diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,4531 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 3.980891719745223, + "eval_steps": 500, + "global_step": 5000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.01, + "learning_rate": 9.196195623374133e-06, + "loss": 0.9546, + "max_memory_allocated (GB)": 82.79, + "memory_allocated (GB)": 23.35, + "step": 10, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.02, + "learning_rate": 1.1961516052046928e-05, + "loss": 0.8455, + "max_memory_allocated (GB)": 82.79, + "memory_allocated (GB)": 23.35, + "step": 20, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.02, + "learning_rate": 1.3579124805298667e-05, + "loss": 0.8202, + "max_memory_allocated (GB)": 82.79, + "memory_allocated (GB)": 23.35, + "step": 30, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.03, + "learning_rate": 1.4726836480719725e-05, + "loss": 0.8082, + "max_memory_allocated (GB)": 82.79, + "memory_allocated (GB)": 23.35, + "step": 40, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.04, + "learning_rate": 1.5617070818075464e-05, + "loss": 0.8079, + "max_memory_allocated (GB)": 82.79, + "memory_allocated (GB)": 23.35, + "step": 50, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.05, + "learning_rate": 1.6344445233971464e-05, + "loss": 0.7957, + "max_memory_allocated (GB)": 82.79, + "memory_allocated (GB)": 23.35, + "step": 60, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.06, + "learning_rate": 1.6959431539875155e-05, + "loss": 0.7643, + "max_memory_allocated (GB)": 82.79, + "memory_allocated (GB)": 23.35, + "step": 70, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.06, + "learning_rate": 1.7492156909392524e-05, + "loss": 0.769, + "max_memory_allocated (GB)": 82.79, + "memory_allocated (GB)": 23.35, + "step": 80, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.07, + "learning_rate": 1.796205398722321e-05, + "loss": 0.7477, + "max_memory_allocated (GB)": 82.79, + "memory_allocated (GB)": 23.35, + "step": 90, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.08, + "learning_rate": 1.8382391246748267e-05, + "loss": 0.752, + "max_memory_allocated (GB)": 82.79, + "memory_allocated (GB)": 23.35, + "step": 100, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.09, + "learning_rate": 1.8762632549988454e-05, + "loss": 0.7523, + "max_memory_allocated (GB)": 82.79, + "memory_allocated (GB)": 23.35, + "step": 110, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.1, + "learning_rate": 1.9109765662644265e-05, + "loss": 0.7313, + "max_memory_allocated (GB)": 82.79, + "memory_allocated (GB)": 23.35, + "step": 120, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.1, + "learning_rate": 1.9429097171021903e-05, + "loss": 0.7147, + "max_memory_allocated (GB)": 82.79, + "memory_allocated (GB)": 23.35, + "step": 130, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.11, + "learning_rate": 1.9724751968547952e-05, + "loss": 0.7018, + "max_memory_allocated (GB)": 82.79, + "memory_allocated (GB)": 23.35, + "step": 140, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.12, + "learning_rate": 2e-05, + "loss": 0.7227, + "max_memory_allocated (GB)": 82.79, + "memory_allocated (GB)": 23.35, + "step": 150, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.13, + "learning_rate": 1.996290515463918e-05, + "loss": 0.7256, + "max_memory_allocated (GB)": 82.79, + "memory_allocated (GB)": 23.35, + "step": 160, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.14, + "learning_rate": 1.9921688659793814e-05, + "loss": 0.7029, + "max_memory_allocated (GB)": 82.79, + "memory_allocated (GB)": 23.35, + "step": 170, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.14, + "learning_rate": 1.9880472164948456e-05, + "loss": 0.6927, + "max_memory_allocated (GB)": 82.79, + "memory_allocated (GB)": 23.35, + "step": 180, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.15, + "learning_rate": 1.9839255670103094e-05, + "loss": 0.6717, + "max_memory_allocated (GB)": 82.79, + "memory_allocated (GB)": 23.35, + "step": 190, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.16, + "learning_rate": 1.9798039175257733e-05, + "loss": 0.685, + "max_memory_allocated (GB)": 82.79, + "memory_allocated (GB)": 23.35, + "step": 200, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.17, + "learning_rate": 1.975682268041237e-05, + "loss": 0.669, + "max_memory_allocated (GB)": 82.79, + "memory_allocated (GB)": 23.35, + "step": 210, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.18, + "learning_rate": 1.971560618556701e-05, + "loss": 0.6623, + "max_memory_allocated (GB)": 82.79, + "memory_allocated (GB)": 23.35, + "step": 220, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.18, + "learning_rate": 1.9674389690721652e-05, + "loss": 0.6552, + "max_memory_allocated (GB)": 82.8, + "memory_allocated (GB)": 23.35, + "step": 230, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.19, + "learning_rate": 1.963317319587629e-05, + "loss": 0.6479, + "max_memory_allocated (GB)": 82.8, + "memory_allocated (GB)": 23.35, + "step": 240, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.2, + "learning_rate": 1.959195670103093e-05, + "loss": 0.653, + "max_memory_allocated (GB)": 82.8, + "memory_allocated (GB)": 23.35, + "step": 250, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.21, + "learning_rate": 1.9550740206185568e-05, + "loss": 0.6267, + "max_memory_allocated (GB)": 82.8, + "memory_allocated (GB)": 23.35, + "step": 260, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.21, + "learning_rate": 1.9509523711340207e-05, + "loss": 0.627, + "max_memory_allocated (GB)": 82.8, + "memory_allocated (GB)": 23.35, + "step": 270, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.22, + "learning_rate": 1.946830721649485e-05, + "loss": 0.6384, + "max_memory_allocated (GB)": 82.8, + "memory_allocated (GB)": 23.35, + "step": 280, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.23, + "learning_rate": 1.9427090721649488e-05, + "loss": 0.6149, + "max_memory_allocated (GB)": 82.8, + "memory_allocated (GB)": 23.35, + "step": 290, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.24, + "learning_rate": 1.9385874226804126e-05, + "loss": 0.608, + "max_memory_allocated (GB)": 82.8, + "memory_allocated (GB)": 23.35, + "step": 300, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.25, + "learning_rate": 1.9344657731958765e-05, + "loss": 0.6031, + "max_memory_allocated (GB)": 82.8, + "memory_allocated (GB)": 23.35, + "step": 310, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.25, + "learning_rate": 1.9303441237113403e-05, + "loss": 0.5952, + "max_memory_allocated (GB)": 82.8, + "memory_allocated (GB)": 23.35, + "step": 320, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.26, + "learning_rate": 1.9262224742268042e-05, + "loss": 0.5995, + "max_memory_allocated (GB)": 82.8, + "memory_allocated (GB)": 23.35, + "step": 330, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.27, + "learning_rate": 1.922100824742268e-05, + "loss": 0.5761, + "max_memory_allocated (GB)": 82.8, + "memory_allocated (GB)": 23.35, + "step": 340, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.28, + "learning_rate": 1.917979175257732e-05, + "loss": 0.5904, + "max_memory_allocated (GB)": 82.8, + "memory_allocated (GB)": 23.35, + "step": 350, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.29, + "learning_rate": 1.913857525773196e-05, + "loss": 0.5747, + "max_memory_allocated (GB)": 82.8, + "memory_allocated (GB)": 23.35, + "step": 360, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.29, + "learning_rate": 1.90973587628866e-05, + "loss": 0.5901, + "max_memory_allocated (GB)": 82.8, + "memory_allocated (GB)": 23.35, + "step": 370, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.3, + "learning_rate": 1.905614226804124e-05, + "loss": 0.5786, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 380, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.31, + "learning_rate": 1.9014925773195877e-05, + "loss": 0.5677, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 390, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.32, + "learning_rate": 1.8973709278350516e-05, + "loss": 0.5579, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 400, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.33, + "learning_rate": 1.8932492783505158e-05, + "loss": 0.5577, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 410, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.33, + "learning_rate": 1.8891276288659797e-05, + "loss": 0.5603, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 420, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.34, + "learning_rate": 1.8850059793814435e-05, + "loss": 0.5421, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 430, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.35, + "learning_rate": 1.8808843298969074e-05, + "loss": 0.5356, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 440, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.36, + "learning_rate": 1.8767626804123712e-05, + "loss": 0.5395, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 450, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.37, + "learning_rate": 1.8726410309278355e-05, + "loss": 0.5466, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 460, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.37, + "learning_rate": 1.868519381443299e-05, + "loss": 0.51, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 470, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.38, + "learning_rate": 1.864397731958763e-05, + "loss": 0.5222, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 480, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.39, + "learning_rate": 1.860276082474227e-05, + "loss": 0.4956, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 490, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.4, + "learning_rate": 1.856154432989691e-05, + "loss": 0.5064, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 500, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.41, + "learning_rate": 1.8520327835051548e-05, + "loss": 0.4982, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 510, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.41, + "learning_rate": 1.8479111340206186e-05, + "loss": 0.4953, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 520, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.42, + "learning_rate": 1.8437894845360825e-05, + "loss": 0.5129, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 530, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.43, + "learning_rate": 1.8396678350515467e-05, + "loss": 0.4947, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 540, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.44, + "learning_rate": 1.8355461855670106e-05, + "loss": 0.4699, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 550, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.45, + "learning_rate": 1.8314245360824744e-05, + "loss": 0.481, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 560, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.45, + "learning_rate": 1.8273028865979383e-05, + "loss": 0.4848, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 570, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.46, + "learning_rate": 1.823181237113402e-05, + "loss": 0.4886, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 580, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.47, + "learning_rate": 1.8190595876288664e-05, + "loss": 0.4774, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 590, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.48, + "learning_rate": 1.8149379381443302e-05, + "loss": 0.4583, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 600, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.49, + "learning_rate": 1.810816288659794e-05, + "loss": 0.4781, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 610, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.49, + "learning_rate": 1.806694639175258e-05, + "loss": 0.4347, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 620, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.5, + "learning_rate": 1.8025729896907218e-05, + "loss": 0.4478, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 630, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.51, + "learning_rate": 1.7984513402061857e-05, + "loss": 0.4488, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 640, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.52, + "learning_rate": 1.7943296907216495e-05, + "loss": 0.441, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 650, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.53, + "learning_rate": 1.7902080412371134e-05, + "loss": 0.4514, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 660, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.53, + "learning_rate": 1.7860863917525776e-05, + "loss": 0.4419, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 670, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.54, + "learning_rate": 1.7819647422680415e-05, + "loss": 0.4446, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 680, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.55, + "learning_rate": 1.7778430927835053e-05, + "loss": 0.4313, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 690, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.56, + "learning_rate": 1.7737214432989692e-05, + "loss": 0.4358, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 700, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.57, + "learning_rate": 1.769599793814433e-05, + "loss": 0.4289, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 710, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.57, + "learning_rate": 1.7654781443298973e-05, + "loss": 0.4178, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 720, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.58, + "learning_rate": 1.761356494845361e-05, + "loss": 0.4108, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 730, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.59, + "learning_rate": 1.757234845360825e-05, + "loss": 0.4269, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 740, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.6, + "learning_rate": 1.753113195876289e-05, + "loss": 0.4093, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 750, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.61, + "learning_rate": 1.7489915463917527e-05, + "loss": 0.4003, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 760, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.61, + "learning_rate": 1.744869896907217e-05, + "loss": 0.4136, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 770, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.62, + "learning_rate": 1.7407482474226804e-05, + "loss": 0.4163, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 780, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.63, + "learning_rate": 1.7366265979381443e-05, + "loss": 0.406, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 790, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.64, + "learning_rate": 1.7325049484536082e-05, + "loss": 0.4142, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 800, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.64, + "learning_rate": 1.7283832989690724e-05, + "loss": 0.3883, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 810, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.65, + "learning_rate": 1.7242616494845362e-05, + "loss": 0.3845, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 820, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.66, + "learning_rate": 1.72014e-05, + "loss": 0.3757, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 830, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.67, + "learning_rate": 1.716018350515464e-05, + "loss": 0.3897, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 840, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.68, + "learning_rate": 1.711896701030928e-05, + "loss": 0.3878, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 850, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.68, + "learning_rate": 1.707775051546392e-05, + "loss": 0.3705, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 860, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.69, + "learning_rate": 1.703653402061856e-05, + "loss": 0.3589, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 870, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.7, + "learning_rate": 1.6995317525773198e-05, + "loss": 0.3626, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 880, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.71, + "learning_rate": 1.6954101030927836e-05, + "loss": 0.3583, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 890, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.72, + "learning_rate": 1.6912884536082478e-05, + "loss": 0.3606, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 900, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.72, + "learning_rate": 1.6871668041237117e-05, + "loss": 0.3677, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 910, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.73, + "learning_rate": 1.6830451546391756e-05, + "loss": 0.3536, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 920, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.74, + "learning_rate": 1.678923505154639e-05, + "loss": 0.3629, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 930, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.75, + "learning_rate": 1.6748018556701033e-05, + "loss": 0.3415, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 940, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.76, + "learning_rate": 1.670680206185567e-05, + "loss": 0.3408, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 950, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.76, + "learning_rate": 1.666558556701031e-05, + "loss": 0.3374, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 960, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.77, + "learning_rate": 1.662436907216495e-05, + "loss": 0.3434, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 970, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.78, + "learning_rate": 1.6583152577319587e-05, + "loss": 0.3447, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 980, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.79, + "learning_rate": 1.654193608247423e-05, + "loss": 0.3333, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 990, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.8, + "learning_rate": 1.6500719587628868e-05, + "loss": 0.333, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 1000, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.8, + "learning_rate": 1.6459503092783507e-05, + "loss": 0.3306, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 1010, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.81, + "learning_rate": 1.6418286597938145e-05, + "loss": 0.3237, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 1020, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.82, + "learning_rate": 1.6377070103092784e-05, + "loss": 0.3262, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 1030, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.83, + "learning_rate": 1.6335853608247426e-05, + "loss": 0.3092, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 1040, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.84, + "learning_rate": 1.6294637113402065e-05, + "loss": 0.3158, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 1050, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.84, + "learning_rate": 1.6253420618556703e-05, + "loss": 0.2999, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 1060, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.85, + "learning_rate": 1.6212204123711342e-05, + "loss": 0.2916, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 1070, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.86, + "learning_rate": 1.617098762886598e-05, + "loss": 0.3108, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 1080, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.87, + "learning_rate": 1.612977113402062e-05, + "loss": 0.2953, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 1090, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.88, + "learning_rate": 1.6088554639175258e-05, + "loss": 0.2897, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 1100, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.88, + "learning_rate": 1.6047338144329896e-05, + "loss": 0.294, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 1110, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.89, + "learning_rate": 1.600612164948454e-05, + "loss": 0.29, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 1120, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.9, + "learning_rate": 1.5964905154639177e-05, + "loss": 0.3022, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 1130, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.91, + "learning_rate": 1.5923688659793816e-05, + "loss": 0.2969, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 1140, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.92, + "learning_rate": 1.5882472164948454e-05, + "loss": 0.2795, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 1150, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.92, + "learning_rate": 1.5841255670103093e-05, + "loss": 0.2814, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 1160, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.93, + "learning_rate": 1.5800039175257735e-05, + "loss": 0.2971, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 1170, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.94, + "learning_rate": 1.5758822680412374e-05, + "loss": 0.2777, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 1180, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.95, + "learning_rate": 1.5717606185567012e-05, + "loss": 0.2691, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 1190, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.96, + "learning_rate": 1.567638969072165e-05, + "loss": 0.2865, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 1200, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.96, + "learning_rate": 1.563517319587629e-05, + "loss": 0.2841, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 1210, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.97, + "learning_rate": 1.559395670103093e-05, + "loss": 0.2662, + "max_memory_allocated (GB)": 82.81, + "memory_allocated (GB)": 23.35, + "step": 1220, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.98, + "learning_rate": 1.5552740206185567e-05, + "loss": 0.2588, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1230, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 0.99, + "learning_rate": 1.5511523711340205e-05, + "loss": 0.2909, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1240, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.0, + "learning_rate": 1.5470307216494848e-05, + "loss": 0.2602, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1250, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.0, + "learning_rate": 1.5429090721649486e-05, + "loss": 0.2276, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1260, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.01, + "learning_rate": 1.5387874226804125e-05, + "loss": 0.1624, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1270, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.02, + "learning_rate": 1.5346657731958763e-05, + "loss": 0.1609, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1280, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.03, + "learning_rate": 1.5305441237113402e-05, + "loss": 0.1531, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1290, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.04, + "learning_rate": 1.526422474226804e-05, + "loss": 0.1596, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1300, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.04, + "learning_rate": 1.5223008247422681e-05, + "loss": 0.1597, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1310, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.05, + "learning_rate": 1.518179175257732e-05, + "loss": 0.1701, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1320, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.06, + "learning_rate": 1.5140575257731958e-05, + "loss": 0.1635, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1330, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.07, + "learning_rate": 1.5099358762886599e-05, + "loss": 0.1522, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1340, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.07, + "learning_rate": 1.5058142268041237e-05, + "loss": 0.1644, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1350, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.08, + "learning_rate": 1.5016925773195878e-05, + "loss": 0.1582, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1360, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.09, + "learning_rate": 1.4975709278350516e-05, + "loss": 0.1613, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1370, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.1, + "learning_rate": 1.4934492783505153e-05, + "loss": 0.1643, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1380, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.11, + "learning_rate": 1.4893276288659794e-05, + "loss": 0.1671, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1390, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.11, + "learning_rate": 1.4852059793814432e-05, + "loss": 0.1636, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1400, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.12, + "learning_rate": 1.481084329896907e-05, + "loss": 0.16, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1410, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.13, + "learning_rate": 1.4769626804123711e-05, + "loss": 0.1646, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1420, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.14, + "learning_rate": 1.472841030927835e-05, + "loss": 0.1587, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1430, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.15, + "learning_rate": 1.468719381443299e-05, + "loss": 0.1667, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1440, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.15, + "learning_rate": 1.4645977319587629e-05, + "loss": 0.1641, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1450, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.16, + "learning_rate": 1.4604760824742267e-05, + "loss": 0.163, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1460, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.17, + "learning_rate": 1.4563544329896908e-05, + "loss": 0.1612, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1470, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.18, + "learning_rate": 1.4522327835051546e-05, + "loss": 0.1643, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1480, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.19, + "learning_rate": 1.4481111340206187e-05, + "loss": 0.1652, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1490, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.19, + "learning_rate": 1.4439894845360825e-05, + "loss": 0.1699, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1500, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.2, + "learning_rate": 1.4398678350515464e-05, + "loss": 0.1527, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1510, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.21, + "learning_rate": 1.4357461855670104e-05, + "loss": 0.1675, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1520, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.22, + "learning_rate": 1.4316245360824743e-05, + "loss": 0.1568, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1530, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.23, + "learning_rate": 1.427502886597938e-05, + "loss": 0.1771, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1540, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.23, + "learning_rate": 1.423381237113402e-05, + "loss": 0.159, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1550, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.24, + "learning_rate": 1.4192595876288659e-05, + "loss": 0.159, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1560, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.25, + "learning_rate": 1.4151379381443299e-05, + "loss": 0.164, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1570, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.26, + "learning_rate": 1.4110162886597938e-05, + "loss": 0.1606, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1580, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.27, + "learning_rate": 1.4068946391752576e-05, + "loss": 0.1659, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1590, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.27, + "learning_rate": 1.4027729896907217e-05, + "loss": 0.1637, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1600, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.28, + "learning_rate": 1.3986513402061855e-05, + "loss": 0.1545, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1610, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.29, + "learning_rate": 1.3945296907216496e-05, + "loss": 0.1627, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1620, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.3, + "learning_rate": 1.3904080412371134e-05, + "loss": 0.1579, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1630, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.31, + "learning_rate": 1.3862863917525773e-05, + "loss": 0.1599, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1640, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.31, + "learning_rate": 1.3821647422680413e-05, + "loss": 0.1504, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1650, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.32, + "learning_rate": 1.3780430927835052e-05, + "loss": 0.1592, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1660, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.33, + "learning_rate": 1.3739214432989692e-05, + "loss": 0.161, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1670, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.34, + "learning_rate": 1.3697997938144331e-05, + "loss": 0.1542, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1680, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.35, + "learning_rate": 1.3656781443298968e-05, + "loss": 0.1579, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1690, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.35, + "learning_rate": 1.3615564948453607e-05, + "loss": 0.1528, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1700, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.36, + "learning_rate": 1.3574348453608247e-05, + "loss": 0.1502, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1710, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.37, + "learning_rate": 1.3533131958762885e-05, + "loss": 0.1558, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1720, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.38, + "learning_rate": 1.3491915463917526e-05, + "loss": 0.1485, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1730, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.39, + "learning_rate": 1.3450698969072164e-05, + "loss": 0.1562, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1740, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.39, + "learning_rate": 1.3409482474226805e-05, + "loss": 0.1511, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1750, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.4, + "learning_rate": 1.3368265979381443e-05, + "loss": 0.1707, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1760, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.41, + "learning_rate": 1.3327049484536082e-05, + "loss": 0.1538, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1770, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.42, + "learning_rate": 1.3285832989690722e-05, + "loss": 0.1561, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1780, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.43, + "learning_rate": 1.3244616494845361e-05, + "loss": 0.1542, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1790, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.43, + "learning_rate": 1.3203400000000001e-05, + "loss": 0.1613, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1800, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.44, + "learning_rate": 1.316218350515464e-05, + "loss": 0.1488, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1810, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.45, + "learning_rate": 1.3120967010309279e-05, + "loss": 0.1537, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1820, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.46, + "learning_rate": 1.3079750515463919e-05, + "loss": 0.143, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1830, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.46, + "learning_rate": 1.3038534020618556e-05, + "loss": 0.16, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1840, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.47, + "learning_rate": 1.2997317525773195e-05, + "loss": 0.151, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1850, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.48, + "learning_rate": 1.2956101030927835e-05, + "loss": 0.1587, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1860, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.49, + "learning_rate": 1.2914884536082474e-05, + "loss": 0.1524, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1870, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.5, + "learning_rate": 1.2873668041237112e-05, + "loss": 0.1583, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1880, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.5, + "learning_rate": 1.2832451546391752e-05, + "loss": 0.1475, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1890, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.51, + "learning_rate": 1.2791235051546391e-05, + "loss": 0.155, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1900, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.52, + "learning_rate": 1.2750018556701031e-05, + "loss": 0.1547, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1910, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.53, + "learning_rate": 1.270880206185567e-05, + "loss": 0.1502, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1920, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.54, + "learning_rate": 1.2667585567010309e-05, + "loss": 0.151, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1930, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.54, + "learning_rate": 1.2626369072164949e-05, + "loss": 0.1441, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1940, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.55, + "learning_rate": 1.2585152577319588e-05, + "loss": 0.1522, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1950, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.56, + "learning_rate": 1.2543936082474228e-05, + "loss": 0.153, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1960, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.57, + "learning_rate": 1.2502719587628867e-05, + "loss": 0.1526, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1970, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.58, + "learning_rate": 1.2461503092783505e-05, + "loss": 0.1558, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1980, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.58, + "learning_rate": 1.2420286597938144e-05, + "loss": 0.1544, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 1990, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.59, + "learning_rate": 1.2379070103092783e-05, + "loss": 0.1431, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2000, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.6, + "learning_rate": 1.2337853608247421e-05, + "loss": 0.1488, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2010, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.61, + "learning_rate": 1.2296637113402062e-05, + "loss": 0.1458, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2020, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.62, + "learning_rate": 1.22554206185567e-05, + "loss": 0.1381, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2030, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.62, + "learning_rate": 1.221420412371134e-05, + "loss": 0.1449, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2040, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.63, + "learning_rate": 1.2172987628865979e-05, + "loss": 0.1509, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2050, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.64, + "learning_rate": 1.2131771134020618e-05, + "loss": 0.134, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2060, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.65, + "learning_rate": 1.2090554639175258e-05, + "loss": 0.1479, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2070, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.66, + "learning_rate": 1.2049338144329897e-05, + "loss": 0.1521, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2080, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.66, + "learning_rate": 1.2008121649484537e-05, + "loss": 0.1492, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2090, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.67, + "learning_rate": 1.1966905154639176e-05, + "loss": 0.1388, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2100, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.68, + "learning_rate": 1.1925688659793814e-05, + "loss": 0.1344, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2110, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.69, + "learning_rate": 1.1884472164948455e-05, + "loss": 0.141, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2120, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.7, + "learning_rate": 1.1843255670103093e-05, + "loss": 0.1447, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2130, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.7, + "learning_rate": 1.180203917525773e-05, + "loss": 0.1408, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2140, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.71, + "learning_rate": 1.176082268041237e-05, + "loss": 0.137, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2150, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.72, + "learning_rate": 1.171960618556701e-05, + "loss": 0.1376, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2160, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.73, + "learning_rate": 1.167838969072165e-05, + "loss": 0.1302, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2170, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.74, + "learning_rate": 1.1637173195876288e-05, + "loss": 0.141, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2180, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.74, + "learning_rate": 1.1595956701030927e-05, + "loss": 0.141, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2190, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.75, + "learning_rate": 1.1554740206185567e-05, + "loss": 0.1384, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2200, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.76, + "learning_rate": 1.1513523711340206e-05, + "loss": 0.138, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2210, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.77, + "learning_rate": 1.1472307216494846e-05, + "loss": 0.1424, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2220, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.78, + "learning_rate": 1.1431090721649485e-05, + "loss": 0.1406, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2230, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.78, + "learning_rate": 1.1389874226804123e-05, + "loss": 0.1369, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2240, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.79, + "learning_rate": 1.1348657731958764e-05, + "loss": 0.1494, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2250, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.8, + "learning_rate": 1.1307441237113402e-05, + "loss": 0.1337, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2260, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.81, + "learning_rate": 1.1266224742268043e-05, + "loss": 0.1473, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2270, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.82, + "learning_rate": 1.1225008247422681e-05, + "loss": 0.1379, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2280, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.82, + "learning_rate": 1.1183791752577318e-05, + "loss": 0.1425, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2290, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.83, + "learning_rate": 1.1142575257731957e-05, + "loss": 0.1436, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2300, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.84, + "learning_rate": 1.1101358762886597e-05, + "loss": 0.1406, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2310, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.85, + "learning_rate": 1.1060142268041236e-05, + "loss": 0.1416, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2320, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.86, + "learning_rate": 1.1018925773195876e-05, + "loss": 0.1343, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2330, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.86, + "learning_rate": 1.0977709278350515e-05, + "loss": 0.1442, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2340, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.87, + "learning_rate": 1.0936492783505155e-05, + "loss": 0.1468, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2350, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.88, + "learning_rate": 1.0895276288659794e-05, + "loss": 0.1267, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2360, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.89, + "learning_rate": 1.0854059793814432e-05, + "loss": 0.1339, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2370, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.89, + "learning_rate": 1.0812843298969073e-05, + "loss": 0.1392, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2380, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.9, + "learning_rate": 1.0771626804123711e-05, + "loss": 0.1379, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2390, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.91, + "learning_rate": 1.0730410309278352e-05, + "loss": 0.1325, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2400, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.92, + "learning_rate": 1.068919381443299e-05, + "loss": 0.129, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2410, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.93, + "learning_rate": 1.0647977319587629e-05, + "loss": 0.1344, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2420, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.93, + "learning_rate": 1.060676082474227e-05, + "loss": 0.1306, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2430, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.94, + "learning_rate": 1.0565544329896906e-05, + "loss": 0.1301, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2440, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.95, + "learning_rate": 1.0524327835051545e-05, + "loss": 0.1327, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2450, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.96, + "learning_rate": 1.0483111340206185e-05, + "loss": 0.1316, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2460, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.97, + "learning_rate": 1.0441894845360824e-05, + "loss": 0.1153, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2470, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.97, + "learning_rate": 1.0400678350515463e-05, + "loss": 0.1412, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2480, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.98, + "learning_rate": 1.0359461855670103e-05, + "loss": 0.1344, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2490, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 1.99, + "learning_rate": 1.0318245360824742e-05, + "loss": 0.144, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2500, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.0, + "learning_rate": 1.0277028865979382e-05, + "loss": 0.1327, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2510, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.01, + "learning_rate": 1.023581237113402e-05, + "loss": 0.0872, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2520, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.01, + "learning_rate": 1.0194595876288659e-05, + "loss": 0.0766, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2530, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.02, + "learning_rate": 1.01533793814433e-05, + "loss": 0.0806, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2540, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.03, + "learning_rate": 1.0112162886597938e-05, + "loss": 0.0779, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2550, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.04, + "learning_rate": 1.0070946391752578e-05, + "loss": 0.0799, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2560, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.05, + "learning_rate": 1.0029729896907217e-05, + "loss": 0.0806, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2570, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.05, + "learning_rate": 9.988513402061856e-06, + "loss": 0.0834, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2580, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.06, + "learning_rate": 9.947296907216494e-06, + "loss": 0.0769, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2590, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.07, + "learning_rate": 9.906080412371135e-06, + "loss": 0.0825, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2600, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.08, + "learning_rate": 9.864863917525773e-06, + "loss": 0.0794, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2610, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.09, + "learning_rate": 9.823647422680412e-06, + "loss": 0.0792, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2620, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.09, + "learning_rate": 9.78243092783505e-06, + "loss": 0.0765, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2630, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.1, + "learning_rate": 9.741214432989691e-06, + "loss": 0.0806, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2640, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.11, + "learning_rate": 9.69999793814433e-06, + "loss": 0.0809, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2650, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.12, + "learning_rate": 9.658781443298968e-06, + "loss": 0.0843, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2660, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.13, + "learning_rate": 9.617564948453609e-06, + "loss": 0.0882, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2670, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.13, + "learning_rate": 9.576348453608247e-06, + "loss": 0.0811, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2680, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.14, + "learning_rate": 9.535131958762888e-06, + "loss": 0.0832, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2690, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.15, + "learning_rate": 9.493915463917526e-06, + "loss": 0.0811, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2700, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.16, + "learning_rate": 9.452698969072165e-06, + "loss": 0.083, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2710, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.17, + "learning_rate": 9.411482474226803e-06, + "loss": 0.0905, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2720, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.17, + "learning_rate": 9.370265979381444e-06, + "loss": 0.088, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2730, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.18, + "learning_rate": 9.329049484536082e-06, + "loss": 0.0863, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2740, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.19, + "learning_rate": 9.287832989690721e-06, + "loss": 0.0848, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2750, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.2, + "learning_rate": 9.246616494845361e-06, + "loss": 0.0808, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2760, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.21, + "learning_rate": 9.2054e-06, + "loss": 0.0806, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2770, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.21, + "learning_rate": 9.16418350515464e-06, + "loss": 0.0848, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2780, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.22, + "learning_rate": 9.122967010309277e-06, + "loss": 0.0845, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2790, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.23, + "learning_rate": 9.081750515463918e-06, + "loss": 0.0817, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2800, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.24, + "learning_rate": 9.040534020618556e-06, + "loss": 0.0822, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2810, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.25, + "learning_rate": 8.999317525773197e-06, + "loss": 0.0873, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2820, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.25, + "learning_rate": 8.958101030927835e-06, + "loss": 0.0808, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2830, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.26, + "learning_rate": 8.916884536082474e-06, + "loss": 0.0882, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2840, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.27, + "learning_rate": 8.875668041237114e-06, + "loss": 0.0843, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2850, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.28, + "learning_rate": 8.834451546391753e-06, + "loss": 0.0894, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2860, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.29, + "learning_rate": 8.793235051546391e-06, + "loss": 0.0868, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2870, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.29, + "learning_rate": 8.75201855670103e-06, + "loss": 0.079, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2880, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.3, + "learning_rate": 8.71080206185567e-06, + "loss": 0.0829, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2890, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.31, + "learning_rate": 8.669585567010309e-06, + "loss": 0.0786, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2900, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.32, + "learning_rate": 8.62836907216495e-06, + "loss": 0.0779, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2910, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.32, + "learning_rate": 8.587152577319588e-06, + "loss": 0.0787, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2920, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.33, + "learning_rate": 8.545936082474227e-06, + "loss": 0.0867, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2930, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.34, + "learning_rate": 8.504719587628865e-06, + "loss": 0.0777, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2940, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.35, + "learning_rate": 8.463503092783506e-06, + "loss": 0.0742, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2950, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.36, + "learning_rate": 8.422286597938144e-06, + "loss": 0.0818, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2960, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.36, + "learning_rate": 8.381070103092783e-06, + "loss": 0.0826, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2970, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.37, + "learning_rate": 8.339853608247423e-06, + "loss": 0.081, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2980, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.38, + "learning_rate": 8.298637113402062e-06, + "loss": 0.0748, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 2990, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.39, + "learning_rate": 8.257420618556702e-06, + "loss": 0.0852, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3000, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.4, + "learning_rate": 8.216204123711339e-06, + "loss": 0.0797, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3010, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.4, + "learning_rate": 8.17498762886598e-06, + "loss": 0.0777, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3020, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.41, + "learning_rate": 8.133771134020618e-06, + "loss": 0.0781, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3030, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.42, + "learning_rate": 8.092554639175257e-06, + "loss": 0.0838, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3040, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.43, + "learning_rate": 8.051338144329897e-06, + "loss": 0.0823, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3050, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.44, + "learning_rate": 8.010121649484536e-06, + "loss": 0.0824, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3060, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.44, + "learning_rate": 7.968905154639176e-06, + "loss": 0.0886, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3070, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.45, + "learning_rate": 7.927688659793815e-06, + "loss": 0.079, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3080, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.46, + "learning_rate": 7.886472164948453e-06, + "loss": 0.0843, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3090, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.47, + "learning_rate": 7.845255670103092e-06, + "loss": 0.0831, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3100, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.48, + "learning_rate": 7.804039175257732e-06, + "loss": 0.0869, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3110, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.48, + "learning_rate": 7.762822680412371e-06, + "loss": 0.079, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3120, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.49, + "learning_rate": 7.72160618556701e-06, + "loss": 0.0846, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3130, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.5, + "learning_rate": 7.68038969072165e-06, + "loss": 0.0814, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3140, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.51, + "learning_rate": 7.63917319587629e-06, + "loss": 0.0839, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3150, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.52, + "learning_rate": 7.597956701030927e-06, + "loss": 0.0803, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3160, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.52, + "learning_rate": 7.556740206185567e-06, + "loss": 0.0816, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3170, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.53, + "learning_rate": 7.515523711340206e-06, + "loss": 0.0814, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3180, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.54, + "learning_rate": 7.474307216494845e-06, + "loss": 0.0785, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3190, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.55, + "learning_rate": 7.433090721649484e-06, + "loss": 0.0815, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3200, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.56, + "learning_rate": 7.391874226804124e-06, + "loss": 0.0797, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3210, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.56, + "learning_rate": 7.350657731958763e-06, + "loss": 0.0807, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3220, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.57, + "learning_rate": 7.309441237113403e-06, + "loss": 0.0889, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3230, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.58, + "learning_rate": 7.2682247422680405e-06, + "loss": 0.0819, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3240, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.59, + "learning_rate": 7.22700824742268e-06, + "loss": 0.0835, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3250, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.6, + "learning_rate": 7.1857917525773195e-06, + "loss": 0.0837, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3260, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.6, + "learning_rate": 7.144575257731959e-06, + "loss": 0.0786, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3270, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.61, + "learning_rate": 7.103358762886598e-06, + "loss": 0.0745, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3280, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.62, + "learning_rate": 7.062142268041237e-06, + "loss": 0.0871, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3290, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.63, + "learning_rate": 7.0209257731958766e-06, + "loss": 0.0832, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3300, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.64, + "learning_rate": 6.979709278350516e-06, + "loss": 0.081, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3310, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.64, + "learning_rate": 6.938492783505154e-06, + "loss": 0.08, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3320, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.65, + "learning_rate": 6.897276288659793e-06, + "loss": 0.0809, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3330, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.66, + "learning_rate": 6.856059793814433e-06, + "loss": 0.0775, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3340, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.67, + "learning_rate": 6.814843298969072e-06, + "loss": 0.0796, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3350, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.68, + "learning_rate": 6.773626804123712e-06, + "loss": 0.0811, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3360, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.68, + "learning_rate": 6.73241030927835e-06, + "loss": 0.083, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3370, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.69, + "learning_rate": 6.69119381443299e-06, + "loss": 0.0797, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3380, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.7, + "learning_rate": 6.6499773195876285e-06, + "loss": 0.0806, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3390, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.71, + "learning_rate": 6.608760824742268e-06, + "loss": 0.0759, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3400, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.71, + "learning_rate": 6.567544329896907e-06, + "loss": 0.0791, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3410, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.72, + "learning_rate": 6.526327835051546e-06, + "loss": 0.0795, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3420, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.73, + "learning_rate": 6.485111340206186e-06, + "loss": 0.0815, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3430, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.74, + "learning_rate": 6.443894845360825e-06, + "loss": 0.0795, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3440, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.75, + "learning_rate": 6.402678350515465e-06, + "loss": 0.078, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3450, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.75, + "learning_rate": 6.361461855670103e-06, + "loss": 0.0779, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3460, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.76, + "learning_rate": 6.320245360824742e-06, + "loss": 0.0796, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3470, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.77, + "learning_rate": 6.279028865979381e-06, + "loss": 0.0841, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3480, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.78, + "learning_rate": 6.23781237113402e-06, + "loss": 0.0795, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3490, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.79, + "learning_rate": 6.1965958762886595e-06, + "loss": 0.0774, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3500, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.79, + "learning_rate": 6.155379381443299e-06, + "loss": 0.0734, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3510, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.8, + "learning_rate": 6.1141628865979384e-06, + "loss": 0.0817, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3520, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.81, + "learning_rate": 6.072946391752578e-06, + "loss": 0.0801, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3530, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.82, + "learning_rate": 6.031729896907216e-06, + "loss": 0.0769, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3540, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.83, + "learning_rate": 5.990513402061855e-06, + "loss": 0.0757, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3550, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.83, + "learning_rate": 5.949296907216495e-06, + "loss": 0.0794, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3560, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.84, + "learning_rate": 5.908080412371134e-06, + "loss": 0.0823, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3570, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.85, + "learning_rate": 5.866863917525773e-06, + "loss": 0.0865, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3580, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.86, + "learning_rate": 5.825647422680412e-06, + "loss": 0.0729, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3590, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.87, + "learning_rate": 5.784430927835052e-06, + "loss": 0.0802, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3600, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.87, + "learning_rate": 5.743214432989691e-06, + "loss": 0.0776, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3610, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.88, + "learning_rate": 5.701997938144329e-06, + "loss": 0.0811, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3620, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.89, + "learning_rate": 5.6607814432989685e-06, + "loss": 0.0744, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3630, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.9, + "learning_rate": 5.619564948453608e-06, + "loss": 0.0743, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3640, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.91, + "learning_rate": 5.5783484536082475e-06, + "loss": 0.0789, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3650, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.91, + "learning_rate": 5.537131958762887e-06, + "loss": 0.0691, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3660, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.92, + "learning_rate": 5.495915463917526e-06, + "loss": 0.0745, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3670, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.93, + "learning_rate": 5.454698969072165e-06, + "loss": 0.0771, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3680, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.94, + "learning_rate": 5.413482474226804e-06, + "loss": 0.0809, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3690, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.95, + "learning_rate": 5.372265979381442e-06, + "loss": 0.0763, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3700, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.95, + "learning_rate": 5.331049484536082e-06, + "loss": 0.0714, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3710, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.96, + "learning_rate": 5.289832989690721e-06, + "loss": 0.0808, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3720, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.97, + "learning_rate": 5.248616494845361e-06, + "loss": 0.0737, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3730, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.98, + "learning_rate": 5.2074e-06, + "loss": 0.0796, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3740, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.99, + "learning_rate": 5.16618350515464e-06, + "loss": 0.0775, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3750, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 2.99, + "learning_rate": 5.1249670103092784e-06, + "loss": 0.0717, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3760, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.0, + "learning_rate": 5.083750515463917e-06, + "loss": 0.0729, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3770, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.01, + "learning_rate": 5.0425340206185566e-06, + "loss": 0.0488, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3780, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.02, + "learning_rate": 5.001317525773195e-06, + "loss": 0.0469, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3790, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.03, + "learning_rate": 4.960101030927835e-06, + "loss": 0.0503, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3800, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.03, + "learning_rate": 4.918884536082474e-06, + "loss": 0.0474, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3810, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.04, + "learning_rate": 4.877668041237114e-06, + "loss": 0.0529, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3820, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.05, + "learning_rate": 4.836451546391752e-06, + "loss": 0.0472, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3830, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.06, + "learning_rate": 4.795235051546392e-06, + "loss": 0.0455, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3840, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.07, + "learning_rate": 4.754018556701031e-06, + "loss": 0.0464, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3850, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.07, + "learning_rate": 4.71280206185567e-06, + "loss": 0.0498, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3860, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.08, + "learning_rate": 4.671585567010309e-06, + "loss": 0.0449, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3870, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.09, + "learning_rate": 4.630369072164948e-06, + "loss": 0.0474, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3880, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.1, + "learning_rate": 4.5891525773195875e-06, + "loss": 0.0487, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3890, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.11, + "learning_rate": 4.547936082474226e-06, + "loss": 0.049, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3900, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.11, + "learning_rate": 4.506719587628866e-06, + "loss": 0.0476, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3910, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.12, + "learning_rate": 4.465503092783505e-06, + "loss": 0.0495, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3920, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.13, + "learning_rate": 4.424286597938145e-06, + "loss": 0.0489, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3930, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.14, + "learning_rate": 4.383070103092783e-06, + "loss": 0.0519, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3940, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.14, + "learning_rate": 4.341853608247423e-06, + "loss": 0.0521, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3950, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.15, + "learning_rate": 4.300637113402062e-06, + "loss": 0.0518, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3960, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.16, + "learning_rate": 4.259420618556701e-06, + "loss": 0.05, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3970, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.17, + "learning_rate": 4.21820412371134e-06, + "loss": 0.0469, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3980, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.18, + "learning_rate": 4.176987628865979e-06, + "loss": 0.0498, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 3990, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.18, + "learning_rate": 4.1357711340206184e-06, + "loss": 0.0514, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4000, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.19, + "learning_rate": 4.094554639175258e-06, + "loss": 0.0475, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4010, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.2, + "learning_rate": 4.0533381443298966e-06, + "loss": 0.0484, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4020, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.21, + "learning_rate": 4.012121649484536e-06, + "loss": 0.0503, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4030, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.22, + "learning_rate": 3.9709051546391755e-06, + "loss": 0.049, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4040, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.22, + "learning_rate": 3.929688659793814e-06, + "loss": 0.0453, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4050, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.23, + "learning_rate": 3.888472164948454e-06, + "loss": 0.0481, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4060, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.24, + "learning_rate": 3.847255670103092e-06, + "loss": 0.0482, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4070, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.25, + "learning_rate": 3.806039175257732e-06, + "loss": 0.0498, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4080, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.26, + "learning_rate": 3.764822680412371e-06, + "loss": 0.0448, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4090, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.26, + "learning_rate": 3.7236061855670103e-06, + "loss": 0.0488, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4100, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.27, + "learning_rate": 3.6823896907216494e-06, + "loss": 0.0499, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4110, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.28, + "learning_rate": 3.641173195876289e-06, + "loss": 0.0542, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4120, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.29, + "learning_rate": 3.5999567010309275e-06, + "loss": 0.0467, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4130, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.3, + "learning_rate": 3.558740206185567e-06, + "loss": 0.0483, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4140, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.3, + "learning_rate": 3.517523711340206e-06, + "loss": 0.0485, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4150, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.31, + "learning_rate": 3.4763072164948455e-06, + "loss": 0.0472, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4160, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.32, + "learning_rate": 3.435090721649484e-06, + "loss": 0.047, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4170, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.33, + "learning_rate": 3.3938742268041237e-06, + "loss": 0.0473, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4180, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.34, + "learning_rate": 3.3526577319587627e-06, + "loss": 0.0495, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4190, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.34, + "learning_rate": 3.311441237113402e-06, + "loss": 0.0486, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4200, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.35, + "learning_rate": 3.270224742268041e-06, + "loss": 0.0495, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4210, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.36, + "learning_rate": 3.2290082474226803e-06, + "loss": 0.0489, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4220, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.37, + "learning_rate": 3.18779175257732e-06, + "loss": 0.051, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4230, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.38, + "learning_rate": 3.1465752577319584e-06, + "loss": 0.0504, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4240, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.38, + "learning_rate": 3.105358762886598e-06, + "loss": 0.0463, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4250, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.39, + "learning_rate": 3.064142268041237e-06, + "loss": 0.0457, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4260, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.4, + "learning_rate": 3.0229257731958765e-06, + "loss": 0.0506, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4270, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.41, + "learning_rate": 2.981709278350515e-06, + "loss": 0.0482, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4280, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.42, + "learning_rate": 2.9404927835051546e-06, + "loss": 0.0491, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4290, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.42, + "learning_rate": 2.8992762886597937e-06, + "loss": 0.0477, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4300, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.43, + "learning_rate": 2.858059793814433e-06, + "loss": 0.0476, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4310, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.44, + "learning_rate": 2.8168432989690718e-06, + "loss": 0.0495, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4320, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.45, + "learning_rate": 2.7756268041237113e-06, + "loss": 0.0479, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4330, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.46, + "learning_rate": 2.7344103092783503e-06, + "loss": 0.0479, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4340, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.46, + "learning_rate": 2.69319381443299e-06, + "loss": 0.0459, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4350, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.47, + "learning_rate": 2.6519773195876284e-06, + "loss": 0.0513, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4360, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.48, + "learning_rate": 2.610760824742268e-06, + "loss": 0.0475, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4370, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.49, + "learning_rate": 2.5695443298969074e-06, + "loss": 0.0438, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4380, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.5, + "learning_rate": 2.5283278350515465e-06, + "loss": 0.0463, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4390, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.5, + "learning_rate": 2.4871113402061855e-06, + "loss": 0.0473, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4400, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.51, + "learning_rate": 2.4458948453608246e-06, + "loss": 0.048, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4410, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.52, + "learning_rate": 2.4046783505154637e-06, + "loss": 0.0481, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4420, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.53, + "learning_rate": 2.363461855670103e-06, + "loss": 0.0522, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4430, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.54, + "learning_rate": 2.322245360824742e-06, + "loss": 0.0473, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4440, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.54, + "learning_rate": 2.2810288659793813e-06, + "loss": 0.0507, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4450, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.55, + "learning_rate": 2.2398123711340203e-06, + "loss": 0.0495, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4460, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.56, + "learning_rate": 2.19859587628866e-06, + "loss": 0.0456, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4470, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.57, + "learning_rate": 2.157379381443299e-06, + "loss": 0.0427, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4480, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.57, + "learning_rate": 2.116162886597938e-06, + "loss": 0.0483, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4490, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.58, + "learning_rate": 2.074946391752577e-06, + "loss": 0.048, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4500, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.59, + "learning_rate": 2.0337298969072165e-06, + "loss": 0.0524, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4510, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.6, + "learning_rate": 1.9925134020618555e-06, + "loss": 0.0466, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4520, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.61, + "learning_rate": 1.951296907216495e-06, + "loss": 0.0485, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4530, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.61, + "learning_rate": 1.910080412371134e-06, + "loss": 0.0457, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4540, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.62, + "learning_rate": 1.8688639175257731e-06, + "loss": 0.0482, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4550, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.63, + "learning_rate": 1.8276474226804122e-06, + "loss": 0.0467, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4560, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.64, + "learning_rate": 1.7864309278350517e-06, + "loss": 0.0495, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4570, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.65, + "learning_rate": 1.7452144329896905e-06, + "loss": 0.048, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4580, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.65, + "learning_rate": 1.7039979381443296e-06, + "loss": 0.0499, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4590, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.66, + "learning_rate": 1.662781443298969e-06, + "loss": 0.0448, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4600, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.67, + "learning_rate": 1.6215649484536081e-06, + "loss": 0.0487, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4610, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.68, + "learning_rate": 1.5803484536082474e-06, + "loss": 0.0434, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4620, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.69, + "learning_rate": 1.5391319587628865e-06, + "loss": 0.0475, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4630, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.69, + "learning_rate": 1.4979154639175257e-06, + "loss": 0.0487, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4640, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.7, + "learning_rate": 1.4566989690721648e-06, + "loss": 0.0476, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4650, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.71, + "learning_rate": 1.415482474226804e-06, + "loss": 0.0468, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4660, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.72, + "learning_rate": 1.3742659793814431e-06, + "loss": 0.0478, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4670, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.73, + "learning_rate": 1.3330494845360824e-06, + "loss": 0.052, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4680, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.73, + "learning_rate": 1.2918329896907215e-06, + "loss": 0.0474, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4690, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.74, + "learning_rate": 1.2506164948453607e-06, + "loss": 0.048, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4700, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.75, + "learning_rate": 1.2093999999999998e-06, + "loss": 0.0507, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4710, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.76, + "learning_rate": 1.168183505154639e-06, + "loss": 0.0473, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4720, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.77, + "learning_rate": 1.1269670103092781e-06, + "loss": 0.0483, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4730, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.77, + "learning_rate": 1.0857505154639174e-06, + "loss": 0.0488, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4740, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.78, + "learning_rate": 1.0445340206185567e-06, + "loss": 0.0461, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4750, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.79, + "learning_rate": 1.0033175257731957e-06, + "loss": 0.0436, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4760, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.8, + "learning_rate": 9.62101030927835e-07, + "loss": 0.0451, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4770, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.81, + "learning_rate": 9.208845360824743e-07, + "loss": 0.0457, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4780, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.81, + "learning_rate": 8.796680412371135e-07, + "loss": 0.0477, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4790, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.82, + "learning_rate": 8.384515463917526e-07, + "loss": 0.0463, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4800, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.83, + "learning_rate": 7.972350515463918e-07, + "loss": 0.0474, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4810, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.84, + "learning_rate": 7.56018556701031e-07, + "loss": 0.0473, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4820, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.85, + "learning_rate": 7.148020618556701e-07, + "loss": 0.0448, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4830, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.85, + "learning_rate": 6.735855670103094e-07, + "loss": 0.0426, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4840, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.86, + "learning_rate": 6.323690721649485e-07, + "loss": 0.0499, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4850, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.87, + "learning_rate": 5.911525773195876e-07, + "loss": 0.0438, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4860, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.88, + "learning_rate": 5.499360824742268e-07, + "loss": 0.0448, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4870, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.89, + "learning_rate": 5.08719587628866e-07, + "loss": 0.0497, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4880, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.89, + "learning_rate": 4.675030927835052e-07, + "loss": 0.0465, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4890, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.9, + "learning_rate": 4.262865979381444e-07, + "loss": 0.0444, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4900, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.91, + "learning_rate": 3.850701030927835e-07, + "loss": 0.043, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4910, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.92, + "learning_rate": 3.438536082474227e-07, + "loss": 0.0489, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4920, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.93, + "learning_rate": 3.0263711340206184e-07, + "loss": 0.043, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4930, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.93, + "learning_rate": 2.6142061855670106e-07, + "loss": 0.0491, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4940, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.94, + "learning_rate": 2.2020412371134023e-07, + "loss": 0.0468, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4950, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.95, + "learning_rate": 1.789876288659794e-07, + "loss": 0.0445, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4960, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.96, + "learning_rate": 1.3777113402061856e-07, + "loss": 0.0454, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4970, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.96, + "learning_rate": 9.655463917525774e-08, + "loss": 0.0456, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4980, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.97, + "learning_rate": 5.533814432989691e-08, + "loss": 0.044, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 4990, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.98, + "learning_rate": 1.4121649484536083e-08, + "loss": 0.0478, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 5000, + "total_memory_available (GB)": 94.57 + }, + { + "epoch": 3.98, + "max_memory_allocated (GB)": 82.87, + "memory_allocated (GB)": 23.35, + "step": 5000, + "total_flos": 0.0, + "total_memory_available (GB)": 94.57, + "train_loss": 0.19149471794962883, + "train_runtime": 16857.8956, + "train_samples_per_second": 37.964, + "train_steps_per_second": 0.297 + } + ], + "logging_steps": 10, + "max_steps": 5000, + "num_train_epochs": 4, + "save_steps": 500, + "total_flos": 0.0, + "trial_name": null, + "trial_params": null +}