| { |
| "best_global_step": 96, |
| "best_metric": 0.9411764705882353, |
| "best_model_checkpoint": "vit-base-patch16-384-finetuned-humid-binary-2/checkpoint-96", |
| "epoch": 50.0, |
| "eval_steps": 500, |
| "global_step": 300, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 1.0, |
| "eval_accuracy": 0.7205882352941176, |
| "eval_f1_dry": 0.6885245901639344, |
| "eval_f1_humid": 0.7466666666666667, |
| "eval_f1_macro": 0.7175956284153006, |
| "eval_loss": 0.614932119846344, |
| "eval_precision_dry": 0.65625, |
| "eval_precision_humid": 0.7777777777777778, |
| "eval_precision_macro": 0.7170138888888888, |
| "eval_recall_dry": 0.7241379310344828, |
| "eval_recall_humid": 0.717948717948718, |
| "eval_recall_macro": 0.7210433244916004, |
| "eval_runtime": 2.2698, |
| "eval_samples_per_second": 29.959, |
| "eval_steps_per_second": 2.203, |
| "step": 6 |
| }, |
| { |
| "epoch": 1.6666666666666665, |
| "grad_norm": 5.209638595581055, |
| "learning_rate": 1.5e-05, |
| "loss": 0.6101, |
| "step": 10 |
| }, |
| { |
| "epoch": 2.0, |
| "eval_accuracy": 0.8823529411764706, |
| "eval_f1_dry": 0.8620689655172413, |
| "eval_f1_humid": 0.8974358974358975, |
| "eval_f1_macro": 0.8797524314765695, |
| "eval_loss": 0.3864850401878357, |
| "eval_precision_dry": 0.8620689655172413, |
| "eval_precision_humid": 0.8974358974358975, |
| "eval_precision_macro": 0.8797524314765695, |
| "eval_recall_dry": 0.8620689655172413, |
| "eval_recall_humid": 0.8974358974358975, |
| "eval_recall_macro": 0.8797524314765695, |
| "eval_runtime": 2.2426, |
| "eval_samples_per_second": 30.322, |
| "eval_steps_per_second": 2.23, |
| "step": 12 |
| }, |
| { |
| "epoch": 3.0, |
| "eval_accuracy": 0.8382352941176471, |
| "eval_f1_dry": 0.8307692307692308, |
| "eval_f1_humid": 0.8450704225352113, |
| "eval_f1_macro": 0.837919826652221, |
| "eval_loss": 0.35972756147384644, |
| "eval_precision_dry": 0.75, |
| "eval_precision_humid": 0.9375, |
| "eval_precision_macro": 0.84375, |
| "eval_recall_dry": 0.9310344827586207, |
| "eval_recall_humid": 0.7692307692307693, |
| "eval_recall_macro": 0.850132625994695, |
| "eval_runtime": 2.2171, |
| "eval_samples_per_second": 30.671, |
| "eval_steps_per_second": 2.255, |
| "step": 18 |
| }, |
| { |
| "epoch": 3.3333333333333335, |
| "grad_norm": 3.4679043292999268, |
| "learning_rate": 3.1666666666666666e-05, |
| "loss": 0.2613, |
| "step": 20 |
| }, |
| { |
| "epoch": 4.0, |
| "eval_accuracy": 0.8235294117647058, |
| "eval_f1_dry": 0.8235294117647058, |
| "eval_f1_humid": 0.8235294117647058, |
| "eval_f1_macro": 0.8235294117647058, |
| "eval_loss": 0.5013691186904907, |
| "eval_precision_dry": 0.717948717948718, |
| "eval_precision_humid": 0.9655172413793104, |
| "eval_precision_macro": 0.8417329796640142, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.717948717948718, |
| "eval_recall_macro": 0.8417329796640142, |
| "eval_runtime": 2.2748, |
| "eval_samples_per_second": 29.893, |
| "eval_steps_per_second": 2.198, |
| "step": 24 |
| }, |
| { |
| "epoch": 5.0, |
| "grad_norm": 4.22995662689209, |
| "learning_rate": 4.8333333333333334e-05, |
| "loss": 0.1505, |
| "step": 30 |
| }, |
| { |
| "epoch": 5.0, |
| "eval_accuracy": 0.8529411764705882, |
| "eval_f1_dry": 0.84375, |
| "eval_f1_humid": 0.8611111111111112, |
| "eval_f1_macro": 0.8524305555555556, |
| "eval_loss": 0.43828415870666504, |
| "eval_precision_dry": 0.7714285714285715, |
| "eval_precision_humid": 0.9393939393939394, |
| "eval_precision_macro": 0.8554112554112554, |
| "eval_recall_dry": 0.9310344827586207, |
| "eval_recall_humid": 0.7948717948717948, |
| "eval_recall_macro": 0.8629531388152077, |
| "eval_runtime": 2.1723, |
| "eval_samples_per_second": 31.303, |
| "eval_steps_per_second": 2.302, |
| "step": 30 |
| }, |
| { |
| "epoch": 6.0, |
| "eval_accuracy": 0.8529411764705882, |
| "eval_f1_dry": 0.8484848484848485, |
| "eval_f1_humid": 0.8571428571428571, |
| "eval_f1_macro": 0.8528138528138528, |
| "eval_loss": 0.4248289167881012, |
| "eval_precision_dry": 0.7567567567567568, |
| "eval_precision_humid": 0.967741935483871, |
| "eval_precision_macro": 0.8622493461203139, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.7692307692307693, |
| "eval_recall_macro": 0.8673740053050398, |
| "eval_runtime": 2.3123, |
| "eval_samples_per_second": 29.408, |
| "eval_steps_per_second": 2.162, |
| "step": 36 |
| }, |
| { |
| "epoch": 6.666666666666667, |
| "grad_norm": 6.771641254425049, |
| "learning_rate": 4.8333333333333334e-05, |
| "loss": 0.1298, |
| "step": 40 |
| }, |
| { |
| "epoch": 7.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8727272727272727, |
| "eval_f1_humid": 0.9135802469135802, |
| "eval_f1_macro": 0.8931537598204264, |
| "eval_loss": 0.39324793219566345, |
| "eval_precision_dry": 0.9230769230769231, |
| "eval_precision_humid": 0.8809523809523809, |
| "eval_precision_macro": 0.902014652014652, |
| "eval_recall_dry": 0.8275862068965517, |
| "eval_recall_humid": 0.9487179487179487, |
| "eval_recall_macro": 0.8881520778072503, |
| "eval_runtime": 2.2011, |
| "eval_samples_per_second": 30.894, |
| "eval_steps_per_second": 2.272, |
| "step": 42 |
| }, |
| { |
| "epoch": 8.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8852459016393442, |
| "eval_f1_humid": 0.9066666666666666, |
| "eval_f1_macro": 0.8959562841530054, |
| "eval_loss": 0.3932932913303375, |
| "eval_precision_dry": 0.84375, |
| "eval_precision_humid": 0.9444444444444444, |
| "eval_precision_macro": 0.8940972222222222, |
| "eval_recall_dry": 0.9310344827586207, |
| "eval_recall_humid": 0.8717948717948718, |
| "eval_recall_macro": 0.9014146772767462, |
| "eval_runtime": 2.1675, |
| "eval_samples_per_second": 31.373, |
| "eval_steps_per_second": 2.307, |
| "step": 48 |
| }, |
| { |
| "epoch": 8.333333333333334, |
| "grad_norm": 2.7544643878936768, |
| "learning_rate": 4.648148148148148e-05, |
| "loss": 0.1033, |
| "step": 50 |
| }, |
| { |
| "epoch": 9.0, |
| "eval_accuracy": 0.7941176470588235, |
| "eval_f1_dry": 0.8055555555555556, |
| "eval_f1_humid": 0.78125, |
| "eval_f1_macro": 0.7934027777777778, |
| "eval_loss": 0.7160156965255737, |
| "eval_precision_dry": 0.6744186046511628, |
| "eval_precision_humid": 1.0, |
| "eval_precision_macro": 0.8372093023255813, |
| "eval_recall_dry": 1.0, |
| "eval_recall_humid": 0.6410256410256411, |
| "eval_recall_macro": 0.8205128205128205, |
| "eval_runtime": 2.1872, |
| "eval_samples_per_second": 31.09, |
| "eval_steps_per_second": 2.286, |
| "step": 54 |
| }, |
| { |
| "epoch": 10.0, |
| "grad_norm": 2.3722591400146484, |
| "learning_rate": 4.462962962962963e-05, |
| "loss": 0.1135, |
| "step": 60 |
| }, |
| { |
| "epoch": 10.0, |
| "eval_accuracy": 0.9264705882352942, |
| "eval_f1_dry": 0.9152542372881356, |
| "eval_f1_humid": 0.935064935064935, |
| "eval_f1_macro": 0.9251595861765354, |
| "eval_loss": 0.40363243222236633, |
| "eval_precision_dry": 0.9, |
| "eval_precision_humid": 0.9473684210526315, |
| "eval_precision_macro": 0.9236842105263158, |
| "eval_recall_dry": 0.9310344827586207, |
| "eval_recall_humid": 0.9230769230769231, |
| "eval_recall_macro": 0.927055702917772, |
| "eval_runtime": 2.1789, |
| "eval_samples_per_second": 31.209, |
| "eval_steps_per_second": 2.295, |
| "step": 60 |
| }, |
| { |
| "epoch": 11.0, |
| "eval_accuracy": 0.9117647058823529, |
| "eval_f1_dry": 0.8928571428571429, |
| "eval_f1_humid": 0.925, |
| "eval_f1_macro": 0.9089285714285715, |
| "eval_loss": 0.4228282868862152, |
| "eval_precision_dry": 0.9259259259259259, |
| "eval_precision_humid": 0.9024390243902439, |
| "eval_precision_macro": 0.9141824751580849, |
| "eval_recall_dry": 0.8620689655172413, |
| "eval_recall_humid": 0.9487179487179487, |
| "eval_recall_macro": 0.905393457117595, |
| "eval_runtime": 2.2933, |
| "eval_samples_per_second": 29.651, |
| "eval_steps_per_second": 2.18, |
| "step": 66 |
| }, |
| { |
| "epoch": 11.666666666666666, |
| "grad_norm": 0.39826828241348267, |
| "learning_rate": 4.277777777777778e-05, |
| "loss": 0.0331, |
| "step": 70 |
| }, |
| { |
| "epoch": 12.0, |
| "eval_accuracy": 0.8823529411764706, |
| "eval_f1_dry": 0.875, |
| "eval_f1_humid": 0.8888888888888888, |
| "eval_f1_macro": 0.8819444444444444, |
| "eval_loss": 0.4820828139781952, |
| "eval_precision_dry": 0.8, |
| "eval_precision_humid": 0.9696969696969697, |
| "eval_precision_macro": 0.8848484848484849, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8205128205128205, |
| "eval_recall_macro": 0.8930150309460654, |
| "eval_runtime": 2.2848, |
| "eval_samples_per_second": 29.762, |
| "eval_steps_per_second": 2.188, |
| "step": 72 |
| }, |
| { |
| "epoch": 13.0, |
| "eval_accuracy": 0.9117647058823529, |
| "eval_f1_dry": 0.9, |
| "eval_f1_humid": 0.9210526315789473, |
| "eval_f1_macro": 0.9105263157894736, |
| "eval_loss": 0.4474344849586487, |
| "eval_precision_dry": 0.8709677419354839, |
| "eval_precision_humid": 0.9459459459459459, |
| "eval_precision_macro": 0.9084568439407149, |
| "eval_recall_dry": 0.9310344827586207, |
| "eval_recall_humid": 0.8974358974358975, |
| "eval_recall_macro": 0.9142351900972591, |
| "eval_runtime": 2.3596, |
| "eval_samples_per_second": 28.819, |
| "eval_steps_per_second": 2.119, |
| "step": 78 |
| }, |
| { |
| "epoch": 13.333333333333334, |
| "grad_norm": 4.4798173904418945, |
| "learning_rate": 4.092592592592593e-05, |
| "loss": 0.0142, |
| "step": 80 |
| }, |
| { |
| "epoch": 14.0, |
| "eval_accuracy": 0.9117647058823529, |
| "eval_f1_dry": 0.896551724137931, |
| "eval_f1_humid": 0.9230769230769231, |
| "eval_f1_macro": 0.9098143236074271, |
| "eval_loss": 0.4478212893009186, |
| "eval_precision_dry": 0.896551724137931, |
| "eval_precision_humid": 0.9230769230769231, |
| "eval_precision_macro": 0.9098143236074271, |
| "eval_recall_dry": 0.896551724137931, |
| "eval_recall_humid": 0.9230769230769231, |
| "eval_recall_macro": 0.9098143236074271, |
| "eval_runtime": 2.1905, |
| "eval_samples_per_second": 31.043, |
| "eval_steps_per_second": 2.283, |
| "step": 84 |
| }, |
| { |
| "epoch": 15.0, |
| "grad_norm": 0.2734902501106262, |
| "learning_rate": 3.9074074074074076e-05, |
| "loss": 0.0074, |
| "step": 90 |
| }, |
| { |
| "epoch": 15.0, |
| "eval_accuracy": 0.8823529411764706, |
| "eval_f1_dry": 0.875, |
| "eval_f1_humid": 0.8888888888888888, |
| "eval_f1_macro": 0.8819444444444444, |
| "eval_loss": 0.6301909685134888, |
| "eval_precision_dry": 0.8, |
| "eval_precision_humid": 0.9696969696969697, |
| "eval_precision_macro": 0.8848484848484849, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8205128205128205, |
| "eval_recall_macro": 0.8930150309460654, |
| "eval_runtime": 2.2, |
| "eval_samples_per_second": 30.909, |
| "eval_steps_per_second": 2.273, |
| "step": 90 |
| }, |
| { |
| "epoch": 16.0, |
| "eval_accuracy": 0.9411764705882353, |
| "eval_f1_dry": 0.9310344827586207, |
| "eval_f1_humid": 0.9487179487179487, |
| "eval_f1_macro": 0.9398762157382847, |
| "eval_loss": 0.4611157774925232, |
| "eval_precision_dry": 0.9310344827586207, |
| "eval_precision_humid": 0.9487179487179487, |
| "eval_precision_macro": 0.9398762157382847, |
| "eval_recall_dry": 0.9310344827586207, |
| "eval_recall_humid": 0.9487179487179487, |
| "eval_recall_macro": 0.9398762157382847, |
| "eval_runtime": 2.2029, |
| "eval_samples_per_second": 30.868, |
| "eval_steps_per_second": 2.27, |
| "step": 96 |
| }, |
| { |
| "epoch": 16.666666666666668, |
| "grad_norm": 0.033265743404626846, |
| "learning_rate": 3.722222222222222e-05, |
| "loss": 0.0114, |
| "step": 100 |
| }, |
| { |
| "epoch": 17.0, |
| "eval_accuracy": 0.8823529411764706, |
| "eval_f1_dry": 0.875, |
| "eval_f1_humid": 0.8888888888888888, |
| "eval_f1_macro": 0.8819444444444444, |
| "eval_loss": 0.6472479104995728, |
| "eval_precision_dry": 0.8, |
| "eval_precision_humid": 0.9696969696969697, |
| "eval_precision_macro": 0.8848484848484849, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8205128205128205, |
| "eval_recall_macro": 0.8930150309460654, |
| "eval_runtime": 2.1737, |
| "eval_samples_per_second": 31.282, |
| "eval_steps_per_second": 2.3, |
| "step": 102 |
| }, |
| { |
| "epoch": 18.0, |
| "eval_accuracy": 0.9117647058823529, |
| "eval_f1_dry": 0.8928571428571429, |
| "eval_f1_humid": 0.925, |
| "eval_f1_macro": 0.9089285714285715, |
| "eval_loss": 0.4689759910106659, |
| "eval_precision_dry": 0.9259259259259259, |
| "eval_precision_humid": 0.9024390243902439, |
| "eval_precision_macro": 0.9141824751580849, |
| "eval_recall_dry": 0.8620689655172413, |
| "eval_recall_humid": 0.9487179487179487, |
| "eval_recall_macro": 0.905393457117595, |
| "eval_runtime": 2.1917, |
| "eval_samples_per_second": 31.027, |
| "eval_steps_per_second": 2.281, |
| "step": 108 |
| }, |
| { |
| "epoch": 18.333333333333332, |
| "grad_norm": 0.9351687431335449, |
| "learning_rate": 3.537037037037037e-05, |
| "loss": 0.0054, |
| "step": 110 |
| }, |
| { |
| "epoch": 19.0, |
| "eval_accuracy": 0.8676470588235294, |
| "eval_f1_dry": 0.8615384615384616, |
| "eval_f1_humid": 0.8732394366197183, |
| "eval_f1_macro": 0.8673889490790899, |
| "eval_loss": 0.7610320448875427, |
| "eval_precision_dry": 0.7777777777777778, |
| "eval_precision_humid": 0.96875, |
| "eval_precision_macro": 0.8732638888888888, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.7948717948717948, |
| "eval_recall_macro": 0.8801945181255526, |
| "eval_runtime": 2.1632, |
| "eval_samples_per_second": 31.435, |
| "eval_steps_per_second": 2.311, |
| "step": 114 |
| }, |
| { |
| "epoch": 20.0, |
| "grad_norm": 0.03664712235331535, |
| "learning_rate": 3.351851851851852e-05, |
| "loss": 0.0024, |
| "step": 120 |
| }, |
| { |
| "epoch": 20.0, |
| "eval_accuracy": 0.9117647058823529, |
| "eval_f1_dry": 0.9, |
| "eval_f1_humid": 0.9210526315789473, |
| "eval_f1_macro": 0.9105263157894736, |
| "eval_loss": 0.5395424365997314, |
| "eval_precision_dry": 0.8709677419354839, |
| "eval_precision_humid": 0.9459459459459459, |
| "eval_precision_macro": 0.9084568439407149, |
| "eval_recall_dry": 0.9310344827586207, |
| "eval_recall_humid": 0.8974358974358975, |
| "eval_recall_macro": 0.9142351900972591, |
| "eval_runtime": 2.1855, |
| "eval_samples_per_second": 31.114, |
| "eval_steps_per_second": 2.288, |
| "step": 120 |
| }, |
| { |
| "epoch": 21.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8852459016393442, |
| "eval_f1_humid": 0.9066666666666666, |
| "eval_f1_macro": 0.8959562841530054, |
| "eval_loss": 0.6274784207344055, |
| "eval_precision_dry": 0.84375, |
| "eval_precision_humid": 0.9444444444444444, |
| "eval_precision_macro": 0.8940972222222222, |
| "eval_recall_dry": 0.9310344827586207, |
| "eval_recall_humid": 0.8717948717948718, |
| "eval_recall_macro": 0.9014146772767462, |
| "eval_runtime": 2.2125, |
| "eval_samples_per_second": 30.734, |
| "eval_steps_per_second": 2.26, |
| "step": 126 |
| }, |
| { |
| "epoch": 21.666666666666668, |
| "grad_norm": 0.053658369928598404, |
| "learning_rate": 3.1666666666666666e-05, |
| "loss": 0.0024, |
| "step": 130 |
| }, |
| { |
| "epoch": 22.0, |
| "eval_accuracy": 0.8676470588235294, |
| "eval_f1_dry": 0.8615384615384616, |
| "eval_f1_humid": 0.8732394366197183, |
| "eval_f1_macro": 0.8673889490790899, |
| "eval_loss": 0.7972272634506226, |
| "eval_precision_dry": 0.7777777777777778, |
| "eval_precision_humid": 0.96875, |
| "eval_precision_macro": 0.8732638888888888, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.7948717948717948, |
| "eval_recall_macro": 0.8801945181255526, |
| "eval_runtime": 2.2057, |
| "eval_samples_per_second": 30.829, |
| "eval_steps_per_second": 2.267, |
| "step": 132 |
| }, |
| { |
| "epoch": 23.0, |
| "eval_accuracy": 0.9117647058823529, |
| "eval_f1_dry": 0.9, |
| "eval_f1_humid": 0.9210526315789473, |
| "eval_f1_macro": 0.9105263157894736, |
| "eval_loss": 0.5465890169143677, |
| "eval_precision_dry": 0.8709677419354839, |
| "eval_precision_humid": 0.9459459459459459, |
| "eval_precision_macro": 0.9084568439407149, |
| "eval_recall_dry": 0.9310344827586207, |
| "eval_recall_humid": 0.8974358974358975, |
| "eval_recall_macro": 0.9142351900972591, |
| "eval_runtime": 2.2093, |
| "eval_samples_per_second": 30.779, |
| "eval_steps_per_second": 2.263, |
| "step": 138 |
| }, |
| { |
| "epoch": 23.333333333333332, |
| "grad_norm": 0.01843937672674656, |
| "learning_rate": 2.981481481481482e-05, |
| "loss": 0.0009, |
| "step": 140 |
| }, |
| { |
| "epoch": 24.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8852459016393442, |
| "eval_f1_humid": 0.9066666666666666, |
| "eval_f1_macro": 0.8959562841530054, |
| "eval_loss": 0.5303817987442017, |
| "eval_precision_dry": 0.84375, |
| "eval_precision_humid": 0.9444444444444444, |
| "eval_precision_macro": 0.8940972222222222, |
| "eval_recall_dry": 0.9310344827586207, |
| "eval_recall_humid": 0.8717948717948718, |
| "eval_recall_macro": 0.9014146772767462, |
| "eval_runtime": 2.1789, |
| "eval_samples_per_second": 31.208, |
| "eval_steps_per_second": 2.295, |
| "step": 144 |
| }, |
| { |
| "epoch": 25.0, |
| "grad_norm": 0.0267292782664299, |
| "learning_rate": 2.7962962962962965e-05, |
| "loss": 0.0012, |
| "step": 150 |
| }, |
| { |
| "epoch": 25.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.656761646270752, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.288, |
| "eval_samples_per_second": 29.721, |
| "eval_steps_per_second": 2.185, |
| "step": 150 |
| }, |
| { |
| "epoch": 26.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.613365650177002, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.3183, |
| "eval_samples_per_second": 29.332, |
| "eval_steps_per_second": 2.157, |
| "step": 156 |
| }, |
| { |
| "epoch": 26.666666666666668, |
| "grad_norm": 0.007465701084583998, |
| "learning_rate": 2.6111111111111114e-05, |
| "loss": 0.0004, |
| "step": 160 |
| }, |
| { |
| "epoch": 27.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.6093538999557495, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.2556, |
| "eval_samples_per_second": 30.148, |
| "eval_steps_per_second": 2.217, |
| "step": 162 |
| }, |
| { |
| "epoch": 28.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.615250825881958, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.1738, |
| "eval_samples_per_second": 31.282, |
| "eval_steps_per_second": 2.3, |
| "step": 168 |
| }, |
| { |
| "epoch": 28.333333333333332, |
| "grad_norm": 0.01064391154795885, |
| "learning_rate": 2.425925925925926e-05, |
| "loss": 0.0003, |
| "step": 170 |
| }, |
| { |
| "epoch": 29.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.6319212913513184, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.1953, |
| "eval_samples_per_second": 30.976, |
| "eval_steps_per_second": 2.278, |
| "step": 174 |
| }, |
| { |
| "epoch": 30.0, |
| "grad_norm": 0.006854104809463024, |
| "learning_rate": 2.240740740740741e-05, |
| "loss": 0.0002, |
| "step": 180 |
| }, |
| { |
| "epoch": 30.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.6456788182258606, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.1711, |
| "eval_samples_per_second": 31.321, |
| "eval_steps_per_second": 2.303, |
| "step": 180 |
| }, |
| { |
| "epoch": 31.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.6529943346977234, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.2044, |
| "eval_samples_per_second": 30.848, |
| "eval_steps_per_second": 2.268, |
| "step": 186 |
| }, |
| { |
| "epoch": 31.666666666666668, |
| "grad_norm": 0.006000952795147896, |
| "learning_rate": 2.0555555555555555e-05, |
| "loss": 0.0002, |
| "step": 190 |
| }, |
| { |
| "epoch": 32.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.6579628586769104, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.1932, |
| "eval_samples_per_second": 31.005, |
| "eval_steps_per_second": 2.28, |
| "step": 192 |
| }, |
| { |
| "epoch": 33.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.6649767160415649, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.1922, |
| "eval_samples_per_second": 31.019, |
| "eval_steps_per_second": 2.281, |
| "step": 198 |
| }, |
| { |
| "epoch": 33.333333333333336, |
| "grad_norm": 0.0073044863529503345, |
| "learning_rate": 1.8703703703703704e-05, |
| "loss": 0.0002, |
| "step": 200 |
| }, |
| { |
| "epoch": 34.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.677439272403717, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.2369, |
| "eval_samples_per_second": 30.4, |
| "eval_steps_per_second": 2.235, |
| "step": 204 |
| }, |
| { |
| "epoch": 35.0, |
| "grad_norm": 0.00664471136406064, |
| "learning_rate": 1.6851851851851853e-05, |
| "loss": 0.0002, |
| "step": 210 |
| }, |
| { |
| "epoch": 35.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.6827277541160583, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.2152, |
| "eval_samples_per_second": 30.697, |
| "eval_steps_per_second": 2.257, |
| "step": 210 |
| }, |
| { |
| "epoch": 36.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.6850156188011169, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.1962, |
| "eval_samples_per_second": 30.963, |
| "eval_steps_per_second": 2.277, |
| "step": 216 |
| }, |
| { |
| "epoch": 36.666666666666664, |
| "grad_norm": 0.003711260389536619, |
| "learning_rate": 1.5e-05, |
| "loss": 0.0001, |
| "step": 220 |
| }, |
| { |
| "epoch": 37.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.684583842754364, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.2135, |
| "eval_samples_per_second": 30.721, |
| "eval_steps_per_second": 2.259, |
| "step": 222 |
| }, |
| { |
| "epoch": 38.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.6853307485580444, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.2406, |
| "eval_samples_per_second": 30.348, |
| "eval_steps_per_second": 2.231, |
| "step": 228 |
| }, |
| { |
| "epoch": 38.333333333333336, |
| "grad_norm": 0.003823125036433339, |
| "learning_rate": 1.3148148148148148e-05, |
| "loss": 0.0001, |
| "step": 230 |
| }, |
| { |
| "epoch": 39.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.6853978633880615, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.1744, |
| "eval_samples_per_second": 31.273, |
| "eval_steps_per_second": 2.299, |
| "step": 234 |
| }, |
| { |
| "epoch": 40.0, |
| "grad_norm": 0.0067051672376692295, |
| "learning_rate": 1.1296296296296297e-05, |
| "loss": 0.0001, |
| "step": 240 |
| }, |
| { |
| "epoch": 40.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.6844778656959534, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.2084, |
| "eval_samples_per_second": 30.791, |
| "eval_steps_per_second": 2.264, |
| "step": 240 |
| }, |
| { |
| "epoch": 41.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.6846556663513184, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.2966, |
| "eval_samples_per_second": 29.61, |
| "eval_steps_per_second": 2.177, |
| "step": 246 |
| }, |
| { |
| "epoch": 41.666666666666664, |
| "grad_norm": 0.005940043367445469, |
| "learning_rate": 9.444444444444445e-06, |
| "loss": 0.0001, |
| "step": 250 |
| }, |
| { |
| "epoch": 42.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.6859800815582275, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.2361, |
| "eval_samples_per_second": 30.41, |
| "eval_steps_per_second": 2.236, |
| "step": 252 |
| }, |
| { |
| "epoch": 43.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.6874513030052185, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.1797, |
| "eval_samples_per_second": 31.196, |
| "eval_steps_per_second": 2.294, |
| "step": 258 |
| }, |
| { |
| "epoch": 43.333333333333336, |
| "grad_norm": 0.003746297210454941, |
| "learning_rate": 7.592592592592593e-06, |
| "loss": 0.0001, |
| "step": 260 |
| }, |
| { |
| "epoch": 44.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.6894810199737549, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.2822, |
| "eval_samples_per_second": 29.796, |
| "eval_steps_per_second": 2.191, |
| "step": 264 |
| }, |
| { |
| "epoch": 45.0, |
| "grad_norm": 0.010260913521051407, |
| "learning_rate": 5.740740740740741e-06, |
| "loss": 0.0001, |
| "step": 270 |
| }, |
| { |
| "epoch": 45.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.6902977824211121, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.2179, |
| "eval_samples_per_second": 30.66, |
| "eval_steps_per_second": 2.254, |
| "step": 270 |
| }, |
| { |
| "epoch": 46.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.691429078578949, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.2021, |
| "eval_samples_per_second": 30.88, |
| "eval_steps_per_second": 2.271, |
| "step": 276 |
| }, |
| { |
| "epoch": 46.666666666666664, |
| "grad_norm": 0.003823869628831744, |
| "learning_rate": 3.888888888888889e-06, |
| "loss": 0.0001, |
| "step": 280 |
| }, |
| { |
| "epoch": 47.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.6919637322425842, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.1849, |
| "eval_samples_per_second": 31.123, |
| "eval_steps_per_second": 2.288, |
| "step": 282 |
| }, |
| { |
| "epoch": 48.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.6921000480651855, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.1918, |
| "eval_samples_per_second": 31.025, |
| "eval_steps_per_second": 2.281, |
| "step": 288 |
| }, |
| { |
| "epoch": 48.333333333333336, |
| "grad_norm": 0.004158890340477228, |
| "learning_rate": 2.0370370370370375e-06, |
| "loss": 0.0001, |
| "step": 290 |
| }, |
| { |
| "epoch": 49.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.6922561526298523, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.2198, |
| "eval_samples_per_second": 30.634, |
| "eval_steps_per_second": 2.252, |
| "step": 294 |
| }, |
| { |
| "epoch": 50.0, |
| "grad_norm": 0.003357668872922659, |
| "learning_rate": 1.851851851851852e-07, |
| "loss": 0.0001, |
| "step": 300 |
| }, |
| { |
| "epoch": 50.0, |
| "eval_accuracy": 0.8970588235294118, |
| "eval_f1_dry": 0.8888888888888888, |
| "eval_f1_humid": 0.9041095890410958, |
| "eval_f1_macro": 0.8964992389649924, |
| "eval_loss": 0.692406177520752, |
| "eval_precision_dry": 0.8235294117647058, |
| "eval_precision_humid": 0.9705882352941176, |
| "eval_precision_macro": 0.8970588235294117, |
| "eval_recall_dry": 0.9655172413793104, |
| "eval_recall_humid": 0.8461538461538461, |
| "eval_recall_macro": 0.9058355437665783, |
| "eval_runtime": 2.2412, |
| "eval_samples_per_second": 30.341, |
| "eval_steps_per_second": 2.231, |
| "step": 300 |
| }, |
| { |
| "epoch": 50.0, |
| "step": 300, |
| "total_flos": 4.3301918339334144e+18, |
| "train_loss": 0.04831398472228708, |
| "train_runtime": 1645.2014, |
| "train_samples_per_second": 11.518, |
| "train_steps_per_second": 0.182 |
| } |
| ], |
| "logging_steps": 10, |
| "max_steps": 300, |
| "num_input_tokens_seen": 0, |
| "num_train_epochs": 50, |
| "save_steps": 500, |
| "stateful_callbacks": { |
| "TrainerControl": { |
| "args": { |
| "should_epoch_stop": false, |
| "should_evaluate": false, |
| "should_log": false, |
| "should_save": true, |
| "should_training_stop": true |
| }, |
| "attributes": {} |
| } |
| }, |
| "total_flos": 4.3301918339334144e+18, |
| "train_batch_size": 16, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
