{ "best_metric": null, "best_model_checkpoint": null, "epoch": 13.0, "eval_steps": 50, "global_step": 884, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 1.0, "grad_norm": 37.9459228515625, "learning_rate": 1.5280898876404495e-05, "loss": 2.6802, "step": 68 }, { "epoch": 1.0, "eval_custom-arc-semantics-data-jp_cosine_accuracy": 0.8529411764705882, "eval_custom-arc-semantics-data-jp_cosine_accuracy_threshold": 0.7407679557800293, "eval_custom-arc-semantics-data-jp_cosine_ap": 0.8861687056931384, "eval_custom-arc-semantics-data-jp_cosine_f1": 0.863013698630137, "eval_custom-arc-semantics-data-jp_cosine_f1_threshold": 0.7407679557800293, "eval_custom-arc-semantics-data-jp_cosine_precision": 0.8873239436619719, "eval_custom-arc-semantics-data-jp_cosine_recall": 0.84, "eval_custom-arc-semantics-data-jp_dot_accuracy": 0.8529411764705882, "eval_custom-arc-semantics-data-jp_dot_accuracy_threshold": 413.34197998046875, "eval_custom-arc-semantics-data-jp_dot_ap": 0.887218522968793, "eval_custom-arc-semantics-data-jp_dot_f1": 0.8648648648648648, "eval_custom-arc-semantics-data-jp_dot_f1_threshold": 413.34197998046875, "eval_custom-arc-semantics-data-jp_dot_precision": 0.8767123287671232, "eval_custom-arc-semantics-data-jp_dot_recall": 0.8533333333333334, "eval_custom-arc-semantics-data-jp_euclidean_accuracy": 0.8602941176470589, "eval_custom-arc-semantics-data-jp_euclidean_accuracy_threshold": 17.441856384277344, "eval_custom-arc-semantics-data-jp_euclidean_ap": 0.8855166425710176, "eval_custom-arc-semantics-data-jp_euclidean_f1": 0.87248322147651, "eval_custom-arc-semantics-data-jp_euclidean_f1_threshold": 17.441856384277344, "eval_custom-arc-semantics-data-jp_euclidean_precision": 0.8783783783783784, "eval_custom-arc-semantics-data-jp_euclidean_recall": 0.8666666666666667, "eval_custom-arc-semantics-data-jp_manhattan_accuracy": 0.8602941176470589, "eval_custom-arc-semantics-data-jp_manhattan_accuracy_threshold": 382.9211730957031, "eval_custom-arc-semantics-data-jp_manhattan_ap": 0.8854876210985647, "eval_custom-arc-semantics-data-jp_manhattan_f1": 0.8741721854304636, "eval_custom-arc-semantics-data-jp_manhattan_f1_threshold": 386.8717956542969, "eval_custom-arc-semantics-data-jp_manhattan_precision": 0.868421052631579, "eval_custom-arc-semantics-data-jp_manhattan_recall": 0.88, "eval_custom-arc-semantics-data-jp_max_accuracy": 0.8602941176470589, "eval_custom-arc-semantics-data-jp_max_accuracy_threshold": 413.34197998046875, "eval_custom-arc-semantics-data-jp_max_ap": 0.887218522968793, "eval_custom-arc-semantics-data-jp_max_f1": 0.8741721854304636, "eval_custom-arc-semantics-data-jp_max_f1_threshold": 413.34197998046875, "eval_custom-arc-semantics-data-jp_max_precision": 0.8873239436619719, "eval_custom-arc-semantics-data-jp_max_recall": 0.88, "eval_loss": 1.7807204723358154, "eval_runtime": 14.8133, "eval_samples_per_second": 9.181, "eval_steps_per_second": 1.148, "step": 68 }, { "epoch": 2.0, "grad_norm": 101.59432983398438, "learning_rate": 1.8817610062893082e-05, "loss": 1.4014, "step": 136 }, { "epoch": 2.0, "eval_custom-arc-semantics-data-jp_cosine_accuracy": 0.8676470588235294, "eval_custom-arc-semantics-data-jp_cosine_accuracy_threshold": 0.656826376914978, "eval_custom-arc-semantics-data-jp_cosine_ap": 0.894049495028942, "eval_custom-arc-semantics-data-jp_cosine_f1": 0.8783783783783784, "eval_custom-arc-semantics-data-jp_cosine_f1_threshold": 0.656826376914978, "eval_custom-arc-semantics-data-jp_cosine_precision": 0.8904109589041096, "eval_custom-arc-semantics-data-jp_cosine_recall": 0.8666666666666667, "eval_custom-arc-semantics-data-jp_dot_accuracy": 0.8602941176470589, "eval_custom-arc-semantics-data-jp_dot_accuracy_threshold": 367.9239501953125, "eval_custom-arc-semantics-data-jp_dot_ap": 0.8915744113817081, "eval_custom-arc-semantics-data-jp_dot_f1": 0.87248322147651, "eval_custom-arc-semantics-data-jp_dot_f1_threshold": 363.2611999511719, "eval_custom-arc-semantics-data-jp_dot_precision": 0.8783783783783784, "eval_custom-arc-semantics-data-jp_dot_recall": 0.8666666666666667, "eval_custom-arc-semantics-data-jp_euclidean_accuracy": 0.8602941176470589, "eval_custom-arc-semantics-data-jp_euclidean_accuracy_threshold": 19.78958511352539, "eval_custom-arc-semantics-data-jp_euclidean_ap": 0.8925974243856872, "eval_custom-arc-semantics-data-jp_euclidean_f1": 0.8819875776397516, "eval_custom-arc-semantics-data-jp_euclidean_f1_threshold": 20.95208740234375, "eval_custom-arc-semantics-data-jp_euclidean_precision": 0.8255813953488372, "eval_custom-arc-semantics-data-jp_euclidean_recall": 0.9466666666666667, "eval_custom-arc-semantics-data-jp_manhattan_accuracy": 0.875, "eval_custom-arc-semantics-data-jp_manhattan_accuracy_threshold": 456.61602783203125, "eval_custom-arc-semantics-data-jp_manhattan_ap": 0.8944639878933996, "eval_custom-arc-semantics-data-jp_manhattan_f1": 0.8930817610062893, "eval_custom-arc-semantics-data-jp_manhattan_f1_threshold": 456.61602783203125, "eval_custom-arc-semantics-data-jp_manhattan_precision": 0.8452380952380952, "eval_custom-arc-semantics-data-jp_manhattan_recall": 0.9466666666666667, "eval_custom-arc-semantics-data-jp_max_accuracy": 0.875, "eval_custom-arc-semantics-data-jp_max_accuracy_threshold": 456.61602783203125, "eval_custom-arc-semantics-data-jp_max_ap": 0.8944639878933996, "eval_custom-arc-semantics-data-jp_max_f1": 0.8930817610062893, "eval_custom-arc-semantics-data-jp_max_f1_threshold": 456.61602783203125, "eval_custom-arc-semantics-data-jp_max_precision": 0.8904109589041096, "eval_custom-arc-semantics-data-jp_max_recall": 0.9466666666666667, "eval_loss": 1.7683357000350952, "eval_runtime": 13.9834, "eval_samples_per_second": 9.726, "eval_steps_per_second": 1.216, "step": 136 }, { "epoch": 3.0, "grad_norm": 102.5489730834961, "learning_rate": 1.7106918238993714e-05, "loss": 0.7937, "step": 204 }, { "epoch": 3.0, "eval_custom-arc-semantics-data-jp_cosine_accuracy": 0.8897058823529411, "eval_custom-arc-semantics-data-jp_cosine_accuracy_threshold": 0.5597226619720459, "eval_custom-arc-semantics-data-jp_cosine_ap": 0.9038906301326961, "eval_custom-arc-semantics-data-jp_cosine_f1": 0.9006622516556291, "eval_custom-arc-semantics-data-jp_cosine_f1_threshold": 0.5076813697814941, "eval_custom-arc-semantics-data-jp_cosine_precision": 0.8947368421052632, "eval_custom-arc-semantics-data-jp_cosine_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_dot_accuracy": 0.8823529411764706, "eval_custom-arc-semantics-data-jp_dot_accuracy_threshold": 314.1024475097656, "eval_custom-arc-semantics-data-jp_dot_ap": 0.9012158574599884, "eval_custom-arc-semantics-data-jp_dot_f1": 0.8904109589041095, "eval_custom-arc-semantics-data-jp_dot_f1_threshold": 295.25762939453125, "eval_custom-arc-semantics-data-jp_dot_precision": 0.9154929577464789, "eval_custom-arc-semantics-data-jp_dot_recall": 0.8666666666666667, "eval_custom-arc-semantics-data-jp_euclidean_accuracy": 0.8970588235294118, "eval_custom-arc-semantics-data-jp_euclidean_accuracy_threshold": 22.511234283447266, "eval_custom-arc-semantics-data-jp_euclidean_ap": 0.9033993255493061, "eval_custom-arc-semantics-data-jp_euclidean_f1": 0.9054054054054055, "eval_custom-arc-semantics-data-jp_euclidean_f1_threshold": 22.511234283447266, "eval_custom-arc-semantics-data-jp_euclidean_precision": 0.9178082191780822, "eval_custom-arc-semantics-data-jp_euclidean_recall": 0.8933333333333333, "eval_custom-arc-semantics-data-jp_manhattan_accuracy": 0.8897058823529411, "eval_custom-arc-semantics-data-jp_manhattan_accuracy_threshold": 480.63677978515625, "eval_custom-arc-semantics-data-jp_manhattan_ap": 0.9033851155012284, "eval_custom-arc-semantics-data-jp_manhattan_f1": 0.8993288590604026, "eval_custom-arc-semantics-data-jp_manhattan_f1_threshold": 502.66156005859375, "eval_custom-arc-semantics-data-jp_manhattan_precision": 0.9054054054054054, "eval_custom-arc-semantics-data-jp_manhattan_recall": 0.8933333333333333, "eval_custom-arc-semantics-data-jp_max_accuracy": 0.8970588235294118, "eval_custom-arc-semantics-data-jp_max_accuracy_threshold": 480.63677978515625, "eval_custom-arc-semantics-data-jp_max_ap": 0.9038906301326961, "eval_custom-arc-semantics-data-jp_max_f1": 0.9054054054054055, "eval_custom-arc-semantics-data-jp_max_f1_threshold": 502.66156005859375, "eval_custom-arc-semantics-data-jp_max_precision": 0.9178082191780822, "eval_custom-arc-semantics-data-jp_max_recall": 0.9066666666666666, "eval_loss": 1.9876813888549805, "eval_runtime": 15.0797, "eval_samples_per_second": 9.019, "eval_steps_per_second": 1.127, "step": 204 }, { "epoch": 4.0, "grad_norm": 54.64971160888672, "learning_rate": 1.539622641509434e-05, "loss": 0.5443, "step": 272 }, { "epoch": 4.0, "eval_custom-arc-semantics-data-jp_cosine_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_cosine_accuracy_threshold": 0.5411365032196045, "eval_custom-arc-semantics-data-jp_cosine_ap": 0.9075251070302803, "eval_custom-arc-semantics-data-jp_cosine_f1": 0.918918918918919, "eval_custom-arc-semantics-data-jp_cosine_f1_threshold": 0.5411365032196045, "eval_custom-arc-semantics-data-jp_cosine_precision": 0.9315068493150684, "eval_custom-arc-semantics-data-jp_cosine_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_dot_accuracy": 0.8897058823529411, "eval_custom-arc-semantics-data-jp_dot_accuracy_threshold": 311.26678466796875, "eval_custom-arc-semantics-data-jp_dot_ap": 0.9040853015076152, "eval_custom-arc-semantics-data-jp_dot_f1": 0.9006622516556291, "eval_custom-arc-semantics-data-jp_dot_f1_threshold": 285.96728515625, "eval_custom-arc-semantics-data-jp_dot_precision": 0.8947368421052632, "eval_custom-arc-semantics-data-jp_dot_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_euclidean_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_euclidean_accuracy_threshold": 22.630477905273438, "eval_custom-arc-semantics-data-jp_euclidean_ap": 0.906411643380887, "eval_custom-arc-semantics-data-jp_euclidean_f1": 0.918918918918919, "eval_custom-arc-semantics-data-jp_euclidean_f1_threshold": 22.630477905273438, "eval_custom-arc-semantics-data-jp_euclidean_precision": 0.9315068493150684, "eval_custom-arc-semantics-data-jp_euclidean_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_manhattan_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_manhattan_accuracy_threshold": 495.93194580078125, "eval_custom-arc-semantics-data-jp_manhattan_ap": 0.9068108312799359, "eval_custom-arc-semantics-data-jp_manhattan_f1": 0.918918918918919, "eval_custom-arc-semantics-data-jp_manhattan_f1_threshold": 495.93194580078125, "eval_custom-arc-semantics-data-jp_manhattan_precision": 0.9315068493150684, "eval_custom-arc-semantics-data-jp_manhattan_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_max_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_max_accuracy_threshold": 495.93194580078125, "eval_custom-arc-semantics-data-jp_max_ap": 0.9075251070302803, "eval_custom-arc-semantics-data-jp_max_f1": 0.918918918918919, "eval_custom-arc-semantics-data-jp_max_f1_threshold": 495.93194580078125, "eval_custom-arc-semantics-data-jp_max_precision": 0.9315068493150684, "eval_custom-arc-semantics-data-jp_max_recall": 0.9066666666666666, "eval_loss": 1.910571813583374, "eval_runtime": 14.3178, "eval_samples_per_second": 9.499, "eval_steps_per_second": 1.187, "step": 272 }, { "epoch": 5.0, "grad_norm": 7.771420955657959, "learning_rate": 1.368553459119497e-05, "loss": 0.4225, "step": 340 }, { "epoch": 5.0, "eval_custom-arc-semantics-data-jp_cosine_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_cosine_accuracy_threshold": 0.5279971361160278, "eval_custom-arc-semantics-data-jp_cosine_ap": 0.9104573720737034, "eval_custom-arc-semantics-data-jp_cosine_f1": 0.918918918918919, "eval_custom-arc-semantics-data-jp_cosine_f1_threshold": 0.5279971361160278, "eval_custom-arc-semantics-data-jp_cosine_precision": 0.9315068493150684, "eval_custom-arc-semantics-data-jp_cosine_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_dot_accuracy": 0.9044117647058824, "eval_custom-arc-semantics-data-jp_dot_accuracy_threshold": 319.6995849609375, "eval_custom-arc-semantics-data-jp_dot_ap": 0.907606464463115, "eval_custom-arc-semantics-data-jp_dot_f1": 0.9115646258503401, "eval_custom-arc-semantics-data-jp_dot_f1_threshold": 290.19317626953125, "eval_custom-arc-semantics-data-jp_dot_precision": 0.9305555555555556, "eval_custom-arc-semantics-data-jp_dot_recall": 0.8933333333333333, "eval_custom-arc-semantics-data-jp_euclidean_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_euclidean_accuracy_threshold": 22.626989364624023, "eval_custom-arc-semantics-data-jp_euclidean_ap": 0.9092245976210003, "eval_custom-arc-semantics-data-jp_euclidean_f1": 0.918918918918919, "eval_custom-arc-semantics-data-jp_euclidean_f1_threshold": 22.626989364624023, "eval_custom-arc-semantics-data-jp_euclidean_precision": 0.9315068493150684, "eval_custom-arc-semantics-data-jp_euclidean_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_manhattan_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_manhattan_accuracy_threshold": 494.6650390625, "eval_custom-arc-semantics-data-jp_manhattan_ap": 0.9108555121538641, "eval_custom-arc-semantics-data-jp_manhattan_f1": 0.92, "eval_custom-arc-semantics-data-jp_manhattan_f1_threshold": 514.7510986328125, "eval_custom-arc-semantics-data-jp_manhattan_precision": 0.92, "eval_custom-arc-semantics-data-jp_manhattan_recall": 0.92, "eval_custom-arc-semantics-data-jp_max_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_max_accuracy_threshold": 494.6650390625, "eval_custom-arc-semantics-data-jp_max_ap": 0.9108555121538641, "eval_custom-arc-semantics-data-jp_max_f1": 0.92, "eval_custom-arc-semantics-data-jp_max_f1_threshold": 514.7510986328125, "eval_custom-arc-semantics-data-jp_max_precision": 0.9315068493150684, "eval_custom-arc-semantics-data-jp_max_recall": 0.92, "eval_loss": 1.9418467283248901, "eval_runtime": 14.4159, "eval_samples_per_second": 9.434, "eval_steps_per_second": 1.179, "step": 340 }, { "epoch": 6.0, "grad_norm": 0.8858943581581116, "learning_rate": 1.1974842767295597e-05, "loss": 0.3347, "step": 408 }, { "epoch": 6.0, "eval_custom-arc-semantics-data-jp_cosine_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_cosine_accuracy_threshold": 0.5068449974060059, "eval_custom-arc-semantics-data-jp_cosine_ap": 0.9088529074244793, "eval_custom-arc-semantics-data-jp_cosine_f1": 0.918918918918919, "eval_custom-arc-semantics-data-jp_cosine_f1_threshold": 0.5068449974060059, "eval_custom-arc-semantics-data-jp_cosine_precision": 0.9315068493150684, "eval_custom-arc-semantics-data-jp_cosine_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_dot_accuracy": 0.8970588235294118, "eval_custom-arc-semantics-data-jp_dot_accuracy_threshold": 322.01324462890625, "eval_custom-arc-semantics-data-jp_dot_ap": 0.9107273300403762, "eval_custom-arc-semantics-data-jp_dot_f1": 0.9054054054054055, "eval_custom-arc-semantics-data-jp_dot_f1_threshold": 276.34893798828125, "eval_custom-arc-semantics-data-jp_dot_precision": 0.9178082191780822, "eval_custom-arc-semantics-data-jp_dot_recall": 0.8933333333333333, "eval_custom-arc-semantics-data-jp_euclidean_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_euclidean_accuracy_threshold": 23.167011260986328, "eval_custom-arc-semantics-data-jp_euclidean_ap": 0.9088745561179659, "eval_custom-arc-semantics-data-jp_euclidean_f1": 0.918918918918919, "eval_custom-arc-semantics-data-jp_euclidean_f1_threshold": 23.167011260986328, "eval_custom-arc-semantics-data-jp_euclidean_precision": 0.9315068493150684, "eval_custom-arc-semantics-data-jp_euclidean_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_manhattan_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_manhattan_accuracy_threshold": 506.7710266113281, "eval_custom-arc-semantics-data-jp_manhattan_ap": 0.9092068556819233, "eval_custom-arc-semantics-data-jp_manhattan_f1": 0.918918918918919, "eval_custom-arc-semantics-data-jp_manhattan_f1_threshold": 506.7710266113281, "eval_custom-arc-semantics-data-jp_manhattan_precision": 0.9315068493150684, "eval_custom-arc-semantics-data-jp_manhattan_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_max_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_max_accuracy_threshold": 506.7710266113281, "eval_custom-arc-semantics-data-jp_max_ap": 0.9107273300403762, "eval_custom-arc-semantics-data-jp_max_f1": 0.918918918918919, "eval_custom-arc-semantics-data-jp_max_f1_threshold": 506.7710266113281, "eval_custom-arc-semantics-data-jp_max_precision": 0.9315068493150684, "eval_custom-arc-semantics-data-jp_max_recall": 0.9066666666666666, "eval_loss": 2.012265920639038, "eval_runtime": 13.9077, "eval_samples_per_second": 9.779, "eval_steps_per_second": 1.222, "step": 408 }, { "epoch": 7.0, "grad_norm": 0.24478188157081604, "learning_rate": 1.0264150943396227e-05, "loss": 0.3425, "step": 476 }, { "epoch": 7.0, "eval_custom-arc-semantics-data-jp_cosine_accuracy": 0.9044117647058824, "eval_custom-arc-semantics-data-jp_cosine_accuracy_threshold": 0.5215883255004883, "eval_custom-arc-semantics-data-jp_cosine_ap": 0.9081430244026829, "eval_custom-arc-semantics-data-jp_cosine_f1": 0.912751677852349, "eval_custom-arc-semantics-data-jp_cosine_f1_threshold": 0.5215883255004883, "eval_custom-arc-semantics-data-jp_cosine_precision": 0.918918918918919, "eval_custom-arc-semantics-data-jp_cosine_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_dot_accuracy": 0.9044117647058824, "eval_custom-arc-semantics-data-jp_dot_accuracy_threshold": 313.0162353515625, "eval_custom-arc-semantics-data-jp_dot_ap": 0.9094090969210814, "eval_custom-arc-semantics-data-jp_dot_f1": 0.9090909090909091, "eval_custom-arc-semantics-data-jp_dot_f1_threshold": 313.0162353515625, "eval_custom-arc-semantics-data-jp_dot_precision": 0.9558823529411765, "eval_custom-arc-semantics-data-jp_dot_recall": 0.8666666666666667, "eval_custom-arc-semantics-data-jp_euclidean_accuracy": 0.9044117647058824, "eval_custom-arc-semantics-data-jp_euclidean_accuracy_threshold": 23.342838287353516, "eval_custom-arc-semantics-data-jp_euclidean_ap": 0.907981722850528, "eval_custom-arc-semantics-data-jp_euclidean_f1": 0.912751677852349, "eval_custom-arc-semantics-data-jp_euclidean_f1_threshold": 23.342838287353516, "eval_custom-arc-semantics-data-jp_euclidean_precision": 0.918918918918919, "eval_custom-arc-semantics-data-jp_euclidean_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_manhattan_accuracy": 0.9044117647058824, "eval_custom-arc-semantics-data-jp_manhattan_accuracy_threshold": 511.61737060546875, "eval_custom-arc-semantics-data-jp_manhattan_ap": 0.9085688725632555, "eval_custom-arc-semantics-data-jp_manhattan_f1": 0.912751677852349, "eval_custom-arc-semantics-data-jp_manhattan_f1_threshold": 511.61737060546875, "eval_custom-arc-semantics-data-jp_manhattan_precision": 0.918918918918919, "eval_custom-arc-semantics-data-jp_manhattan_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_max_accuracy": 0.9044117647058824, "eval_custom-arc-semantics-data-jp_max_accuracy_threshold": 511.61737060546875, "eval_custom-arc-semantics-data-jp_max_ap": 0.9094090969210814, "eval_custom-arc-semantics-data-jp_max_f1": 0.912751677852349, "eval_custom-arc-semantics-data-jp_max_f1_threshold": 511.61737060546875, "eval_custom-arc-semantics-data-jp_max_precision": 0.9558823529411765, "eval_custom-arc-semantics-data-jp_max_recall": 0.9066666666666666, "eval_loss": 2.038686752319336, "eval_runtime": 14.7627, "eval_samples_per_second": 9.212, "eval_steps_per_second": 1.152, "step": 476 }, { "epoch": 8.0, "grad_norm": 0.09618318825960159, "learning_rate": 8.553459119496857e-06, "loss": 0.2427, "step": 544 }, { "epoch": 8.0, "eval_custom-arc-semantics-data-jp_cosine_accuracy": 0.9044117647058824, "eval_custom-arc-semantics-data-jp_cosine_accuracy_threshold": 0.5501157641410828, "eval_custom-arc-semantics-data-jp_cosine_ap": 0.9086817589029806, "eval_custom-arc-semantics-data-jp_cosine_f1": 0.912751677852349, "eval_custom-arc-semantics-data-jp_cosine_f1_threshold": 0.5119773149490356, "eval_custom-arc-semantics-data-jp_cosine_precision": 0.918918918918919, "eval_custom-arc-semantics-data-jp_cosine_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_dot_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_dot_accuracy_threshold": 294.31005859375, "eval_custom-arc-semantics-data-jp_dot_ap": 0.9096226292878296, "eval_custom-arc-semantics-data-jp_dot_f1": 0.9166666666666666, "eval_custom-arc-semantics-data-jp_dot_f1_threshold": 294.31005859375, "eval_custom-arc-semantics-data-jp_dot_precision": 0.9565217391304348, "eval_custom-arc-semantics-data-jp_dot_recall": 0.88, "eval_custom-arc-semantics-data-jp_euclidean_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_euclidean_accuracy_threshold": 23.28342056274414, "eval_custom-arc-semantics-data-jp_euclidean_ap": 0.9089227316320502, "eval_custom-arc-semantics-data-jp_euclidean_f1": 0.918918918918919, "eval_custom-arc-semantics-data-jp_euclidean_f1_threshold": 23.28342056274414, "eval_custom-arc-semantics-data-jp_euclidean_precision": 0.9315068493150684, "eval_custom-arc-semantics-data-jp_euclidean_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_manhattan_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_manhattan_accuracy_threshold": 509.5169677734375, "eval_custom-arc-semantics-data-jp_manhattan_ap": 0.9102569832574411, "eval_custom-arc-semantics-data-jp_manhattan_f1": 0.918918918918919, "eval_custom-arc-semantics-data-jp_manhattan_f1_threshold": 509.5169677734375, "eval_custom-arc-semantics-data-jp_manhattan_precision": 0.9315068493150684, "eval_custom-arc-semantics-data-jp_manhattan_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_max_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_max_accuracy_threshold": 509.5169677734375, "eval_custom-arc-semantics-data-jp_max_ap": 0.9102569832574411, "eval_custom-arc-semantics-data-jp_max_f1": 0.918918918918919, "eval_custom-arc-semantics-data-jp_max_f1_threshold": 509.5169677734375, "eval_custom-arc-semantics-data-jp_max_precision": 0.9565217391304348, "eval_custom-arc-semantics-data-jp_max_recall": 0.9066666666666666, "eval_loss": 1.9878324270248413, "eval_runtime": 13.7764, "eval_samples_per_second": 9.872, "eval_steps_per_second": 1.234, "step": 544 }, { "epoch": 9.0, "grad_norm": 0.046123113483190536, "learning_rate": 6.842767295597485e-06, "loss": 0.2412, "step": 612 }, { "epoch": 9.0, "eval_custom-arc-semantics-data-jp_cosine_accuracy": 0.9044117647058824, "eval_custom-arc-semantics-data-jp_cosine_accuracy_threshold": 0.5531296730041504, "eval_custom-arc-semantics-data-jp_cosine_ap": 0.9088232956862466, "eval_custom-arc-semantics-data-jp_cosine_f1": 0.912751677852349, "eval_custom-arc-semantics-data-jp_cosine_f1_threshold": 0.49595877528190613, "eval_custom-arc-semantics-data-jp_cosine_precision": 0.918918918918919, "eval_custom-arc-semantics-data-jp_cosine_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_dot_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_dot_accuracy_threshold": 295.7380676269531, "eval_custom-arc-semantics-data-jp_dot_ap": 0.9178358242971421, "eval_custom-arc-semantics-data-jp_dot_f1": 0.9166666666666666, "eval_custom-arc-semantics-data-jp_dot_f1_threshold": 295.7380676269531, "eval_custom-arc-semantics-data-jp_dot_precision": 0.9565217391304348, "eval_custom-arc-semantics-data-jp_dot_recall": 0.88, "eval_custom-arc-semantics-data-jp_euclidean_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_euclidean_accuracy_threshold": 23.565818786621094, "eval_custom-arc-semantics-data-jp_euclidean_ap": 0.9092338106583002, "eval_custom-arc-semantics-data-jp_euclidean_f1": 0.918918918918919, "eval_custom-arc-semantics-data-jp_euclidean_f1_threshold": 23.565818786621094, "eval_custom-arc-semantics-data-jp_euclidean_precision": 0.9315068493150684, "eval_custom-arc-semantics-data-jp_euclidean_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_manhattan_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_manhattan_accuracy_threshold": 515.80322265625, "eval_custom-arc-semantics-data-jp_manhattan_ap": 0.9102278723159124, "eval_custom-arc-semantics-data-jp_manhattan_f1": 0.918918918918919, "eval_custom-arc-semantics-data-jp_manhattan_f1_threshold": 515.80322265625, "eval_custom-arc-semantics-data-jp_manhattan_precision": 0.9315068493150684, "eval_custom-arc-semantics-data-jp_manhattan_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_max_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_max_accuracy_threshold": 515.80322265625, "eval_custom-arc-semantics-data-jp_max_ap": 0.9178358242971421, "eval_custom-arc-semantics-data-jp_max_f1": 0.918918918918919, "eval_custom-arc-semantics-data-jp_max_f1_threshold": 515.80322265625, "eval_custom-arc-semantics-data-jp_max_precision": 0.9565217391304348, "eval_custom-arc-semantics-data-jp_max_recall": 0.9066666666666666, "eval_loss": 2.0423502922058105, "eval_runtime": 18.8817, "eval_samples_per_second": 7.203, "eval_steps_per_second": 0.9, "step": 612 }, { "epoch": 10.0, "grad_norm": 0.029908249154686928, "learning_rate": 5.1320754716981136e-06, "loss": 0.1623, "step": 680 }, { "epoch": 10.0, "eval_custom-arc-semantics-data-jp_cosine_accuracy": 0.9044117647058824, "eval_custom-arc-semantics-data-jp_cosine_accuracy_threshold": 0.5516438484191895, "eval_custom-arc-semantics-data-jp_cosine_ap": 0.9085774598037802, "eval_custom-arc-semantics-data-jp_cosine_f1": 0.9103448275862068, "eval_custom-arc-semantics-data-jp_cosine_f1_threshold": 0.519783079624176, "eval_custom-arc-semantics-data-jp_cosine_precision": 0.9428571428571428, "eval_custom-arc-semantics-data-jp_cosine_recall": 0.88, "eval_custom-arc-semantics-data-jp_dot_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_dot_accuracy_threshold": 294.6772766113281, "eval_custom-arc-semantics-data-jp_dot_ap": 0.918847565389113, "eval_custom-arc-semantics-data-jp_dot_f1": 0.9166666666666666, "eval_custom-arc-semantics-data-jp_dot_f1_threshold": 294.6772766113281, "eval_custom-arc-semantics-data-jp_dot_precision": 0.9565217391304348, "eval_custom-arc-semantics-data-jp_dot_recall": 0.88, "eval_custom-arc-semantics-data-jp_euclidean_accuracy": 0.9044117647058824, "eval_custom-arc-semantics-data-jp_euclidean_accuracy_threshold": 21.938934326171875, "eval_custom-arc-semantics-data-jp_euclidean_ap": 0.9089511554768472, "eval_custom-arc-semantics-data-jp_euclidean_f1": 0.912751677852349, "eval_custom-arc-semantics-data-jp_euclidean_f1_threshold": 23.713035583496094, "eval_custom-arc-semantics-data-jp_euclidean_precision": 0.918918918918919, "eval_custom-arc-semantics-data-jp_euclidean_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_manhattan_accuracy": 0.9044117647058824, "eval_custom-arc-semantics-data-jp_manhattan_accuracy_threshold": 482.1662292480469, "eval_custom-arc-semantics-data-jp_manhattan_ap": 0.910254731797207, "eval_custom-arc-semantics-data-jp_manhattan_f1": 0.912751677852349, "eval_custom-arc-semantics-data-jp_manhattan_f1_threshold": 524.188232421875, "eval_custom-arc-semantics-data-jp_manhattan_precision": 0.918918918918919, "eval_custom-arc-semantics-data-jp_manhattan_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_max_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_max_accuracy_threshold": 482.1662292480469, "eval_custom-arc-semantics-data-jp_max_ap": 0.918847565389113, "eval_custom-arc-semantics-data-jp_max_f1": 0.9166666666666666, "eval_custom-arc-semantics-data-jp_max_f1_threshold": 524.188232421875, "eval_custom-arc-semantics-data-jp_max_precision": 0.9565217391304348, "eval_custom-arc-semantics-data-jp_max_recall": 0.9066666666666666, "eval_loss": 2.0273122787475586, "eval_runtime": 20.5658, "eval_samples_per_second": 6.613, "eval_steps_per_second": 0.827, "step": 680 }, { "epoch": 11.0, "grad_norm": 0.007907501421868801, "learning_rate": 3.4213836477987424e-06, "loss": 0.1909, "step": 748 }, { "epoch": 11.0, "eval_custom-arc-semantics-data-jp_cosine_accuracy": 0.9044117647058824, "eval_custom-arc-semantics-data-jp_cosine_accuracy_threshold": 0.550028383731842, "eval_custom-arc-semantics-data-jp_cosine_ap": 0.9090819039672458, "eval_custom-arc-semantics-data-jp_cosine_f1": 0.912751677852349, "eval_custom-arc-semantics-data-jp_cosine_f1_threshold": 0.48157936334609985, "eval_custom-arc-semantics-data-jp_cosine_precision": 0.918918918918919, "eval_custom-arc-semantics-data-jp_cosine_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_dot_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_dot_accuracy_threshold": 293.80010986328125, "eval_custom-arc-semantics-data-jp_dot_ap": 0.9219827879602106, "eval_custom-arc-semantics-data-jp_dot_f1": 0.9166666666666666, "eval_custom-arc-semantics-data-jp_dot_f1_threshold": 293.80010986328125, "eval_custom-arc-semantics-data-jp_dot_precision": 0.9565217391304348, "eval_custom-arc-semantics-data-jp_dot_recall": 0.88, "eval_custom-arc-semantics-data-jp_euclidean_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_euclidean_accuracy_threshold": 23.849559783935547, "eval_custom-arc-semantics-data-jp_euclidean_ap": 0.90922024992935, "eval_custom-arc-semantics-data-jp_euclidean_f1": 0.918918918918919, "eval_custom-arc-semantics-data-jp_euclidean_f1_threshold": 23.849559783935547, "eval_custom-arc-semantics-data-jp_euclidean_precision": 0.9315068493150684, "eval_custom-arc-semantics-data-jp_euclidean_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_manhattan_accuracy": 0.9044117647058824, "eval_custom-arc-semantics-data-jp_manhattan_accuracy_threshold": 482.9535827636719, "eval_custom-arc-semantics-data-jp_manhattan_ap": 0.910689468531078, "eval_custom-arc-semantics-data-jp_manhattan_f1": 0.912751677852349, "eval_custom-arc-semantics-data-jp_manhattan_f1_threshold": 527.540771484375, "eval_custom-arc-semantics-data-jp_manhattan_precision": 0.918918918918919, "eval_custom-arc-semantics-data-jp_manhattan_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_max_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_max_accuracy_threshold": 482.9535827636719, "eval_custom-arc-semantics-data-jp_max_ap": 0.9219827879602106, "eval_custom-arc-semantics-data-jp_max_f1": 0.918918918918919, "eval_custom-arc-semantics-data-jp_max_f1_threshold": 527.540771484375, "eval_custom-arc-semantics-data-jp_max_precision": 0.9565217391304348, "eval_custom-arc-semantics-data-jp_max_recall": 0.9066666666666666, "eval_loss": 2.0955376625061035, "eval_runtime": 20.4117, "eval_samples_per_second": 6.663, "eval_steps_per_second": 0.833, "step": 748 }, { "epoch": 12.0, "grad_norm": 0.3338662087917328, "learning_rate": 1.7106918238993712e-06, "loss": 0.1507, "step": 816 }, { "epoch": 12.0, "eval_custom-arc-semantics-data-jp_cosine_accuracy": 0.9044117647058824, "eval_custom-arc-semantics-data-jp_cosine_accuracy_threshold": 0.5501536726951599, "eval_custom-arc-semantics-data-jp_cosine_ap": 0.9084179566135925, "eval_custom-arc-semantics-data-jp_cosine_f1": 0.912751677852349, "eval_custom-arc-semantics-data-jp_cosine_f1_threshold": 0.4790937304496765, "eval_custom-arc-semantics-data-jp_cosine_precision": 0.918918918918919, "eval_custom-arc-semantics-data-jp_cosine_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_dot_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_dot_accuracy_threshold": 294.13421630859375, "eval_custom-arc-semantics-data-jp_dot_ap": 0.915716305189008, "eval_custom-arc-semantics-data-jp_dot_f1": 0.9166666666666666, "eval_custom-arc-semantics-data-jp_dot_f1_threshold": 294.13421630859375, "eval_custom-arc-semantics-data-jp_dot_precision": 0.9565217391304348, "eval_custom-arc-semantics-data-jp_dot_recall": 0.88, "eval_custom-arc-semantics-data-jp_euclidean_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_euclidean_accuracy_threshold": 23.818954467773438, "eval_custom-arc-semantics-data-jp_euclidean_ap": 0.9093211275077335, "eval_custom-arc-semantics-data-jp_euclidean_f1": 0.918918918918919, "eval_custom-arc-semantics-data-jp_euclidean_f1_threshold": 23.818954467773438, "eval_custom-arc-semantics-data-jp_euclidean_precision": 0.9315068493150684, "eval_custom-arc-semantics-data-jp_euclidean_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_manhattan_accuracy": 0.9044117647058824, "eval_custom-arc-semantics-data-jp_manhattan_accuracy_threshold": 482.6566162109375, "eval_custom-arc-semantics-data-jp_manhattan_ap": 0.9104676924615509, "eval_custom-arc-semantics-data-jp_manhattan_f1": 0.913907284768212, "eval_custom-arc-semantics-data-jp_manhattan_f1_threshold": 532.9744262695312, "eval_custom-arc-semantics-data-jp_manhattan_precision": 0.9078947368421053, "eval_custom-arc-semantics-data-jp_manhattan_recall": 0.92, "eval_custom-arc-semantics-data-jp_max_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_max_accuracy_threshold": 482.6566162109375, "eval_custom-arc-semantics-data-jp_max_ap": 0.915716305189008, "eval_custom-arc-semantics-data-jp_max_f1": 0.918918918918919, "eval_custom-arc-semantics-data-jp_max_f1_threshold": 532.9744262695312, "eval_custom-arc-semantics-data-jp_max_precision": 0.9565217391304348, "eval_custom-arc-semantics-data-jp_max_recall": 0.92, "eval_loss": 2.2124059200286865, "eval_runtime": 20.6137, "eval_samples_per_second": 6.598, "eval_steps_per_second": 0.825, "step": 816 }, { "epoch": 13.0, "grad_norm": 0.01023839507251978, "learning_rate": 0.0, "loss": 0.1406, "step": 884 }, { "epoch": 13.0, "eval_custom-arc-semantics-data-jp_cosine_accuracy": 0.9044117647058824, "eval_custom-arc-semantics-data-jp_cosine_accuracy_threshold": 0.5485918521881104, "eval_custom-arc-semantics-data-jp_cosine_ap": 0.9088999169341241, "eval_custom-arc-semantics-data-jp_cosine_f1": 0.912751677852349, "eval_custom-arc-semantics-data-jp_cosine_f1_threshold": 0.47659817337989807, "eval_custom-arc-semantics-data-jp_cosine_precision": 0.918918918918919, "eval_custom-arc-semantics-data-jp_cosine_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_dot_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_dot_accuracy_threshold": 293.22845458984375, "eval_custom-arc-semantics-data-jp_dot_ap": 0.9171086358892895, "eval_custom-arc-semantics-data-jp_dot_f1": 0.9166666666666666, "eval_custom-arc-semantics-data-jp_dot_f1_threshold": 293.22845458984375, "eval_custom-arc-semantics-data-jp_dot_precision": 0.9565217391304348, "eval_custom-arc-semantics-data-jp_dot_recall": 0.88, "eval_custom-arc-semantics-data-jp_euclidean_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_euclidean_accuracy_threshold": 23.82940673828125, "eval_custom-arc-semantics-data-jp_euclidean_ap": 0.9094221163568814, "eval_custom-arc-semantics-data-jp_euclidean_f1": 0.918918918918919, "eval_custom-arc-semantics-data-jp_euclidean_f1_threshold": 23.82940673828125, "eval_custom-arc-semantics-data-jp_euclidean_precision": 0.9315068493150684, "eval_custom-arc-semantics-data-jp_euclidean_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_manhattan_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_manhattan_accuracy_threshold": 524.0676879882812, "eval_custom-arc-semantics-data-jp_manhattan_ap": 0.9111567321590129, "eval_custom-arc-semantics-data-jp_manhattan_f1": 0.918918918918919, "eval_custom-arc-semantics-data-jp_manhattan_f1_threshold": 524.0676879882812, "eval_custom-arc-semantics-data-jp_manhattan_precision": 0.9315068493150684, "eval_custom-arc-semantics-data-jp_manhattan_recall": 0.9066666666666666, "eval_custom-arc-semantics-data-jp_max_accuracy": 0.9117647058823529, "eval_custom-arc-semantics-data-jp_max_accuracy_threshold": 524.0676879882812, "eval_custom-arc-semantics-data-jp_max_ap": 0.9171086358892895, "eval_custom-arc-semantics-data-jp_max_f1": 0.918918918918919, "eval_custom-arc-semantics-data-jp_max_f1_threshold": 524.0676879882812, "eval_custom-arc-semantics-data-jp_max_precision": 0.9565217391304348, "eval_custom-arc-semantics-data-jp_max_recall": 0.9066666666666666, "eval_loss": 2.2125871181488037, "eval_runtime": 20.4613, "eval_samples_per_second": 6.647, "eval_steps_per_second": 0.831, "step": 884 } ], "logging_steps": 500, "max_steps": 884, "num_input_tokens_seen": 0, "num_train_epochs": 13, "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": 0.0, "train_batch_size": 8, "trial_name": null, "trial_params": null }