diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,2827 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.0, + "global_step": 2105, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 1.9824163913726807, + "epoch": 0.0, + "learning_rate": 1.99144893111639e-05, + "loss": 1.8791, + "step": 10, + "task_loss": 0.6514205932617188 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 1.9337328672409058, + "epoch": 0.01, + "learning_rate": 1.9819477434679337e-05, + "loss": 1.8084, + "step": 20, + "task_loss": 0.68878173828125 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 1.7075116634368896, + "epoch": 0.01, + "learning_rate": 1.9724465558194775e-05, + "loss": 1.7447, + "step": 30, + "task_loss": 0.5935821533203125 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 1.5379259586334229, + "epoch": 0.02, + "learning_rate": 1.9629453681710216e-05, + "loss": 1.4292, + "step": 40, + "task_loss": 0.5262355804443359 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.7455981969833374, + "epoch": 0.02, + "learning_rate": 1.954394299287411e-05, + "loss": 1.2067, + "step": 50, + "task_loss": 0.20388412475585938 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 1.1262179613113403, + "epoch": 0.03, + "learning_rate": 1.944893111638955e-05, + "loss": 0.9282, + "step": 60, + "task_loss": 0.45258522033691406 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 1.2520625591278076, + "epoch": 0.03, + "learning_rate": 1.935391923990499e-05, + "loss": 0.9656, + "step": 70, + "task_loss": 0.6228516101837158 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.8546397686004639, + "epoch": 0.04, + "learning_rate": 1.9268408551068884e-05, + "loss": 0.9789, + "step": 80, + "task_loss": 0.31339454650878906 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 1.1221280097961426, + "epoch": 0.04, + "learning_rate": 1.9173396674584325e-05, + "loss": 0.8387, + "step": 90, + "task_loss": 0.44170093536376953 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.4717516005039215, + "epoch": 0.05, + "learning_rate": 1.9078384798099766e-05, + "loss": 0.6816, + "step": 100, + "task_loss": 0.15700101852416992 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.7137082815170288, + "epoch": 0.05, + "learning_rate": 1.8983372921615203e-05, + "loss": 0.6424, + "step": 110, + "task_loss": 0.3361825942993164 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.7122435569763184, + "epoch": 0.06, + "learning_rate": 1.8888361045130644e-05, + "loss": 0.7162, + "step": 120, + "task_loss": 0.5846670866012573 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5021461248397827, + "epoch": 0.06, + "learning_rate": 1.8793349168646082e-05, + "loss": 0.5908, + "step": 130, + "task_loss": 0.18646156787872314 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.33661046624183655, + "epoch": 0.07, + "learning_rate": 1.8698337292161523e-05, + "loss": 0.8316, + "step": 140, + "task_loss": 0.11925870180130005 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.4071422219276428, + "epoch": 0.07, + "learning_rate": 1.860332541567696e-05, + "loss": 0.6662, + "step": 150, + "task_loss": 0.38656365871429443 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.3433850407600403, + "epoch": 0.08, + "learning_rate": 1.8508313539192398e-05, + "loss": 0.6492, + "step": 160, + "task_loss": 0.1292821764945984 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.853670597076416, + "epoch": 0.08, + "learning_rate": 1.8413301662707842e-05, + "loss": 0.6022, + "step": 170, + "task_loss": 0.4025900363922119 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.6042287945747375, + "epoch": 0.09, + "learning_rate": 1.831828978622328e-05, + "loss": 0.7525, + "step": 180, + "task_loss": 0.23989427089691162 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 1.0951387882232666, + "epoch": 0.09, + "learning_rate": 1.8223277909738718e-05, + "loss": 0.7468, + "step": 190, + "task_loss": 0.5307990908622742 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.27730026841163635, + "epoch": 0.1, + "learning_rate": 1.812826603325416e-05, + "loss": 0.5985, + "step": 200, + "task_loss": 0.05895298719406128 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.9861497282981873, + "epoch": 0.1, + "learning_rate": 1.8033254156769596e-05, + "loss": 0.6073, + "step": 210, + "task_loss": 0.5837754011154175 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.532859206199646, + "epoch": 0.1, + "learning_rate": 1.7938242280285037e-05, + "loss": 0.5815, + "step": 220, + "task_loss": 0.2384859323501587 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.7019934058189392, + "epoch": 0.11, + "learning_rate": 1.7843230403800475e-05, + "loss": 0.778, + "step": 230, + "task_loss": 0.3135228157043457 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.9965037703514099, + "epoch": 0.11, + "learning_rate": 1.7748218527315916e-05, + "loss": 0.6839, + "step": 240, + "task_loss": 0.4901553988456726 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5550552606582642, + "epoch": 0.12, + "learning_rate": 1.7653206650831357e-05, + "loss": 0.4129, + "step": 250, + "task_loss": 0.2684090733528137 + }, + { + "epoch": 0.12, + "eval_accuracy": 0.8761467889908257, + "eval_loss": 0.44158032536506653, + "eval_runtime": 21.7125, + "eval_samples_per_second": 40.161, + "eval_steps_per_second": 5.02, + "step": 250 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5287630558013916, + "epoch": 0.12, + "learning_rate": 1.7558194774346795e-05, + "loss": 0.5833, + "step": 260, + "task_loss": 0.2730463147163391 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.9515942335128784, + "epoch": 0.13, + "learning_rate": 1.7463182897862236e-05, + "loss": 0.6329, + "step": 270, + "task_loss": 0.4847598969936371 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5145795345306396, + "epoch": 0.13, + "learning_rate": 1.7368171021377673e-05, + "loss": 0.6194, + "step": 280, + "task_loss": 0.4189565181732178 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.7319836020469666, + "epoch": 0.14, + "learning_rate": 1.727315914489311e-05, + "loss": 0.4816, + "step": 290, + "task_loss": 0.3963426649570465 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.4611780047416687, + "epoch": 0.14, + "learning_rate": 1.7178147268408552e-05, + "loss": 0.5048, + "step": 300, + "task_loss": 0.19796603918075562 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5850609540939331, + "epoch": 0.15, + "learning_rate": 1.7083135391923993e-05, + "loss": 0.5864, + "step": 310, + "task_loss": 0.2181399166584015 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.4029274880886078, + "epoch": 0.15, + "learning_rate": 1.698812351543943e-05, + "loss": 0.534, + "step": 320, + "task_loss": 0.1707906723022461 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5725364685058594, + "epoch": 0.16, + "learning_rate": 1.689311163895487e-05, + "loss": 0.6037, + "step": 330, + "task_loss": 0.23323196172714233 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.6093335747718811, + "epoch": 0.16, + "learning_rate": 1.679809976247031e-05, + "loss": 0.7098, + "step": 340, + "task_loss": 0.4206717610359192 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 1.0119699239730835, + "epoch": 0.17, + "learning_rate": 1.670308788598575e-05, + "loss": 0.4015, + "step": 350, + "task_loss": 0.6245448589324951 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.6928797960281372, + "epoch": 0.17, + "learning_rate": 1.6608076009501188e-05, + "loss": 0.6882, + "step": 360, + "task_loss": 0.32340091466903687 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.2802911400794983, + "epoch": 0.18, + "learning_rate": 1.651306413301663e-05, + "loss": 0.5088, + "step": 370, + "task_loss": 0.12505391240119934 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.6381040811538696, + "epoch": 0.18, + "learning_rate": 1.641805225653207e-05, + "loss": 0.5281, + "step": 380, + "task_loss": 0.25593793392181396 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 1.2553234100341797, + "epoch": 0.19, + "learning_rate": 1.6323040380047507e-05, + "loss": 0.4673, + "step": 390, + "task_loss": 0.7861297130584717 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.32122546434402466, + "epoch": 0.19, + "learning_rate": 1.622802850356295e-05, + "loss": 0.5358, + "step": 400, + "task_loss": 0.14392141997814178 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.4124768078327179, + "epoch": 0.19, + "learning_rate": 1.6133016627078386e-05, + "loss": 0.388, + "step": 410, + "task_loss": 0.18850776553153992 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5564821362495422, + "epoch": 0.2, + "learning_rate": 1.6038004750593824e-05, + "loss": 0.4706, + "step": 420, + "task_loss": 0.35438424348831177 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.8114709854125977, + "epoch": 0.2, + "learning_rate": 1.5942992874109265e-05, + "loss": 0.3992, + "step": 430, + "task_loss": 0.5884829163551331 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.9019944071769714, + "epoch": 0.21, + "learning_rate": 1.5847980997624702e-05, + "loss": 0.5377, + "step": 440, + "task_loss": 0.4972737431526184 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.4662189781665802, + "epoch": 0.21, + "learning_rate": 1.5752969121140143e-05, + "loss": 0.4676, + "step": 450, + "task_loss": 0.22651702165603638 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.7801036834716797, + "epoch": 0.22, + "learning_rate": 1.5657957244655584e-05, + "loss": 0.5115, + "step": 460, + "task_loss": 0.3273088335990906 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.24446932971477509, + "epoch": 0.22, + "learning_rate": 1.5562945368171022e-05, + "loss": 0.4323, + "step": 470, + "task_loss": 0.28846049308776855 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.31131500005722046, + "epoch": 0.23, + "learning_rate": 1.5467933491686463e-05, + "loss": 0.4727, + "step": 480, + "task_loss": 0.06133585423231125 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.2505212426185608, + "epoch": 0.23, + "learning_rate": 1.53729216152019e-05, + "loss": 0.6103, + "step": 490, + "task_loss": 0.16016395390033722 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.6978294849395752, + "epoch": 0.24, + "learning_rate": 1.527790973871734e-05, + "loss": 0.412, + "step": 500, + "task_loss": 0.3384208679199219 + }, + { + "epoch": 0.24, + "eval_accuracy": 0.8899082568807339, + "eval_loss": 0.49690014123916626, + "eval_runtime": 21.7581, + "eval_samples_per_second": 40.077, + "eval_steps_per_second": 5.01, + "step": 500 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5271720886230469, + "epoch": 0.24, + "learning_rate": 1.5182897862232779e-05, + "loss": 0.3288, + "step": 510, + "task_loss": 0.3533409833908081 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.39492568373680115, + "epoch": 0.25, + "learning_rate": 1.508788598574822e-05, + "loss": 0.3691, + "step": 520, + "task_loss": 0.17543694376945496 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5424761772155762, + "epoch": 0.25, + "learning_rate": 1.499287410926366e-05, + "loss": 0.5957, + "step": 530, + "task_loss": 0.35330963134765625 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.2259323000907898, + "epoch": 0.26, + "learning_rate": 1.4897862232779099e-05, + "loss": 0.4177, + "step": 540, + "task_loss": 0.05318892002105713 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.912689208984375, + "epoch": 0.26, + "learning_rate": 1.4802850356294538e-05, + "loss": 0.5001, + "step": 550, + "task_loss": 0.447698175907135 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.7216652631759644, + "epoch": 0.27, + "learning_rate": 1.4707838479809977e-05, + "loss": 0.5728, + "step": 560, + "task_loss": 0.5098874568939209 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.2023361772298813, + "epoch": 0.27, + "learning_rate": 1.4612826603325417e-05, + "loss": 0.3707, + "step": 570, + "task_loss": 0.22737565636634827 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.46668243408203125, + "epoch": 0.28, + "learning_rate": 1.4517814726840856e-05, + "loss": 0.4068, + "step": 580, + "task_loss": 0.2710917294025421 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5557296276092529, + "epoch": 0.28, + "learning_rate": 1.4422802850356297e-05, + "loss": 0.3836, + "step": 590, + "task_loss": 0.43243685364723206 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.6273843050003052, + "epoch": 0.29, + "learning_rate": 1.4327790973871736e-05, + "loss": 0.5479, + "step": 600, + "task_loss": 0.2661285400390625 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.3024005889892578, + "epoch": 0.29, + "learning_rate": 1.4232779097387176e-05, + "loss": 0.3326, + "step": 610, + "task_loss": 0.11362183094024658 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.6180804967880249, + "epoch": 0.29, + "learning_rate": 1.4137767220902613e-05, + "loss": 0.2927, + "step": 620, + "task_loss": 0.3157673478126526 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.2821282148361206, + "epoch": 0.3, + "learning_rate": 1.4042755344418053e-05, + "loss": 0.3051, + "step": 630, + "task_loss": 0.1337086707353592 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5159429907798767, + "epoch": 0.3, + "learning_rate": 1.3947743467933492e-05, + "loss": 0.4209, + "step": 640, + "task_loss": 0.35967183113098145 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.8615990281105042, + "epoch": 0.31, + "learning_rate": 1.3852731591448931e-05, + "loss": 0.4166, + "step": 650, + "task_loss": 0.6024147868156433 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.1939113736152649, + "epoch": 0.31, + "learning_rate": 1.3757719714964372e-05, + "loss": 0.3193, + "step": 660, + "task_loss": 0.023959562182426453 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.31010207533836365, + "epoch": 0.32, + "learning_rate": 1.3662707838479811e-05, + "loss": 0.3788, + "step": 670, + "task_loss": 0.17067261040210724 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.6690168976783752, + "epoch": 0.32, + "learning_rate": 1.356769596199525e-05, + "loss": 0.3763, + "step": 680, + "task_loss": 0.355002760887146 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.23823483288288116, + "epoch": 0.33, + "learning_rate": 1.347268408551069e-05, + "loss": 0.4861, + "step": 690, + "task_loss": 0.0851842388510704 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.19940122961997986, + "epoch": 0.33, + "learning_rate": 1.337767220902613e-05, + "loss": 0.3373, + "step": 700, + "task_loss": 0.1375439167022705 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.34697651863098145, + "epoch": 0.34, + "learning_rate": 1.3282660332541569e-05, + "loss": 0.2368, + "step": 710, + "task_loss": 0.32895755767822266 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5809268951416016, + "epoch": 0.34, + "learning_rate": 1.3187648456057008e-05, + "loss": 0.4335, + "step": 720, + "task_loss": 0.350343257188797 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.4058021605014801, + "epoch": 0.35, + "learning_rate": 1.3092636579572449e-05, + "loss": 0.3762, + "step": 730, + "task_loss": 0.4189315438270569 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5029650330543518, + "epoch": 0.35, + "learning_rate": 1.2997624703087888e-05, + "loss": 0.3535, + "step": 740, + "task_loss": 0.352708101272583 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.15115290880203247, + "epoch": 0.36, + "learning_rate": 1.2902612826603326e-05, + "loss": 0.3191, + "step": 750, + "task_loss": 0.05512949079275131 + }, + { + "epoch": 0.36, + "eval_accuracy": 0.9162844036697247, + "eval_loss": 0.2716875970363617, + "eval_runtime": 21.7417, + "eval_samples_per_second": 40.107, + "eval_steps_per_second": 5.013, + "step": 750 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.42094355821609497, + "epoch": 0.36, + "learning_rate": 1.2807600950118765e-05, + "loss": 0.3572, + "step": 760, + "task_loss": 0.22914057970046997 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.22227834165096283, + "epoch": 0.37, + "learning_rate": 1.2712589073634205e-05, + "loss": 0.5129, + "step": 770, + "task_loss": 0.13911421597003937 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.6405156254768372, + "epoch": 0.37, + "learning_rate": 1.2617577197149644e-05, + "loss": 0.4576, + "step": 780, + "task_loss": 0.4629451036453247 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.432763010263443, + "epoch": 0.38, + "learning_rate": 1.2522565320665083e-05, + "loss": 0.3085, + "step": 790, + "task_loss": 0.36907055974006653 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.30914443731307983, + "epoch": 0.38, + "learning_rate": 1.2427553444180524e-05, + "loss": 0.4019, + "step": 800, + "task_loss": 0.21935078501701355 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5716268420219421, + "epoch": 0.38, + "learning_rate": 1.2332541567695964e-05, + "loss": 0.3996, + "step": 810, + "task_loss": 0.4366046190261841 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.20762380957603455, + "epoch": 0.39, + "learning_rate": 1.2237529691211403e-05, + "loss": 0.4031, + "step": 820, + "task_loss": 0.07009849697351456 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.48847681283950806, + "epoch": 0.39, + "learning_rate": 1.2142517814726842e-05, + "loss": 0.4559, + "step": 830, + "task_loss": 0.20559999346733093 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.38028720021247864, + "epoch": 0.4, + "learning_rate": 1.2047505938242281e-05, + "loss": 0.2783, + "step": 840, + "task_loss": 0.12343298643827438 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.6831973791122437, + "epoch": 0.4, + "learning_rate": 1.195249406175772e-05, + "loss": 0.4664, + "step": 850, + "task_loss": 0.6523771286010742 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.21925556659698486, + "epoch": 0.41, + "learning_rate": 1.1857482185273158e-05, + "loss": 0.366, + "step": 860, + "task_loss": 0.15919412672519684 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.07446402311325073, + "epoch": 0.41, + "learning_rate": 1.1762470308788601e-05, + "loss": 0.3713, + "step": 870, + "task_loss": 0.014185778796672821 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.31408172845840454, + "epoch": 0.42, + "learning_rate": 1.166745843230404e-05, + "loss": 0.3861, + "step": 880, + "task_loss": 0.13161490857601166 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.14646156132221222, + "epoch": 0.42, + "learning_rate": 1.1572446555819478e-05, + "loss": 0.2613, + "step": 890, + "task_loss": 0.07529384642839432 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.05397426709532738, + "epoch": 0.43, + "learning_rate": 1.1477434679334917e-05, + "loss": 0.3604, + "step": 900, + "task_loss": 0.043072961270809174 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.12949523329734802, + "epoch": 0.43, + "learning_rate": 1.1382422802850357e-05, + "loss": 0.2215, + "step": 910, + "task_loss": 0.029655031859874725 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5533275008201599, + "epoch": 0.44, + "learning_rate": 1.1287410926365796e-05, + "loss": 0.3736, + "step": 920, + "task_loss": 0.3869485855102539 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.06997716426849365, + "epoch": 0.44, + "learning_rate": 1.1192399049881235e-05, + "loss": 0.2925, + "step": 930, + "task_loss": 0.17637060582637787 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5217384099960327, + "epoch": 0.45, + "learning_rate": 1.1097387173396676e-05, + "loss": 0.3657, + "step": 940, + "task_loss": 0.35078302025794983 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.20354126393795013, + "epoch": 0.45, + "learning_rate": 1.1002375296912116e-05, + "loss": 0.2885, + "step": 950, + "task_loss": 0.33367669582366943 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.6612205505371094, + "epoch": 0.46, + "learning_rate": 1.0907363420427555e-05, + "loss": 0.4112, + "step": 960, + "task_loss": 0.38991737365722656 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5148907899856567, + "epoch": 0.46, + "learning_rate": 1.0812351543942994e-05, + "loss": 0.3226, + "step": 970, + "task_loss": 0.2609829604625702 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.6710062026977539, + "epoch": 0.47, + "learning_rate": 1.0717339667458434e-05, + "loss": 0.2813, + "step": 980, + "task_loss": 0.34536463022232056 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.35892608761787415, + "epoch": 0.47, + "learning_rate": 1.0622327790973871e-05, + "loss": 0.3694, + "step": 990, + "task_loss": 0.190689355134964 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.1477653980255127, + "epoch": 0.48, + "learning_rate": 1.052731591448931e-05, + "loss": 0.2688, + "step": 1000, + "task_loss": 0.04818693548440933 + }, + { + "epoch": 0.48, + "eval_accuracy": 0.911697247706422, + "eval_loss": 0.24315589666366577, + "eval_runtime": 21.9413, + "eval_samples_per_second": 39.742, + "eval_steps_per_second": 4.968, + "step": 1000 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.47181323170661926, + "epoch": 0.48, + "learning_rate": 1.0432304038004753e-05, + "loss": 0.2546, + "step": 1010, + "task_loss": 0.2418743222951889 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.21640026569366455, + "epoch": 0.48, + "learning_rate": 1.033729216152019e-05, + "loss": 0.3818, + "step": 1020, + "task_loss": 0.14040601253509521 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.27131223678588867, + "epoch": 0.49, + "learning_rate": 1.024228028503563e-05, + "loss": 0.253, + "step": 1030, + "task_loss": 0.1497097909450531 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.7797483801841736, + "epoch": 0.49, + "learning_rate": 1.014726840855107e-05, + "loss": 0.4507, + "step": 1040, + "task_loss": 0.46980684995651245 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.11180303990840912, + "epoch": 0.5, + "learning_rate": 1.0052256532066509e-05, + "loss": 0.2814, + "step": 1050, + "task_loss": 0.4403308629989624 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.18791627883911133, + "epoch": 0.5, + "learning_rate": 9.95724465558195e-06, + "loss": 0.387, + "step": 1060, + "task_loss": 0.06846746802330017 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.16468051075935364, + "epoch": 0.51, + "learning_rate": 9.862232779097387e-06, + "loss": 0.3529, + "step": 1070, + "task_loss": 0.18533943593502045 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.42276692390441895, + "epoch": 0.51, + "learning_rate": 9.767220902612827e-06, + "loss": 0.2331, + "step": 1080, + "task_loss": 0.23515379428863525 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.31465187668800354, + "epoch": 0.52, + "learning_rate": 9.672209026128266e-06, + "loss": 0.2775, + "step": 1090, + "task_loss": 0.2029043436050415 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.7025846838951111, + "epoch": 0.52, + "learning_rate": 9.577197149643707e-06, + "loss": 0.3614, + "step": 1100, + "task_loss": 0.43116331100463867 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.43838998675346375, + "epoch": 0.53, + "learning_rate": 9.482185273159146e-06, + "loss": 0.2565, + "step": 1110, + "task_loss": 0.24070000648498535 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.526702880859375, + "epoch": 0.53, + "learning_rate": 9.387173396674586e-06, + "loss": 0.3354, + "step": 1120, + "task_loss": 0.28051871061325073 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.471240758895874, + "epoch": 0.54, + "learning_rate": 9.292161520190025e-06, + "loss": 0.3342, + "step": 1130, + "task_loss": 0.2174275815486908 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.13051442801952362, + "epoch": 0.54, + "learning_rate": 9.197149643705464e-06, + "loss": 0.2616, + "step": 1140, + "task_loss": 0.08567549288272858 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5350298881530762, + "epoch": 0.55, + "learning_rate": 9.102137767220904e-06, + "loss": 0.3387, + "step": 1150, + "task_loss": 0.3132125437259674 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.32484662532806396, + "epoch": 0.55, + "learning_rate": 9.007125890736343e-06, + "loss": 0.1753, + "step": 1160, + "task_loss": 0.18460990488529205 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.3009415864944458, + "epoch": 0.56, + "learning_rate": 8.912114014251782e-06, + "loss": 0.3577, + "step": 1170, + "task_loss": 0.22530747950077057 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.4420413672924042, + "epoch": 0.56, + "learning_rate": 8.817102137767222e-06, + "loss": 0.3324, + "step": 1180, + "task_loss": 0.4033554196357727 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.06501191854476929, + "epoch": 0.57, + "learning_rate": 8.722090261282661e-06, + "loss": 0.3307, + "step": 1190, + "task_loss": 0.012013241648674011 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.2965516448020935, + "epoch": 0.57, + "learning_rate": 8.6270783847981e-06, + "loss": 0.3437, + "step": 1200, + "task_loss": 0.13238894939422607 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.36214399337768555, + "epoch": 0.57, + "learning_rate": 8.53206650831354e-06, + "loss": 0.3629, + "step": 1210, + "task_loss": 0.26368093490600586 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.3077784478664398, + "epoch": 0.58, + "learning_rate": 8.437054631828979e-06, + "loss": 0.3036, + "step": 1220, + "task_loss": 0.0927947610616684 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.20123688876628876, + "epoch": 0.58, + "learning_rate": 8.342042755344418e-06, + "loss": 0.2894, + "step": 1230, + "task_loss": 0.010961085557937622 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5563689470291138, + "epoch": 0.59, + "learning_rate": 8.247030878859859e-06, + "loss": 0.3299, + "step": 1240, + "task_loss": 0.3913767337799072 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.471743643283844, + "epoch": 0.59, + "learning_rate": 8.152019002375298e-06, + "loss": 0.3306, + "step": 1250, + "task_loss": 0.19782987236976624 + }, + { + "epoch": 0.59, + "eval_accuracy": 0.9243119266055045, + "eval_loss": 0.20327819883823395, + "eval_runtime": 21.7723, + "eval_samples_per_second": 40.051, + "eval_steps_per_second": 5.006, + "step": 1250 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.01976114884018898, + "epoch": 0.6, + "learning_rate": 8.057007125890736e-06, + "loss": 0.2307, + "step": 1260, + "task_loss": 0.006360933184623718 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.2867504358291626, + "epoch": 0.6, + "learning_rate": 7.961995249406177e-06, + "loss": 0.4469, + "step": 1270, + "task_loss": 0.1376619189977646 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.3130514919757843, + "epoch": 0.61, + "learning_rate": 7.866983372921616e-06, + "loss": 0.2792, + "step": 1280, + "task_loss": 0.30938223004341125 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.21932274103164673, + "epoch": 0.61, + "learning_rate": 7.771971496437056e-06, + "loss": 0.2458, + "step": 1290, + "task_loss": 0.07220742851495743 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.2027844786643982, + "epoch": 0.62, + "learning_rate": 7.676959619952495e-06, + "loss": 0.2947, + "step": 1300, + "task_loss": 0.1091059073805809 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.4004560708999634, + "epoch": 0.62, + "learning_rate": 7.581947743467934e-06, + "loss": 0.437, + "step": 1310, + "task_loss": 0.13185200095176697 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.11762793362140656, + "epoch": 0.63, + "learning_rate": 7.486935866983374e-06, + "loss": 0.2396, + "step": 1320, + "task_loss": 0.035176947712898254 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.38001149892807007, + "epoch": 0.63, + "learning_rate": 7.391923990498813e-06, + "loss": 0.225, + "step": 1330, + "task_loss": 0.49264460802078247 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.16133537888526917, + "epoch": 0.64, + "learning_rate": 7.296912114014253e-06, + "loss": 0.2908, + "step": 1340, + "task_loss": 0.05434826388955116 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5223551988601685, + "epoch": 0.64, + "learning_rate": 7.201900237529692e-06, + "loss": 0.2895, + "step": 1350, + "task_loss": 0.27843406796455383 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.09743548929691315, + "epoch": 0.65, + "learning_rate": 7.106888361045131e-06, + "loss": 0.2579, + "step": 1360, + "task_loss": 0.014732744544744492 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.06801208853721619, + "epoch": 0.65, + "learning_rate": 7.01187648456057e-06, + "loss": 0.2038, + "step": 1370, + "task_loss": 0.01958051323890686 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.49920564889907837, + "epoch": 0.66, + "learning_rate": 6.91686460807601e-06, + "loss": 0.3022, + "step": 1380, + "task_loss": 0.27152562141418457 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.07850401848554611, + "epoch": 0.66, + "learning_rate": 6.82185273159145e-06, + "loss": 0.1903, + "step": 1390, + "task_loss": 0.19040407240390778 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.05175900086760521, + "epoch": 0.67, + "learning_rate": 6.726840855106889e-06, + "loss": 0.1951, + "step": 1400, + "task_loss": 0.0510752871632576 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.11907292902469635, + "epoch": 0.67, + "learning_rate": 6.631828978622329e-06, + "loss": 0.1969, + "step": 1410, + "task_loss": 0.10641947388648987 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.3327619433403015, + "epoch": 0.67, + "learning_rate": 6.536817102137768e-06, + "loss": 0.2099, + "step": 1420, + "task_loss": 0.3468553125858307 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.2979893088340759, + "epoch": 0.68, + "learning_rate": 6.441805225653207e-06, + "loss": 0.3192, + "step": 1430, + "task_loss": 0.26997804641723633 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.03590167686343193, + "epoch": 0.68, + "learning_rate": 6.346793349168646e-06, + "loss": 0.179, + "step": 1440, + "task_loss": 0.00364762544631958 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.16363513469696045, + "epoch": 0.69, + "learning_rate": 6.251781472684086e-06, + "loss": 0.3228, + "step": 1450, + "task_loss": 0.20806732773780823 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.28741005063056946, + "epoch": 0.69, + "learning_rate": 6.156769596199526e-06, + "loss": 0.2762, + "step": 1460, + "task_loss": 0.12442530691623688 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.49749070405960083, + "epoch": 0.7, + "learning_rate": 6.061757719714965e-06, + "loss": 0.3035, + "step": 1470, + "task_loss": 0.22260920703411102 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5554820895195007, + "epoch": 0.7, + "learning_rate": 5.9667458432304035e-06, + "loss": 0.3454, + "step": 1480, + "task_loss": 0.2797449231147766 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.13691575825214386, + "epoch": 0.71, + "learning_rate": 5.871733966745844e-06, + "loss": 0.2283, + "step": 1490, + "task_loss": 0.18624231219291687 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.80184406042099, + "epoch": 0.71, + "learning_rate": 5.776722090261283e-06, + "loss": 0.224, + "step": 1500, + "task_loss": 0.4933280646800995 + }, + { + "epoch": 0.71, + "eval_accuracy": 0.9243119266055045, + "eval_loss": 0.2382841557264328, + "eval_runtime": 21.7644, + "eval_samples_per_second": 40.065, + "eval_steps_per_second": 5.008, + "step": 1500 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.07949607074260712, + "epoch": 0.72, + "learning_rate": 5.681710213776722e-06, + "loss": 0.2137, + "step": 1510, + "task_loss": 0.043377894908189774 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.42658352851867676, + "epoch": 0.72, + "learning_rate": 5.5866983372921624e-06, + "loss": 0.2807, + "step": 1520, + "task_loss": 0.24954530596733093 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.244135782122612, + "epoch": 0.73, + "learning_rate": 5.491686460807602e-06, + "loss": 0.3032, + "step": 1530, + "task_loss": 0.5667487382888794 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.12168382853269577, + "epoch": 0.73, + "learning_rate": 5.39667458432304e-06, + "loss": 0.2224, + "step": 1540, + "task_loss": 0.051823940128088 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5252543091773987, + "epoch": 0.74, + "learning_rate": 5.3016627078384795e-06, + "loss": 0.3264, + "step": 1550, + "task_loss": 0.5932276248931885 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.10500893741846085, + "epoch": 0.74, + "learning_rate": 5.20665083135392e-06, + "loss": 0.2, + "step": 1560, + "task_loss": 0.06464926898479462 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.46214228868484497, + "epoch": 0.75, + "learning_rate": 5.111638954869359e-06, + "loss": 0.3135, + "step": 1570, + "task_loss": 0.27241963148117065 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.41102948784828186, + "epoch": 0.75, + "learning_rate": 5.016627078384798e-06, + "loss": 0.2825, + "step": 1580, + "task_loss": 0.25513461232185364 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.0665275901556015, + "epoch": 0.76, + "learning_rate": 4.921615201900238e-06, + "loss": 0.3124, + "step": 1590, + "task_loss": 0.01647743582725525 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.20230551064014435, + "epoch": 0.76, + "learning_rate": 4.826603325415678e-06, + "loss": 0.2472, + "step": 1600, + "task_loss": 0.6591448783874512 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.1525014340877533, + "epoch": 0.76, + "learning_rate": 4.731591448931116e-06, + "loss": 0.1319, + "step": 1610, + "task_loss": 0.06440484523773193 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.38346511125564575, + "epoch": 0.77, + "learning_rate": 4.636579572446556e-06, + "loss": 0.2337, + "step": 1620, + "task_loss": 0.27461180090904236 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.26302579045295715, + "epoch": 0.77, + "learning_rate": 4.541567695961996e-06, + "loss": 0.2821, + "step": 1630, + "task_loss": 0.1236814334988594 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.624555230140686, + "epoch": 0.78, + "learning_rate": 4.446555819477435e-06, + "loss": 0.3081, + "step": 1640, + "task_loss": 0.3441321849822998 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.04459763318300247, + "epoch": 0.78, + "learning_rate": 4.351543942992874e-06, + "loss": 0.1871, + "step": 1650, + "task_loss": 0.009879574179649353 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.47281283140182495, + "epoch": 0.79, + "learning_rate": 4.256532066508314e-06, + "loss": 0.3221, + "step": 1660, + "task_loss": 0.3637796640396118 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.2519533634185791, + "epoch": 0.79, + "learning_rate": 4.161520190023753e-06, + "loss": 0.2995, + "step": 1670, + "task_loss": 0.24460454285144806 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.3187342584133148, + "epoch": 0.8, + "learning_rate": 4.066508313539192e-06, + "loss": 0.3526, + "step": 1680, + "task_loss": 0.25094783306121826 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.3183995485305786, + "epoch": 0.8, + "learning_rate": 3.9714964370546325e-06, + "loss": 0.2348, + "step": 1690, + "task_loss": 0.16011971235275269 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5012897849082947, + "epoch": 0.81, + "learning_rate": 3.876484560570072e-06, + "loss": 0.2874, + "step": 1700, + "task_loss": 0.30354738235473633 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.09144863486289978, + "epoch": 0.81, + "learning_rate": 3.781472684085511e-06, + "loss": 0.2732, + "step": 1710, + "task_loss": 0.11224113404750824 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.39011454582214355, + "epoch": 0.82, + "learning_rate": 3.6864608076009504e-06, + "loss": 0.205, + "step": 1720, + "task_loss": 0.3499948978424072 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.0968744307756424, + "epoch": 0.82, + "learning_rate": 3.5914489311163897e-06, + "loss": 0.2088, + "step": 1730, + "task_loss": 0.03987715393304825 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.49030616879463196, + "epoch": 0.83, + "learning_rate": 3.4964370546318295e-06, + "loss": 0.3065, + "step": 1740, + "task_loss": 0.33337563276290894 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.3354378938674927, + "epoch": 0.83, + "learning_rate": 3.4014251781472683e-06, + "loss": 0.2082, + "step": 1750, + "task_loss": 0.2503882646560669 + }, + { + "epoch": 0.83, + "eval_accuracy": 0.9254587155963303, + "eval_loss": 0.22330859303474426, + "eval_runtime": 21.7241, + "eval_samples_per_second": 40.14, + "eval_steps_per_second": 5.017, + "step": 1750 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.07079672068357468, + "epoch": 0.84, + "learning_rate": 3.306413301662708e-06, + "loss": 0.3192, + "step": 1760, + "task_loss": 0.006712011992931366 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.12471574544906616, + "epoch": 0.84, + "learning_rate": 3.211401425178148e-06, + "loss": 0.1666, + "step": 1770, + "task_loss": 0.047819193452596664 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.11018840968608856, + "epoch": 0.85, + "learning_rate": 3.1163895486935867e-06, + "loss": 0.2473, + "step": 1780, + "task_loss": 0.029587876051664352 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.27604231238365173, + "epoch": 0.85, + "learning_rate": 3.0213776722090264e-06, + "loss": 0.299, + "step": 1790, + "task_loss": 0.12574651837348938 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.19024261832237244, + "epoch": 0.86, + "learning_rate": 2.9263657957244658e-06, + "loss": 0.2579, + "step": 1800, + "task_loss": 0.11378154903650284 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.24252384901046753, + "epoch": 0.86, + "learning_rate": 2.8313539192399055e-06, + "loss": 0.2374, + "step": 1810, + "task_loss": 0.08886312693357468 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.40556859970092773, + "epoch": 0.86, + "learning_rate": 2.7363420427553444e-06, + "loss": 0.2556, + "step": 1820, + "task_loss": 0.16999585926532745 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.6070685386657715, + "epoch": 0.87, + "learning_rate": 2.641330166270784e-06, + "loss": 0.3513, + "step": 1830, + "task_loss": 0.21770261228084564 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.11093200743198395, + "epoch": 0.87, + "learning_rate": 2.546318289786224e-06, + "loss": 0.2251, + "step": 1840, + "task_loss": 0.07486331462860107 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.15582458674907684, + "epoch": 0.88, + "learning_rate": 2.4513064133016627e-06, + "loss": 0.3013, + "step": 1850, + "task_loss": 0.2854090631008148 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5564573407173157, + "epoch": 0.88, + "learning_rate": 2.356294536817102e-06, + "loss": 0.3546, + "step": 1860, + "task_loss": 0.2976299524307251 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.35662102699279785, + "epoch": 0.89, + "learning_rate": 2.261282660332542e-06, + "loss": 0.1928, + "step": 1870, + "task_loss": 0.17799919843673706 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.07805321365594864, + "epoch": 0.89, + "learning_rate": 2.166270783847981e-06, + "loss": 0.273, + "step": 1880, + "task_loss": 0.032489631325006485 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.0970858484506607, + "epoch": 0.9, + "learning_rate": 2.071258907363421e-06, + "loss": 0.2183, + "step": 1890, + "task_loss": 0.011649325489997864 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.06496274471282959, + "epoch": 0.9, + "learning_rate": 1.97624703087886e-06, + "loss": 0.3183, + "step": 1900, + "task_loss": 0.09150334447622299 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.15299738943576813, + "epoch": 0.91, + "learning_rate": 1.8812351543942995e-06, + "loss": 0.2088, + "step": 1910, + "task_loss": 0.06896167993545532 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.3061752915382385, + "epoch": 0.91, + "learning_rate": 1.7862232779097388e-06, + "loss": 0.1873, + "step": 1920, + "task_loss": 0.14114254713058472 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.04780574142932892, + "epoch": 0.92, + "learning_rate": 1.691211401425178e-06, + "loss": 0.1784, + "step": 1930, + "task_loss": 0.009304128587245941 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.08279172331094742, + "epoch": 0.92, + "learning_rate": 1.5961995249406176e-06, + "loss": 0.2511, + "step": 1940, + "task_loss": 0.017669253051280975 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.033537302166223526, + "epoch": 0.93, + "learning_rate": 1.5011876484560572e-06, + "loss": 0.2502, + "step": 1950, + "task_loss": 0.04457058385014534 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.1866932362318039, + "epoch": 0.93, + "learning_rate": 1.4061757719714967e-06, + "loss": 0.251, + "step": 1960, + "task_loss": 0.1178937703371048 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.3302549719810486, + "epoch": 0.94, + "learning_rate": 1.311163895486936e-06, + "loss": 0.1995, + "step": 1970, + "task_loss": 0.2511519193649292 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.2090863585472107, + "epoch": 0.94, + "learning_rate": 1.2161520190023753e-06, + "loss": 0.2799, + "step": 1980, + "task_loss": 0.14533783495426178 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.064464271068573, + "epoch": 0.95, + "learning_rate": 1.1211401425178148e-06, + "loss": 0.3001, + "step": 1990, + "task_loss": 0.11410848796367645 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.5515795946121216, + "epoch": 0.95, + "learning_rate": 1.0261282660332544e-06, + "loss": 0.2161, + "step": 2000, + "task_loss": 0.21116505563259125 + }, + { + "epoch": 0.95, + "eval_accuracy": 0.9254587155963303, + "eval_loss": 0.22065171599388123, + "eval_runtime": 21.7453, + "eval_samples_per_second": 40.101, + "eval_steps_per_second": 5.013, + "step": 2000 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.33025607466697693, + "epoch": 0.95, + "learning_rate": 9.311163895486937e-07, + "loss": 0.1835, + "step": 2010, + "task_loss": 0.4712632894515991 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.30521997809410095, + "epoch": 0.96, + "learning_rate": 8.361045130641331e-07, + "loss": 0.2597, + "step": 2020, + "task_loss": 0.2004762589931488 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.10356535017490387, + "epoch": 0.96, + "learning_rate": 7.410926365795724e-07, + "loss": 0.2461, + "step": 2030, + "task_loss": 0.07815247774124146 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.07229259610176086, + "epoch": 0.97, + "learning_rate": 6.460807600950119e-07, + "loss": 0.1703, + "step": 2040, + "task_loss": 0.188236802816391 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.8154426217079163, + "epoch": 0.97, + "learning_rate": 5.510688836104513e-07, + "loss": 0.2093, + "step": 2050, + "task_loss": 0.6037241220474243 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.26825618743896484, + "epoch": 0.98, + "learning_rate": 4.560570071258908e-07, + "loss": 0.2883, + "step": 2060, + "task_loss": 0.31348007917404175 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.7826902270317078, + "epoch": 0.98, + "learning_rate": 3.610451306413302e-07, + "loss": 0.3093, + "step": 2070, + "task_loss": 0.6269980669021606 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.39573800563812256, + "epoch": 0.99, + "learning_rate": 2.660332541567696e-07, + "loss": 0.3111, + "step": 2080, + "task_loss": 0.24385812878608704 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.23440049588680267, + "epoch": 0.99, + "learning_rate": 1.7102137767220902e-07, + "loss": 0.2771, + "step": 2090, + "task_loss": 0.12204352021217346 + }, + { + "compression/movement_sparsity/importance_regularization_factor": 0.0, + "compression/movement_sparsity/importance_threshold": -Infinity, + "compression/movement_sparsity/linear_layer_sparsity": 0.0, + "compression/movement_sparsity/model_sparsity": 0.0, + "compression_loss": 0.0, + "distillation_loss": 0.2844744920730591, + "epoch": 1.0, + "learning_rate": 7.600950118764846e-08, + "loss": 0.2458, + "step": 2100, + "task_loss": 0.3573310077190399 + }, + { + "epoch": 1.0, + "step": 2105, + "total_flos": 4441630972486656.0, + "train_loss": 0.40093172477146793, + "train_runtime": 1341.7612, + "train_samples_per_second": 50.194, + "train_steps_per_second": 1.569 + } + ], + "max_steps": 2105, + "num_train_epochs": 1, + "total_flos": 4441630972486656.0, + "trial_name": null, + "trial_params": null +}