.2663579962945205e-05, |
| "loss": 0.3334, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31803208589553833, |
| "step": 1920, |
| "valid_targets_mean": 5532.6, |
| "valid_targets_min": 1069 |
| }, |
| { |
| "epoch": 3.08, |
| "grad_norm": 0.4306145547858713, |
| "learning_rate": 2.2571296414676503e-05, |
| "loss": 0.3521, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36826592683792114, |
| "step": 1925, |
| "valid_targets_mean": 5122.6, |
| "valid_targets_min": 2221 |
| }, |
| { |
| "epoch": 3.088, |
| "grad_norm": 0.4228288875025722, |
| "learning_rate": 2.2478957167900038e-05, |
| "loss": 0.3557, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.371548056602478, |
| "step": 1930, |
| "valid_targets_mean": 5281.6, |
| "valid_targets_min": 1580 |
| }, |
| { |
| "epoch": 3.096, |
| "grad_norm": 0.400976033839328, |
| "learning_rate": 2.23865642228357e-05, |
| "loss": 0.3423, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32284289598464966, |
| "step": 1935, |
| "valid_targets_mean": 5280.7, |
| "valid_targets_min": 2676 |
| }, |
| { |
| "epoch": 3.104, |
| "grad_norm": 0.40709750858408195, |
| "learning_rate": 2.2294119580866592e-05, |
| "loss": 0.3487, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3363198935985565, |
| "step": 1940, |
| "valid_targets_mean": 5185.8, |
| "valid_targets_min": 1540 |
| }, |
| { |
| "epoch": 3.112, |
| "grad_norm": 0.425727985644393, |
| "learning_rate": 2.2201625244495646e-05, |
| "loss": 0.3419, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3281407356262207, |
| "step": 1945, |
| "valid_targets_mean": 4658.4, |
| "valid_targets_min": 2559 |
| }, |
| { |
| "epoch": 3.12, |
| "grad_norm": 0.5660942021382636, |
| "learning_rate": 2.2109083217302242e-05, |
| "loss": 0.3452, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3341931700706482, |
| "step": 1950, |
| "valid_targets_mean": 5100.8, |
| "valid_targets_min": 2819 |
| }, |
| { |
| "epoch": 3.128, |
| "grad_norm": 0.42427883262380733, |
| "learning_rate": 2.201649550389885e-05, |
| "loss": 0.3571, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35237109661102295, |
| "step": 1955, |
| "valid_targets_mean": 5110.9, |
| "valid_targets_min": 3501 |
| }, |
| { |
| "epoch": 3.136, |
| "grad_norm": 0.4307316579133534, |
| "learning_rate": 2.1923864109887556e-05, |
| "loss": 0.3501, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35384735465049744, |
| "step": 1960, |
| "valid_targets_mean": 5255.4, |
| "valid_targets_min": 3088 |
| }, |
| { |
| "epoch": 3.144, |
| "grad_norm": 0.379905209562253, |
| "learning_rate": 2.1831191041816652e-05, |
| "loss": 0.3312, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3215791583061218, |
| "step": 1965, |
| "valid_targets_mean": 5907.9, |
| "valid_targets_min": 1945 |
| }, |
| { |
| "epoch": 3.152, |
| "grad_norm": 0.4232475478486178, |
| "learning_rate": 2.173847830713715e-05, |
| "loss": 0.3505, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33959484100341797, |
| "step": 1970, |
| "valid_targets_mean": 5273.2, |
| "valid_targets_min": 2058 |
| }, |
| { |
| "epoch": 3.16, |
| "grad_norm": 0.49388248846829574, |
| "learning_rate": 2.1645727914159315e-05, |
| "loss": 0.356, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38517916202545166, |
| "step": 1975, |
| "valid_targets_mean": 4690.1, |
| "valid_targets_min": 1882 |
| }, |
| { |
| "epoch": 3.168, |
| "grad_norm": 0.5057159469960442, |
| "learning_rate": 2.1552941872009144e-05, |
| "loss": 0.3477, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3522062301635742, |
| "step": 1980, |
| "valid_targets_mean": 5137.5, |
| "valid_targets_min": 2132 |
| }, |
| { |
| "epoch": 3.176, |
| "grad_norm": 0.4382052003517721, |
| "learning_rate": 2.1460122190584868e-05, |
| "loss": 0.3533, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3731418251991272, |
| "step": 1985, |
| "valid_targets_mean": 4892.1, |
| "valid_targets_min": 2628 |
| }, |
| { |
| "epoch": 3.184, |
| "grad_norm": 0.41802499179365327, |
| "learning_rate": 2.1367270880513377e-05, |
| "loss": 0.3547, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3516068756580353, |
| "step": 1990, |
| "valid_targets_mean": 5159.9, |
| "valid_targets_min": 3982 |
| }, |
| { |
| "epoch": 3.192, |
| "grad_norm": 0.41688188196439396, |
| "learning_rate": 2.127438995310671e-05, |
| "loss": 0.3685, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3436731696128845, |
| "step": 1995, |
| "valid_targets_mean": 5668.6, |
| "valid_targets_min": 3059 |
| }, |
| { |
| "epoch": 3.2, |
| "grad_norm": 0.4408406470930303, |
| "learning_rate": 2.118148142031846e-05, |
| "loss": 0.3573, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3509392738342285, |
| "step": 2000, |
| "valid_targets_mean": 4855.9, |
| "valid_targets_min": 1797 |
| }, |
| { |
| "epoch": 3.208, |
| "grad_norm": 0.40291006212039276, |
| "learning_rate": 2.1088547294700182e-05, |
| "loss": 0.35, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3529694080352783, |
| "step": 2005, |
| "valid_targets_mean": 5605.9, |
| "valid_targets_min": 3049 |
| }, |
| { |
| "epoch": 3.216, |
| "grad_norm": 0.46932499166234165, |
| "learning_rate": 2.0995589589357846e-05, |
| "loss": 0.3531, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3627557158470154, |
| "step": 2010, |
| "valid_targets_mean": 4759.1, |
| "valid_targets_min": 2729 |
| }, |
| { |
| "epoch": 3.224, |
| "grad_norm": 0.413162793631976, |
| "learning_rate": 2.0902610317908175e-05, |
| "loss": 0.3429, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32285478711128235, |
| "step": 2015, |
| "valid_targets_mean": 5271.2, |
| "valid_targets_min": 2008 |
| }, |
| { |
| "epoch": 3.232, |
| "grad_norm": 0.4113897597805335, |
| "learning_rate": 2.080961149443505e-05, |
| "loss": 0.3478, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3467862904071808, |
| "step": 2020, |
| "valid_targets_mean": 4974.3, |
| "valid_targets_min": 1621 |
| }, |
| { |
| "epoch": 3.24, |
| "grad_norm": 0.41925043364100606, |
| "learning_rate": 2.071659513344589e-05, |
| "loss": 0.3416, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3169032335281372, |
| "step": 2025, |
| "valid_targets_mean": 5479.2, |
| "valid_targets_min": 3003 |
| }, |
| { |
| "epoch": 3.248, |
| "grad_norm": 0.4273601009040025, |
| "learning_rate": 2.0623563249828e-05, |
| "loss": 0.3513, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3677609860897064, |
| "step": 2030, |
| "valid_targets_mean": 5218.2, |
| "valid_targets_min": 3579 |
| }, |
| { |
| "epoch": 3.2560000000000002, |
| "grad_norm": 0.6013082607337529, |
| "learning_rate": 2.053051785880492e-05, |
| "loss": 0.3549, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3681921362876892, |
| "step": 2035, |
| "valid_targets_mean": 4732.3, |
| "valid_targets_min": 1653 |
| }, |
| { |
| "epoch": 3.2640000000000002, |
| "grad_norm": 0.47615810779044365, |
| "learning_rate": 2.0437460975892814e-05, |
| "loss": 0.3562, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3696865141391754, |
| "step": 2040, |
| "valid_targets_mean": 4245.9, |
| "valid_targets_min": 1366 |
| }, |
| { |
| "epoch": 3.2720000000000002, |
| "grad_norm": 0.46226390133742384, |
| "learning_rate": 2.0344394616856736e-05, |
| "loss": 0.363, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3763802647590637, |
| "step": 2045, |
| "valid_targets_mean": 4421.6, |
| "valid_targets_min": 2042 |
| }, |
| { |
| "epoch": 3.2800000000000002, |
| "grad_norm": 0.4353466690086545, |
| "learning_rate": 2.0251320797667056e-05, |
| "loss": 0.3433, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3462734520435333, |
| "step": 2050, |
| "valid_targets_mean": 4847.9, |
| "valid_targets_min": 2518 |
| }, |
| { |
| "epoch": 3.288, |
| "grad_norm": 0.4623668917956047, |
| "learning_rate": 2.01582415344557e-05, |
| "loss": 0.3565, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34538695216178894, |
| "step": 2055, |
| "valid_targets_mean": 5000.6, |
| "valid_targets_min": 1941 |
| }, |
| { |
| "epoch": 3.296, |
| "grad_norm": 0.5367762552173402, |
| "learning_rate": 2.006515884347255e-05, |
| "loss": 0.35, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35341116786003113, |
| "step": 2060, |
| "valid_targets_mean": 5066.4, |
| "valid_targets_min": 2966 |
| }, |
| { |
| "epoch": 3.304, |
| "grad_norm": 0.42311967873379336, |
| "learning_rate": 1.9972074741041712e-05, |
| "loss": 0.3492, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37139976024627686, |
| "step": 2065, |
| "valid_targets_mean": 5594.8, |
| "valid_targets_min": 3438 |
| }, |
| { |
| "epoch": 3.312, |
| "grad_norm": 0.4448223581845237, |
| "learning_rate": 1.9878991243517913e-05, |
| "loss": 0.3447, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33316367864608765, |
| "step": 2070, |
| "valid_targets_mean": 5806.3, |
| "valid_targets_min": 4256 |
| }, |
| { |
| "epoch": 3.32, |
| "grad_norm": 0.4599986041487345, |
| "learning_rate": 1.9785910367242712e-05, |
| "loss": 0.3694, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3815212547779083, |
| "step": 2075, |
| "valid_targets_mean": 4435.9, |
| "valid_targets_min": 2232 |
| }, |
| { |
| "epoch": 3.328, |
| "grad_norm": 0.4273445474456081, |
| "learning_rate": 1.969283412850094e-05, |
| "loss": 0.3508, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35738271474838257, |
| "step": 2080, |
| "valid_targets_mean": 5494.7, |
| "valid_targets_min": 1768 |
| }, |
| { |
| "epoch": 3.336, |
| "grad_norm": 0.43547801715101797, |
| "learning_rate": 1.959976454347696e-05, |
| "loss": 0.3505, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35320064425468445, |
| "step": 2085, |
| "valid_targets_mean": 5078.2, |
| "valid_targets_min": 2697 |
| }, |
| { |
| "epoch": 3.344, |
| "grad_norm": 0.4653153905277336, |
| "learning_rate": 1.950670362821098e-05, |
| "loss": 0.3501, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35863110423088074, |
| "step": 2090, |
| "valid_targets_mean": 5025.6, |
| "valid_targets_min": 2999 |
| }, |
| { |
| "epoch": 3.352, |
| "grad_norm": 0.41291837548650984, |
| "learning_rate": 1.9413653398555437e-05, |
| "loss": 0.3546, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3500649631023407, |
| "step": 2095, |
| "valid_targets_mean": 5031.2, |
| "valid_targets_min": 1504 |
| }, |
| { |
| "epoch": 3.36, |
| "grad_norm": 0.45572891912635116, |
| "learning_rate": 1.9320615870131282e-05, |
| "loss": 0.3521, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35971707105636597, |
| "step": 2100, |
| "valid_targets_mean": 4820.2, |
| "valid_targets_min": 2351 |
| }, |
| { |
| "epoch": 3.368, |
| "grad_norm": 0.4154574465602922, |
| "learning_rate": 1.9227593058284343e-05, |
| "loss": 0.3489, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34268325567245483, |
| "step": 2105, |
| "valid_targets_mean": 5277.4, |
| "valid_targets_min": 3470 |
| }, |
| { |
| "epoch": 3.376, |
| "grad_norm": 0.42100800650352777, |
| "learning_rate": 1.9134586978041663e-05, |
| "loss": 0.3492, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3382658064365387, |
| "step": 2110, |
| "valid_targets_mean": 5261.9, |
| "valid_targets_min": 1858 |
| }, |
| { |
| "epoch": 3.384, |
| "grad_norm": 0.49220213775808336, |
| "learning_rate": 1.9041599644067846e-05, |
| "loss": 0.3515, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33838170766830444, |
| "step": 2115, |
| "valid_targets_mean": 4577.6, |
| "valid_targets_min": 2345 |
| }, |
| { |
| "epoch": 3.392, |
| "grad_norm": 0.4524113735647787, |
| "learning_rate": 1.8948633070621433e-05, |
| "loss": 0.364, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3661993741989136, |
| "step": 2120, |
| "valid_targets_mean": 5524.1, |
| "valid_targets_min": 3719 |
| }, |
| { |
| "epoch": 3.4, |
| "grad_norm": 0.4882018618884875, |
| "learning_rate": 1.885568927151124e-05, |
| "loss": 0.3663, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37692198157310486, |
| "step": 2125, |
| "valid_targets_mean": 4274.7, |
| "valid_targets_min": 1986 |
| }, |
| { |
| "epoch": 3.408, |
| "grad_norm": 0.4657765317884049, |
| "learning_rate": 1.8762770260052773e-05, |
| "loss": 0.359, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3934306502342224, |
| "step": 2130, |
| "valid_targets_mean": 4857.9, |
| "valid_targets_min": 1280 |
| }, |
| { |
| "epoch": 3.416, |
| "grad_norm": 0.413351441761086, |
| "learning_rate": 1.8669878049024575e-05, |
| "loss": 0.3581, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35687702894210815, |
| "step": 2135, |
| "valid_targets_mean": 5524.4, |
| "valid_targets_min": 3546 |
| }, |
| { |
| "epoch": 3.424, |
| "grad_norm": 0.4472078942301431, |
| "learning_rate": 1.857701465062467e-05, |
| "loss": 0.3594, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3557502031326294, |
| "step": 2140, |
| "valid_targets_mean": 5125.4, |
| "valid_targets_min": 3405 |
| }, |
| { |
| "epoch": 3.432, |
| "grad_norm": 0.4166965567589518, |
| "learning_rate": 1.848418207642693e-05, |
| "loss": 0.345, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34031885862350464, |
| "step": 2145, |
| "valid_targets_mean": 4965.1, |
| "valid_targets_min": 1780 |
| }, |
| { |
| "epoch": 3.44, |
| "grad_norm": 0.44076564501237603, |
| "learning_rate": 1.8391382337337548e-05, |
| "loss": 0.3529, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32455456256866455, |
| "step": 2150, |
| "valid_targets_mean": 4592.2, |
| "valid_targets_min": 1855 |
| }, |
| { |
| "epoch": 3.448, |
| "grad_norm": 0.4240515596807026, |
| "learning_rate": 1.829861744355144e-05, |
| "loss": 0.3516, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34144943952560425, |
| "step": 2155, |
| "valid_targets_mean": 5466.1, |
| "valid_targets_min": 3089 |
| }, |
| { |
| "epoch": 3.456, |
| "grad_norm": 0.45084998575917856, |
| "learning_rate": 1.820588940450872e-05, |
| "loss": 0.3524, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3536096215248108, |
| "step": 2160, |
| "valid_targets_mean": 5376.4, |
| "valid_targets_min": 1688 |
| }, |
| { |
| "epoch": 3.464, |
| "grad_norm": 0.4558038392947665, |
| "learning_rate": 1.8113200228851163e-05, |
| "loss": 0.3534, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37722301483154297, |
| "step": 2165, |
| "valid_targets_mean": 5039.1, |
| "valid_targets_min": 3626 |
| }, |
| { |
| "epoch": 3.472, |
| "grad_norm": 0.5548099785021319, |
| "learning_rate": 1.80205519243787e-05, |
| "loss": 0.3475, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36542123556137085, |
| "step": 2170, |
| "valid_targets_mean": 4514.8, |
| "valid_targets_min": 2124 |
| }, |
| { |
| "epoch": 3.48, |
| "grad_norm": 0.4184419885975123, |
| "learning_rate": 1.7927946498005934e-05, |
| "loss": 0.3566, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3498879075050354, |
| "step": 2175, |
| "valid_targets_mean": 5055.1, |
| "valid_targets_min": 1836 |
| }, |
| { |
| "epoch": 3.488, |
| "grad_norm": 0.4483872416462883, |
| "learning_rate": 1.7835385955718653e-05, |
| "loss": 0.3513, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35198789834976196, |
| "step": 2180, |
| "valid_targets_mean": 4430.7, |
| "valid_targets_min": 2312 |
| }, |
| { |
| "epoch": 3.496, |
| "grad_norm": 0.4487827693128441, |
| "learning_rate": 1.7742872302530366e-05, |
| "loss": 0.3499, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34633007645606995, |
| "step": 2185, |
| "valid_targets_mean": 4793.9, |
| "valid_targets_min": 2378 |
| }, |
| { |
| "epoch": 3.504, |
| "grad_norm": 0.4217630344402312, |
| "learning_rate": 1.765040754243892e-05, |
| "loss": 0.3371, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33683687448501587, |
| "step": 2190, |
| "valid_targets_mean": 5274.2, |
| "valid_targets_min": 1487 |
| }, |
| { |
| "epoch": 3.512, |
| "grad_norm": 0.4363895619120034, |
| "learning_rate": 1.755799367838302e-05, |
| "loss": 0.3517, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3614504337310791, |
| "step": 2195, |
| "valid_targets_mean": 5223.8, |
| "valid_targets_min": 3300 |
| }, |
| { |
| "epoch": 3.52, |
| "grad_norm": 0.5150415668681367, |
| "learning_rate": 1.746563271219891e-05, |
| "loss": 0.3489, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3470648527145386, |
| "step": 2200, |
| "valid_targets_mean": 4654.9, |
| "valid_targets_min": 2198 |
| }, |
| { |
| "epoch": 3.528, |
| "grad_norm": 0.4275870908063472, |
| "learning_rate": 1.7373326644576965e-05, |
| "loss": 0.3432, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3559647798538208, |
| "step": 2205, |
| "valid_targets_mean": 5303.2, |
| "valid_targets_min": 3382 |
| }, |
| { |
| "epoch": 3.536, |
| "grad_norm": 0.41880213200746397, |
| "learning_rate": 1.728107747501836e-05, |
| "loss": 0.3572, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3397200107574463, |
| "step": 2210, |
| "valid_targets_mean": 5187.6, |
| "valid_targets_min": 3027 |
| }, |
| { |
| "epoch": 3.544, |
| "grad_norm": 0.4574320678007308, |
| "learning_rate": 1.7188887201791785e-05, |
| "loss": 0.3462, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3712041974067688, |
| "step": 2215, |
| "valid_targets_mean": 5076.4, |
| "valid_targets_min": 1751 |
| }, |
| { |
| "epoch": 3.552, |
| "grad_norm": 0.41277208099586044, |
| "learning_rate": 1.7096757821890117e-05, |
| "loss": 0.353, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3571726083755493, |
| "step": 2220, |
| "valid_targets_mean": 5582.5, |
| "valid_targets_min": 3702 |
| }, |
| { |
| "epoch": 3.56, |
| "grad_norm": 0.3986134166462501, |
| "learning_rate": 1.7004691330987196e-05, |
| "loss": 0.3498, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35364773869514465, |
| "step": 2225, |
| "valid_targets_mean": 5962.1, |
| "valid_targets_min": 2658 |
| }, |
| { |
| "epoch": 3.568, |
| "grad_norm": 0.38027822317353077, |
| "learning_rate": 1.691268972339458e-05, |
| "loss": 0.3492, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32820767164230347, |
| "step": 2230, |
| "valid_targets_mean": 6134.9, |
| "valid_targets_min": 3501 |
| }, |
| { |
| "epoch": 3.576, |
| "grad_norm": 0.4474020657920458, |
| "learning_rate": 1.6820754992018344e-05, |
| "loss": 0.3625, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3796349763870239, |
| "step": 2235, |
| "valid_targets_mean": 4794.5, |
| "valid_targets_min": 2284 |
| }, |
| { |
| "epoch": 3.584, |
| "grad_norm": 0.45157232424321236, |
| "learning_rate": 1.6728889128315932e-05, |
| "loss": 0.3423, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3563898801803589, |
| "step": 2240, |
| "valid_targets_mean": 4597.2, |
| "valid_targets_min": 2710 |
| }, |
| { |
| "epoch": 3.592, |
| "grad_norm": 0.462425496164009, |
| "learning_rate": 1.663709412225297e-05, |
| "loss": 0.347, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3441668152809143, |
| "step": 2245, |
| "valid_targets_mean": 4952.3, |
| "valid_targets_min": 1954 |
| }, |
| { |
| "epoch": 3.6, |
| "grad_norm": 0.514260537687594, |
| "learning_rate": 1.654537196226022e-05, |
| "loss": 0.3518, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3405025005340576, |
| "step": 2250, |
| "valid_targets_mean": 4896.5, |
| "valid_targets_min": 1819 |
| }, |
| { |
| "epoch": 3.608, |
| "grad_norm": 0.4300961949801748, |
| "learning_rate": 1.6453724635190455e-05, |
| "loss": 0.3437, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3618677258491516, |
| "step": 2255, |
| "valid_targets_mean": 5453.6, |
| "valid_targets_min": 3307 |
| }, |
| { |
| "epoch": 3.616, |
| "grad_norm": 0.43333079674636255, |
| "learning_rate": 1.6362154126275467e-05, |
| "loss": 0.3408, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35662803053855896, |
| "step": 2260, |
| "valid_targets_mean": 5232.2, |
| "valid_targets_min": 3411 |
| }, |
| { |
| "epoch": 3.624, |
| "grad_norm": 1.2318807368014002, |
| "learning_rate": 1.6270662419083018e-05, |
| "loss": 0.3531, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35945945978164673, |
| "step": 2265, |
| "valid_targets_mean": 5503.7, |
| "valid_targets_min": 3052 |
| }, |
| { |
| "epoch": 3.632, |
| "grad_norm": 0.4089063473461896, |
| "learning_rate": 1.617925149547391e-05, |
| "loss": 0.3732, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3592604696750641, |
| "step": 2270, |
| "valid_targets_mean": 5226.4, |
| "valid_targets_min": 3282 |
| }, |
| { |
| "epoch": 3.64, |
| "grad_norm": 0.47541282987296213, |
| "learning_rate": 1.608792333555904e-05, |
| "loss": 0.3578, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34865763783454895, |
| "step": 2275, |
| "valid_targets_mean": 3965.5, |
| "valid_targets_min": 2104 |
| }, |
| { |
| "epoch": 3.648, |
| "grad_norm": 0.44342197844592535, |
| "learning_rate": 1.5996679917656492e-05, |
| "loss": 0.3562, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3967997431755066, |
| "step": 2280, |
| "valid_targets_mean": 4911.2, |
| "valid_targets_min": 2949 |
| }, |
| { |
| "epoch": 3.656, |
| "grad_norm": 0.4949337159754104, |
| "learning_rate": 1.5905523218248723e-05, |
| "loss": 0.3572, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36941060423851013, |
| "step": 2285, |
| "valid_targets_mean": 4507.0, |
| "valid_targets_min": 1979 |
| }, |
| { |
| "epoch": 3.664, |
| "grad_norm": 0.4255536989518861, |
| "learning_rate": 1.5814455211939698e-05, |
| "loss": 0.3501, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3457421362400055, |
| "step": 2290, |
| "valid_targets_mean": 5655.9, |
| "valid_targets_min": 2491 |
| }, |
| { |
| "epoch": 3.672, |
| "grad_norm": 0.41944624092055655, |
| "learning_rate": 1.5723477871412168e-05, |
| "loss": 0.3406, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3456742763519287, |
| "step": 2295, |
| "valid_targets_mean": 5660.8, |
| "valid_targets_min": 3346 |
| }, |
| { |
| "epoch": 3.68, |
| "grad_norm": 0.40453467445473806, |
| "learning_rate": 1.56325931673849e-05, |
| "loss": 0.3675, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3466411232948303, |
| "step": 2300, |
| "valid_targets_mean": 5072.2, |
| "valid_targets_min": 1944 |
| }, |
| { |
| "epoch": 3.6879999999999997, |
| "grad_norm": 0.4358519774776647, |
| "learning_rate": 1.5541803068569993e-05, |
| "loss": 0.3536, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3639441132545471, |
| "step": 2305, |
| "valid_targets_mean": 5320.5, |
| "valid_targets_min": 2618 |
| }, |
| { |
| "epoch": 3.6959999999999997, |
| "grad_norm": 0.41147457059571874, |
| "learning_rate": 1.5451109541630275e-05, |
| "loss": 0.3517, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3438965678215027, |
| "step": 2310, |
| "valid_targets_mean": 5867.8, |
| "valid_targets_min": 4022 |
| }, |
| { |
| "epoch": 3.7039999999999997, |
| "grad_norm": 0.4489903036345103, |
| "learning_rate": 1.536051455113663e-05, |
| "loss": 0.3502, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33912402391433716, |
| "step": 2315, |
| "valid_targets_mean": 5059.9, |
| "valid_targets_min": 2707 |
| }, |
| { |
| "epoch": 3.7119999999999997, |
| "grad_norm": 0.4152918841251614, |
| "learning_rate": 1.527002005952551e-05, |
| "loss": 0.353, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35494786500930786, |
| "step": 2320, |
| "valid_targets_mean": 5683.9, |
| "valid_targets_min": 2065 |
| }, |
| { |
| "epoch": 3.7199999999999998, |
| "grad_norm": 0.5238688759146345, |
| "learning_rate": 1.5179628027056373e-05, |
| "loss": 0.3455, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3303306996822357, |
| "step": 2325, |
| "valid_targets_mean": 4786.8, |
| "valid_targets_min": 2361 |
| }, |
| { |
| "epoch": 3.7279999999999998, |
| "grad_norm": 0.4059798982895525, |
| "learning_rate": 1.5089340411769257e-05, |
| "loss": 0.3511, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3390509784221649, |
| "step": 2330, |
| "valid_targets_mean": 5375.0, |
| "valid_targets_min": 2743 |
| }, |
| { |
| "epoch": 3.7359999999999998, |
| "grad_norm": 0.40827993378436633, |
| "learning_rate": 1.499915916944236e-05, |
| "loss": 0.3496, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34825414419174194, |
| "step": 2335, |
| "valid_targets_mean": 5299.3, |
| "valid_targets_min": 2173 |
| }, |
| { |
| "epoch": 3.7439999999999998, |
| "grad_norm": 0.4275972461676449, |
| "learning_rate": 1.490908625354964e-05, |
| "loss": 0.3601, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34784191846847534, |
| "step": 2340, |
| "valid_targets_mean": 5317.0, |
| "valid_targets_min": 3006 |
| }, |
| { |
| "epoch": 3.752, |
| "grad_norm": 0.45286219555156115, |
| "learning_rate": 1.4819123615218556e-05, |
| "loss": 0.3438, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3748186528682709, |
| "step": 2345, |
| "valid_targets_mean": 5492.8, |
| "valid_targets_min": 3816 |
| }, |
| { |
| "epoch": 3.76, |
| "grad_norm": 0.4324888989714492, |
| "learning_rate": 1.472927320318775e-05, |
| "loss": 0.3514, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3797706365585327, |
| "step": 2350, |
| "valid_targets_mean": 5235.1, |
| "valid_targets_min": 3278 |
| }, |
| { |
| "epoch": 3.768, |
| "grad_norm": 0.44505086916508524, |
| "learning_rate": 1.4639536963764878e-05, |
| "loss": 0.3281, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34119608998298645, |
| "step": 2355, |
| "valid_targets_mean": 4604.6, |
| "valid_targets_min": 2075 |
| }, |
| { |
| "epoch": 3.776, |
| "grad_norm": 0.48617979367842906, |
| "learning_rate": 1.4549916840784409e-05, |
| "loss": 0.3471, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3472432494163513, |
| "step": 2360, |
| "valid_targets_mean": 5265.9, |
| "valid_targets_min": 2876 |
| }, |
| { |
| "epoch": 3.784, |
| "grad_norm": 0.4465609937625221, |
| "learning_rate": 1.4460414775565555e-05, |
| "loss": 0.3429, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37058374285697937, |
| "step": 2365, |
| "valid_targets_mean": 4591.2, |
| "valid_targets_min": 1854 |
| }, |
| { |
| "epoch": 3.792, |
| "grad_norm": 0.46792461451915623, |
| "learning_rate": 1.43710327068702e-05, |
| "loss": 0.3588, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3647359013557434, |
| "step": 2370, |
| "valid_targets_mean": 4611.2, |
| "valid_targets_min": 2034 |
| }, |
| { |
| "epoch": 3.8, |
| "grad_norm": 0.436600344830401, |
| "learning_rate": 1.4281772570860897e-05, |
| "loss": 0.3463, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3691827058792114, |
| "step": 2375, |
| "valid_targets_mean": 4832.8, |
| "valid_targets_min": 2162 |
| }, |
| { |
| "epoch": 3.808, |
| "grad_norm": 0.4350289458556257, |
| "learning_rate": 1.4192636301058952e-05, |
| "loss": 0.3433, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34155136346817017, |
| "step": 2380, |
| "valid_targets_mean": 5288.4, |
| "valid_targets_min": 2130 |
| }, |
| { |
| "epoch": 3.816, |
| "grad_norm": 0.46315899331541166, |
| "learning_rate": 1.4103625828302508e-05, |
| "loss": 0.349, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3611070513725281, |
| "step": 2385, |
| "valid_targets_mean": 4848.5, |
| "valid_targets_min": 2740 |
| }, |
| { |
| "epoch": 3.824, |
| "grad_norm": 0.43021409702834346, |
| "learning_rate": 1.4014743080704743e-05, |
| "loss": 0.3517, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3638980984687805, |
| "step": 2390, |
| "valid_targets_mean": 5335.2, |
| "valid_targets_min": 3195 |
| }, |
| { |
| "epoch": 3.832, |
| "grad_norm": 0.40844590516723256, |
| "learning_rate": 1.3925989983612118e-05, |
| "loss": 0.3473, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3579707741737366, |
| "step": 2395, |
| "valid_targets_mean": 5167.5, |
| "valid_targets_min": 3532 |
| }, |
| { |
| "epoch": 3.84, |
| "grad_norm": 0.42151240277254826, |
| "learning_rate": 1.383736845956261e-05, |
| "loss": 0.3497, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33858180046081543, |
| "step": 2400, |
| "valid_targets_mean": 5275.1, |
| "valid_targets_min": 2427 |
| }, |
| { |
| "epoch": 3.848, |
| "grad_norm": 0.48222299533812996, |
| "learning_rate": 1.3748880428244154e-05, |
| "loss": 0.3408, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35040467977523804, |
| "step": 2405, |
| "valid_targets_mean": 5353.0, |
| "valid_targets_min": 2540 |
| }, |
| { |
| "epoch": 3.856, |
| "grad_norm": 0.4020519339172705, |
| "learning_rate": 1.3660527806452965e-05, |
| "loss": 0.3431, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.350327730178833, |
| "step": 2410, |
| "valid_targets_mean": 5579.1, |
| "valid_targets_min": 2531 |
| }, |
| { |
| "epoch": 3.864, |
| "grad_norm": 0.5085198879205962, |
| "learning_rate": 1.3572312508052118e-05, |
| "loss": 0.3495, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36508017778396606, |
| "step": 2415, |
| "valid_targets_mean": 3858.1, |
| "valid_targets_min": 1579 |
| }, |
| { |
| "epoch": 3.872, |
| "grad_norm": 0.4183209237198874, |
| "learning_rate": 1.3484236443929982e-05, |
| "loss": 0.3513, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3143860101699829, |
| "step": 2420, |
| "valid_targets_mean": 5434.0, |
| "valid_targets_min": 3140 |
| }, |
| { |
| "epoch": 3.88, |
| "grad_norm": 0.4096711410714446, |
| "learning_rate": 1.3396301521958926e-05, |
| "loss": 0.3509, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3651965856552124, |
| "step": 2425, |
| "valid_targets_mean": 5408.8, |
| "valid_targets_min": 1740 |
| }, |
| { |
| "epoch": 3.888, |
| "grad_norm": 0.4112083143736238, |
| "learning_rate": 1.3308509646953934e-05, |
| "loss": 0.3527, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3459765613079071, |
| "step": 2430, |
| "valid_targets_mean": 5369.6, |
| "valid_targets_min": 2192 |
| }, |
| { |
| "epoch": 3.896, |
| "grad_norm": 0.3677776926427397, |
| "learning_rate": 1.3220862720631349e-05, |
| "loss": 0.3502, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31041914224624634, |
| "step": 2435, |
| "valid_targets_mean": 6079.5, |
| "valid_targets_min": 3186 |
| }, |
| { |
| "epoch": 3.904, |
| "grad_norm": 0.41903548107966454, |
| "learning_rate": 1.3133362641567697e-05, |
| "loss": 0.3469, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3313430845737457, |
| "step": 2440, |
| "valid_targets_mean": 5767.3, |
| "valid_targets_min": 2754 |
| }, |
| { |
| "epoch": 3.912, |
| "grad_norm": 0.4570504709610015, |
| "learning_rate": 1.3046011305158546e-05, |
| "loss": 0.363, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3728758990764618, |
| "step": 2445, |
| "valid_targets_mean": 5060.2, |
| "valid_targets_min": 3078 |
| }, |
| { |
| "epoch": 3.92, |
| "grad_norm": 0.39488174999109327, |
| "learning_rate": 1.2958810603577456e-05, |
| "loss": 0.333, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3275885581970215, |
| "step": 2450, |
| "valid_targets_mean": 5300.8, |
| "valid_targets_min": 2121 |
| }, |
| { |
| "epoch": 3.928, |
| "grad_norm": 0.423197246952939, |
| "learning_rate": 1.2871762425734989e-05, |
| "loss": 0.3381, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3365005552768707, |
| "step": 2455, |
| "valid_targets_mean": 5088.2, |
| "valid_targets_min": 3577 |
| }, |
| { |
| "epoch": 3.936, |
| "grad_norm": 0.4442823306007306, |
| "learning_rate": 1.278486865723779e-05, |
| "loss": 0.3486, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35032421350479126, |
| "step": 2460, |
| "valid_targets_mean": 4681.9, |
| "valid_targets_min": 2547 |
| }, |
| { |
| "epoch": 3.944, |
| "grad_norm": 0.4949569884565107, |
| "learning_rate": 1.269813118034775e-05, |
| "loss": 0.341, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35367071628570557, |
| "step": 2465, |
| "valid_targets_mean": 4858.9, |
| "valid_targets_min": 2016 |
| }, |
| { |
| "epoch": 3.952, |
| "grad_norm": 0.47090349069650783, |
| "learning_rate": 1.2611551873941213e-05, |
| "loss": 0.3511, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33958202600479126, |
| "step": 2470, |
| "valid_targets_mean": 5593.2, |
| "valid_targets_min": 2883 |
| }, |
| { |
| "epoch": 3.96, |
| "grad_norm": 0.4117528257305293, |
| "learning_rate": 1.2525132613468309e-05, |
| "loss": 0.3446, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3318389654159546, |
| "step": 2475, |
| "valid_targets_mean": 5085.9, |
| "valid_targets_min": 3081 |
| }, |
| { |
| "epoch": 3.968, |
| "grad_norm": 0.4355964002455219, |
| "learning_rate": 1.2438875270912294e-05, |
| "loss": 0.3399, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35395902395248413, |
| "step": 2480, |
| "valid_targets_mean": 4881.1, |
| "valid_targets_min": 1794 |
| }, |
| { |
| "epoch": 3.976, |
| "grad_norm": 0.4473670632692554, |
| "learning_rate": 1.2352781714749016e-05, |
| "loss": 0.3486, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38158032298088074, |
| "step": 2485, |
| "valid_targets_mean": 4953.5, |
| "valid_targets_min": 3239 |
| }, |
| { |
| "epoch": 3.984, |
| "grad_norm": 0.3973659500753273, |
| "learning_rate": 1.2266853809906469e-05, |
| "loss": 0.3488, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3382183313369751, |
| "step": 2490, |
| "valid_targets_mean": 5590.4, |
| "valid_targets_min": 2991 |
| }, |
| { |
| "epoch": 3.992, |
| "grad_norm": 0.4710455576260363, |
| "learning_rate": 1.2181093417724317e-05, |
| "loss": 0.3485, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3666461110115051, |
| "step": 2495, |
| "valid_targets_mean": 4910.9, |
| "valid_targets_min": 3382 |
| }, |
| { |
| "epoch": 4.0, |
| "grad_norm": 0.44066943755115956, |
| "learning_rate": 1.2095502395913676e-05, |
| "loss": 0.3469, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3465125858783722, |
| "step": 2500, |
| "valid_targets_mean": 5251.6, |
| "valid_targets_min": 3506 |
| }, |
| { |
| "epoch": 4.008, |
| "grad_norm": 0.4427749517617702, |
| "learning_rate": 1.2010082598516775e-05, |
| "loss": 0.3384, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33215904235839844, |
| "step": 2505, |
| "valid_targets_mean": 4974.8, |
| "valid_targets_min": 1979 |
| }, |
| { |
| "epoch": 4.016, |
| "grad_norm": 0.44953105232660073, |
| "learning_rate": 1.1924835875866884e-05, |
| "loss": 0.3451, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35312649607658386, |
| "step": 2510, |
| "valid_targets_mean": 5263.2, |
| "valid_targets_min": 3158 |
| }, |
| { |
| "epoch": 4.024, |
| "grad_norm": 0.4443113294031506, |
| "learning_rate": 1.1839764074548145e-05, |
| "loss": 0.36, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3570879101753235, |
| "step": 2515, |
| "valid_targets_mean": 5002.6, |
| "valid_targets_min": 3202 |
| }, |
| { |
| "epoch": 4.032, |
| "grad_norm": 0.42273114814017915, |
| "learning_rate": 1.1754869037355659e-05, |
| "loss": 0.3494, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33162546157836914, |
| "step": 2520, |
| "valid_targets_mean": 5304.5, |
| "valid_targets_min": 3212 |
| }, |
| { |
| "epoch": 4.04, |
| "grad_norm": 0.4178054974882896, |
| "learning_rate": 1.1670152603255504e-05, |
| "loss": 0.3379, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34266212582588196, |
| "step": 2525, |
| "valid_targets_mean": 5299.8, |
| "valid_targets_min": 3034 |
| }, |
| { |
| "epoch": 4.048, |
| "grad_norm": 0.5121566519360325, |
| "learning_rate": 1.1585616607344909e-05, |
| "loss": 0.3437, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3702877163887024, |
| "step": 2530, |
| "valid_targets_mean": 4934.2, |
| "valid_targets_min": 3374 |
| }, |
| { |
| "epoch": 4.056, |
| "grad_norm": 0.4501602566423154, |
| "learning_rate": 1.1501262880812547e-05, |
| "loss": 0.3437, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3468663692474365, |
| "step": 2535, |
| "valid_targets_mean": 4796.7, |
| "valid_targets_min": 2716 |
| }, |
| { |
| "epoch": 4.064, |
| "grad_norm": 0.40240850175186393, |
| "learning_rate": 1.141709325089881e-05, |
| "loss": 0.3424, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33826762437820435, |
| "step": 2540, |
| "valid_targets_mean": 5921.9, |
| "valid_targets_min": 3318 |
| }, |
| { |
| "epoch": 4.072, |
| "grad_norm": 0.42067028180373117, |
| "learning_rate": 1.1333109540856257e-05, |
| "loss": 0.3401, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3362520933151245, |
| "step": 2545, |
| "valid_targets_mean": 5226.8, |
| "valid_targets_min": 3783 |
| }, |
| { |
| "epoch": 4.08, |
| "grad_norm": 0.45758348729992443, |
| "learning_rate": 1.1249313569910143e-05, |
| "loss": 0.3395, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34522631764411926, |
| "step": 2550, |
| "valid_targets_mean": 4955.1, |
| "valid_targets_min": 2752 |
| }, |
| { |
| "epoch": 4.088, |
| "grad_norm": 0.46260791432906484, |
| "learning_rate": 1.1165707153218942e-05, |
| "loss": 0.3447, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37173616886138916, |
| "step": 2555, |
| "valid_targets_mean": 4847.7, |
| "valid_targets_min": 3039 |
| }, |
| { |
| "epoch": 4.096, |
| "grad_norm": 0.45178101383894775, |
| "learning_rate": 1.1082292101835121e-05, |
| "loss": 0.3365, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33136144280433655, |
| "step": 2560, |
| "valid_targets_mean": 5177.0, |
| "valid_targets_min": 3080 |
| }, |
| { |
| "epoch": 4.104, |
| "grad_norm": 0.4231361554022084, |
| "learning_rate": 1.099907022266582e-05, |
| "loss": 0.3251, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3230976462364197, |
| "step": 2565, |
| "valid_targets_mean": 4989.7, |
| "valid_targets_min": 1487 |
| }, |
| { |
| "epoch": 4.112, |
| "grad_norm": 0.4277751599231597, |
| "learning_rate": 1.0916043318433767e-05, |
| "loss": 0.3345, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34056150913238525, |
| "step": 2570, |
| "valid_targets_mean": 5540.6, |
| "valid_targets_min": 1751 |
| }, |
| { |
| "epoch": 4.12, |
| "grad_norm": 0.3998179438154001, |
| "learning_rate": 1.0833213187638203e-05, |
| "loss": 0.3333, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31875139474868774, |
| "step": 2575, |
| "valid_targets_mean": 5543.2, |
| "valid_targets_min": 3226 |
| }, |
| { |
| "epoch": 4.128, |
| "grad_norm": 0.4745880449051221, |
| "learning_rate": 1.0750581624515957e-05, |
| "loss": 0.3447, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3355952501296997, |
| "step": 2580, |
| "valid_targets_mean": 5081.5, |
| "valid_targets_min": 2531 |
| }, |
| { |
| "epoch": 4.136, |
| "grad_norm": 0.44499903268408586, |
| "learning_rate": 1.0668150419002527e-05, |
| "loss": 0.3485, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35789117217063904, |
| "step": 2585, |
| "valid_targets_mean": 5335.2, |
| "valid_targets_min": 3387 |
| }, |
| { |
| "epoch": 4.144, |
| "grad_norm": 0.4197341708863458, |
| "learning_rate": 1.0585921356693349e-05, |
| "loss": 0.3346, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3353608250617981, |
| "step": 2590, |
| "valid_targets_mean": 5839.6, |
| "valid_targets_min": 3317 |
| }, |
| { |
| "epoch": 4.152, |
| "grad_norm": 0.409411375443442, |
| "learning_rate": 1.0503896218805112e-05, |
| "loss": 0.3297, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30914759635925293, |
| "step": 2595, |
| "valid_targets_mean": 5053.8, |
| "valid_targets_min": 1464 |
| }, |
| { |
| "epoch": 4.16, |
| "grad_norm": 0.3925462046782356, |
| "learning_rate": 1.0422076782137155e-05, |
| "loss": 0.3361, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30242836475372314, |
| "step": 2600, |
| "valid_targets_mean": 5689.5, |
| "valid_targets_min": 2949 |
| }, |
| { |
| "epoch": 4.168, |
| "grad_norm": 0.4277957873278747, |
| "learning_rate": 1.0340464819032991e-05, |
| "loss": 0.3401, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3455298840999603, |
| "step": 2605, |
| "valid_targets_mean": 5142.5, |
| "valid_targets_min": 2023 |
| }, |
| { |
| "epoch": 4.176, |
| "grad_norm": 0.38074315117848584, |
| "learning_rate": 1.0259062097341911e-05, |
| "loss": 0.326, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32840967178344727, |
| "step": 2610, |
| "valid_targets_mean": 5884.4, |
| "valid_targets_min": 3399 |
| }, |
| { |
| "epoch": 4.184, |
| "grad_norm": 0.4548307845141411, |
| "learning_rate": 1.017787038038071e-05, |
| "loss": 0.3521, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3642027974128723, |
| "step": 2615, |
| "valid_targets_mean": 5079.9, |
| "valid_targets_min": 3305 |
| }, |
| { |
| "epoch": 4.192, |
| "grad_norm": 0.4083245961179395, |
| "learning_rate": 1.0096891426895476e-05, |
| "loss": 0.3296, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3238316774368286, |
| "step": 2620, |
| "valid_targets_mean": 5573.8, |
| "valid_targets_min": 3595 |
| }, |
| { |
| "epoch": 4.2, |
| "grad_norm": 0.46589638331084376, |
| "learning_rate": 1.0016126991023447e-05, |
| "loss": 0.3348, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3392515182495117, |
| "step": 2625, |
| "valid_targets_mean": 5265.4, |
| "valid_targets_min": 3135 |
| }, |
| { |
| "epoch": 4.208, |
| "grad_norm": 0.38531739809636123, |
| "learning_rate": 9.935578822255113e-06, |
| "loss": 0.3338, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32173964381217957, |
| "step": 2630, |
| "valid_targets_mean": 5669.6, |
| "valid_targets_min": 2117 |
| }, |
| { |
| "epoch": 4.216, |
| "grad_norm": 0.41920020972401806, |
| "learning_rate": 9.855248665396218e-06, |
| "loss": 0.3377, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33915600180625916, |
| "step": 2635, |
| "valid_targets_mean": 5117.1, |
| "valid_targets_min": 2132 |
| }, |
| { |
| "epoch": 4.224, |
| "grad_norm": 0.44221634770458634, |
| "learning_rate": 9.775138260530046e-06, |
| "loss": 0.3352, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.334242582321167, |
| "step": 2640, |
| "valid_targets_mean": 4853.6, |
| "valid_targets_min": 2720 |
| }, |
| { |
| "epoch": 4.232, |
| "grad_norm": 0.4405776685046209, |
| "learning_rate": 9.695249342979667e-06, |
| "loss": 0.3377, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.355324923992157, |
| "step": 2645, |
| "valid_targets_mean": 5592.2, |
| "valid_targets_min": 3977 |
| }, |
| { |
| "epoch": 4.24, |
| "grad_norm": 0.4334275151976404, |
| "learning_rate": 9.615583643270371e-06, |
| "loss": 0.3396, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.341583788394928, |
| "step": 2650, |
| "valid_targets_mean": 5136.8, |
| "valid_targets_min": 3091 |
| }, |
| { |
| "epoch": 4.248, |
| "grad_norm": 0.446139566395475, |
| "learning_rate": 9.536142887092208e-06, |
| "loss": 0.3418, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3564864993095398, |
| "step": 2655, |
| "valid_targets_mean": 4890.9, |
| "valid_targets_min": 2962 |
| }, |
| { |
| "epoch": 4.256, |
| "grad_norm": 0.4386885933858475, |
| "learning_rate": 9.456928795262552e-06, |
| "loss": 0.3393, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36785727739334106, |
| "step": 2660, |
| "valid_targets_mean": 5116.8, |
| "valid_targets_min": 2240 |
| }, |
| { |
| "epoch": 4.264, |
| "grad_norm": 0.596320499454288, |
| "learning_rate": 9.377943083688873e-06, |
| "loss": 0.3446, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.361165314912796, |
| "step": 2665, |
| "valid_targets_mean": 4523.8, |
| "valid_targets_min": 1960 |
| }, |
| { |
| "epoch": 4.272, |
| "grad_norm": 0.39665797700039535, |
| "learning_rate": 9.29918746333153e-06, |
| "loss": 0.3373, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32250356674194336, |
| "step": 2670, |
| "valid_targets_mean": 5857.8, |
| "valid_targets_min": 3837 |
| }, |
| { |
| "epoch": 4.28, |
| "grad_norm": 0.43318545823286686, |
| "learning_rate": 9.220663640166756e-06, |
| "loss": 0.3476, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36065182089805603, |
| "step": 2675, |
| "valid_targets_mean": 5157.8, |
| "valid_targets_min": 3049 |
| }, |
| { |
| "epoch": 4.288, |
| "grad_norm": 0.46945049959680446, |
| "learning_rate": 9.142373315149655e-06, |
| "loss": 0.3333, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3442700207233429, |
| "step": 2680, |
| "valid_targets_mean": 5421.9, |
| "valid_targets_min": 3347 |
| }, |
| { |
| "epoch": 4.296, |
| "grad_norm": 0.40931060278446235, |
| "learning_rate": 9.064318184177373e-06, |
| "loss": 0.3406, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34707844257354736, |
| "step": 2685, |
| "valid_targets_mean": 5549.7, |
| "valid_targets_min": 3187 |
| }, |
| { |
| "epoch": 4.304, |
| "grad_norm": 0.42908936298576517, |
| "learning_rate": 8.986499938052396e-06, |
| "loss": 0.3432, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3396552503108978, |
| "step": 2690, |
| "valid_targets_mean": 5102.5, |
| "valid_targets_min": 3434 |
| }, |
| { |
| "epoch": 4.312, |
| "grad_norm": 0.4148684358509639, |
| "learning_rate": 8.908920262445859e-06, |
| "loss": 0.3421, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34144869446754456, |
| "step": 2695, |
| "valid_targets_mean": 5402.9, |
| "valid_targets_min": 2782 |
| }, |
| { |
| "epoch": 4.32, |
| "grad_norm": 0.42447500131216004, |
| "learning_rate": 8.831580837861082e-06, |
| "loss": 0.3385, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3340325653553009, |
| "step": 2700, |
| "valid_targets_mean": 5383.0, |
| "valid_targets_min": 3420 |
| }, |
| { |
| "epoch": 4.328, |
| "grad_norm": 0.4372036756530608, |
| "learning_rate": 8.754483339597166e-06, |
| "loss": 0.3458, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3363448977470398, |
| "step": 2705, |
| "valid_targets_mean": 4504.7, |
| "valid_targets_min": 1485 |
| }, |
| { |
| "epoch": 4.336, |
| "grad_norm": 0.4555122263985765, |
| "learning_rate": 8.677629437712665e-06, |
| "loss": 0.3383, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32118475437164307, |
| "step": 2710, |
| "valid_targets_mean": 4833.8, |
| "valid_targets_min": 1280 |
| }, |
| { |
| "epoch": 4.344, |
| "grad_norm": 0.4149086469223621, |
| "learning_rate": 8.601020796989467e-06, |
| "loss": 0.3453, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35315921902656555, |
| "step": 2715, |
| "valid_targets_mean": 5461.7, |
| "valid_targets_min": 2602 |
| }, |
| { |
| "epoch": 4.352, |
| "grad_norm": 0.4435557477075673, |
| "learning_rate": 8.524659076896656e-06, |
| "loss": 0.3436, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3623204827308655, |
| "step": 2720, |
| "valid_targets_mean": 5235.9, |
| "valid_targets_min": 3886 |
| }, |
| { |
| "epoch": 4.36, |
| "grad_norm": 0.37505229747997204, |
| "learning_rate": 8.448545931554652e-06, |
| "loss": 0.3391, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3214714527130127, |
| "step": 2725, |
| "valid_targets_mean": 6011.2, |
| "valid_targets_min": 3746 |
| }, |
| { |
| "epoch": 4.368, |
| "grad_norm": 0.43822655934463034, |
| "learning_rate": 8.372683009699307e-06, |
| "loss": 0.3441, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3378896713256836, |
| "step": 2730, |
| "valid_targets_mean": 5091.6, |
| "valid_targets_min": 2447 |
| }, |
| { |
| "epoch": 4.376, |
| "grad_norm": 0.4071844114631333, |
| "learning_rate": 8.297071954646248e-06, |
| "loss": 0.345, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33859431743621826, |
| "step": 2735, |
| "valid_targets_mean": 5246.2, |
| "valid_targets_min": 1960 |
| }, |
| { |
| "epoch": 4.384, |
| "grad_norm": 0.3997148655696296, |
| "learning_rate": 8.22171440425523e-06, |
| "loss": 0.3405, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3419719338417053, |
| "step": 2740, |
| "valid_targets_mean": 5360.1, |
| "valid_targets_min": 2472 |
| }, |
| { |
| "epoch": 4.392, |
| "grad_norm": 0.4441497428999854, |
| "learning_rate": 8.146611990894683e-06, |
| "loss": 0.3464, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34048211574554443, |
| "step": 2745, |
| "valid_targets_mean": 4843.6, |
| "valid_targets_min": 2864 |
| }, |
| { |
| "epoch": 4.4, |
| "grad_norm": 0.4343331057640048, |
| "learning_rate": 8.071766341406363e-06, |
| "loss": 0.341, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3464965224266052, |
| "step": 2750, |
| "valid_targets_mean": 5031.8, |
| "valid_targets_min": 2780 |
| }, |
| { |
| "epoch": 4.408, |
| "grad_norm": 0.40403499778520163, |
| "learning_rate": 7.997179077070092e-06, |
| "loss": 0.3396, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32712116837501526, |
| "step": 2755, |
| "valid_targets_mean": 5296.3, |
| "valid_targets_min": 3197 |
| }, |
| { |
| "epoch": 4.416, |
| "grad_norm": 0.455322315703816, |
| "learning_rate": 7.92285181356864e-06, |
| "loss": 0.3546, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34141358733177185, |
| "step": 2760, |
| "valid_targets_mean": 4961.4, |
| "valid_targets_min": 3605 |
| }, |
| { |
| "epoch": 4.424, |
| "grad_norm": 0.4292368218051709, |
| "learning_rate": 7.848786160952726e-06, |
| "loss": 0.3389, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33180510997772217, |
| "step": 2765, |
| "valid_targets_mean": 5140.6, |
| "valid_targets_min": 2121 |
| }, |
| { |
| "epoch": 4.432, |
| "grad_norm": 0.46757747415225315, |
| "learning_rate": 7.77498372360617e-06, |
| "loss": 0.333, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35177403688430786, |
| "step": 2770, |
| "valid_targets_mean": 4805.9, |
| "valid_targets_min": 2116 |
| }, |
| { |
| "epoch": 4.44, |
| "grad_norm": 0.3869878122428203, |
| "learning_rate": 7.701446100211095e-06, |
| "loss": 0.3341, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3173353970050812, |
| "step": 2775, |
| "valid_targets_mean": 5524.0, |
| "valid_targets_min": 2646 |
| }, |
| { |
| "epoch": 4.448, |
| "grad_norm": 0.4930829439512979, |
| "learning_rate": 7.628174883713322e-06, |
| "loss": 0.3425, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2954050302505493, |
| "step": 2780, |
| "valid_targets_mean": 5265.7, |
| "valid_targets_min": 1562 |
| }, |
| { |
| "epoch": 4.456, |
| "grad_norm": 0.39707382706871663, |
| "learning_rate": 7.555171661287875e-06, |
| "loss": 0.3401, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33590763807296753, |
| "step": 2785, |
| "valid_targets_mean": 6042.8, |
| "valid_targets_min": 3399 |
| }, |
| { |
| "epoch": 4.464, |
| "grad_norm": 0.4230374276638538, |
| "learning_rate": 7.482438014304567e-06, |
| "loss": 0.335, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34652984142303467, |
| "step": 2790, |
| "valid_targets_mean": 5315.8, |
| "valid_targets_min": 1957 |
| }, |
| { |
| "epoch": 4.4719999999999995, |
| "grad_norm": 0.44991921021306885, |
| "learning_rate": 7.4099755182937685e-06, |
| "loss": 0.3466, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3313753008842468, |
| "step": 2795, |
| "valid_targets_mean": 4746.2, |
| "valid_targets_min": 2949 |
| }, |
| { |
| "epoch": 4.48, |
| "grad_norm": 0.4523465443983067, |
| "learning_rate": 7.337785742912289e-06, |
| "loss": 0.3433, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3567259609699249, |
| "step": 2800, |
| "valid_targets_mean": 4966.8, |
| "valid_targets_min": 1653 |
| }, |
| { |
| "epoch": 4.4879999999999995, |
| "grad_norm": 0.482523966075641, |
| "learning_rate": 7.265870251909335e-06, |
| "loss": 0.3479, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3563911020755768, |
| "step": 2805, |
| "valid_targets_mean": 4374.0, |
| "valid_targets_min": 2278 |
| }, |
| { |
| "epoch": 4.496, |
| "grad_norm": 0.42094895293775425, |
| "learning_rate": 7.194230603092697e-06, |
| "loss": 0.339, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3318529427051544, |
| "step": 2810, |
| "valid_targets_mean": 5356.8, |
| "valid_targets_min": 2427 |
| }, |
| { |
| "epoch": 4.504, |
| "grad_norm": 0.42265403007187885, |
| "learning_rate": 7.122868348294927e-06, |
| "loss": 0.3388, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33271998167037964, |
| "step": 2815, |
| "valid_targets_mean": 5330.4, |
| "valid_targets_min": 2997 |
| }, |
| { |
| "epoch": 4.5120000000000005, |
| "grad_norm": 0.4266127970193255, |
| "learning_rate": 7.051785033339804e-06, |
| "loss": 0.3438, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3605421185493469, |
| "step": 2820, |
| "valid_targets_mean": 5375.0, |
| "valid_targets_min": 2082 |
| }, |
| { |
| "epoch": 4.52, |
| "grad_norm": 0.462563157244385, |
| "learning_rate": 6.980982198008785e-06, |
| "loss": 0.3371, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36290299892425537, |
| "step": 2825, |
| "valid_targets_mean": 4983.0, |
| "valid_targets_min": 3381 |
| }, |
| { |
| "epoch": 4.5280000000000005, |
| "grad_norm": 0.4315839383928626, |
| "learning_rate": 6.910461376007704e-06, |
| "loss": 0.3442, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3388533592224121, |
| "step": 2830, |
| "valid_targets_mean": 5020.9, |
| "valid_targets_min": 2911 |
| }, |
| { |
| "epoch": 4.536, |
| "grad_norm": 0.4295397939899924, |
| "learning_rate": 6.840224094933501e-06, |
| "loss": 0.3382, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3397209644317627, |
| "step": 2835, |
| "valid_targets_mean": 5689.8, |
| "valid_targets_min": 2310 |
| }, |
| { |
| "epoch": 4.5440000000000005, |
| "grad_norm": 0.4170631216405348, |
| "learning_rate": 6.7702718762411505e-06, |
| "loss": 0.3349, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3343551754951477, |
| "step": 2840, |
| "valid_targets_mean": 5557.9, |
| "valid_targets_min": 1754 |
| }, |
| { |
| "epoch": 4.552, |
| "grad_norm": 0.4517885270081739, |
| "learning_rate": 6.700606235210731e-06, |
| "loss": 0.3347, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35840559005737305, |
| "step": 2845, |
| "valid_targets_mean": 4803.7, |
| "valid_targets_min": 1528 |
| }, |
| { |
| "epoch": 4.5600000000000005, |
| "grad_norm": 0.3932310236336448, |
| "learning_rate": 6.631228680914558e-06, |
| "loss": 0.3282, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31834396719932556, |
| "step": 2850, |
| "valid_targets_mean": 5674.2, |
| "valid_targets_min": 3117 |
| }, |
| { |
| "epoch": 4.568, |
| "grad_norm": 0.4059544181591978, |
| "learning_rate": 6.562140716184515e-06, |
| "loss": 0.3358, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3435199558734894, |
| "step": 2855, |
| "valid_targets_mean": 5475.6, |
| "valid_targets_min": 3619 |
| }, |
| { |
| "epoch": 4.576, |
| "grad_norm": 0.4451688063550575, |
| "learning_rate": 6.493343837579511e-06, |
| "loss": 0.3483, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35447821021080017, |
| "step": 2860, |
| "valid_targets_mean": 5314.1, |
| "valid_targets_min": 1848 |
| }, |
| { |
| "epoch": 4.584, |
| "grad_norm": 0.4684663558838747, |
| "learning_rate": 6.424839535353045e-06, |
| "loss": 0.3377, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34784889221191406, |
| "step": 2865, |
| "valid_targets_mean": 4624.0, |
| "valid_targets_min": 2330 |
| }, |
| { |
| "epoch": 4.592, |
| "grad_norm": 0.43755096560108775, |
| "learning_rate": 6.356629293420926e-06, |
| "loss": 0.3332, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32967817783355713, |
| "step": 2870, |
| "valid_targets_mean": 4628.1, |
| "valid_targets_min": 2412 |
| }, |
| { |
| "epoch": 4.6, |
| "grad_norm": 0.4250819161634356, |
| "learning_rate": 6.28871458932913e-06, |
| "loss": 0.3355, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3466927409172058, |
| "step": 2875, |
| "valid_targets_mean": 5005.2, |
| "valid_targets_min": 2235 |
| }, |
| { |
| "epoch": 4.608, |
| "grad_norm": 0.43930760572460914, |
| "learning_rate": 6.2210968942218206e-06, |
| "loss": 0.3529, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36589741706848145, |
| "step": 2880, |
| "valid_targets_mean": 5353.2, |
| "valid_targets_min": 2245 |
| }, |
| { |
| "epoch": 4.616, |
| "grad_norm": 0.4144766134536535, |
| "learning_rate": 6.153777672809438e-06, |
| "loss": 0.3444, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3390125036239624, |
| "step": 2885, |
| "valid_targets_mean": 5239.1, |
| "valid_targets_min": 3500 |
| }, |
| { |
| "epoch": 4.624, |
| "grad_norm": 0.4195309742877588, |
| "learning_rate": 6.086758383336984e-06, |
| "loss": 0.3473, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3362986147403717, |
| "step": 2890, |
| "valid_targets_mean": 5146.6, |
| "valid_targets_min": 1468 |
| }, |
| { |
| "epoch": 4.632, |
| "grad_norm": 0.4639884860136478, |
| "learning_rate": 6.0200404775524715e-06, |
| "loss": 0.33, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3373122215270996, |
| "step": 2895, |
| "valid_targets_mean": 5300.8, |
| "valid_targets_min": 2327 |
| }, |
| { |
| "epoch": 4.64, |
| "grad_norm": 0.45382665243137615, |
| "learning_rate": 5.9536254006754155e-06, |
| "loss": 0.3507, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34784841537475586, |
| "step": 2900, |
| "valid_targets_mean": 5321.5, |
| "valid_targets_min": 3978 |
| }, |
| { |
| "epoch": 4.648, |
| "grad_norm": 0.46084097471225915, |
| "learning_rate": 5.887514591365593e-06, |
| "loss": 0.3525, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36149275302886963, |
| "step": 2905, |
| "valid_targets_mean": 4977.4, |
| "valid_targets_min": 2192 |
| }, |
| { |
| "epoch": 4.656, |
| "grad_norm": 0.4089187550531159, |
| "learning_rate": 5.821709481691798e-06, |
| "loss": 0.3295, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31506332755088806, |
| "step": 2910, |
| "valid_targets_mean": 5682.9, |
| "valid_targets_min": 2859 |
| }, |
| { |
| "epoch": 4.664, |
| "grad_norm": 0.4480481539993405, |
| "learning_rate": 5.75621149710091e-06, |
| "loss": 0.3421, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35835838317871094, |
| "step": 2915, |
| "valid_targets_mean": 5031.6, |
| "valid_targets_min": 2398 |
| }, |
| { |
| "epoch": 4.672, |
| "grad_norm": 0.41794722900175096, |
| "learning_rate": 5.691022056386961e-06, |
| "loss": 0.3316, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31518641114234924, |
| "step": 2920, |
| "valid_targets_mean": 4875.4, |
| "valid_targets_min": 2539 |
| }, |
| { |
| "epoch": 4.68, |
| "grad_norm": 0.4519718447657964, |
| "learning_rate": 5.6261425716604136e-06, |
| "loss": 0.3385, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3236932158470154, |
| "step": 2925, |
| "valid_targets_mean": 4779.7, |
| "valid_targets_min": 2029 |
| }, |
| { |
| "epoch": 4.688, |
| "grad_norm": 0.47607631242056503, |
| "learning_rate": 5.56157444831757e-06, |
| "loss": 0.3467, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3667972683906555, |
| "step": 2930, |
| "valid_targets_mean": 4450.4, |
| "valid_targets_min": 1599 |
| }, |
| { |
| "epoch": 4.696, |
| "grad_norm": 0.40825427127361363, |
| "learning_rate": 5.4973190850101334e-06, |
| "loss": 0.3414, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3439498245716095, |
| "step": 2935, |
| "valid_targets_mean": 5499.9, |
| "valid_targets_min": 2961 |
| }, |
| { |
| "epoch": 4.704, |
| "grad_norm": 0.6336087943351952, |
| "learning_rate": 5.433377873614925e-06, |
| "loss": 0.3425, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34850215911865234, |
| "step": 2940, |
| "valid_targets_mean": 4769.9, |
| "valid_targets_min": 3052 |
| }, |
| { |
| "epoch": 4.712, |
| "grad_norm": 0.4257858163844753, |
| "learning_rate": 5.369752199203702e-06, |
| "loss": 0.3453, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.357036292552948, |
| "step": 2945, |
| "valid_targets_mean": 5721.9, |
| "valid_targets_min": 3712 |
| }, |
| { |
| "epoch": 4.72, |
| "grad_norm": 0.4188351519357239, |
| "learning_rate": 5.306443440013171e-06, |
| "loss": 0.3407, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3333517014980316, |
| "step": 2950, |
| "valid_targets_mean": 5174.9, |
| "valid_targets_min": 3042 |
| }, |
| { |
| "epoch": 4.728, |
| "grad_norm": 0.4001929726754892, |
| "learning_rate": 5.243452967415155e-06, |
| "loss": 0.3372, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3302866220474243, |
| "step": 2955, |
| "valid_targets_mean": 5438.0, |
| "valid_targets_min": 3074 |
| }, |
| { |
| "epoch": 4.736, |
| "grad_norm": 0.47203382510856906, |
| "learning_rate": 5.180782145886846e-06, |
| "loss": 0.3488, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34411707520484924, |
| "step": 2960, |
| "valid_targets_mean": 4999.2, |
| "valid_targets_min": 2989 |
| }, |
| { |
| "epoch": 4.744, |
| "grad_norm": 0.4550716248013781, |
| "learning_rate": 5.118432332981273e-06, |
| "loss": 0.3479, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35314249992370605, |
| "step": 2965, |
| "valid_targets_mean": 5829.4, |
| "valid_targets_min": 3356 |
| }, |
| { |
| "epoch": 4.752, |
| "grad_norm": 0.5680722390477754, |
| "learning_rate": 5.056404879297887e-06, |
| "loss": 0.3435, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33602648973464966, |
| "step": 2970, |
| "valid_targets_mean": 4552.8, |
| "valid_targets_min": 2142 |
| }, |
| { |
| "epoch": 4.76, |
| "grad_norm": 0.4457215079955225, |
| "learning_rate": 4.994701128453325e-06, |
| "loss": 0.3376, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34097105264663696, |
| "step": 2975, |
| "valid_targets_mean": 5039.1, |
| "valid_targets_min": 2723 |
| }, |
| { |
| "epoch": 4.768, |
| "grad_norm": 0.4316979469769324, |
| "learning_rate": 4.933322417052269e-06, |
| "loss": 0.3362, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3383985161781311, |
| "step": 2980, |
| "valid_targets_mean": 5467.0, |
| "valid_targets_min": 3350 |
| }, |
| { |
| "epoch": 4.776, |
| "grad_norm": 0.4400319258814758, |
| "learning_rate": 4.8722700746585135e-06, |
| "loss": 0.3419, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35413968563079834, |
| "step": 2985, |
| "valid_targets_mean": 5308.8, |
| "valid_targets_min": 2985 |
| }, |
| { |
| "epoch": 4.784, |
| "grad_norm": 0.4175790907420823, |
| "learning_rate": 4.811545423766184e-06, |
| "loss": 0.3402, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3484022915363312, |
| "step": 2990, |
| "valid_targets_mean": 5538.1, |
| "valid_targets_min": 2809 |
| }, |
| { |
| "epoch": 4.792, |
| "grad_norm": 0.44908397928819016, |
| "learning_rate": 4.75114977977104e-06, |
| "loss": 0.3413, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35889238119125366, |
| "step": 2995, |
| "valid_targets_mean": 5153.4, |
| "valid_targets_min": 3031 |
| }, |
| { |
| "epoch": 4.8, |
| "grad_norm": 0.44695349658875827, |
| "learning_rate": 4.691084450942047e-06, |
| "loss": 0.3255, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3283981680870056, |
| "step": 3000, |
| "valid_targets_mean": 5413.8, |
| "valid_targets_min": 3613 |
| }, |
| { |
| "epoch": 4.808, |
| "grad_norm": 0.45777245915222337, |
| "learning_rate": 4.631350738392955e-06, |
| "loss": 0.3477, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36985111236572266, |
| "step": 3005, |
| "valid_targets_mean": 4923.0, |
| "valid_targets_min": 3070 |
| }, |
| { |
| "epoch": 4.816, |
| "grad_norm": 0.48730862029313277, |
| "learning_rate": 4.571949936054197e-06, |
| "loss": 0.3539, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3493921160697937, |
| "step": 3010, |
| "valid_targets_mean": 4505.7, |
| "valid_targets_min": 1715 |
| }, |
| { |
| "epoch": 4.824, |
| "grad_norm": 0.45860105820398944, |
| "learning_rate": 4.512883330644815e-06, |
| "loss": 0.3474, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3336547017097473, |
| "step": 3015, |
| "valid_targets_mean": 4725.6, |
| "valid_targets_min": 2016 |
| }, |
| { |
| "epoch": 4.832, |
| "grad_norm": 0.4059212694144429, |
| "learning_rate": 4.454152201644591e-06, |
| "loss": 0.3463, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3586534261703491, |
| "step": 3020, |
| "valid_targets_mean": 5550.7, |
| "valid_targets_min": 3995 |
| }, |
| { |
| "epoch": 4.84, |
| "grad_norm": 0.4314362146670214, |
| "learning_rate": 4.395757821266333e-06, |
| "loss": 0.3339, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34733009338378906, |
| "step": 3025, |
| "valid_targets_mean": 4733.9, |
| "valid_targets_min": 3248 |
| }, |
| { |
| "epoch": 4.848, |
| "grad_norm": 0.4312884228428965, |
| "learning_rate": 4.337701454428322e-06, |
| "loss": 0.3455, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36343497037887573, |
| "step": 3030, |
| "valid_targets_mean": 5482.3, |
| "valid_targets_min": 1704 |
| }, |
| { |
| "epoch": 4.856, |
| "grad_norm": 0.47332787710939644, |
| "learning_rate": 4.279984358726925e-06, |
| "loss": 0.3347, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37544023990631104, |
| "step": 3035, |
| "valid_targets_mean": 4763.6, |
| "valid_targets_min": 1540 |
| }, |
| { |
| "epoch": 4.864, |
| "grad_norm": 0.4266282371228391, |
| "learning_rate": 4.2226077844093205e-06, |
| "loss": 0.3436, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33394739031791687, |
| "step": 3040, |
| "valid_targets_mean": 5358.6, |
| "valid_targets_min": 1740 |
| }, |
| { |
| "epoch": 4.872, |
| "grad_norm": 0.42908213889086566, |
| "learning_rate": 4.165572974346435e-06, |
| "loss": 0.3443, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3351410925388336, |
| "step": 3045, |
| "valid_targets_mean": 5059.8, |
| "valid_targets_min": 1700 |
| }, |
| { |
| "epoch": 4.88, |
| "grad_norm": 0.4722849165601408, |
| "learning_rate": 4.108881164006033e-06, |
| "loss": 0.335, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3244861960411072, |
| "step": 3050, |
| "valid_targets_mean": 4402.8, |
| "valid_targets_min": 1694 |
| }, |
| { |
| "epoch": 4.888, |
| "grad_norm": 0.41132931755077545, |
| "learning_rate": 4.05253358142593e-06, |
| "loss": 0.3422, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3330739140510559, |
| "step": 3055, |
| "valid_targets_mean": 5242.5, |
| "valid_targets_min": 2427 |
| }, |
| { |
| "epoch": 4.896, |
| "grad_norm": 0.4226283526620901, |
| "learning_rate": 3.9965314471874035e-06, |
| "loss": 0.3438, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3348574638366699, |
| "step": 3060, |
| "valid_targets_mean": 5464.9, |
| "valid_targets_min": 2156 |
| }, |
| { |
| "epoch": 4.904, |
| "grad_norm": 0.4246285967034345, |
| "learning_rate": 3.940875974388749e-06, |
| "loss": 0.3388, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3279423415660858, |
| "step": 3065, |
| "valid_targets_mean": 4891.8, |
| "valid_targets_min": 1448 |
| }, |
| { |
| "epoch": 4.912, |
| "grad_norm": 0.4424935546768732, |
| "learning_rate": 3.885568368619013e-06, |
| "loss": 0.3367, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34123337268829346, |
| "step": 3070, |
| "valid_targets_mean": 5148.7, |
| "valid_targets_min": 1665 |
| }, |
| { |
| "epoch": 4.92, |
| "grad_norm": 0.40371364466737, |
| "learning_rate": 3.830609827931877e-06, |
| "loss": 0.3409, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34550178050994873, |
| "step": 3075, |
| "valid_targets_mean": 5743.9, |
| "valid_targets_min": 2022 |
| }, |
| { |
| "epoch": 4.928, |
| "grad_norm": 0.3996534126721198, |
| "learning_rate": 3.7760015428196694e-06, |
| "loss": 0.3234, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3405061960220337, |
| "step": 3080, |
| "valid_targets_mean": 6064.2, |
| "valid_targets_min": 3202 |
| }, |
| { |
| "epoch": 4.936, |
| "grad_norm": 0.4055299751437052, |
| "learning_rate": 3.7217446961876413e-06, |
| "loss": 0.3408, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32878968119621277, |
| "step": 3085, |
| "valid_targets_mean": 5562.4, |
| "valid_targets_min": 1580 |
| }, |
| { |
| "epoch": 4.944, |
| "grad_norm": 0.4181909911730216, |
| "learning_rate": 3.6678404633282826e-06, |
| "loss": 0.3459, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33040690422058105, |
| "step": 3090, |
| "valid_targets_mean": 4960.6, |
| "valid_targets_min": 2980 |
| }, |
| { |
| "epoch": 4.952, |
| "grad_norm": 0.39690077382738215, |
| "learning_rate": 3.6142900118959158e-06, |
| "loss": 0.3412, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.333196222782135, |
| "step": 3095, |
| "valid_targets_mean": 5628.4, |
| "valid_targets_min": 3818 |
| }, |
| { |
| "epoch": 4.96, |
| "grad_norm": 0.4223495412899275, |
| "learning_rate": 3.561094501881339e-06, |
| "loss": 0.3438, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3505244255065918, |
| "step": 3100, |
| "valid_targets_mean": 5111.3, |
| "valid_targets_min": 3194 |
| }, |
| { |
| "epoch": 4.968, |
| "grad_norm": 0.44224825626361164, |
| "learning_rate": 3.5082550855867693e-06, |
| "loss": 0.3492, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3330351710319519, |
| "step": 3105, |
| "valid_targets_mean": 4783.5, |
| "valid_targets_min": 1819 |
| }, |
| { |
| "epoch": 4.976, |
| "grad_norm": 0.431986746484784, |
| "learning_rate": 3.455772907600841e-06, |
| "loss": 0.3443, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3432115316390991, |
| "step": 3110, |
| "valid_targets_mean": 4827.2, |
| "valid_targets_min": 2792 |
| }, |
| { |
| "epoch": 4.984, |
| "grad_norm": 0.42667723199101093, |
| "learning_rate": 3.4036491047738075e-06, |
| "loss": 0.3464, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3512681722640991, |
| "step": 3115, |
| "valid_targets_mean": 5096.8, |
| "valid_targets_min": 1483 |
| }, |
| { |
| "epoch": 4.992, |
| "grad_norm": 0.45459305786249576, |
| "learning_rate": 3.351884806192933e-06, |
| "loss": 0.3415, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.347655713558197, |
| "step": 3120, |
| "valid_targets_mean": 5322.4, |
| "valid_targets_min": 1738 |
| }, |
| { |
| "epoch": 5.0, |
| "grad_norm": 0.42328532457656726, |
| "learning_rate": 3.3004811331580268e-06, |
| "loss": 0.3365, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3429858684539795, |
| "step": 3125, |
| "valid_targets_mean": 5141.6, |
| "valid_targets_min": 2704 |
| }, |
| { |
| "epoch": 5.008, |
| "grad_norm": 0.42901308239886793, |
| "learning_rate": 3.249439199157167e-06, |
| "loss": 0.3363, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3417869210243225, |
| "step": 3130, |
| "valid_targets_mean": 5391.9, |
| "valid_targets_min": 2202 |
| }, |
| { |
| "epoch": 5.016, |
| "grad_norm": 0.4260420348876251, |
| "learning_rate": 3.198760109842558e-06, |
| "loss": 0.3238, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33276575803756714, |
| "step": 3135, |
| "valid_targets_mean": 5223.7, |
| "valid_targets_min": 3576 |
| }, |
| { |
| "epoch": 5.024, |
| "grad_norm": 0.4408675885387519, |
| "learning_rate": 3.1484449630065894e-06, |
| "loss": 0.3337, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32523873448371887, |
| "step": 3140, |
| "valid_targets_mean": 4703.8, |
| "valid_targets_min": 2436 |
| }, |
| { |
| "epoch": 5.032, |
| "grad_norm": 0.4405975436295544, |
| "learning_rate": 3.0984948485580736e-06, |
| "loss": 0.3294, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3462005853652954, |
| "step": 3145, |
| "valid_targets_mean": 5158.6, |
| "valid_targets_min": 3611 |
| }, |
| { |
| "epoch": 5.04, |
| "grad_norm": 0.5693774880949125, |
| "learning_rate": 3.048910848498605e-06, |
| "loss": 0.3382, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3288172483444214, |
| "step": 3150, |
| "valid_targets_mean": 5607.6, |
| "valid_targets_min": 2552 |
| }, |
| { |
| "epoch": 5.048, |
| "grad_norm": 0.46949876746611796, |
| "learning_rate": 2.9996940368991477e-06, |
| "loss": 0.3227, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29781395196914673, |
| "step": 3155, |
| "valid_targets_mean": 4516.3, |
| "valid_targets_min": 1999 |
| }, |
| { |
| "epoch": 5.056, |
| "grad_norm": 0.4540981739707222, |
| "learning_rate": 2.9508454798767516e-06, |
| "loss": 0.3325, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3395349085330963, |
| "step": 3160, |
| "valid_targets_mean": 4758.6, |
| "valid_targets_min": 3420 |
| }, |
| { |
| "epoch": 5.064, |
| "grad_norm": 0.45632001933391947, |
| "learning_rate": 2.9023662355714766e-06, |
| "loss": 0.3398, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35087907314300537, |
| "step": 3165, |
| "valid_targets_mean": 5128.9, |
| "valid_targets_min": 1826 |
| }, |
| { |
| "epoch": 5.072, |
| "grad_norm": 0.5095903464942326, |
| "learning_rate": 2.8542573541234707e-06, |
| "loss": 0.3366, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37809988856315613, |
| "step": 3170, |
| "valid_targets_mean": 4477.9, |
| "valid_targets_min": 2625 |
| }, |
| { |
| "epoch": 5.08, |
| "grad_norm": 0.4631906138996648, |
| "learning_rate": 2.80651987765018e-06, |
| "loss": 0.3338, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3382030129432678, |
| "step": 3175, |
| "valid_targets_mean": 4418.7, |
| "valid_targets_min": 2559 |
| }, |
| { |
| "epoch": 5.088, |
| "grad_norm": 0.42944950286845207, |
| "learning_rate": 2.759154840223843e-06, |
| "loss": 0.3302, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3201037645339966, |
| "step": 3180, |
| "valid_targets_mean": 5604.8, |
| "valid_targets_min": 2712 |
| }, |
| { |
| "epoch": 5.096, |
| "grad_norm": 0.4697840819263023, |
| "learning_rate": 2.7121632678490327e-06, |
| "loss": 0.347, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3684273958206177, |
| "step": 3185, |
| "valid_targets_mean": 5030.7, |
| "valid_targets_min": 1726 |
| }, |
| { |
| "epoch": 5.104, |
| "grad_norm": 0.39839274047490786, |
| "learning_rate": 2.6655461784404768e-06, |
| "loss": 0.3281, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30260369181632996, |
| "step": 3190, |
| "valid_targets_mean": 5406.9, |
| "valid_targets_min": 2982 |
| }, |
| { |
| "epoch": 5.112, |
| "grad_norm": 0.4290841031387009, |
| "learning_rate": 2.6193045818009654e-06, |
| "loss": 0.3304, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33319029211997986, |
| "step": 3195, |
| "valid_targets_mean": 4860.5, |
| "valid_targets_min": 2359 |
| }, |
| { |
| "epoch": 5.12, |
| "grad_norm": 0.41492477955573337, |
| "learning_rate": 2.5734394795995066e-06, |
| "loss": 0.3323, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3239573538303375, |
| "step": 3200, |
| "valid_targets_mean": 5293.9, |
| "valid_targets_min": 3077 |
| }, |
| { |
| "epoch": 5.128, |
| "grad_norm": 0.44835767237189234, |
| "learning_rate": 2.5279518653496272e-06, |
| "loss": 0.3439, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35685932636260986, |
| "step": 3205, |
| "valid_targets_mean": 5400.4, |
| "valid_targets_min": 2999 |
| }, |
| { |
| "epoch": 5.136, |
| "grad_norm": 0.4306740454907328, |
| "learning_rate": 2.4828427243878307e-06, |
| "loss": 0.3446, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33979862928390503, |
| "step": 3210, |
| "valid_targets_mean": 5140.8, |
| "valid_targets_min": 3029 |
| }, |
| { |
| "epoch": 5.144, |
| "grad_norm": 0.40333852322383756, |
| "learning_rate": 2.4381130338522762e-06, |
| "loss": 0.331, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3204619884490967, |
| "step": 3215, |
| "valid_targets_mean": 5399.6, |
| "valid_targets_min": 3262 |
| }, |
| { |
| "epoch": 5.152, |
| "grad_norm": 0.6147130315217775, |
| "learning_rate": 2.393763762661596e-06, |
| "loss": 0.3298, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3417896628379822, |
| "step": 3220, |
| "valid_targets_mean": 5373.1, |
| "valid_targets_min": 2192 |
| }, |
| { |
| "epoch": 5.16, |
| "grad_norm": 0.42715298177641725, |
| "learning_rate": 2.349795871493925e-06, |
| "loss": 0.3305, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32653748989105225, |
| "step": 3225, |
| "valid_targets_mean": 5330.1, |
| "valid_targets_min": 2083 |
| }, |
| { |
| "epoch": 5.168, |
| "grad_norm": 0.38590516643672373, |
| "learning_rate": 2.3062103127660686e-06, |
| "loss": 0.3359, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2938699722290039, |
| "step": 3230, |
| "valid_targets_mean": 5427.9, |
| "valid_targets_min": 2720 |
| }, |
| { |
| "epoch": 5.176, |
| "grad_norm": 0.42113088597857545, |
| "learning_rate": 2.2630080306128833e-06, |
| "loss": 0.3362, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3201785087585449, |
| "step": 3235, |
| "valid_targets_mean": 5269.2, |
| "valid_targets_min": 2977 |
| }, |
| { |
| "epoch": 5.184, |
| "grad_norm": 0.4504212617119637, |
| "learning_rate": 2.2201899608668365e-06, |
| "loss": 0.349, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3692011535167694, |
| "step": 3240, |
| "valid_targets_mean": 5372.2, |
| "valid_targets_min": 2572 |
| }, |
| { |
| "epoch": 5.192, |
| "grad_norm": 0.39345202632833703, |
| "learning_rate": 2.1777570310377084e-06, |
| "loss": 0.3309, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3296594023704529, |
| "step": 3245, |
| "valid_targets_mean": 6190.1, |
| "valid_targets_min": 3662 |
| }, |
| { |
| "epoch": 5.2, |
| "grad_norm": 0.45006063103269195, |
| "learning_rate": 2.1357101602925323e-06, |
| "loss": 0.3373, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34597110748291016, |
| "step": 3250, |
| "valid_targets_mean": 4728.8, |
| "valid_targets_min": 1854 |
| }, |
| { |
| "epoch": 5.208, |
| "grad_norm": 0.410840610702434, |
| "learning_rate": 2.0940502594356427e-06, |
| "loss": 0.3384, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3230043649673462, |
| "step": 3255, |
| "valid_targets_mean": 5769.7, |
| "valid_targets_min": 3432 |
| }, |
| { |
| "epoch": 5.216, |
| "grad_norm": 0.4194223804825284, |
| "learning_rate": 2.052778230888994e-06, |
| "loss": 0.341, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35491862893104553, |
| "step": 3260, |
| "valid_targets_mean": 5402.6, |
| "valid_targets_min": 3006 |
| }, |
| { |
| "epoch": 5.224, |
| "grad_norm": 0.42904255574296907, |
| "learning_rate": 2.0118949686725786e-06, |
| "loss": 0.329, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3379438519477844, |
| "step": 3265, |
| "valid_targets_mean": 5162.3, |
| "valid_targets_min": 3253 |
| }, |
| { |
| "epoch": 5.232, |
| "grad_norm": 0.4610036531252929, |
| "learning_rate": 1.971401358385072e-06, |
| "loss": 0.3334, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34576213359832764, |
| "step": 3270, |
| "valid_targets_mean": 4440.6, |
| "valid_targets_min": 1715 |
| }, |
| { |
| "epoch": 5.24, |
| "grad_norm": 0.4649416211831189, |
| "learning_rate": 1.9312982771846435e-06, |
| "loss": 0.339, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3332167863845825, |
| "step": 3275, |
| "valid_targets_mean": 5475.2, |
| "valid_targets_min": 2901 |
| }, |
| { |
| "epoch": 5.248, |
| "grad_norm": 0.4838663899585091, |
| "learning_rate": 1.8915865937699652e-06, |
| "loss": 0.3207, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32056039571762085, |
| "step": 3280, |
| "valid_targets_mean": 5782.2, |
| "valid_targets_min": 3351 |
| }, |
| { |
| "epoch": 5.256, |
| "grad_norm": 0.4302361357428255, |
| "learning_rate": 1.8522671683613946e-06, |
| "loss": 0.3307, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3357390761375427, |
| "step": 3285, |
| "valid_targets_mean": 5174.5, |
| "valid_targets_min": 2042 |
| }, |
| { |
| "epoch": 5.264, |
| "grad_norm": 0.40975263187293187, |
| "learning_rate": 1.8133408526823283e-06, |
| "loss": 0.3352, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.335743248462677, |
| "step": 3290, |
| "valid_targets_mean": 5680.5, |
| "valid_targets_min": 1775 |
| }, |
| { |
| "epoch": 5.272, |
| "grad_norm": 0.43590281536820424, |
| "learning_rate": 1.7748084899407558e-06, |
| "loss": 0.3332, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33717116713523865, |
| "step": 3295, |
| "valid_targets_mean": 5221.1, |
| "valid_targets_min": 2330 |
| }, |
| { |
| "epoch": 5.28, |
| "grad_norm": 0.4386100154019209, |
| "learning_rate": 1.7366709148110118e-06, |
| "loss": 0.3318, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.334185391664505, |
| "step": 3300, |
| "valid_targets_mean": 5279.2, |
| "valid_targets_min": 2876 |
| }, |
| { |
| "epoch": 5.288, |
| "grad_norm": 0.43370175874435607, |
| "learning_rate": 1.698928953415675e-06, |
| "loss": 0.3393, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.343210369348526, |
| "step": 3305, |
| "valid_targets_mean": 5037.2, |
| "valid_targets_min": 2009 |
| }, |
| { |
| "epoch": 5.296, |
| "grad_norm": 0.4663989553416844, |
| "learning_rate": 1.6615834233076756e-06, |
| "loss": 0.3256, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3293733596801758, |
| "step": 3310, |
| "valid_targets_mean": 4645.7, |
| "valid_targets_min": 2048 |
| }, |
| { |
| "epoch": 5.304, |
| "grad_norm": 0.39861935934082426, |
| "learning_rate": 1.6246351334525944e-06, |
| "loss": 0.3393, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32675302028656006, |
| "step": 3315, |
| "valid_targets_mean": 5744.1, |
| "valid_targets_min": 3102 |
| }, |
| { |
| "epoch": 5.312, |
| "grad_norm": 0.414449341067629, |
| "learning_rate": 1.5880848842111362e-06, |
| "loss": 0.3329, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3292161226272583, |
| "step": 3320, |
| "valid_targets_mean": 5460.0, |
| "valid_targets_min": 3156 |
| }, |
| { |
| "epoch": 5.32, |
| "grad_norm": 0.4483666526489511, |
| "learning_rate": 1.5519334673218023e-06, |
| "loss": 0.3383, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34213170409202576, |
| "step": 3325, |
| "valid_targets_mean": 4839.4, |
| "valid_targets_min": 2443 |
| }, |
| { |
| "epoch": 5.328, |
| "grad_norm": 0.3998701056137738, |
| "learning_rate": 1.5161816658837002e-06, |
| "loss": 0.3313, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3217969536781311, |
| "step": 3330, |
| "valid_targets_mean": 5386.6, |
| "valid_targets_min": 3684 |
| }, |
| { |
| "epoch": 5.336, |
| "grad_norm": 0.5526444877716085, |
| "learning_rate": 1.4808302543396423e-06, |
| "loss": 0.341, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35861143469810486, |
| "step": 3335, |
| "valid_targets_mean": 4977.2, |
| "valid_targets_min": 2658 |
| }, |
| { |
| "epoch": 5.344, |
| "grad_norm": 0.4354406527246741, |
| "learning_rate": 1.445879998459314e-06, |
| "loss": 0.3372, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.317405104637146, |
| "step": 3340, |
| "valid_targets_mean": 4936.8, |
| "valid_targets_min": 1841 |
| }, |
| { |
| "epoch": 5.352, |
| "grad_norm": 0.37861489851542923, |
| "learning_rate": 1.4113316553227296e-06, |
| "loss": 0.3415, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30580228567123413, |
| "step": 3345, |
| "valid_targets_mean": 5747.1, |
| "valid_targets_min": 4342 |
| }, |
| { |
| "epoch": 5.36, |
| "grad_norm": 0.4811114265901374, |
| "learning_rate": 1.3771859733037896e-06, |
| "loss": 0.3358, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3479752540588379, |
| "step": 3350, |
| "valid_targets_mean": 4715.6, |
| "valid_targets_min": 2566 |
| }, |
| { |
| "epoch": 5.368, |
| "grad_norm": 0.4357664605682178, |
| "learning_rate": 1.3434436920541072e-06, |
| "loss": 0.3422, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35315829515457153, |
| "step": 3355, |
| "valid_targets_mean": 5134.8, |
| "valid_targets_min": 3655 |
| }, |
| { |
| "epoch": 5.376, |
| "grad_norm": 0.4298990977461586, |
| "learning_rate": 1.3101055424869768e-06, |
| "loss": 0.3305, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32195818424224854, |
| "step": 3360, |
| "valid_targets_mean": 5151.8, |
| "valid_targets_min": 3079 |
| }, |
| { |
| "epoch": 5.384, |
| "grad_norm": 0.44250892332451125, |
| "learning_rate": 1.2771722467615266e-06, |
| "loss": 0.3373, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34324002265930176, |
| "step": 3365, |
| "valid_targets_mean": 4999.4, |
| "valid_targets_min": 2239 |
| }, |
| { |
| "epoch": 5.392, |
| "grad_norm": 0.4937466929385468, |
| "learning_rate": 1.2446445182670818e-06, |
| "loss": 0.3429, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34580278396606445, |
| "step": 3370, |
| "valid_targets_mean": 4824.5, |
| "valid_targets_min": 3223 |
| }, |
| { |
| "epoch": 5.4, |
| "grad_norm": 0.49044996996004026, |
| "learning_rate": 1.21252306160772e-06, |
| "loss": 0.3287, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33593958616256714, |
| "step": 3375, |
| "valid_targets_mean": 5087.5, |
| "valid_targets_min": 3117 |
| }, |
| { |
| "epoch": 5.408, |
| "grad_norm": 0.40609643033822934, |
| "learning_rate": 1.1808085725870088e-06, |
| "loss": 0.3279, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3253338038921356, |
| "step": 3380, |
| "valid_targets_mean": 5542.4, |
| "valid_targets_min": 2134 |
| }, |
| { |
| "epoch": 5.416, |
| "grad_norm": 0.4000751197765485, |
| "learning_rate": 1.1495017381929197e-06, |
| "loss": 0.3427, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.325018972158432, |
| "step": 3385, |
| "valid_targets_mean": 5634.6, |
| "valid_targets_min": 2858 |
| }, |
| { |
| "epoch": 5.424, |
| "grad_norm": 0.4465052541179308, |
| "learning_rate": 1.1186032365829514e-06, |
| "loss": 0.3387, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3391878604888916, |
| "step": 3390, |
| "valid_targets_mean": 4559.3, |
| "valid_targets_min": 2104 |
| }, |
| { |
| "epoch": 5.432, |
| "grad_norm": 0.44942703799888495, |
| "learning_rate": 1.088113737069456e-06, |
| "loss": 0.3299, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34843721985816956, |
| "step": 3395, |
| "valid_targets_mean": 4866.8, |
| "valid_targets_min": 2029 |
| }, |
| { |
| "epoch": 5.44, |
| "grad_norm": 0.43457035557377144, |
| "learning_rate": 1.0580339001051153e-06, |
| "loss": 0.3478, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35081779956817627, |
| "step": 3400, |
| "valid_targets_mean": 5510.8, |
| "valid_targets_min": 3243 |
| }, |
| { |
| "epoch": 5.448, |
| "grad_norm": 0.44909758534896727, |
| "learning_rate": 1.0283643772686535e-06, |
| "loss": 0.35, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35025009512901306, |
| "step": 3405, |
| "valid_targets_mean": 5027.0, |
| "valid_targets_min": 1858 |
| }, |
| { |
| "epoch": 5.456, |
| "grad_norm": 0.41998770573331656, |
| "learning_rate": 9.991058112507113e-07, |
| "loss": 0.3344, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3351707458496094, |
| "step": 3410, |
| "valid_targets_mean": 5479.9, |
| "valid_targets_min": 2278 |
| }, |
| { |
| "epoch": 5.464, |
| "grad_norm": 0.4239706844384994, |
| "learning_rate": 9.702588358399345e-07, |
| "loss": 0.3452, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3386349081993103, |
| "step": 3415, |
| "valid_targets_mean": 5666.3, |
| "valid_targets_min": 1862 |
| }, |
| { |
| "epoch": 5.4719999999999995, |
| "grad_norm": 0.40756673803240767, |
| "learning_rate": 9.418240759092434e-07, |
| "loss": 0.3297, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32479339838027954, |
| "step": 3420, |
| "valid_targets_mean": 5275.1, |
| "valid_targets_min": 2618 |
| }, |
| { |
| "epoch": 5.48, |
| "grad_norm": 0.43323981728956723, |
| "learning_rate": 9.138021474022763e-07, |
| "loss": 0.3322, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3363816440105438, |
| "step": 3425, |
| "valid_targets_mean": 5225.4, |
| "valid_targets_min": 2591 |
| }, |
| { |
| "epoch": 5.4879999999999995, |
| "grad_norm": 0.42414795356970714, |
| "learning_rate": 8.861936573200825e-07, |
| "loss": 0.3384, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33706793189048767, |
| "step": 3430, |
| "valid_targets_mean": 5631.1, |
| "valid_targets_min": 3621 |
| }, |
| { |
| "epoch": 5.496, |
| "grad_norm": 0.38589520593550974, |
| "learning_rate": 8.58999203707942e-07, |
| "loss": 0.3408, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3141070604324341, |
| "step": 3435, |
| "valid_targets_mean": 5707.1, |
| "valid_targets_min": 2130 |
| }, |
| { |
| "epoch": 5.504, |
| "grad_norm": 0.4467140030228506, |
| "learning_rate": 8.322193756424401e-07, |
| "loss": 0.3487, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33098894357681274, |
| "step": 3440, |
| "valid_targets_mean": 5112.6, |
| "valid_targets_min": 1763 |
| }, |
| { |
| "epoch": 5.5120000000000005, |
| "grad_norm": 0.4103877467060428, |
| "learning_rate": 8.058547532186667e-07, |
| "loss": 0.3296, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3255775272846222, |
| "step": 3445, |
| "valid_targets_mean": 5470.9, |
| "valid_targets_min": 2179 |
| }, |
| { |
| "epoch": 5.52, |
| "grad_norm": 0.4164193127019014, |
| "learning_rate": 7.799059075376991e-07, |
| "loss": 0.3347, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3454502522945404, |
| "step": 3450, |
| "valid_targets_mean": 5424.9, |
| "valid_targets_min": 3481 |
| }, |
| { |
| "epoch": 5.5280000000000005, |
| "grad_norm": 0.4127346903435722, |
| "learning_rate": 7.54373400694195e-07, |
| "loss": 0.3404, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33300644159317017, |
| "step": 3455, |
| "valid_targets_mean": 5427.4, |
| "valid_targets_min": 1280 |
| }, |
| { |
| "epoch": 5.536, |
| "grad_norm": 0.4318762137134822, |
| "learning_rate": 7.292577857642302e-07, |
| "loss": 0.3339, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3523763120174408, |
| "step": 3460, |
| "valid_targets_mean": 5254.2, |
| "valid_targets_min": 2240 |
| }, |
| { |
| "epoch": 5.5440000000000005, |
| "grad_norm": 0.44346247664838123, |
| "learning_rate": 7.045596067933158e-07, |
| "loss": 0.3473, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32930928468704224, |
| "step": 3465, |
| "valid_targets_mean": 5555.8, |
| "valid_targets_min": 3208 |
| }, |
| { |
| "epoch": 5.552, |
| "grad_norm": 0.42918039529783797, |
| "learning_rate": 6.80279398784609e-07, |
| "loss": 0.3432, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3481246829032898, |
| "step": 3470, |
| "valid_targets_mean": 5283.9, |
| "valid_targets_min": 2848 |
| }, |
| { |
| "epoch": 5.5600000000000005, |
| "grad_norm": 0.39577646570047353, |
| "learning_rate": 6.56417687687343e-07, |
| "loss": 0.3277, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3224024772644043, |
| "step": 3475, |
| "valid_targets_mean": 5485.9, |
| "valid_targets_min": 3284 |
| }, |
| { |
| "epoch": 5.568, |
| "grad_norm": 0.46256825487509723, |
| "learning_rate": 6.329749903854066e-07, |
| "loss": 0.3345, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34906747937202454, |
| "step": 3480, |
| "valid_targets_mean": 5249.0, |
| "valid_targets_min": 2911 |
| }, |
| { |
| "epoch": 5.576, |
| "grad_norm": 0.4240344406212783, |
| "learning_rate": 6.099518146861628e-07, |
| "loss": 0.3294, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33610647916793823, |
| "step": 3485, |
| "valid_targets_mean": 5354.5, |
| "valid_targets_min": 3600 |
| }, |
| { |
| "epoch": 5.584, |
| "grad_norm": 0.45711353794813675, |
| "learning_rate": 5.873486593094546e-07, |
| "loss": 0.3482, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.372916042804718, |
| "step": 3490, |
| "valid_targets_mean": 5177.9, |
| "valid_targets_min": 3226 |
| }, |
| { |
| "epoch": 5.592, |
| "grad_norm": 0.42735953681908784, |
| "learning_rate": 5.651660138767834e-07, |
| "loss": 0.3284, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32060104608535767, |
| "step": 3495, |
| "valid_targets_mean": 4939.6, |
| "valid_targets_min": 1932 |
| }, |
| { |
| "epoch": 5.6, |
| "grad_norm": 0.3985707143593975, |
| "learning_rate": 5.434043589007232e-07, |
| "loss": 0.3373, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3353392779827118, |
| "step": 3500, |
| "valid_targets_mean": 5613.9, |
| "valid_targets_min": 1675 |
| }, |
| { |
| "epoch": 5.608, |
| "grad_norm": 0.4243231115561023, |
| "learning_rate": 5.220641657744963e-07, |
| "loss": 0.3367, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3319443464279175, |
| "step": 3505, |
| "valid_targets_mean": 5297.3, |
| "valid_targets_min": 3160 |
| }, |
| { |
| "epoch": 5.616, |
| "grad_norm": 0.45917430276358706, |
| "learning_rate": 5.0114589676177e-07, |
| "loss": 0.3364, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3552168309688568, |
| "step": 3510, |
| "valid_targets_mean": 4615.1, |
| "valid_targets_min": 1609 |
| }, |
| { |
| "epoch": 5.624, |
| "grad_norm": 0.41450041959797973, |
| "learning_rate": 4.806500049866492e-07, |
| "loss": 0.3314, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31722262501716614, |
| "step": 3515, |
| "valid_targets_mean": 5304.6, |
| "valid_targets_min": 946 |
| }, |
| { |
| "epoch": 5.632, |
| "grad_norm": 0.4445078069051559, |
| "learning_rate": 4.6057693442383756e-07, |
| "loss": 0.3347, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34328895807266235, |
| "step": 3520, |
| "valid_targets_mean": 4784.5, |
| "valid_targets_min": 2117 |
| }, |
| { |
| "epoch": 5.64, |
| "grad_norm": 0.42969429606338644, |
| "learning_rate": 4.409271198890519e-07, |
| "loss": 0.3333, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3385939598083496, |
| "step": 3525, |
| "valid_targets_mean": 5221.7, |
| "valid_targets_min": 2038 |
| }, |
| { |
| "epoch": 5.648, |
| "grad_norm": 0.42627065119720203, |
| "learning_rate": 4.217009870295763e-07, |
| "loss": 0.3305, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.344087690114975, |
| "step": 3530, |
| "valid_targets_mean": 5177.1, |
| "valid_targets_min": 3077 |
| }, |
| { |
| "epoch": 5.656, |
| "grad_norm": 0.4755693161082013, |
| "learning_rate": 4.028989523150628e-07, |
| "loss": 0.3323, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32462215423583984, |
| "step": 3535, |
| "valid_targets_mean": 4869.7, |
| "valid_targets_min": 2321 |
| }, |
| { |
| "epoch": 5.664, |
| "grad_norm": 0.4605204468942749, |
| "learning_rate": 3.8452142302849216e-07, |
| "loss": 0.3335, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3514576554298401, |
| "step": 3540, |
| "valid_targets_mean": 4608.3, |
| "valid_targets_min": 1362 |
| }, |
| { |
| "epoch": 5.672, |
| "grad_norm": 0.413253306286827, |
| "learning_rate": 3.665687972573606e-07, |
| "loss": 0.3402, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3194332718849182, |
| "step": 3545, |
| "valid_targets_mean": 5139.2, |
| "valid_targets_min": 2427 |
| }, |
| { |
| "epoch": 5.68, |
| "grad_norm": 0.43476403350120596, |
| "learning_rate": 3.4904146388506475e-07, |
| "loss": 0.3443, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3476591408252716, |
| "step": 3550, |
| "valid_targets_mean": 5013.3, |
| "valid_targets_min": 2748 |
| }, |
| { |
| "epoch": 5.688, |
| "grad_norm": 0.45750002615755175, |
| "learning_rate": 3.319398025824572e-07, |
| "loss": 0.33, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31479451060295105, |
| "step": 3555, |
| "valid_targets_mean": 5306.8, |
| "valid_targets_min": 2966 |
| }, |
| { |
| "epoch": 5.696, |
| "grad_norm": 0.41444237474505013, |
| "learning_rate": 3.152641837996373e-07, |
| "loss": 0.3301, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33273518085479736, |
| "step": 3560, |
| "valid_targets_mean": 5168.8, |
| "valid_targets_min": 2036 |
| }, |
| { |
| "epoch": 5.704, |
| "grad_norm": 0.4246773187925928, |
| "learning_rate": 2.990149687579247e-07, |
| "loss": 0.3261, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3242553770542145, |
| "step": 3565, |
| "valid_targets_mean": 5141.4, |
| "valid_targets_min": 2449 |
| }, |
| { |
| "epoch": 5.712, |
| "grad_norm": 0.40536798596968243, |
| "learning_rate": 2.8319250944203625e-07, |
| "loss": 0.3341, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3195679783821106, |
| "step": 3570, |
| "valid_targets_mean": 5465.1, |
| "valid_targets_min": 3245 |
| }, |
| { |
| "epoch": 5.72, |
| "grad_norm": 0.45769806298438864, |
| "learning_rate": 2.677971485924502e-07, |
| "loss": 0.3484, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.347002238035202, |
| "step": 3575, |
| "valid_targets_mean": 4445.2, |
| "valid_targets_min": 1849 |
| }, |
| { |
| "epoch": 5.728, |
| "grad_norm": 0.41315830430739403, |
| "learning_rate": 2.52829219697992e-07, |
| "loss": 0.3388, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3243047595024109, |
| "step": 3580, |
| "valid_targets_mean": 5430.5, |
| "valid_targets_min": 1838 |
| }, |
| { |
| "epoch": 5.736, |
| "grad_norm": 0.4601699730593337, |
| "learning_rate": 2.3828904698861565e-07, |
| "loss": 0.3325, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3690045475959778, |
| "step": 3585, |
| "valid_targets_mean": 5184.4, |
| "valid_targets_min": 1483 |
| }, |
| { |
| "epoch": 5.744, |
| "grad_norm": 0.4236356917181989, |
| "learning_rate": 2.2417694542836489e-07, |
| "loss": 0.3345, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3260542154312134, |
| "step": 3590, |
| "valid_targets_mean": 5006.3, |
| "valid_targets_min": 2373 |
| }, |
| { |
| "epoch": 5.752, |
| "grad_norm": 0.48218766925520656, |
| "learning_rate": 2.104932207085586e-07, |
| "loss": 0.3484, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33884555101394653, |
| "step": 3595, |
| "valid_targets_mean": 5235.7, |
| "valid_targets_min": 3180 |
| }, |
| { |
| "epoch": 5.76, |
| "grad_norm": 0.40212889894019854, |
| "learning_rate": 1.97238169241174e-07, |
| "loss": 0.3351, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32871001958847046, |
| "step": 3600, |
| "valid_targets_mean": 5231.4, |
| "valid_targets_min": 3278 |
| }, |
| { |
| "epoch": 5.768, |
| "grad_norm": 0.39293874785889826, |
| "learning_rate": 1.8441207815241613e-07, |
| "loss": 0.3277, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32871419191360474, |
| "step": 3605, |
| "valid_targets_mean": 5769.8, |
| "valid_targets_min": 1923 |
| }, |
| { |
| "epoch": 5.776, |
| "grad_norm": 0.4457537107907888, |
| "learning_rate": 1.720152252765095e-07, |
| "loss": 0.3389, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33081740140914917, |
| "step": 3610, |
| "valid_targets_mean": 5351.0, |
| "valid_targets_min": 2837 |
| }, |
| { |
| "epoch": 5.784, |
| "grad_norm": 0.43419880912940195, |
| "learning_rate": 1.600478791496629e-07, |
| "loss": 0.3441, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.347716748714447, |
| "step": 3615, |
| "valid_targets_mean": 5300.3, |
| "valid_targets_min": 3114 |
| }, |
| { |
| "epoch": 5.792, |
| "grad_norm": 0.4417519399125093, |
| "learning_rate": 1.4851029900427415e-07, |
| "loss": 0.3383, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34869933128356934, |
| "step": 3620, |
| "valid_targets_mean": 5072.6, |
| "valid_targets_min": 2547 |
| }, |
| { |
| "epoch": 5.8, |
| "grad_norm": 0.4877179861280195, |
| "learning_rate": 1.3740273476329224e-07, |
| "loss": 0.3482, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35640770196914673, |
| "step": 3625, |
| "valid_targets_mean": 4348.1, |
| "valid_targets_min": 2065 |
| }, |
| { |
| "epoch": 5.808, |
| "grad_norm": 0.5203236919701949, |
| "learning_rate": 1.2672542703482616e-07, |
| "loss": 0.3417, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3371855914592743, |
| "step": 3630, |
| "valid_targets_mean": 4832.5, |
| "valid_targets_min": 3077 |
| }, |
| { |
| "epoch": 5.816, |
| "grad_norm": 0.4862950759578312, |
| "learning_rate": 1.164786071069135e-07, |
| "loss": 0.3413, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36151158809661865, |
| "step": 3635, |
| "valid_targets_mean": 4490.9, |
| "valid_targets_min": 2180 |
| }, |
| { |
| "epoch": 5.824, |
| "grad_norm": 0.4234984919555524, |
| "learning_rate": 1.0666249694251785e-07, |
| "loss": 0.338, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.339851438999176, |
| "step": 3640, |
| "valid_targets_mean": 5232.4, |
| "valid_targets_min": 1847 |
| }, |
| { |
| "epoch": 5.832, |
| "grad_norm": 0.4593105927366553, |
| "learning_rate": 9.72773091747281e-08, |
| "loss": 0.3411, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36562031507492065, |
| "step": 3645, |
| "valid_targets_mean": 4687.6, |
| "valid_targets_min": 1794 |
| }, |
| { |
| "epoch": 5.84, |
| "grad_norm": 0.4144645037193504, |
| "learning_rate": 8.832324710214002e-08, |
| "loss": 0.3369, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32951825857162476, |
| "step": 3650, |
| "valid_targets_mean": 5630.6, |
| "valid_targets_min": 2644 |
| }, |
| { |
| "epoch": 5.848, |
| "grad_norm": 0.430846518153487, |
| "learning_rate": 7.980050468445744e-08, |
| "loss": 0.3429, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3294622302055359, |
| "step": 3655, |
| "valid_targets_mean": 4923.2, |
| "valid_targets_min": 3227 |
| }, |
| { |
| "epoch": 5.856, |
| "grad_norm": 0.43710694219115687, |
| "learning_rate": 7.170926653829347e-08, |
| "loss": 0.3423, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34210455417633057, |
| "step": 3660, |
| "valid_targets_mean": 5407.6, |
| "valid_targets_min": 2908 |
| }, |
| { |
| "epoch": 5.864, |
| "grad_norm": 0.4406013261763632, |
| "learning_rate": 6.404970793317145e-08, |
| "loss": 0.3364, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33530712127685547, |
| "step": 3665, |
| "valid_targets_mean": 4958.0, |
| "valid_targets_min": 3270 |
| }, |
| { |
| "epoch": 5.872, |
| "grad_norm": 0.41259383033209157, |
| "learning_rate": 5.682199478772133e-08, |
| "loss": 0.3337, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32724255323410034, |
| "step": 3670, |
| "valid_targets_mean": 5494.9, |
| "valid_targets_min": 2009 |
| }, |
| { |
| "epoch": 5.88, |
| "grad_norm": 0.4044773772148501, |
| "learning_rate": 5.0026283666093635e-08, |
| "loss": 0.3374, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32856377959251404, |
| "step": 3675, |
| "valid_targets_mean": 5556.9, |
| "valid_targets_min": 3823 |
| }, |
| { |
| "epoch": 5.888, |
| "grad_norm": 0.4401471474120136, |
| "learning_rate": 4.366272177456665e-08, |
| "loss": 0.3348, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34074699878692627, |
| "step": 3680, |
| "valid_targets_mean": 5127.6, |
| "valid_targets_min": 2445 |
| }, |
| { |
| "epoch": 5.896, |
| "grad_norm": 0.4251523914995826, |
| "learning_rate": 3.773144695834674e-08, |
| "loss": 0.3345, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32303211092948914, |
| "step": 3685, |
| "valid_targets_mean": 5276.8, |
| "valid_targets_min": 3276 |
| }, |
| { |
| "epoch": 5.904, |
| "grad_norm": 0.40487180968369096, |
| "learning_rate": 3.223258769860405e-08, |
| "loss": 0.3133, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30014678835868835, |
| "step": 3690, |
| "valid_targets_mean": 5441.4, |
| "valid_targets_min": 2332 |
| }, |
| { |
| "epoch": 5.912, |
| "grad_norm": 0.4246242694501025, |
| "learning_rate": 2.716626310966808e-08, |
| "loss": 0.3341, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34587806463241577, |
| "step": 3695, |
| "valid_targets_mean": 5514.6, |
| "valid_targets_min": 2365 |
| }, |
| { |
| "epoch": 5.92, |
| "grad_norm": 0.4327180894649795, |
| "learning_rate": 2.253258293645866e-08, |
| "loss": 0.3387, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31725406646728516, |
| "step": 3700, |
| "valid_targets_mean": 4581.5, |
| "valid_targets_min": 3141 |
| }, |
| { |
| "epoch": 5.928, |
| "grad_norm": 0.4204611319442943, |
| "learning_rate": 1.8331647552110033e-08, |
| "loss": 0.326, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32098111510276794, |
| "step": 3705, |
| "valid_targets_mean": 5609.5, |
| "valid_targets_min": 2205 |
| }, |
| { |
| "epoch": 5.936, |
| "grad_norm": 0.46245026359207037, |
| "learning_rate": 1.456354795578374e-08, |
| "loss": 0.3273, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3033859133720398, |
| "step": 3710, |
| "valid_targets_mean": 4053.8, |
| "valid_targets_min": 1880 |
| }, |
| { |
| "epoch": 5.944, |
| "grad_norm": 0.4218061428121794, |
| "learning_rate": 1.1228365770714622e-08, |
| "loss": 0.3323, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3287618160247803, |
| "step": 3715, |
| "valid_targets_mean": 5388.4, |
| "valid_targets_min": 2903 |
| }, |
| { |
| "epoch": 5.952, |
| "grad_norm": 0.45611992270488994, |
| "learning_rate": 8.326173242432233e-09, |
| "loss": 0.3297, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34398141503334045, |
| "step": 3720, |
| "valid_targets_mean": 4851.6, |
| "valid_targets_min": 3039 |
| }, |
| { |
| "epoch": 5.96, |
| "grad_norm": 0.42708125554785126, |
| "learning_rate": 5.857033237199883e-09, |
| "loss": 0.3271, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30770760774612427, |
| "step": 3725, |
| "valid_targets_mean": 4662.5, |
| "valid_targets_min": 1593 |
| }, |
| { |
| "epoch": 5.968, |
| "grad_norm": 0.37754464860250025, |
| "learning_rate": 3.820999240644608e-09, |
| "loss": 0.3375, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32217705249786377, |
| "step": 3730, |
| "valid_targets_mean": 5964.7, |
| "valid_targets_min": 4112 |
| }, |
| { |
| "epoch": 5.976, |
| "grad_norm": 0.473167254804087, |
| "learning_rate": 2.2181153566158687e-09, |
| "loss": 0.3351, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3374801576137543, |
| "step": 3735, |
| "valid_targets_mean": 4968.4, |
| "valid_targets_min": 2368 |
| }, |
| { |
| "epoch": 5.984, |
| "grad_norm": 0.44323430453403967, |
| "learning_rate": 1.0484163062107755e-09, |
| "loss": 0.3462, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38353562355041504, |
| "step": 3740, |
| "valid_targets_mean": 5432.4, |
| "valid_targets_min": 3244 |
| }, |
| { |
| "epoch": 5.992, |
| "grad_norm": 0.42772788172019927, |
| "learning_rate": 3.11927427034675e-10, |
| "loss": 0.3273, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33847105503082275, |
| "step": 3745, |
| "valid_targets_mean": 5236.5, |
| "valid_targets_min": 3052 |
| }, |
| { |
| "epoch": 6.0, |
| "grad_norm": 0.47314210224549785, |
| "learning_rate": 8.664672648261985e-12, |
| "loss": 0.3403, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.366549551486969, |
| "step": 3750, |
| "valid_targets_mean": 5138.3, |
| "valid_targets_min": 2565 |
| }, |
| { |
| "epoch": 6.0, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.366549551486969, |
| "step": 3750, |
| "total_flos": 1872142477819904.0, |
| "train_loss": 0.36636094560623167, |
| "train_runtime": 26323.5361, |
| "train_samples_per_second": 2.279, |
| "train_steps_per_second": 0.142, |
| "valid_targets_mean": 5138.3, |
| "valid_targets_min": 2565 |
| } |
| ], |
| "logging_steps": 5, |
| "max_steps": 3750, |
| "num_input_tokens_seen": 0, |
| "num_train_epochs": 6, |
| "save_steps": 100, |
| "stateful_callbacks": { |
| "TrainerControl": { |
| "args": { |
| "should_epoch_stop": false, |
| "should_evaluate": false, |
| "should_log": false, |
| "should_save": true, |
| "should_training_stop": true |
| }, |
| "attributes": {} |
| } |
| }, |
| "total_flos": 1872142477819904.0, |
| "train_batch_size": 1, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|