{ "best_global_step": null, "best_metric": null, "best_model_checkpoint": null, "epoch": 2.843601895734597, "eval_steps": 500, "global_step": 1200, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0, "step": 0, "train/alpha": 0.2, "train/clean_ce_loss": 0.203125, "train/clean_loss_scale": 4.511524677276611, "train/clean_loss_weight": 0.9023049473762512, "train/combined_loss": 1.0996841192245483, "train/pos_loss/pos0": 0.49648502469062805, "train/pos_loss/pos1": 0.9778217673301697, "train/pos_loss/pos2": 1.389070749282837, "train/pos_loss/pos3": 1.8537260293960571, "train/pos_loss/pos4": 2.3314146995544434, "train/pos_loss/pos5": 2.723902463912964, "train/pos_loss/pos6": 3.1192147731781006, "train/pos_loss/weight0": 3.563014268875122, "train/pos_loss/weight1": 1.7744934558868408, "train/pos_loss/weight2": 0.8802329301834106, "train/pos_loss/weight3": 0.43310263752937317, "train/pos_loss/weight4": 0.20953750610351562, "train/pos_loss/weight5": 0.09775494784116745, "train/pos_loss/weight6": 0.041863664984703064, "train/pos_weight_progress": 0.0, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9164034128189087 }, { "epoch": 0, "step": 0, "train/alpha": 0.2, "train/clean_ce_loss": 0.1982421875, "train/clean_loss_scale": 4.078215599060059, "train/clean_loss_weight": 0.8156431317329407, "train/combined_loss": 0.9701693058013916, "train/pos_loss/pos0": 0.41213464736938477, "train/pos_loss/pos1": 0.7705235481262207, "train/pos_loss/pos2": 1.1630730628967285, "train/pos_loss/pos3": 1.6438443660736084, "train/pos_loss/pos4": 2.10780668258667, "train/pos_loss/pos5": 2.556405782699585, "train/pos_loss/pos6": 2.780921697616577, "train/pos_loss/weight0": 3.0614287853240967, "train/pos_loss/weight1": 1.9781510829925537, "train/pos_loss/weight2": 1.1244336366653442, "train/pos_loss/weight3": 0.527822732925415, "train/pos_loss/weight4": 0.2122650295495987, "train/pos_loss/weight5": 0.07148314267396927, "train/pos_loss/weight6": 0.024415219202637672, "train/pos_weight_progress": 0.0, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8084744215011597 }, { "epoch": 0, "step": 0, "train/alpha": 0.2, "train/clean_ce_loss": 0.25, "train/clean_loss_scale": 3.7402517795562744, "train/clean_loss_weight": 0.7480503916740417, "train/combined_loss": 1.1220755577087402, "train/pos_loss/pos0": 0.4840814173221588, "train/pos_loss/pos1": 0.8135052919387817, "train/pos_loss/pos2": 1.2519017457962036, "train/pos_loss/pos3": 1.6865466833114624, "train/pos_loss/pos4": 2.0643720626831055, "train/pos_loss/pos5": 2.5065393447875977, "train/pos_loss/pos6": 2.85445237159729, "train/pos_loss/weight0": 2.7850470542907715, "train/pos_loss/weight1": 1.919108271598816, "train/pos_loss/weight2": 1.176357626914978, "train/pos_loss/weight3": 0.6424422264099121, "train/pos_loss/weight4": 0.29898470640182495, "train/pos_loss/weight5": 0.12511736154556274, "train/pos_loss/weight6": 0.052942827343940735, "train/pos_weight_progress": 0.0, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9350629448890686 }, { "epoch": 0, "step": 0, "train/alpha": 0.2, "train/clean_ce_loss": 0.19140625, "train/clean_loss_scale": 3.394860029220581, "train/clean_loss_weight": 0.6789720058441162, "train/combined_loss": 0.7797569036483765, "train/pos_loss/pos0": 0.3453536629676819, "train/pos_loss/pos1": 0.5214296579360962, "train/pos_loss/pos2": 0.8380880355834961, "train/pos_loss/pos3": 1.3898426294326782, "train/pos_loss/pos4": 1.7260123491287231, "train/pos_loss/pos5": 2.138331651687622, "train/pos_loss/pos6": 2.642733335494995, "train/pos_loss/weight0": 2.9282236099243164, "train/pos_loss/weight1": 1.9519730806350708, "train/pos_loss/weight2": 1.1453887224197388, "train/pos_loss/weight3": 0.5809472799301147, "train/pos_loss/weight4": 0.25822827219963074, "train/pos_loss/weight5": 0.0967177152633667, "train/pos_loss/weight6": 0.03852115571498871, "train/pos_weight_progress": 0.0, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6497974395751953 }, { "epoch": 0.002369668246445498, "step": 1, "train/alpha": 0.2, "train/clean_ce_loss": 0.2255859375, "train/clean_loss_scale": 4.100978374481201, "train/clean_loss_weight": 0.8201956748962402, "train/combined_loss": 1.1101475954055786, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.00023696682464454977, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9251230359077454 }, { "epoch": 0.002369668246445498, "step": 1, "train/alpha": 0.2, "train/clean_ce_loss": 0.275390625, "train/clean_loss_scale": 5.162907123565674, "train/clean_loss_weight": 1.0325814485549927, "train/combined_loss": 1.7061794996261597, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.00023696682464454977, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4218162298202515 }, { "epoch": 0.002369668246445498, "step": 1, "train/alpha": 0.2, "train/clean_ce_loss": 0.26953125, "train/clean_loss_scale": 3.930144786834717, "train/clean_loss_weight": 0.7860289812088013, "train/combined_loss": 1.2711561918258667, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.00023696682464454977, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0592968463897705 }, { "epoch": 0.002369668246445498, "step": 1, "train/alpha": 0.2, "train/clean_ce_loss": 0.25, "train/clean_loss_scale": 3.717956304550171, "train/clean_loss_weight": 0.7435912489891052, "train/combined_loss": 1.1153868436813354, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.00023696682464454977, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9294890761375427 }, { "epoch": 0.004739336492890996, "step": 2, "train/alpha": 0.2, "train/clean_ce_loss": 0.1787109375, "train/clean_loss_scale": 4.358402729034424, "train/clean_loss_weight": 0.8716805577278137, "train/combined_loss": 0.9346730709075928, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.00047393364928909954, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7788942456245422 }, { "epoch": 0.004739336492890996, "step": 2, "train/alpha": 0.2, "train/clean_ce_loss": 0.4140625, "train/clean_loss_scale": 4.275359153747559, "train/clean_loss_weight": 0.8550718426704407, "train/combined_loss": 2.124319076538086, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.00047393364928909954, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.770265817642212 }, { "epoch": 0.004739336492890996, "step": 2, "train/alpha": 0.2, "train/clean_ce_loss": 0.2890625, "train/clean_loss_scale": 3.844139337539673, "train/clean_loss_weight": 0.7688278555870056, "train/combined_loss": 1.3334357738494873, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.00047393364928909954, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.111196517944336 }, { "epoch": 0.004739336492890996, "step": 2, "train/alpha": 0.2, "train/clean_ce_loss": 0.3203125, "train/clean_loss_scale": 4.010780334472656, "train/clean_loss_weight": 0.8021560907363892, "train/combined_loss": 1.5416436195373535, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.00047393364928909954, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.284703016281128 }, { "epoch": 0.0071090047393364926, "step": 3, "train/alpha": 0.2, "train/clean_ce_loss": 0.43359375, "train/clean_loss_scale": 4.109915733337402, "train/clean_loss_weight": 0.8219831585884094, "train/combined_loss": 2.1384406089782715, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.0007109004739336493, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.7820338010787964 }, { "epoch": 0.0071090047393364926, "step": 3, "train/alpha": 0.2, "train/clean_ce_loss": 0.2373046875, "train/clean_loss_scale": 3.8449079990386963, "train/clean_loss_weight": 0.7689816355705261, "train/combined_loss": 1.0948976278305054, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.0007109004739336493, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9124146699905396 }, { "epoch": 0.0071090047393364926, "step": 3, "train/alpha": 0.2, "train/clean_ce_loss": 0.34765625, "train/clean_loss_scale": 4.112997531890869, "train/clean_loss_weight": 0.8225995302200317, "train/combined_loss": 1.7158912420272827, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.0007109004739336493, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.429909348487854 }, { "epoch": 0.0071090047393364926, "step": 3, "train/alpha": 0.2, "train/clean_ce_loss": 0.212890625, "train/clean_loss_scale": 4.001126289367676, "train/clean_loss_weight": 0.8002252578735352, "train/combined_loss": 1.0221627950668335, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.0007109004739336493, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8518022894859314 }, { "epoch": 0.009478672985781991, "step": 4, "train/alpha": 0.2, "train/clean_ce_loss": 0.1796875, "train/clean_loss_scale": 3.6903421878814697, "train/clean_loss_weight": 0.7380684614181519, "train/combined_loss": 0.7957299947738647, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.0009478672985781991, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6631083488464355 }, { "epoch": 0.009478672985781991, "step": 4, "train/alpha": 0.2, "train/clean_ce_loss": 0.2578125, "train/clean_loss_scale": 4.4380693435668945, "train/clean_loss_weight": 0.8876138925552368, "train/combined_loss": 1.3730276823043823, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.0009478672985781991, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.144189715385437 }, { "epoch": 0.009478672985781991, "step": 4, "train/alpha": 0.2, "train/clean_ce_loss": 0.25390625, "train/clean_loss_scale": 4.644913673400879, "train/clean_loss_weight": 0.9289827346801758, "train/combined_loss": 1.4152472019195557, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.0009478672985781991, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1793726682662964 }, { "epoch": 0.009478672985781991, "step": 4, "train/alpha": 0.2, "train/clean_ce_loss": 0.1767578125, "train/clean_loss_scale": 4.233400344848633, "train/clean_loss_weight": 0.8466801047325134, "train/combined_loss": 0.897943913936615, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.0009478672985781991, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7482866048812866 }, { "epoch": 0.011848341232227487, "grad_norm": 37.255760192871094, "learning_rate": 6.299212598425197e-08, "loss": 5.0158, "step": 5 }, { "epoch": 0.011848341232227487, "step": 5, "train/alpha": 0.2, "train/clean_ce_loss": 0.279296875, "train/clean_loss_scale": 3.5288825035095215, "train/clean_loss_weight": 0.7057765126228333, "train/combined_loss": 1.1827269792556763, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.001184834123222749, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9856058359146118 }, { "epoch": 0.011848341232227487, "step": 5, "train/alpha": 0.2, "train/clean_ce_loss": 0.53125, "train/clean_loss_scale": 3.3020496368408203, "train/clean_loss_weight": 0.6604099273681641, "train/combined_loss": 2.1050565242767334, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.001184834123222749, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.754213809967041 }, { "epoch": 0.011848341232227487, "step": 5, "train/alpha": 0.2, "train/clean_ce_loss": 0.1884765625, "train/clean_loss_scale": 4.619018077850342, "train/clean_loss_weight": 0.9238036274909973, "train/combined_loss": 1.0446919202804565, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.001184834123222749, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8705766201019287 }, { "epoch": 0.011848341232227487, "step": 5, "train/alpha": 0.2, "train/clean_ce_loss": 0.248046875, "train/clean_loss_scale": 4.91286563873291, "train/clean_loss_weight": 0.9825731515884399, "train/combined_loss": 1.4623452425003052, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.001184834123222749, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.218621015548706 }, { "epoch": 0.014218009478672985, "step": 6, "train/alpha": 0.2, "train/clean_ce_loss": 0.31640625, "train/clean_loss_scale": 3.9748952388763428, "train/clean_loss_weight": 0.7949790358543396, "train/combined_loss": 1.5092180967330933, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.0014218009478672985, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2576817274093628 }, { "epoch": 0.014218009478672985, "step": 6, "train/alpha": 0.2, "train/clean_ce_loss": 0.380859375, "train/clean_loss_scale": 3.8665542602539062, "train/clean_loss_weight": 0.7733108401298523, "train/combined_loss": 1.7671360969543457, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.0014218009478672985, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4726134538650513 }, { "epoch": 0.014218009478672985, "step": 6, "train/alpha": 0.2, "train/clean_ce_loss": 0.275390625, "train/clean_loss_scale": 3.049146890640259, "train/clean_loss_weight": 0.6098293662071228, "train/combined_loss": 1.0076477527618408, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.0014218009478672985, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8397064805030823 }, { "epoch": 0.014218009478672985, "step": 6, "train/alpha": 0.2, "train/clean_ce_loss": 0.412109375, "train/clean_loss_scale": 3.340461015701294, "train/clean_loss_weight": 0.6680921912193298, "train/combined_loss": 1.651962399482727, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.0014218009478672985, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3766353130340576 }, { "epoch": 0.016587677725118485, "step": 7, "train/alpha": 0.2, "train/clean_ce_loss": 0.2890625, "train/clean_loss_scale": 4.570075988769531, "train/clean_loss_weight": 0.9140152335166931, "train/combined_loss": 1.585245132446289, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.0016587677725118483, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3210375308990479 }, { "epoch": 0.016587677725118485, "step": 7, "train/alpha": 0.2, "train/clean_ce_loss": 0.287109375, "train/clean_loss_scale": 3.873563051223755, "train/clean_loss_weight": 0.7747126221656799, "train/combined_loss": 1.3345634937286377, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.0016587677725118483, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1121362447738647 }, { "epoch": 0.016587677725118485, "step": 7, "train/alpha": 0.2, "train/clean_ce_loss": 0.267578125, "train/clean_loss_scale": 4.206150054931641, "train/clean_loss_weight": 0.841230034828186, "train/combined_loss": 1.3505685329437256, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.0016587677725118483, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1254737377166748 }, { "epoch": 0.016587677725118485, "step": 7, "train/alpha": 0.2, "train/clean_ce_loss": 0.42578125, "train/clean_loss_scale": 3.209045648574829, "train/clean_loss_weight": 0.6418091654777527, "train/combined_loss": 1.6396217346191406, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.0016587677725118483, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3663514852523804 }, { "epoch": 0.018957345971563982, "step": 8, "train/alpha": 0.2, "train/clean_ce_loss": 0.3359375, "train/clean_loss_scale": 3.1581194400787354, "train/clean_loss_weight": 0.6316239237785339, "train/combined_loss": 1.2731168270111084, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.0018957345971563982, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0609307289123535 }, { "epoch": 0.018957345971563982, "step": 8, "train/alpha": 0.2, "train/clean_ce_loss": 0.515625, "train/clean_loss_scale": 3.420492172241211, "train/clean_loss_weight": 0.6840984225273132, "train/combined_loss": 2.116429567337036, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.0018957345971563982, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.7636913061141968 }, { "epoch": 0.018957345971563982, "step": 8, "train/alpha": 0.2, "train/clean_ce_loss": 0.302734375, "train/clean_loss_scale": 3.9883882999420166, "train/clean_loss_weight": 0.7976776957511902, "train/combined_loss": 1.448906660079956, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.0018957345971563982, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2074222564697266 }, { "epoch": 0.018957345971563982, "step": 8, "train/alpha": 0.2, "train/clean_ce_loss": 0.41796875, "train/clean_loss_scale": 2.9982070922851562, "train/clean_loss_weight": 0.5996414422988892, "train/combined_loss": 1.5037882328033447, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.0018957345971563982, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2531569004058838 }, { "epoch": 0.02132701421800948, "step": 9, "train/alpha": 0.2, "train/clean_ce_loss": 0.353515625, "train/clean_loss_scale": 2.697275161743164, "train/clean_loss_weight": 0.5394550561904907, "train/combined_loss": 1.1442346572875977, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.0021327014218009478, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.953528881072998 }, { "epoch": 0.02132701421800948, "step": 9, "train/alpha": 0.2, "train/clean_ce_loss": 0.330078125, "train/clean_loss_scale": 4.386131286621094, "train/clean_loss_weight": 0.8772262930870056, "train/combined_loss": 1.7373192310333252, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.0021327014218009478, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4477660655975342 }, { "epoch": 0.02132701421800948, "step": 9, "train/alpha": 0.2, "train/clean_ce_loss": 0.197265625, "train/clean_loss_scale": 3.9224536418914795, "train/clean_loss_weight": 0.7844907641410828, "train/combined_loss": 0.9285182952880859, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.0021327014218009478, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7737652659416199 }, { "epoch": 0.02132701421800948, "step": 9, "train/alpha": 0.2, "train/clean_ce_loss": 0.431640625, "train/clean_loss_scale": 3.2716400623321533, "train/clean_loss_weight": 0.6543280482292175, "train/combined_loss": 1.6946073770523071, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.0021327014218009478, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.412172794342041 }, { "epoch": 0.023696682464454975, "grad_norm": 36.731754302978516, "learning_rate": 1.4173228346456693e-07, "loss": 5.076, "step": 10 }, { "epoch": 0.023696682464454975, "step": 10, "train/alpha": 0.2, "train/clean_ce_loss": 0.29296875, "train/clean_loss_scale": 3.7969274520874023, "train/clean_loss_weight": 0.7593855261802673, "train/combined_loss": 1.3348573446273804, "train/pos_loss/pos0": 0.5758299827575684, "train/pos_loss/pos1": 0.9634164571762085, "train/pos_loss/pos2": 1.3915501832962036, "train/pos_loss/pos3": 1.778464913368225, "train/pos_loss/pos4": 2.2532596588134766, "train/pos_loss/pos5": 2.4743094444274902, "train/pos_loss/pos6": 2.919783353805542, "train/pos_loss/weight0": 2.5950448513031006, "train/pos_loss/weight1": 1.8284908533096313, "train/pos_loss/weight2": 1.197707176208496, "train/pos_loss/weight3": 0.7197592258453369, "train/pos_loss/weight4": 0.382377028465271, "train/pos_loss/weight5": 0.1820412576198578, "train/pos_loss/weight6": 0.09458011388778687, "train/pos_weight_progress": 0.002369668246445498, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.112381100654602 }, { "epoch": 0.023696682464454975, "step": 10, "train/alpha": 0.2, "train/clean_ce_loss": 0.29296875, "train/clean_loss_scale": 3.676830530166626, "train/clean_loss_weight": 0.7353661060333252, "train/combined_loss": 1.2926357984542847, "train/pos_loss/pos0": 0.5737395286560059, "train/pos_loss/pos1": 1.04105544090271, "train/pos_loss/pos2": 1.5792064666748047, "train/pos_loss/pos3": 2.0897610187530518, "train/pos_loss/pos4": 2.6679186820983887, "train/pos_loss/pos5": 3.2177841663360596, "train/pos_loss/pos6": 3.465186357498169, "train/pos_loss/weight0": 3.0999295711517334, "train/pos_loss/weight1": 1.962023138999939, "train/pos_loss/weight2": 1.1067910194396973, "train/pos_loss/weight3": 0.5222539305686951, "train/pos_loss/weight4": 0.21178999543190002, "train/pos_loss/weight5": 0.07089737057685852, "train/pos_loss/weight6": 0.026314985007047653, "train/pos_weight_progress": 0.002369668246445498, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.077196478843689 }, { "epoch": 0.023696682464454975, "step": 10, "train/alpha": 0.2, "train/clean_ce_loss": 0.30859375, "train/clean_loss_scale": 2.953160047531128, "train/clean_loss_weight": 0.5906320214271545, "train/combined_loss": 1.0935920476913452, "train/pos_loss/pos0": 0.4624355137348175, "train/pos_loss/pos1": 1.0069913864135742, "train/pos_loss/pos2": 1.4482991695404053, "train/pos_loss/pos3": 1.798815131187439, "train/pos_loss/pos4": 2.297410488128662, "train/pos_loss/pos5": 2.777390956878662, "train/pos_loss/pos6": 3.4249112606048584, "train/pos_loss/weight0": 3.2777037620544434, "train/pos_loss/weight1": 2.010021924972534, "train/pos_loss/weight2": 1.0574122667312622, "train/pos_loss/weight3": 0.4486446976661682, "train/pos_loss/weight4": 0.15699903666973114, "train/pos_loss/weight5": 0.03956563025712967, "train/pos_loss/weight6": 0.009652948938310146, "train/pos_weight_progress": 0.002369668246445498, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9113267064094543 }, { "epoch": 0.023696682464454975, "step": 10, "train/alpha": 0.2, "train/clean_ce_loss": 0.2275390625, "train/clean_loss_scale": 3.74646258354187, "train/clean_loss_weight": 0.7492925524711609, "train/combined_loss": 1.0229599475860596, "train/pos_loss/pos0": 0.43258315324783325, "train/pos_loss/pos1": 0.8089306354522705, "train/pos_loss/pos2": 1.2269717454910278, "train/pos_loss/pos3": 1.704798936843872, "train/pos_loss/pos4": 2.1838667392730713, "train/pos_loss/pos5": 2.7343857288360596, "train/pos_loss/pos6": 3.2143990993499756, "train/pos_loss/weight0": 3.026073694229126, "train/pos_loss/weight1": 2.017758846282959, "train/pos_loss/weight2": 1.1057112216949463, "train/pos_loss/weight3": 0.5307342410087585, "train/pos_loss/weight4": 0.2197258472442627, "train/pos_loss/weight5": 0.07338833808898926, "train/pos_loss/weight6": 0.026607180014252663, "train/pos_weight_progress": 0.002369668246445498, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8524665832519531 }, { "epoch": 0.026066350710900472, "step": 11, "train/alpha": 0.2, "train/clean_ce_loss": 0.2275390625, "train/clean_loss_scale": 3.892887830734253, "train/clean_loss_weight": 0.7785775661468506, "train/combined_loss": 1.0629408359527588, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.0026066350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8857840299606323 }, { "epoch": 0.026066350710900472, "step": 11, "train/alpha": 0.2, "train/clean_ce_loss": 0.361328125, "train/clean_loss_scale": 4.03380823135376, "train/clean_loss_weight": 0.8067616820335388, "train/combined_loss": 1.74903404712677, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.0026066350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4575283527374268 }, { "epoch": 0.026066350710900472, "step": 11, "train/alpha": 0.2, "train/clean_ce_loss": 0.32421875, "train/clean_loss_scale": 3.2932887077331543, "train/clean_loss_weight": 0.6586577296257019, "train/combined_loss": 1.2812950611114502, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.0026066350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0677459239959717 }, { "epoch": 0.026066350710900472, "step": 11, "train/alpha": 0.2, "train/clean_ce_loss": 0.2890625, "train/clean_loss_scale": 3.482861042022705, "train/clean_loss_weight": 0.6965722441673279, "train/combined_loss": 1.2081174850463867, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.0026066350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.006764531135559 }, { "epoch": 0.02843601895734597, "step": 12, "train/alpha": 0.2, "train/clean_ce_loss": 0.4765625, "train/clean_loss_scale": 3.004070997238159, "train/clean_loss_weight": 0.6008142232894897, "train/combined_loss": 1.7179532051086426, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.002843601895734597, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.431627631187439 }, { "epoch": 0.02843601895734597, "step": 12, "train/alpha": 0.2, "train/clean_ce_loss": 0.302734375, "train/clean_loss_scale": 3.205136775970459, "train/clean_loss_weight": 0.6410273909568787, "train/combined_loss": 1.1643661260604858, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.002843601895734597, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9703050851821899 }, { "epoch": 0.02843601895734597, "step": 12, "train/alpha": 0.2, "train/clean_ce_loss": 0.216796875, "train/clean_loss_scale": 3.4303531646728516, "train/clean_loss_weight": 0.6860706210136414, "train/combined_loss": 0.8924278020858765, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.002843601895734597, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7436898350715637 }, { "epoch": 0.02843601895734597, "step": 12, "train/alpha": 0.2, "train/clean_ce_loss": 0.2890625, "train/clean_loss_scale": 3.4786715507507324, "train/clean_loss_weight": 0.6957343220710754, "train/combined_loss": 1.2066642045974731, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.002843601895734597, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0055534839630127 }, { "epoch": 0.030805687203791468, "step": 13, "train/alpha": 0.2, "train/clean_ce_loss": 0.486328125, "train/clean_loss_scale": 2.6385130882263184, "train/clean_loss_weight": 0.5277026295661926, "train/combined_loss": 1.5398197174072266, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.003080568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2831830978393555 }, { "epoch": 0.030805687203791468, "step": 13, "train/alpha": 0.2, "train/clean_ce_loss": 0.189453125, "train/clean_loss_scale": 3.8044755458831787, "train/clean_loss_weight": 0.7608951330184937, "train/combined_loss": 0.8649237155914307, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.003080568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7207697629928589 }, { "epoch": 0.030805687203791468, "step": 13, "train/alpha": 0.2, "train/clean_ce_loss": 0.30078125, "train/clean_loss_scale": 5.081211566925049, "train/clean_loss_weight": 1.0162423849105835, "train/combined_loss": 1.8339998722076416, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.003080568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5283331871032715 }, { "epoch": 0.030805687203791468, "step": 13, "train/alpha": 0.2, "train/clean_ce_loss": 0.291015625, "train/clean_loss_scale": 2.4446167945861816, "train/clean_loss_weight": 0.4889233708381653, "train/combined_loss": 0.8537060022354126, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.003080568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7114216685295105 }, { "epoch": 0.03317535545023697, "step": 14, "train/alpha": 0.2, "train/clean_ce_loss": 0.3046875, "train/clean_loss_scale": 3.8476622104644775, "train/clean_loss_weight": 0.7695324420928955, "train/combined_loss": 1.406801462173462, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.0033175355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1723345518112183 }, { "epoch": 0.03317535545023697, "step": 14, "train/alpha": 0.2, "train/clean_ce_loss": 0.3046875, "train/clean_loss_scale": 2.8889429569244385, "train/clean_loss_weight": 0.5777885913848877, "train/combined_loss": 1.0562697649002075, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.0033175355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8802248239517212 }, { "epoch": 0.03317535545023697, "step": 14, "train/alpha": 0.2, "train/clean_ce_loss": 0.37890625, "train/clean_loss_scale": 3.776972532272339, "train/clean_loss_weight": 0.7553945183753967, "train/combined_loss": 1.7173421382904053, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.0033175355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4311184883117676 }, { "epoch": 0.03317535545023697, "step": 14, "train/alpha": 0.2, "train/clean_ce_loss": 0.328125, "train/clean_loss_scale": 3.190887451171875, "train/clean_loss_weight": 0.6381775140762329, "train/combined_loss": 1.2564119100570679, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.0033175355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0470099449157715 }, { "epoch": 0.035545023696682464, "grad_norm": 37.208248138427734, "learning_rate": 2.2047244094488187e-07, "loss": 5.2878, "step": 15 }, { "epoch": 0.035545023696682464, "step": 15, "train/alpha": 0.2, "train/clean_ce_loss": 0.388671875, "train/clean_loss_scale": 3.656477689743042, "train/clean_loss_weight": 0.7312955260276794, "train/combined_loss": 1.7054040431976318, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.0035545023696682463, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4211699962615967 }, { "epoch": 0.035545023696682464, "step": 15, "train/alpha": 0.2, "train/clean_ce_loss": 0.244140625, "train/clean_loss_scale": 3.4583065509796143, "train/clean_loss_weight": 0.6916612982749939, "train/combined_loss": 1.0131757259368896, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.0035545023696682463, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8443131446838379 }, { "epoch": 0.035545023696682464, "step": 15, "train/alpha": 0.2, "train/clean_ce_loss": 0.44921875, "train/clean_loss_scale": 3.1653168201446533, "train/clean_loss_weight": 0.6330633759498596, "train/combined_loss": 1.706303596496582, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.0035545023696682463, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4219197034835815 }, { "epoch": 0.035545023696682464, "step": 15, "train/alpha": 0.2, "train/clean_ce_loss": 0.359375, "train/clean_loss_scale": 2.8364503383636475, "train/clean_loss_weight": 0.5672900676727295, "train/combined_loss": 1.2232191562652588, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.0035545023696682463, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0193493366241455 }, { "epoch": 0.037914691943127965, "step": 16, "train/alpha": 0.2, "train/clean_ce_loss": 0.1533203125, "train/clean_loss_scale": 4.0196661949157715, "train/clean_loss_weight": 0.8039332628250122, "train/combined_loss": 0.7395557761192322, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.0037914691943127963, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6162964701652527 }, { "epoch": 0.037914691943127965, "step": 16, "train/alpha": 0.2, "train/clean_ce_loss": 0.341796875, "train/clean_loss_scale": 3.8650662899017334, "train/clean_loss_weight": 0.7730132937431335, "train/combined_loss": 1.5852811336517334, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.0037914691943127963, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3210675716400146 }, { "epoch": 0.037914691943127965, "step": 16, "train/alpha": 0.2, "train/clean_ce_loss": 0.2373046875, "train/clean_loss_scale": 3.612661600112915, "train/clean_loss_weight": 0.722532331943512, "train/combined_loss": 1.028761863708496, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.0037914691943127963, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8573015332221985 }, { "epoch": 0.037914691943127965, "step": 16, "train/alpha": 0.2, "train/clean_ce_loss": 0.1630859375, "train/clean_loss_scale": 3.2798702716827393, "train/clean_loss_weight": 0.6559740900993347, "train/combined_loss": 0.6418808698654175, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.0037914691943127963, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5349007248878479 }, { "epoch": 0.04028436018957346, "step": 17, "train/alpha": 0.2, "train/clean_ce_loss": 0.28515625, "train/clean_loss_scale": 3.354091167449951, "train/clean_loss_weight": 0.6708182692527771, "train/combined_loss": 1.1477280855178833, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.004028436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9564400315284729 }, { "epoch": 0.04028436018957346, "step": 17, "train/alpha": 0.2, "train/clean_ce_loss": 0.294921875, "train/clean_loss_scale": 3.5184452533721924, "train/clean_loss_weight": 0.7036890387535095, "train/combined_loss": 1.2451996803283691, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.004028436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0376664400100708 }, { "epoch": 0.04028436018957346, "step": 17, "train/alpha": 0.2, "train/clean_ce_loss": 0.224609375, "train/clean_loss_scale": 3.8835184574127197, "train/clean_loss_weight": 0.7767037153244019, "train/combined_loss": 1.046729564666748, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.004028436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.87227463722229 }, { "epoch": 0.04028436018957346, "step": 17, "train/alpha": 0.2, "train/clean_ce_loss": 0.23046875, "train/clean_loss_scale": 3.197021961212158, "train/clean_loss_weight": 0.6394044160842896, "train/combined_loss": 0.8841763734817505, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.004028436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7368136644363403 }, { "epoch": 0.04265402843601896, "step": 18, "train/alpha": 0.2, "train/clean_ce_loss": 0.21875, "train/clean_loss_scale": 3.6059420108795166, "train/clean_loss_weight": 0.7211884260177612, "train/combined_loss": 0.9465597867965698, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.0042654028436018955, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7887998223304749 }, { "epoch": 0.04265402843601896, "step": 18, "train/alpha": 0.2, "train/clean_ce_loss": 0.44140625, "train/clean_loss_scale": 4.137649059295654, "train/clean_loss_weight": 0.8275298476219177, "train/combined_loss": 2.1916611194610596, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.0042654028436018955, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.82638418674469 }, { "epoch": 0.04265402843601896, "step": 18, "train/alpha": 0.2, "train/clean_ce_loss": 0.34375, "train/clean_loss_scale": 4.096887588500977, "train/clean_loss_weight": 0.8193775415420532, "train/combined_loss": 1.689966082572937, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.0042654028436018955, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.408305048942566 }, { "epoch": 0.04265402843601896, "step": 18, "train/alpha": 0.2, "train/clean_ce_loss": 0.2373046875, "train/clean_loss_scale": 4.283703327178955, "train/clean_loss_weight": 0.8567406535148621, "train/combined_loss": 1.2198514938354492, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.0042654028436018955, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.016542911529541 }, { "epoch": 0.045023696682464455, "step": 19, "train/alpha": 0.2, "train/clean_ce_loss": 0.40625, "train/clean_loss_scale": 2.8238601684570312, "train/clean_loss_weight": 0.5647720694541931, "train/combined_loss": 1.3766318559646606, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.004502369668246445, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.147193193435669 }, { "epoch": 0.045023696682464455, "step": 19, "train/alpha": 0.2, "train/clean_ce_loss": 0.365234375, "train/clean_loss_scale": 3.698171615600586, "train/clean_loss_weight": 0.7396343350410461, "train/combined_loss": 1.6208393573760986, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.004502369668246445, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3506994247436523 }, { "epoch": 0.045023696682464455, "step": 19, "train/alpha": 0.2, "train/clean_ce_loss": 0.51953125, "train/clean_loss_scale": 2.6169769763946533, "train/clean_loss_weight": 0.5233954191207886, "train/combined_loss": 1.631521463394165, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.004502369668246445, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3596012592315674 }, { "epoch": 0.045023696682464455, "step": 19, "train/alpha": 0.2, "train/clean_ce_loss": 0.328125, "train/clean_loss_scale": 2.891909599304199, "train/clean_loss_weight": 0.5783819556236267, "train/combined_loss": 1.138689398765564, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.004502369668246445, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9489078521728516 }, { "epoch": 0.04739336492890995, "grad_norm": 36.09403610229492, "learning_rate": 2.9921259842519687e-07, "loss": 5.2845, "step": 20 }, { "epoch": 0.04739336492890995, "step": 20, "train/alpha": 0.2, "train/clean_ce_loss": 0.1796875, "train/clean_loss_scale": 3.561680555343628, "train/clean_loss_weight": 0.7123361229896545, "train/combined_loss": 0.7679873704910278, "train/pos_loss/pos0": 0.3402285873889923, "train/pos_loss/pos1": 0.561901867389679, "train/pos_loss/pos2": 0.9612217545509338, "train/pos_loss/pos3": 1.2779461145401, "train/pos_loss/pos4": 1.6872928142547607, "train/pos_loss/pos5": 2.1343705654144287, "train/pos_loss/pos6": 2.5080697536468506, "train/pos_loss/weight0": 3.044431447982788, "train/pos_loss/weight1": 1.9900749921798706, "train/pos_loss/weight2": 1.1248925924301147, "train/pos_loss/weight3": 0.53574538230896, "train/pos_loss/weight4": 0.21542660892009735, "train/pos_loss/weight5": 0.06788691133260727, "train/pos_loss/weight6": 0.02154218964278698, "train/pos_weight_progress": 0.004739336492890996, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6399894952774048 }, { "epoch": 0.04739336492890995, "step": 20, "train/alpha": 0.2, "train/clean_ce_loss": 0.3046875, "train/clean_loss_scale": 4.07739782333374, "train/clean_loss_weight": 0.815479576587677, "train/combined_loss": 1.4907984733581543, "train/pos_loss/pos0": 0.5997536182403564, "train/pos_loss/pos1": 1.034315586090088, "train/pos_loss/pos2": 1.4800167083740234, "train/pos_loss/pos3": 2.1915953159332275, "train/pos_loss/pos4": 2.6927878856658936, "train/pos_loss/pos5": 2.993361711502075, "train/pos_loss/pos6": 3.443601131439209, "train/pos_loss/weight0": 2.576962471008301, "train/pos_loss/weight1": 1.8318238258361816, "train/pos_loss/weight2": 1.219946265220642, "train/pos_loss/weight3": 0.7159668207168579, "train/pos_loss/weight4": 0.38104918599128723, "train/pos_loss/weight5": 0.18251904845237732, "train/pos_loss/weight6": 0.09173264354467392, "train/pos_weight_progress": 0.004739336492890996, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.242332100868225 }, { "epoch": 0.04739336492890995, "step": 20, "train/alpha": 0.2, "train/clean_ce_loss": 0.53125, "train/clean_loss_scale": 2.270054817199707, "train/clean_loss_weight": 0.4540109634399414, "train/combined_loss": 1.4471598863601685, "train/pos_loss/pos0": 0.6599676012992859, "train/pos_loss/pos1": 1.1563862562179565, "train/pos_loss/pos2": 1.8893051147460938, "train/pos_loss/pos3": 2.6318464279174805, "train/pos_loss/pos4": 3.0524332523345947, "train/pos_loss/pos5": 4.02109956741333, "train/pos_loss/pos6": 4.025102615356445, "train/pos_loss/weight0": 3.2320425510406494, "train/pos_loss/weight1": 2.0004067420959473, "train/pos_loss/weight2": 1.0471161603927612, "train/pos_loss/weight3": 0.49354037642478943, "train/pos_loss/weight4": 0.16411623358726501, "train/pos_loss/weight5": 0.04829803481698036, "train/pos_loss/weight6": 0.014479982666671276, "train/pos_weight_progress": 0.004739336492890996, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.205966591835022 }, { "epoch": 0.04739336492890995, "step": 20, "train/alpha": 0.2, "train/clean_ce_loss": 0.484375, "train/clean_loss_scale": 2.7021188735961914, "train/clean_loss_weight": 0.5404238104820251, "train/combined_loss": 1.5706065893173218, "train/pos_loss/pos0": 1.0320359468460083, "train/pos_loss/pos1": 1.4271601438522339, "train/pos_loss/pos2": 1.391964316368103, "train/pos_loss/pos3": 2.1367979049682617, "train/pos_loss/pos4": 2.838198184967041, "train/pos_loss/pos5": 3.871478796005249, "train/pos_loss/pos6": 3.7595014572143555, "train/pos_loss/weight0": 3.318842887878418, "train/pos_loss/weight1": 2.019369602203369, "train/pos_loss/weight2": 1.0651012659072876, "train/pos_loss/weight3": 0.44572386145591736, "train/pos_loss/weight4": 0.1201852485537529, "train/pos_loss/weight5": 0.02640704996883869, "train/pos_loss/weight6": 0.004370104521512985, "train/pos_weight_progress": 0.004739336492890996, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3088388442993164 }, { "epoch": 0.04976303317535545, "step": 21, "train/alpha": 0.2, "train/clean_ce_loss": 0.267578125, "train/clean_loss_scale": 3.9235281944274902, "train/clean_loss_weight": 0.784705638885498, "train/combined_loss": 1.2598204612731934, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.004976303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.049850344657898 }, { "epoch": 0.04976303317535545, "step": 21, "train/alpha": 0.2, "train/clean_ce_loss": 0.318359375, "train/clean_loss_scale": 2.683980703353882, "train/clean_loss_weight": 0.5367961525917053, "train/combined_loss": 1.0253645181655884, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.004976303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.854470431804657 }, { "epoch": 0.04976303317535545, "step": 21, "train/alpha": 0.2, "train/clean_ce_loss": 0.39453125, "train/clean_loss_scale": 3.1474549770355225, "train/clean_loss_weight": 0.6294910311698914, "train/combined_loss": 1.4901231527328491, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.004976303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2417693138122559 }, { "epoch": 0.04976303317535545, "step": 21, "train/alpha": 0.2, "train/clean_ce_loss": 0.337890625, "train/clean_loss_scale": 4.309828281402588, "train/clean_loss_weight": 0.8619656562805176, "train/combined_loss": 1.7475006580352783, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.004976303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.456250548362732 }, { "epoch": 0.052132701421800945, "step": 22, "train/alpha": 0.2, "train/clean_ce_loss": 0.197265625, "train/clean_loss_scale": 3.9808855056762695, "train/clean_loss_weight": 0.796177089214325, "train/combined_loss": 0.9423502087593079, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.005213270142180095, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.785291850566864 }, { "epoch": 0.052132701421800945, "step": 22, "train/alpha": 0.2, "train/clean_ce_loss": 0.1962890625, "train/clean_loss_scale": 3.0840377807617188, "train/clean_loss_weight": 0.6168075799942017, "train/combined_loss": 0.7264354825019836, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.005213270142180095, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6053628921508789 }, { "epoch": 0.052132701421800945, "step": 22, "train/alpha": 0.2, "train/clean_ce_loss": 0.251953125, "train/clean_loss_scale": 2.5587470531463623, "train/clean_loss_weight": 0.5117494463920593, "train/combined_loss": 0.7736212015151978, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.005213270142180095, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6446843147277832 }, { "epoch": 0.052132701421800945, "step": 22, "train/alpha": 0.2, "train/clean_ce_loss": 0.2373046875, "train/clean_loss_scale": 3.203573226928711, "train/clean_loss_weight": 0.6407146453857422, "train/combined_loss": 0.9122675657272339, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.005213270142180095, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7602229714393616 }, { "epoch": 0.054502369668246446, "step": 23, "train/alpha": 0.2, "train/clean_ce_loss": 0.5546875, "train/clean_loss_scale": 3.134568929672241, "train/clean_loss_weight": 0.6269137859344482, "train/combined_loss": 2.0864474773406982, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.0054502369668246444, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.7387062311172485 }, { "epoch": 0.054502369668246446, "step": 23, "train/alpha": 0.2, "train/clean_ce_loss": 0.37890625, "train/clean_loss_scale": 2.7918455600738525, "train/clean_loss_weight": 0.5583691000938416, "train/combined_loss": 1.2694172859191895, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.0054502369668246444, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0578477382659912 }, { "epoch": 0.054502369668246446, "step": 23, "train/alpha": 0.2, "train/clean_ce_loss": 0.2158203125, "train/clean_loss_scale": 4.040778160095215, "train/clean_loss_weight": 0.8081556558609009, "train/combined_loss": 1.046498417854309, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.0054502369668246444, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8720820546150208 }, { "epoch": 0.054502369668246446, "step": 23, "train/alpha": 0.2, "train/clean_ce_loss": 0.251953125, "train/clean_loss_scale": 3.0199997425079346, "train/clean_loss_weight": 0.6039999723434448, "train/combined_loss": 0.9130780100822449, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.0054502369668246444, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7608983516693115 }, { "epoch": 0.05687203791469194, "step": 24, "train/alpha": 0.2, "train/clean_ce_loss": 0.306640625, "train/clean_loss_scale": 2.940384864807129, "train/clean_loss_weight": 0.5880770087242126, "train/combined_loss": 1.0819697380065918, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.005687203791469194, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9016414880752563 }, { "epoch": 0.05687203791469194, "step": 24, "train/alpha": 0.2, "train/clean_ce_loss": 0.248046875, "train/clean_loss_scale": 3.36142897605896, "train/clean_loss_weight": 0.672285795211792, "train/combined_loss": 1.000550389289856, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.005687203791469194, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.833791971206665 }, { "epoch": 0.05687203791469194, "step": 24, "train/alpha": 0.2, "train/clean_ce_loss": 0.279296875, "train/clean_loss_scale": 3.222710132598877, "train/clean_loss_weight": 0.6445420384407043, "train/combined_loss": 1.0801115036010742, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.005687203791469194, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9000928997993469 }, { "epoch": 0.05687203791469194, "step": 24, "train/alpha": 0.2, "train/clean_ce_loss": 0.20703125, "train/clean_loss_scale": 3.887105941772461, "train/clean_loss_weight": 0.7774211764335632, "train/combined_loss": 0.9657028913497925, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.005687203791469194, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8047524094581604 }, { "epoch": 0.05924170616113744, "grad_norm": 35.133663177490234, "learning_rate": 3.779527559055118e-07, "loss": 5.1166, "step": 25 }, { "epoch": 0.05924170616113744, "step": 25, "train/alpha": 0.2, "train/clean_ce_loss": 0.29296875, "train/clean_loss_scale": 3.801551103591919, "train/clean_loss_weight": 0.7603102326393127, "train/combined_loss": 1.3364827632904053, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.005924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1137356758117676 }, { "epoch": 0.05924170616113744, "step": 25, "train/alpha": 0.2, "train/clean_ce_loss": 0.162109375, "train/clean_loss_scale": 2.708181381225586, "train/clean_loss_weight": 0.5416362881660461, "train/combined_loss": 0.5268259048461914, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.005924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43902158737182617 }, { "epoch": 0.05924170616113744, "step": 25, "train/alpha": 0.2, "train/clean_ce_loss": 0.384765625, "train/clean_loss_scale": 2.9237005710601807, "train/clean_loss_weight": 0.5847401022911072, "train/combined_loss": 1.349927306175232, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.005924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1249394416809082 }, { "epoch": 0.05924170616113744, "step": 25, "train/alpha": 0.2, "train/clean_ce_loss": 0.458984375, "train/clean_loss_scale": 2.609081745147705, "train/clean_loss_weight": 0.5218163728713989, "train/combined_loss": 1.4370332956314087, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.005924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1975277662277222 }, { "epoch": 0.061611374407582936, "step": 26, "train/alpha": 0.2, "train/clean_ce_loss": 0.34765625, "train/clean_loss_scale": 2.843949794769287, "train/clean_loss_weight": 0.5687899589538574, "train/combined_loss": 1.186460256576538, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.006161137440758294, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9887169003486633 }, { "epoch": 0.061611374407582936, "step": 26, "train/alpha": 0.2, "train/clean_ce_loss": 0.3203125, "train/clean_loss_scale": 2.8346002101898193, "train/clean_loss_weight": 0.5669200420379639, "train/combined_loss": 1.0895495414733887, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.006161137440758294, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.907957911491394 }, { "epoch": 0.061611374407582936, "step": 26, "train/alpha": 0.2, "train/clean_ce_loss": 0.373046875, "train/clean_loss_scale": 3.9793500900268555, "train/clean_loss_weight": 0.7958700060844421, "train/combined_loss": 1.7813808917999268, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.006161137440758294, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.484484076499939 }, { "epoch": 0.061611374407582936, "step": 26, "train/alpha": 0.2, "train/clean_ce_loss": 0.23828125, "train/clean_loss_scale": 4.107112407684326, "train/clean_loss_weight": 0.8214225172996521, "train/combined_loss": 1.17437744140625, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.006161137440758294, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9786478877067566 }, { "epoch": 0.06398104265402843, "step": 27, "train/alpha": 0.2, "train/clean_ce_loss": 0.34765625, "train/clean_loss_scale": 2.9895741939544678, "train/clean_loss_weight": 0.5979148745536804, "train/combined_loss": 1.2472130060195923, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.006398104265402844, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0393441915512085 }, { "epoch": 0.06398104265402843, "step": 27, "train/alpha": 0.2, "train/clean_ce_loss": 0.251953125, "train/clean_loss_scale": 3.4980621337890625, "train/clean_loss_weight": 0.6996124386787415, "train/combined_loss": 1.0576171875, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.006398104265402844, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.88134765625 }, { "epoch": 0.06398104265402843, "step": 27, "train/alpha": 0.2, "train/clean_ce_loss": 0.275390625, "train/clean_loss_scale": 4.213562965393066, "train/clean_loss_weight": 0.8427125811576843, "train/combined_loss": 1.3924508094787598, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.006398104265402844, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.160375714302063 }, { "epoch": 0.06398104265402843, "step": 27, "train/alpha": 0.2, "train/clean_ce_loss": 0.462890625, "train/clean_loss_scale": 3.382047414779663, "train/clean_loss_weight": 0.6764094829559326, "train/combined_loss": 1.8786215782165527, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.006398104265402844, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5655180215835571 }, { "epoch": 0.06635071090047394, "step": 28, "train/alpha": 0.2, "train/clean_ce_loss": 0.220703125, "train/clean_loss_scale": 3.0243473052978516, "train/clean_loss_weight": 0.6048694849014282, "train/combined_loss": 0.800979495048523, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.006635071090047393, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6674829125404358 }, { "epoch": 0.06635071090047394, "step": 28, "train/alpha": 0.2, "train/clean_ce_loss": 0.322265625, "train/clean_loss_scale": 2.9726662635803223, "train/clean_loss_weight": 0.5945332646369934, "train/combined_loss": 1.1495857238769531, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.006635071090047393, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9579881429672241 }, { "epoch": 0.06635071090047394, "step": 28, "train/alpha": 0.2, "train/clean_ce_loss": 0.1796875, "train/clean_loss_scale": 3.948019504547119, "train/clean_loss_weight": 0.7896038889884949, "train/combined_loss": 0.8512917160987854, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.006635071090047393, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.709409773349762 }, { "epoch": 0.06635071090047394, "step": 28, "train/alpha": 0.2, "train/clean_ce_loss": 0.2236328125, "train/clean_loss_scale": 3.4581282138824463, "train/clean_loss_weight": 0.6916256546974182, "train/combined_loss": 0.9280211329460144, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.006635071090047393, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7733509540557861 }, { "epoch": 0.06872037914691943, "step": 29, "train/alpha": 0.2, "train/clean_ce_loss": 0.28515625, "train/clean_loss_scale": 3.263744354248047, "train/clean_loss_weight": 0.6527488827705383, "train/combined_loss": 1.1168125867843628, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.006872037914691943, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9306771159172058 }, { "epoch": 0.06872037914691943, "step": 29, "train/alpha": 0.2, "train/clean_ce_loss": 0.384765625, "train/clean_loss_scale": 3.269423484802246, "train/clean_loss_weight": 0.6538847088813782, "train/combined_loss": 1.509554147720337, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.006872037914691943, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2579617500305176 }, { "epoch": 0.06872037914691943, "step": 29, "train/alpha": 0.2, "train/clean_ce_loss": 0.30078125, "train/clean_loss_scale": 4.002274990081787, "train/clean_loss_weight": 0.8004550337791443, "train/combined_loss": 1.4445711374282837, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.006872037914691943, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2038092613220215 }, { "epoch": 0.06872037914691943, "step": 29, "train/alpha": 0.2, "train/clean_ce_loss": 0.30078125, "train/clean_loss_scale": 2.850430965423584, "train/clean_loss_weight": 0.5700861811637878, "train/combined_loss": 1.028827428817749, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.006872037914691943, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8573561906814575 }, { "epoch": 0.07109004739336493, "grad_norm": 32.8569450378418, "learning_rate": 4.5669291338582676e-07, "loss": 5.2203, "step": 30 }, { "epoch": 0.07109004739336493, "step": 30, "train/alpha": 0.2, "train/clean_ce_loss": 0.349609375, "train/clean_loss_scale": 2.799133062362671, "train/clean_loss_weight": 0.5598266124725342, "train/combined_loss": 1.1743237972259521, "train/pos_loss/pos0": 0.48509299755096436, "train/pos_loss/pos1": 0.8966527581214905, "train/pos_loss/pos2": 1.6142513751983643, "train/pos_loss/pos3": 2.3710641860961914, "train/pos_loss/pos4": 2.9271175861358643, "train/pos_loss/pos5": 3.3268072605133057, "train/pos_loss/pos6": 3.621769666671753, "train/pos_loss/weight0": 3.265589714050293, "train/pos_loss/weight1": 2.0128378868103027, "train/pos_loss/weight2": 1.0128909349441528, "train/pos_loss/weight3": 0.47770264744758606, "train/pos_loss/weight4": 0.17416313290596008, "train/pos_loss/weight5": 0.04502259939908981, "train/pos_loss/weight6": 0.011793537065386772, "train/pos_weight_progress": 0.0071090047393364926, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.978603184223175 }, { "epoch": 0.07109004739336493, "step": 30, "train/alpha": 0.2, "train/clean_ce_loss": 0.30859375, "train/clean_loss_scale": 3.321442127227783, "train/clean_loss_weight": 0.6642884612083435, "train/combined_loss": 1.2299715280532837, "train/pos_loss/pos0": 0.6666896343231201, "train/pos_loss/pos1": 1.023169994354248, "train/pos_loss/pos2": 1.43348228931427, "train/pos_loss/pos3": 2.0252888202667236, "train/pos_loss/pos4": 2.36393404006958, "train/pos_loss/pos5": 2.9029901027679443, "train/pos_loss/pos6": 3.3646912574768066, "train/pos_loss/weight0": 3.243509531021118, "train/pos_loss/weight1": 2.040316104888916, "train/pos_loss/weight2": 1.0741088390350342, "train/pos_loss/weight3": 0.4601147770881653, "train/pos_loss/weight4": 0.141342893242836, "train/pos_loss/weight5": 0.03287818282842636, "train/pos_loss/weight6": 0.007729212753474712, "train/pos_weight_progress": 0.0071090047393364926, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0249762535095215 }, { "epoch": 0.07109004739336493, "step": 30, "train/alpha": 0.2, "train/clean_ce_loss": 0.2421875, "train/clean_loss_scale": 3.5263564586639404, "train/clean_loss_weight": 0.705271303653717, "train/combined_loss": 1.0248472690582275, "train/pos_loss/pos0": 0.4905744194984436, "train/pos_loss/pos1": 0.8310745358467102, "train/pos_loss/pos2": 1.2766056060791016, "train/pos_loss/pos3": 1.8061896562576294, "train/pos_loss/pos4": 2.2480721473693848, "train/pos_loss/pos5": 2.7955758571624756, "train/pos_loss/pos6": 3.174128532409668, "train/pos_loss/weight0": 3.2611563205718994, "train/pos_loss/weight1": 2.00546932220459, "train/pos_loss/weight2": 1.0475083589553833, "train/pos_loss/weight3": 0.45846855640411377, "train/pos_loss/weight4": 0.16611221432685852, "train/pos_loss/weight5": 0.04739413410425186, "train/pos_loss/weight6": 0.013891191221773624, "train/pos_weight_progress": 0.0071090047393364926, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8540394306182861 }, { "epoch": 0.07109004739336493, "step": 30, "train/alpha": 0.2, "train/clean_ce_loss": 0.51171875, "train/clean_loss_scale": 2.755385160446167, "train/clean_loss_weight": 0.5510770678520203, "train/combined_loss": 1.6919786930084229, "train/pos_loss/pos0": 0.7155395746231079, "train/pos_loss/pos1": 1.301934838294983, "train/pos_loss/pos2": 2.0426173210144043, "train/pos_loss/pos3": 3.229888439178467, "train/pos_loss/pos4": 3.7969863414764404, "train/pos_loss/pos5": 4.5116424560546875, "train/pos_loss/pos6": 4.60114860534668, "train/pos_loss/weight0": 3.1131696701049805, "train/pos_loss/weight1": 1.9624335765838623, "train/pos_loss/weight2": 1.118654489517212, "train/pos_loss/weight3": 0.5252212882041931, "train/pos_loss/weight4": 0.1977626383304596, "train/pos_loss/weight5": 0.0627361610531807, "train/pos_loss/weight6": 0.020022675395011902, "train/pos_weight_progress": 0.0071090047393364926, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4099822044372559 }, { "epoch": 0.07345971563981042, "step": 31, "train/alpha": 0.2, "train/clean_ce_loss": 0.34765625, "train/clean_loss_scale": 2.938076972961426, "train/clean_loss_weight": 0.587615430355072, "train/combined_loss": 1.225728988647461, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.007345971563981043, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.021440863609314 }, { "epoch": 0.07345971563981042, "step": 31, "train/alpha": 0.2, "train/clean_ce_loss": 0.1767578125, "train/clean_loss_scale": 2.797614336013794, "train/clean_loss_weight": 0.5595228672027588, "train/combined_loss": 0.593400239944458, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.007345971563981043, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.49450019001960754 }, { "epoch": 0.07345971563981042, "step": 31, "train/alpha": 0.2, "train/clean_ce_loss": 0.244140625, "train/clean_loss_scale": 2.9808602333068848, "train/clean_loss_weight": 0.596172034740448, "train/combined_loss": 0.8732989430427551, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.007345971563981043, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7277491092681885 }, { "epoch": 0.07345971563981042, "step": 31, "train/alpha": 0.2, "train/clean_ce_loss": 0.404296875, "train/clean_loss_scale": 2.8149213790893555, "train/clean_loss_weight": 0.5629842877388, "train/combined_loss": 1.3656766414642334, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.007345971563981043, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.138063907623291 }, { "epoch": 0.07582938388625593, "step": 32, "train/alpha": 0.2, "train/clean_ce_loss": 0.34765625, "train/clean_loss_scale": 4.540036678314209, "train/clean_loss_weight": 0.9080073237419128, "train/combined_loss": 1.8940465450286865, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.007582938388625593, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5783721208572388 }, { "epoch": 0.07582938388625593, "step": 32, "train/alpha": 0.2, "train/clean_ce_loss": 0.17578125, "train/clean_loss_scale": 3.6931447982788086, "train/clean_loss_weight": 0.7386289834976196, "train/combined_loss": 0.7790226936340332, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.007582938388625593, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6491855978965759 }, { "epoch": 0.07582938388625593, "step": 32, "train/alpha": 0.2, "train/clean_ce_loss": 0.2109375, "train/clean_loss_scale": 3.321687698364258, "train/clean_loss_weight": 0.6643375754356384, "train/combined_loss": 0.8408021926879883, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.007582938388625593, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7006685137748718 }, { "epoch": 0.07582938388625593, "step": 32, "train/alpha": 0.2, "train/clean_ce_loss": 0.2177734375, "train/clean_loss_scale": 3.4117512702941895, "train/clean_loss_weight": 0.6823502779006958, "train/combined_loss": 0.8915866017341614, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.007582938388625593, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7429888248443604 }, { "epoch": 0.07819905213270142, "step": 33, "train/alpha": 0.2, "train/clean_ce_loss": 0.1708984375, "train/clean_loss_scale": 3.287465810775757, "train/clean_loss_weight": 0.6574931740760803, "train/combined_loss": 0.6741873025894165, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.007819905213270141, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.561822772026062 }, { "epoch": 0.07819905213270142, "step": 33, "train/alpha": 0.2, "train/clean_ce_loss": 0.33984375, "train/clean_loss_scale": 3.201535701751709, "train/clean_loss_weight": 0.6403071284294128, "train/combined_loss": 1.3056262731552124, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.007819905213270141, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0880218744277954 }, { "epoch": 0.07819905213270142, "step": 33, "train/alpha": 0.2, "train/clean_ce_loss": 0.423828125, "train/clean_loss_scale": 3.30041241645813, "train/clean_loss_weight": 0.6600825190544128, "train/combined_loss": 1.6785691976547241, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.007819905213270141, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3988076448440552 }, { "epoch": 0.07819905213270142, "step": 33, "train/alpha": 0.2, "train/clean_ce_loss": 0.31640625, "train/clean_loss_scale": 3.321892738342285, "train/clean_loss_weight": 0.6643785834312439, "train/combined_loss": 1.261281132698059, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.007819905213270141, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.051067590713501 }, { "epoch": 0.08056872037914692, "step": 34, "train/alpha": 0.2, "train/clean_ce_loss": 0.33203125, "train/clean_loss_scale": 3.2291579246520996, "train/clean_loss_weight": 0.6458315849304199, "train/combined_loss": 1.286617636680603, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.008056872037914692, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0721813440322876 }, { "epoch": 0.08056872037914692, "step": 34, "train/alpha": 0.2, "train/clean_ce_loss": 0.333984375, "train/clean_loss_scale": 2.5050301551818848, "train/clean_loss_weight": 0.5010060667991638, "train/combined_loss": 1.0039690732955933, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.008056872037914692, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8366408944129944 }, { "epoch": 0.08056872037914692, "step": 34, "train/alpha": 0.2, "train/clean_ce_loss": 0.23046875, "train/clean_loss_scale": 2.010258674621582, "train/clean_loss_weight": 0.40205174684524536, "train/combined_loss": 0.5559621453285217, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.008056872037914692, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46330177783966064 }, { "epoch": 0.08056872037914692, "step": 34, "train/alpha": 0.2, "train/clean_ce_loss": 0.47265625, "train/clean_loss_scale": 2.8936350345611572, "train/clean_loss_weight": 0.5787270069122314, "train/combined_loss": 1.6412336826324463, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.008056872037914692, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3676947355270386 }, { "epoch": 0.08293838862559241, "grad_norm": 26.05545425415039, "learning_rate": 5.354330708661418e-07, "loss": 5.0718, "step": 35 }, { "epoch": 0.08293838862559241, "step": 35, "train/alpha": 0.2, "train/clean_ce_loss": 0.28515625, "train/clean_loss_scale": 3.4550750255584717, "train/clean_loss_weight": 0.6910150051116943, "train/combined_loss": 1.1822835206985474, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.008293838862559242, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9852362275123596 }, { "epoch": 0.08293838862559241, "step": 35, "train/alpha": 0.2, "train/clean_ce_loss": 0.34375, "train/clean_loss_scale": 3.99985408782959, "train/clean_loss_weight": 0.799970805644989, "train/combined_loss": 1.649939775466919, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.008293838862559242, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3749498128890991 }, { "epoch": 0.08293838862559241, "step": 35, "train/alpha": 0.2, "train/clean_ce_loss": 0.298828125, "train/clean_loss_scale": 4.524815082550049, "train/clean_loss_weight": 0.9049630165100098, "train/combined_loss": 1.6225703954696655, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.008293838862559242, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3521419763565063 }, { "epoch": 0.08293838862559241, "step": 35, "train/alpha": 0.2, "train/clean_ce_loss": 0.328125, "train/clean_loss_scale": 4.049854278564453, "train/clean_loss_weight": 0.8099708557128906, "train/combined_loss": 1.5946300029754639, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.008293838862559242, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3288583755493164 }, { "epoch": 0.08530805687203792, "step": 36, "train/alpha": 0.2, "train/clean_ce_loss": 0.51171875, "train/clean_loss_scale": 2.4080307483673096, "train/clean_loss_weight": 0.4816061556339264, "train/combined_loss": 1.4786813259124756, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.008530805687203791, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2322344779968262 }, { "epoch": 0.08530805687203792, "step": 36, "train/alpha": 0.2, "train/clean_ce_loss": 0.2578125, "train/clean_loss_scale": 2.672245502471924, "train/clean_loss_weight": 0.5344491004943848, "train/combined_loss": 0.826725959777832, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.008530805687203791, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.688938319683075 }, { "epoch": 0.08530805687203792, "step": 36, "train/alpha": 0.2, "train/clean_ce_loss": 0.2333984375, "train/clean_loss_scale": 3.44071888923645, "train/clean_loss_weight": 0.688143789768219, "train/combined_loss": 0.9636700749397278, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.008530805687203791, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.803058385848999 }, { "epoch": 0.08530805687203792, "step": 36, "train/alpha": 0.2, "train/clean_ce_loss": 0.2392578125, "train/clean_loss_scale": 4.265698432922363, "train/clean_loss_weight": 0.8531396985054016, "train/combined_loss": 1.224721908569336, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.008530805687203791, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0206016302108765 }, { "epoch": 0.08767772511848342, "step": 37, "train/alpha": 0.2, "train/clean_ce_loss": 0.1240234375, "train/clean_loss_scale": 3.37379789352417, "train/clean_loss_weight": 0.674759566783905, "train/combined_loss": 0.5021160244941711, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.008767772511848342, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.41843000054359436 }, { "epoch": 0.08767772511848342, "step": 37, "train/alpha": 0.2, "train/clean_ce_loss": 0.3203125, "train/clean_loss_scale": 4.885049819946289, "train/clean_loss_weight": 0.9770099520683289, "train/combined_loss": 1.8776909112930298, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.008767772511848342, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5647424459457397 }, { "epoch": 0.08767772511848342, "step": 37, "train/alpha": 0.2, "train/clean_ce_loss": 0.2333984375, "train/clean_loss_scale": 3.577622652053833, "train/clean_loss_weight": 0.7155245542526245, "train/combined_loss": 1.002013921737671, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.008767772511848342, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8350115418434143 }, { "epoch": 0.08767772511848342, "step": 37, "train/alpha": 0.2, "train/clean_ce_loss": 0.447265625, "train/clean_loss_scale": 3.022101640701294, "train/clean_loss_weight": 0.6044203639030457, "train/combined_loss": 1.622018575668335, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.008767772511848342, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.351682186126709 }, { "epoch": 0.09004739336492891, "step": 38, "train/alpha": 0.2, "train/clean_ce_loss": 0.373046875, "train/clean_loss_scale": 2.671877145767212, "train/clean_loss_weight": 0.5343754291534424, "train/combined_loss": 1.1960824728012085, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.00900473933649289, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9967353940010071 }, { "epoch": 0.09004739336492891, "step": 38, "train/alpha": 0.2, "train/clean_ce_loss": 0.1884765625, "train/clean_loss_scale": 3.381464958190918, "train/clean_loss_weight": 0.6762930154800415, "train/combined_loss": 0.764792263507843, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.00900473933649289, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6373268961906433 }, { "epoch": 0.09004739336492891, "step": 38, "train/alpha": 0.2, "train/clean_ce_loss": 0.2412109375, "train/clean_loss_scale": 3.6355419158935547, "train/clean_loss_weight": 0.7271084189414978, "train/combined_loss": 1.052319049835205, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.00900473933649289, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8769325017929077 }, { "epoch": 0.09004739336492891, "step": 38, "train/alpha": 0.2, "train/clean_ce_loss": 0.263671875, "train/clean_loss_scale": 2.7479794025421143, "train/clean_loss_weight": 0.5495958924293518, "train/combined_loss": 0.8694778084754944, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.00900473933649289, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7245648503303528 }, { "epoch": 0.0924170616113744, "step": 39, "train/alpha": 0.2, "train/clean_ce_loss": 0.2412109375, "train/clean_loss_scale": 3.443570613861084, "train/clean_loss_weight": 0.6887141466140747, "train/combined_loss": 0.9967522621154785, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.00924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.830626904964447 }, { "epoch": 0.0924170616113744, "step": 39, "train/alpha": 0.2, "train/clean_ce_loss": 0.3125, "train/clean_loss_scale": 2.8702454566955566, "train/clean_loss_weight": 0.5740491151809692, "train/combined_loss": 1.0763421058654785, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.00924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8969517350196838 }, { "epoch": 0.0924170616113744, "step": 39, "train/alpha": 0.2, "train/clean_ce_loss": 0.1748046875, "train/clean_loss_scale": 3.851308822631836, "train/clean_loss_weight": 0.7702617645263672, "train/combined_loss": 0.8078721761703491, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.00924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6732268333435059 }, { "epoch": 0.0924170616113744, "step": 39, "train/alpha": 0.2, "train/clean_ce_loss": 0.486328125, "train/clean_loss_scale": 2.6913952827453613, "train/clean_loss_weight": 0.5382790565490723, "train/combined_loss": 1.5706814527511597, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.00924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3089011907577515 }, { "epoch": 0.0947867298578199, "grad_norm": 24.87386703491211, "learning_rate": 6.141732283464567e-07, "loss": 5.2969, "step": 40 }, { "epoch": 0.0947867298578199, "step": 40, "train/alpha": 0.2, "train/clean_ce_loss": 0.41015625, "train/clean_loss_scale": 2.7036149501800537, "train/clean_loss_weight": 0.5407230257987976, "train/combined_loss": 1.3306854963302612, "train/pos_loss/pos0": 0.7127829194068909, "train/pos_loss/pos1": 1.194064974784851, "train/pos_loss/pos2": 1.6652531623840332, "train/pos_loss/pos3": 2.6449332237243652, "train/pos_loss/pos4": 3.3216209411621094, "train/pos_loss/pos5": 3.957294464111328, "train/pos_loss/pos6": 3.959927797317505, "train/pos_loss/weight0": 3.4918980598449707, "train/pos_loss/weight1": 2.1509151458740234, "train/pos_loss/weight2": 0.9529327750205994, "train/pos_loss/weight3": 0.31830018758773804, "train/pos_loss/weight4": 0.0740472599864006, "train/pos_loss/weight5": 0.01075594313442707, "train/pos_loss/weight6": 0.0011507711606100202, "train/pos_weight_progress": 0.009478672985781991, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1089046001434326 }, { "epoch": 0.0947867298578199, "step": 40, "train/alpha": 0.2, "train/clean_ce_loss": 0.376953125, "train/clean_loss_scale": 3.464970350265503, "train/clean_loss_weight": 0.6929940581321716, "train/combined_loss": 1.5673576593399048, "train/pos_loss/pos0": 0.7672763466835022, "train/pos_loss/pos1": 1.3189376592636108, "train/pos_loss/pos2": 1.9619187116622925, "train/pos_loss/pos3": 2.504124879837036, "train/pos_loss/pos4": 3.1787495613098145, "train/pos_loss/pos5": 3.525470495223999, "train/pos_loss/pos6": 3.5467493534088135, "train/pos_loss/weight0": 3.187248706817627, "train/pos_loss/weight1": 2.0153398513793945, "train/pos_loss/weight2": 1.084816575050354, "train/pos_loss/weight3": 0.494334876537323, "train/pos_loss/weight4": 0.16617847979068756, "train/pos_loss/weight5": 0.04167242720723152, "train/pos_loss/weight6": 0.010408652946352959, "train/pos_weight_progress": 0.009478672985781991, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.306131362915039 }, { "epoch": 0.0947867298578199, "step": 40, "train/alpha": 0.2, "train/clean_ce_loss": 0.197265625, "train/clean_loss_scale": 4.279987812042236, "train/clean_loss_weight": 0.8559975624084473, "train/combined_loss": 1.0131534337997437, "train/pos_loss/pos0": 0.4444923400878906, "train/pos_loss/pos1": 0.8566749095916748, "train/pos_loss/pos2": 1.323577880859375, "train/pos_loss/pos3": 1.8239285945892334, "train/pos_loss/pos4": 2.266613006591797, "train/pos_loss/pos5": 2.6534554958343506, "train/pos_loss/pos6": 2.9986491203308105, "train/pos_loss/weight0": 3.208470582962036, "train/pos_loss/weight1": 2.0690619945526123, "train/pos_loss/weight2": 1.0752365589141846, "train/pos_loss/weight3": 0.44996851682662964, "train/pos_loss/weight4": 0.1462966799736023, "train/pos_loss/weight5": 0.04086339473724365, "train/pos_loss/weight6": 0.010101948864758015, "train/pos_weight_progress": 0.009478672985781991, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8442944884300232 }, { "epoch": 0.0947867298578199, "step": 40, "train/alpha": 0.2, "train/clean_ce_loss": 0.63671875, "train/clean_loss_scale": 2.6967215538024902, "train/clean_loss_weight": 0.539344310760498, "train/combined_loss": 2.0604639053344727, "train/pos_loss/pos0": 1.1133203506469727, "train/pos_loss/pos1": 1.7799454927444458, "train/pos_loss/pos2": 2.367408275604248, "train/pos_loss/pos3": 2.8148341178894043, "train/pos_loss/pos4": 3.114771842956543, "train/pos_loss/pos5": 3.6522364616394043, "train/pos_loss/pos6": 3.6168224811553955, "train/pos_loss/weight0": 3.0902082920074463, "train/pos_loss/weight1": 1.9802818298339844, "train/pos_loss/weight2": 1.1248011589050293, "train/pos_loss/weight3": 0.5233244895935059, "train/pos_loss/weight4": 0.19993235170841217, "train/pos_loss/weight5": 0.06171341985464096, "train/pos_loss/weight6": 0.019738376140594482, "train/pos_weight_progress": 0.009478672985781991, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.7170531749725342 }, { "epoch": 0.0971563981042654, "step": 41, "train/alpha": 0.2, "train/clean_ce_loss": 0.1943359375, "train/clean_loss_scale": 4.14866828918457, "train/clean_loss_weight": 0.829733669757843, "train/combined_loss": 0.9674824476242065, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.00971563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8062353730201721 }, { "epoch": 0.0971563981042654, "step": 41, "train/alpha": 0.2, "train/clean_ce_loss": 0.2578125, "train/clean_loss_scale": 2.75913143157959, "train/clean_loss_weight": 0.5518262982368469, "train/combined_loss": 0.8536062836647034, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.00971563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7113385796546936 }, { "epoch": 0.0971563981042654, "step": 41, "train/alpha": 0.2, "train/clean_ce_loss": 0.2255859375, "train/clean_loss_scale": 3.8520114421844482, "train/clean_loss_weight": 0.7704023122787476, "train/combined_loss": 1.0427515506744385, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.00971563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8689596056938171 }, { "epoch": 0.0971563981042654, "step": 41, "train/alpha": 0.2, "train/clean_ce_loss": 0.2158203125, "train/clean_loss_scale": 3.9737706184387207, "train/clean_loss_weight": 0.794754147529602, "train/combined_loss": 1.029144525527954, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.00971563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8576204180717468 }, { "epoch": 0.0995260663507109, "step": 42, "train/alpha": 0.2, "train/clean_ce_loss": 0.376953125, "train/clean_loss_scale": 3.3435280323028564, "train/clean_loss_weight": 0.6687056422233582, "train/combined_loss": 1.5124239921569824, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.00995260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2603533267974854 }, { "epoch": 0.0995260663507109, "step": 42, "train/alpha": 0.2, "train/clean_ce_loss": 0.408203125, "train/clean_loss_scale": 3.9127066135406494, "train/clean_loss_weight": 0.7825413346290588, "train/combined_loss": 1.9166148900985718, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.00995260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5971790552139282 }, { "epoch": 0.0995260663507109, "step": 42, "train/alpha": 0.2, "train/clean_ce_loss": 0.1826171875, "train/clean_loss_scale": 3.3941421508789062, "train/clean_loss_weight": 0.6788284182548523, "train/combined_loss": 0.7437944412231445, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.00995260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6198287010192871 }, { "epoch": 0.0995260663507109, "step": 42, "train/alpha": 0.2, "train/clean_ce_loss": 0.4453125, "train/clean_loss_scale": 3.18780255317688, "train/clean_loss_weight": 0.6375605463981628, "train/combined_loss": 1.703481912612915, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.00995260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4195683002471924 }, { "epoch": 0.1018957345971564, "step": 43, "train/alpha": 0.2, "train/clean_ce_loss": 0.353515625, "train/clean_loss_scale": 3.267165422439575, "train/clean_loss_weight": 0.653433084487915, "train/combined_loss": 1.3859927654266357, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.01018957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.154994010925293 }, { "epoch": 0.1018957345971564, "step": 43, "train/alpha": 0.2, "train/clean_ce_loss": 0.1708984375, "train/clean_loss_scale": 4.236515998840332, "train/clean_loss_weight": 0.8473032116889954, "train/combined_loss": 0.8688167333602905, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.01018957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7240139245986938 }, { "epoch": 0.1018957345971564, "step": 43, "train/alpha": 0.2, "train/clean_ce_loss": 0.4375, "train/clean_loss_scale": 3.6311182975769043, "train/clean_loss_weight": 0.7262236475944519, "train/combined_loss": 1.906337022781372, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.01018957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5886142253875732 }, { "epoch": 0.1018957345971564, "step": 43, "train/alpha": 0.2, "train/clean_ce_loss": 0.1982421875, "train/clean_loss_scale": 3.5441887378692627, "train/clean_loss_weight": 0.7088377475738525, "train/combined_loss": 0.8431292772293091, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.01018957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7026077508926392 }, { "epoch": 0.10426540284360189, "step": 44, "train/alpha": 0.2, "train/clean_ce_loss": 0.18359375, "train/clean_loss_scale": 4.041802883148193, "train/clean_loss_weight": 0.8083605766296387, "train/combined_loss": 0.8904597163200378, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.01042654028436019, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7420497536659241 }, { "epoch": 0.10426540284360189, "step": 44, "train/alpha": 0.2, "train/clean_ce_loss": 0.1328125, "train/clean_loss_scale": 3.536482572555542, "train/clean_loss_weight": 0.7072965502738953, "train/combined_loss": 0.5636269450187683, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.01042654028436019, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46968910098075867 }, { "epoch": 0.10426540284360189, "step": 44, "train/alpha": 0.2, "train/clean_ce_loss": 0.2275390625, "train/clean_loss_scale": 3.4579498767852783, "train/clean_loss_weight": 0.6915900111198425, "train/combined_loss": 0.9441823959350586, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.01042654028436019, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7868186831474304 }, { "epoch": 0.10426540284360189, "step": 44, "train/alpha": 0.2, "train/clean_ce_loss": 0.28515625, "train/clean_loss_scale": 4.448174476623535, "train/clean_loss_weight": 0.889634907245636, "train/combined_loss": 1.5221097469329834, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.01042654028436019, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2684247493743896 }, { "epoch": 0.1066350710900474, "grad_norm": 18.6380558013916, "learning_rate": 6.929133858267716e-07, "loss": 4.8367, "step": 45 }, { "epoch": 0.1066350710900474, "step": 45, "train/alpha": 0.2, "train/clean_ce_loss": 0.21875, "train/clean_loss_scale": 3.882791757583618, "train/clean_loss_weight": 0.7765583395957947, "train/combined_loss": 1.0192328691482544, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.01066350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8493607044219971 }, { "epoch": 0.1066350710900474, "step": 45, "train/alpha": 0.2, "train/clean_ce_loss": 0.291015625, "train/clean_loss_scale": 4.02040433883667, "train/clean_loss_weight": 0.8040809035301208, "train/combined_loss": 1.4040005207061768, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.01066350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.170000433921814 }, { "epoch": 0.1066350710900474, "step": 45, "train/alpha": 0.2, "train/clean_ce_loss": 0.22265625, "train/clean_loss_scale": 3.3513519763946533, "train/clean_loss_weight": 0.6702703833580017, "train/combined_loss": 0.8954393863677979, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.01066350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7461994886398315 }, { "epoch": 0.1066350710900474, "step": 45, "train/alpha": 0.2, "train/clean_ce_loss": 0.158203125, "train/clean_loss_scale": 4.52177619934082, "train/clean_loss_weight": 0.9043552279472351, "train/combined_loss": 0.8584309220314026, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.01066350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.715359091758728 }, { "epoch": 0.10900473933649289, "step": 46, "train/alpha": 0.2, "train/clean_ce_loss": 0.189453125, "train/clean_loss_scale": 3.9495346546173096, "train/clean_loss_weight": 0.789906919002533, "train/combined_loss": 0.8979020118713379, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.010900473933649289, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7482516765594482 }, { "epoch": 0.10900473933649289, "step": 46, "train/alpha": 0.2, "train/clean_ce_loss": 0.166015625, "train/clean_loss_scale": 3.6585750579833984, "train/clean_loss_weight": 0.7317150235176086, "train/combined_loss": 0.7288567423820496, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.010900473933649289, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6073806285858154 }, { "epoch": 0.10900473933649289, "step": 46, "train/alpha": 0.2, "train/clean_ce_loss": 0.2060546875, "train/clean_loss_scale": 4.047502517700195, "train/clean_loss_weight": 0.809500515460968, "train/combined_loss": 1.0008082389831543, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.010900473933649289, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8340068459510803 }, { "epoch": 0.10900473933649289, "step": 46, "train/alpha": 0.2, "train/clean_ce_loss": 0.1396484375, "train/clean_loss_scale": 3.8938803672790527, "train/clean_loss_weight": 0.7787761092185974, "train/combined_loss": 0.6525291800498962, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.010900473933649289, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5437743067741394 }, { "epoch": 0.11137440758293839, "step": 47, "train/alpha": 0.2, "train/clean_ce_loss": 0.259765625, "train/clean_loss_scale": 3.3898701667785645, "train/clean_loss_weight": 0.6779740452766418, "train/combined_loss": 1.0566860437393188, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.01113744075829384, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.880571722984314 }, { "epoch": 0.11137440758293839, "step": 47, "train/alpha": 0.2, "train/clean_ce_loss": 0.2734375, "train/clean_loss_scale": 4.907901287078857, "train/clean_loss_weight": 0.9815802574157715, "train/combined_loss": 1.6104052066802979, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.01113744075829384, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3420042991638184 }, { "epoch": 0.11137440758293839, "step": 47, "train/alpha": 0.2, "train/clean_ce_loss": 0.140625, "train/clean_loss_scale": 4.328933238983154, "train/clean_loss_weight": 0.8657866716384888, "train/combined_loss": 0.730507493019104, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.01113744075829384, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6087562441825867 }, { "epoch": 0.11137440758293839, "step": 47, "train/alpha": 0.2, "train/clean_ce_loss": 0.1484375, "train/clean_loss_scale": 3.7934529781341553, "train/clean_loss_weight": 0.758690595626831, "train/combined_loss": 0.6757088303565979, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.01113744075829384, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5630906820297241 }, { "epoch": 0.11374407582938388, "step": 48, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 4.02006721496582, "train/clean_loss_weight": 0.8040134310722351, "train/combined_loss": 0.40279191732406616, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.011374407582938388, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.335659921169281 }, { "epoch": 0.11374407582938388, "step": 48, "train/alpha": 0.2, "train/clean_ce_loss": 0.169921875, "train/clean_loss_scale": 3.6296727657318115, "train/clean_loss_weight": 0.7259345650672913, "train/combined_loss": 0.7401129603385925, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.011374407582938388, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.616760790348053 }, { "epoch": 0.11374407582938388, "step": 48, "train/alpha": 0.2, "train/clean_ce_loss": 0.17578125, "train/clean_loss_scale": 4.434303283691406, "train/clean_loss_weight": 0.8868606686592102, "train/combined_loss": 0.9353609085083008, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.011374407582938388, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7794674038887024 }, { "epoch": 0.11374407582938388, "step": 48, "train/alpha": 0.2, "train/clean_ce_loss": 0.1904296875, "train/clean_loss_scale": 3.9879140853881836, "train/clean_loss_weight": 0.7975828051567078, "train/combined_loss": 0.9113006591796875, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.011374407582938388, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7594172358512878 }, { "epoch": 0.11611374407582939, "step": 49, "train/alpha": 0.2, "train/clean_ce_loss": 0.109375, "train/clean_loss_scale": 3.7804665565490723, "train/clean_loss_weight": 0.7560933232307434, "train/combined_loss": 0.4961862564086914, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.011611374407582939, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4134885370731354 }, { "epoch": 0.11611374407582939, "step": 49, "train/alpha": 0.2, "train/clean_ce_loss": 0.291015625, "train/clean_loss_scale": 3.9716849327087402, "train/clean_loss_weight": 0.7943369746208191, "train/combined_loss": 1.3869868516921997, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.011611374407582939, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1558223962783813 }, { "epoch": 0.11611374407582939, "step": 49, "train/alpha": 0.2, "train/clean_ce_loss": 0.12109375, "train/clean_loss_scale": 4.258702754974365, "train/clean_loss_weight": 0.8517405390739441, "train/combined_loss": 0.6188427805900574, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.011611374407582939, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5157023072242737 }, { "epoch": 0.11611374407582939, "step": 49, "train/alpha": 0.2, "train/clean_ce_loss": 0.265625, "train/clean_loss_scale": 4.777732849121094, "train/clean_loss_weight": 0.9555465579032898, "train/combined_loss": 1.5229023694992065, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.011611374407582939, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2690852880477905 }, { "epoch": 0.11848341232227488, "grad_norm": 15.85562801361084, "learning_rate": 7.716535433070865e-07, "loss": 4.2504, "step": 50 }, { "epoch": 0.11848341232227488, "step": 50, "train/alpha": 0.2, "train/clean_ce_loss": 0.248046875, "train/clean_loss_scale": 3.6632981300354004, "train/clean_loss_weight": 0.732659637928009, "train/combined_loss": 1.0904035568237305, "train/pos_loss/pos0": 0.519919216632843, "train/pos_loss/pos1": 0.9736855626106262, "train/pos_loss/pos2": 1.3785731792449951, "train/pos_loss/pos3": 2.0462019443511963, "train/pos_loss/pos4": 2.7600011825561523, "train/pos_loss/pos5": 3.463080883026123, "train/pos_loss/pos6": 3.6780688762664795, "train/pos_loss/weight0": 3.4737653732299805, "train/pos_loss/weight1": 1.9829097986221313, "train/pos_loss/weight2": 1.0052123069763184, "train/pos_loss/weight3": 0.3781760632991791, "train/pos_loss/weight4": 0.1256696879863739, "train/pos_loss/weight5": 0.027812175452709198, "train/pos_loss/weight6": 0.006454944610595703, "train/pos_weight_progress": 0.011848341232227487, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.908669650554657 }, { "epoch": 0.11848341232227488, "step": 50, "train/alpha": 0.2, "train/clean_ce_loss": 0.2490234375, "train/clean_loss_scale": 5.7795233726501465, "train/clean_loss_weight": 1.1559046506881714, "train/combined_loss": 1.7270841598510742, "train/pos_loss/pos0": 0.8587399125099182, "train/pos_loss/pos1": 1.6276229619979858, "train/pos_loss/pos2": 2.1058566570281982, "train/pos_loss/pos3": 2.5552284717559814, "train/pos_loss/pos4": 3.277350425720215, "train/pos_loss/pos5": 3.5076849460601807, "train/pos_loss/pos6": 3.6826326847076416, "train/pos_loss/weight0": 3.2779998779296875, "train/pos_loss/weight1": 2.049229145050049, "train/pos_loss/weight2": 1.0339062213897705, "train/pos_loss/weight3": 0.44550758600234985, "train/pos_loss/weight4": 0.147862508893013, "train/pos_loss/weight5": 0.036486003547906876, "train/pos_loss/weight6": 0.009008659981191158, "train/pos_weight_progress": 0.011848341232227487, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4392367601394653 }, { "epoch": 0.11848341232227488, "step": 50, "train/alpha": 0.2, "train/clean_ce_loss": 0.1611328125, "train/clean_loss_scale": 4.016469955444336, "train/clean_loss_weight": 0.8032940030097961, "train/combined_loss": 0.7766221165657043, "train/pos_loss/pos0": 0.4064059555530548, "train/pos_loss/pos1": 0.6489284038543701, "train/pos_loss/pos2": 0.9769259095191956, "train/pos_loss/pos3": 1.5132116079330444, "train/pos_loss/pos4": 1.971595287322998, "train/pos_loss/pos5": 2.3314907550811768, "train/pos_loss/pos6": 2.750753164291382, "train/pos_loss/weight0": 3.432121515274048, "train/pos_loss/weight1": 2.091555595397949, "train/pos_loss/weight2": 0.9834204912185669, "train/pos_loss/weight3": 0.35278332233428955, "train/pos_loss/weight4": 0.11219155788421631, "train/pos_loss/weight5": 0.022931434214115143, "train/pos_loss/weight6": 0.004996026400476694, "train/pos_weight_progress": 0.011848341232227487, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6471850872039795 }, { "epoch": 0.11848341232227488, "step": 50, "train/alpha": 0.2, "train/clean_ce_loss": 0.1328125, "train/clean_loss_scale": 4.984748840332031, "train/clean_loss_weight": 0.9969497919082642, "train/combined_loss": 0.7944443225860596, "train/pos_loss/pos0": 0.25088489055633545, "train/pos_loss/pos1": 0.5733867883682251, "train/pos_loss/pos2": 0.8579292297363281, "train/pos_loss/pos3": 1.3337054252624512, "train/pos_loss/pos4": 1.7986699342727661, "train/pos_loss/pos5": 2.3095510005950928, "train/pos_loss/pos6": 2.653576135635376, "train/pos_loss/weight0": 2.807197093963623, "train/pos_loss/weight1": 1.9088671207427979, "train/pos_loss/weight2": 1.1564913988113403, "train/pos_loss/weight3": 0.6358646750450134, "train/pos_loss/weight4": 0.31078773736953735, "train/pos_loss/weight5": 0.12558792531490326, "train/pos_loss/weight6": 0.055203963071107864, "train/pos_weight_progress": 0.011848341232227487, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6620369553565979 }, { "epoch": 0.12085308056872038, "step": 51, "train/alpha": 0.2, "train/clean_ce_loss": 0.2578125, "train/clean_loss_scale": 6.133994102478027, "train/clean_loss_weight": 1.2267988920211792, "train/combined_loss": 1.8977043628692627, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.012085308056872038, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5814203023910522 }, { "epoch": 0.12085308056872038, "step": 51, "train/alpha": 0.2, "train/clean_ce_loss": 0.1552734375, "train/clean_loss_scale": 4.956117153167725, "train/clean_loss_weight": 0.9912234544754028, "train/combined_loss": 0.9234640598297119, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.012085308056872038, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7695533633232117 }, { "epoch": 0.12085308056872038, "step": 51, "train/alpha": 0.2, "train/clean_ce_loss": 0.1259765625, "train/clean_loss_scale": 4.514373302459717, "train/clean_loss_weight": 0.9028746485710144, "train/combined_loss": 0.6824462413787842, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.012085308056872038, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5687052011489868 }, { "epoch": 0.12085308056872038, "step": 51, "train/alpha": 0.2, "train/clean_ce_loss": 0.15234375, "train/clean_loss_scale": 5.125781536102295, "train/clean_loss_weight": 1.0251563787460327, "train/combined_loss": 0.9370568990707397, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.012085308056872038, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7808807492256165 }, { "epoch": 0.12322274881516587, "step": 52, "train/alpha": 0.2, "train/clean_ce_loss": 0.2060546875, "train/clean_loss_scale": 4.647298812866211, "train/clean_loss_weight": 0.9294597506523132, "train/combined_loss": 1.1491172313690186, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.012322274881516588, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9575976729393005 }, { "epoch": 0.12322274881516587, "step": 52, "train/alpha": 0.2, "train/clean_ce_loss": 0.1630859375, "train/clean_loss_scale": 4.967060565948486, "train/clean_loss_weight": 0.9934121370315552, "train/combined_loss": 0.972069263458252, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.012322274881516588, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8100576996803284 }, { "epoch": 0.12322274881516587, "step": 52, "train/alpha": 0.2, "train/clean_ce_loss": 0.1904296875, "train/clean_loss_scale": 4.793014049530029, "train/clean_loss_weight": 0.9586028456687927, "train/combined_loss": 1.0952786207199097, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.012322274881516588, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9127321839332581 }, { "epoch": 0.12322274881516587, "step": 52, "train/alpha": 0.2, "train/clean_ce_loss": 0.150390625, "train/clean_loss_scale": 5.421943664550781, "train/clean_loss_weight": 1.0843887329101562, "train/combined_loss": 0.9784913659095764, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.012322274881516588, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8154094815254211 }, { "epoch": 0.12559241706161137, "step": 53, "train/alpha": 0.2, "train/clean_ce_loss": 0.279296875, "train/clean_loss_scale": 5.4970574378967285, "train/clean_loss_weight": 1.0994114875793457, "train/combined_loss": 1.8423731327056885, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.012559241706161137, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.535310983657837 }, { "epoch": 0.12559241706161137, "step": 53, "train/alpha": 0.2, "train/clean_ce_loss": 0.1669921875, "train/clean_loss_scale": 5.51105260848999, "train/clean_loss_weight": 1.102210521697998, "train/combined_loss": 1.1043633222579956, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.012559241706161137, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9203027486801147 }, { "epoch": 0.12559241706161137, "step": 53, "train/alpha": 0.2, "train/clean_ce_loss": 0.28515625, "train/clean_loss_scale": 5.1149702072143555, "train/clean_loss_weight": 1.022994041442871, "train/combined_loss": 1.7502788305282593, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.012559241706161137, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4585657119750977 }, { "epoch": 0.12559241706161137, "step": 53, "train/alpha": 0.2, "train/clean_ce_loss": 0.1650390625, "train/clean_loss_scale": 5.245106220245361, "train/clean_loss_weight": 1.0490212440490723, "train/combined_loss": 1.0387768745422363, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.012559241706161137, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8656473755836487 }, { "epoch": 0.12796208530805686, "step": 54, "train/alpha": 0.2, "train/clean_ce_loss": 0.31640625, "train/clean_loss_scale": 5.270715236663818, "train/clean_loss_weight": 1.0541430711746216, "train/combined_loss": 2.0012247562408447, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.012796208530805687, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.6676872968673706 }, { "epoch": 0.12796208530805686, "step": 54, "train/alpha": 0.2, "train/clean_ce_loss": 0.15234375, "train/clean_loss_scale": 5.714585304260254, "train/clean_loss_weight": 1.1429170370101929, "train/combined_loss": 1.044697642326355, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.012796208530805687, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.870581328868866 }, { "epoch": 0.12796208530805686, "step": 54, "train/alpha": 0.2, "train/clean_ce_loss": 0.162109375, "train/clean_loss_scale": 5.299238681793213, "train/clean_loss_weight": 1.0598477125167847, "train/combined_loss": 1.030867576599121, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.012796208530805687, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.859056293964386 }, { "epoch": 0.12796208530805686, "step": 54, "train/alpha": 0.2, "train/clean_ce_loss": 0.193359375, "train/clean_loss_scale": 5.423685550689697, "train/clean_loss_weight": 1.0847371816635132, "train/combined_loss": 1.2584645748138428, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.012796208530805687, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0487204790115356 }, { "epoch": 0.13033175355450238, "grad_norm": 19.032474517822266, "learning_rate": 8.503937007874016e-07, "loss": 4.5353, "step": 55 }, { "epoch": 0.13033175355450238, "step": 55, "train/alpha": 0.2, "train/clean_ce_loss": 0.1767578125, "train/clean_loss_scale": 4.212305068969727, "train/clean_loss_weight": 0.8424610495567322, "train/combined_loss": 0.8934694528579712, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.013033175355450236, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7445578575134277 }, { "epoch": 0.13033175355450238, "step": 55, "train/alpha": 0.2, "train/clean_ce_loss": 0.1591796875, "train/clean_loss_scale": 5.696670055389404, "train/clean_loss_weight": 1.1393340826034546, "train/combined_loss": 1.0881530046463013, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.013033175355450236, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9067941308021545 }, { "epoch": 0.13033175355450238, "step": 55, "train/alpha": 0.2, "train/clean_ce_loss": 0.10009765625, "train/clean_loss_scale": 5.850931167602539, "train/clean_loss_weight": 1.1701862812042236, "train/combined_loss": 0.7027974128723145, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.013033175355450236, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5856645107269287 }, { "epoch": 0.13033175355450238, "step": 55, "train/alpha": 0.2, "train/clean_ce_loss": 0.1044921875, "train/clean_loss_scale": 7.3378376960754395, "train/clean_loss_weight": 1.4675675630569458, "train/combined_loss": 0.9200960397720337, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.013033175355450236, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7667466998100281 }, { "epoch": 0.13270142180094788, "step": 56, "train/alpha": 0.2, "train/clean_ce_loss": 0.291015625, "train/clean_loss_scale": 5.803935527801514, "train/clean_loss_weight": 1.1607871055603027, "train/combined_loss": 2.0268430709838867, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.013270142180094787, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.6890358924865723 }, { "epoch": 0.13270142180094788, "step": 56, "train/alpha": 0.2, "train/clean_ce_loss": 0.1484375, "train/clean_loss_scale": 5.30999755859375, "train/clean_loss_weight": 1.0619995594024658, "train/combined_loss": 0.9458433389663696, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.013270142180094787, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7882027626037598 }, { "epoch": 0.13270142180094788, "step": 56, "train/alpha": 0.2, "train/clean_ce_loss": 0.1484375, "train/clean_loss_scale": 5.841833114624023, "train/clean_loss_weight": 1.1683666706085205, "train/combined_loss": 1.040576457977295, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.013270142180094787, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8671470880508423 }, { "epoch": 0.13270142180094788, "step": 56, "train/alpha": 0.2, "train/clean_ce_loss": 0.1865234375, "train/clean_loss_scale": 4.807586669921875, "train/clean_loss_weight": 0.961517333984375, "train/combined_loss": 1.0760730504989624, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.013270142180094787, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8967275619506836 }, { "epoch": 0.13507109004739337, "step": 57, "train/alpha": 0.2, "train/clean_ce_loss": 0.171875, "train/clean_loss_scale": 5.271163463592529, "train/clean_loss_weight": 1.0542327165603638, "train/combined_loss": 1.0871773958206177, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.013507109004739337, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.905981183052063 }, { "epoch": 0.13507109004739337, "step": 57, "train/alpha": 0.2, "train/clean_ce_loss": 0.16796875, "train/clean_loss_scale": 5.485191345214844, "train/clean_loss_weight": 1.0970382690429688, "train/combined_loss": 1.1056088209152222, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.013507109004739337, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9213407039642334 }, { "epoch": 0.13507109004739337, "step": 57, "train/alpha": 0.2, "train/clean_ce_loss": 0.142578125, "train/clean_loss_scale": 6.65493631362915, "train/clean_loss_weight": 1.3309873342514038, "train/combined_loss": 1.138617992401123, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.013507109004739337, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9488483667373657 }, { "epoch": 0.13507109004739337, "step": 57, "train/alpha": 0.2, "train/clean_ce_loss": 0.16796875, "train/clean_loss_scale": 6.839920520782471, "train/clean_loss_weight": 1.3679841756820679, "train/combined_loss": 1.378671407699585, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.013507109004739337, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.148892879486084 }, { "epoch": 0.13744075829383887, "step": 58, "train/alpha": 0.2, "train/clean_ce_loss": 0.1015625, "train/clean_loss_scale": 6.665782451629639, "train/clean_loss_weight": 1.3331564664840698, "train/combined_loss": 0.8123922348022461, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.013744075829383886, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6769935488700867 }, { "epoch": 0.13744075829383887, "step": 58, "train/alpha": 0.2, "train/clean_ce_loss": 0.189453125, "train/clean_loss_scale": 4.806124210357666, "train/clean_loss_weight": 0.9612248539924622, "train/combined_loss": 1.0926423072814941, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.013744075829383886, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9105352759361267 }, { "epoch": 0.13744075829383887, "step": 58, "train/alpha": 0.2, "train/clean_ce_loss": 0.099609375, "train/clean_loss_scale": 5.448939800262451, "train/clean_loss_weight": 1.0897879600524902, "train/combined_loss": 0.6513186097145081, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.013744075829383886, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5427654981613159 }, { "epoch": 0.13744075829383887, "step": 58, "train/alpha": 0.2, "train/clean_ce_loss": 0.16015625, "train/clean_loss_scale": 5.2675604820251465, "train/clean_loss_weight": 1.0535120964050293, "train/combined_loss": 1.0123592615127563, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.013744075829383886, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8436327576637268 }, { "epoch": 0.13981042654028436, "step": 59, "train/alpha": 0.2, "train/clean_ce_loss": 0.125, "train/clean_loss_scale": 7.249770164489746, "train/clean_loss_weight": 1.4499540328979492, "train/combined_loss": 1.087465524673462, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.013981042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9062212705612183 }, { "epoch": 0.13981042654028436, "step": 59, "train/alpha": 0.2, "train/clean_ce_loss": 0.11767578125, "train/clean_loss_scale": 6.772454738616943, "train/clean_loss_weight": 1.3544909954071045, "train/combined_loss": 0.956344723701477, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.013981042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7969539165496826 }, { "epoch": 0.13981042654028436, "step": 59, "train/alpha": 0.2, "train/clean_ce_loss": 0.138671875, "train/clean_loss_scale": 5.718717575073242, "train/clean_loss_weight": 1.1437435150146484, "train/combined_loss": 0.9516303539276123, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.013981042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7930253148078918 }, { "epoch": 0.13981042654028436, "step": 59, "train/alpha": 0.2, "train/clean_ce_loss": 0.2216796875, "train/clean_loss_scale": 6.506667137145996, "train/clean_loss_weight": 1.3013334274291992, "train/combined_loss": 1.7308751344680786, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.013981042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4423959255218506 }, { "epoch": 0.14218009478672985, "grad_norm": 16.403072357177734, "learning_rate": 9.291338582677165e-07, "loss": 4.3403, "step": 60 }, { "epoch": 0.14218009478672985, "step": 60, "train/alpha": 0.2, "train/clean_ce_loss": 0.10302734375, "train/clean_loss_scale": 6.369451999664307, "train/clean_loss_weight": 1.2738903760910034, "train/combined_loss": 0.7874732613563538, "train/pos_loss/pos0": 0.2409505695104599, "train/pos_loss/pos1": 0.5454679727554321, "train/pos_loss/pos2": 0.855732798576355, "train/pos_loss/pos3": 1.2816108465194702, "train/pos_loss/pos4": 1.6885756254196167, "train/pos_loss/pos5": 2.1668872833251953, "train/pos_loss/pos6": 2.499373197555542, "train/pos_loss/weight0": 2.675713062286377, "train/pos_loss/weight1": 1.9094401597976685, "train/pos_loss/weight2": 1.2102693319320679, "train/pos_loss/weight3": 0.6786504983901978, "train/pos_loss/weight4": 0.3289879560470581, "train/pos_loss/weight5": 0.1371336579322815, "train/pos_loss/weight6": 0.059805091470479965, "train/pos_weight_progress": 0.014218009478672985, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.656227707862854 }, { "epoch": 0.14218009478672985, "step": 60, "train/alpha": 0.2, "train/clean_ce_loss": 0.12353515625, "train/clean_loss_scale": 5.991231918334961, "train/clean_loss_weight": 1.1982463598251343, "train/combined_loss": 0.8881533145904541, "train/pos_loss/pos0": 0.2746502757072449, "train/pos_loss/pos1": 0.6033614873886108, "train/pos_loss/pos2": 0.9139565229415894, "train/pos_loss/pos3": 1.2903234958648682, "train/pos_loss/pos4": 1.7475098371505737, "train/pos_loss/pos5": 2.140153646469116, "train/pos_loss/pos6": 2.5702450275421143, "train/pos_loss/weight0": 2.540465831756592, "train/pos_loss/weight1": 1.8443551063537598, "train/pos_loss/weight2": 1.220036506652832, "train/pos_loss/weight3": 0.7258066534996033, "train/pos_loss/weight4": 0.3866976797580719, "train/pos_loss/weight5": 0.19016315042972565, "train/pos_loss/weight6": 0.0924757868051529, "train/pos_weight_progress": 0.014218009478672985, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7401277422904968 }, { "epoch": 0.14218009478672985, "step": 60, "train/alpha": 0.2, "train/clean_ce_loss": 0.1181640625, "train/clean_loss_scale": 6.630100250244141, "train/clean_loss_weight": 1.3260201215744019, "train/combined_loss": 0.9401274919509888, "train/pos_loss/pos0": 0.3114432990550995, "train/pos_loss/pos1": 0.5487316846847534, "train/pos_loss/pos2": 1.0686525106430054, "train/pos_loss/pos3": 1.4635307788848877, "train/pos_loss/pos4": 1.959611177444458, "train/pos_loss/pos5": 2.1677582263946533, "train/pos_loss/pos6": 2.6400699615478516, "train/pos_loss/weight0": 2.5805294513702393, "train/pos_loss/weight1": 1.8730921745300293, "train/pos_loss/weight2": 1.2033212184906006, "train/pos_loss/weight3": 0.716966986656189, "train/pos_loss/weight4": 0.36853641271591187, "train/pos_loss/weight5": 0.1722179651260376, "train/pos_loss/weight6": 0.08533576130867004, "train/pos_weight_progress": 0.014218009478672985, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.783439576625824 }, { "epoch": 0.14218009478672985, "step": 60, "train/alpha": 0.2, "train/clean_ce_loss": 0.1953125, "train/clean_loss_scale": 6.7355756759643555, "train/clean_loss_weight": 1.347115159034729, "train/combined_loss": 1.5786504745483398, "train/pos_loss/pos0": 0.565870463848114, "train/pos_loss/pos1": 1.0982129573822021, "train/pos_loss/pos2": 1.8323498964309692, "train/pos_loss/pos3": 2.412663459777832, "train/pos_loss/pos4": 2.882214069366455, "train/pos_loss/pos5": 3.183298110961914, "train/pos_loss/pos6": 3.8641655445098877, "train/pos_loss/weight0": 2.662947654724121, "train/pos_loss/weight1": 1.8504834175109863, "train/pos_loss/weight2": 1.2239069938659668, "train/pos_loss/weight3": 0.6872218251228333, "train/pos_loss/weight4": 0.34852612018585205, "train/pos_loss/weight5": 0.15284785628318787, "train/pos_loss/weight6": 0.07406651973724365, "train/pos_weight_progress": 0.014218009478672985, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3155421018600464 }, { "epoch": 0.14454976303317535, "step": 61, "train/alpha": 0.2, "train/clean_ce_loss": 0.16796875, "train/clean_loss_scale": 4.756119251251221, "train/clean_loss_weight": 0.9512238502502441, "train/combined_loss": 0.9586553573608398, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.014454976303317536, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7988794445991516 }, { "epoch": 0.14454976303317535, "step": 61, "train/alpha": 0.2, "train/clean_ce_loss": 0.1181640625, "train/clean_loss_scale": 4.575212001800537, "train/clean_loss_weight": 0.9150424003601074, "train/combined_loss": 0.6487507820129395, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.014454976303317536, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5406256318092346 }, { "epoch": 0.14454976303317535, "step": 61, "train/alpha": 0.2, "train/clean_ce_loss": 0.2333984375, "train/clean_loss_scale": 5.418076992034912, "train/clean_loss_weight": 1.0836154222488403, "train/combined_loss": 1.5174849033355713, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.014454976303317536, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.264570713043213 }, { "epoch": 0.14454976303317535, "step": 61, "train/alpha": 0.2, "train/clean_ce_loss": 0.1640625, "train/clean_loss_scale": 6.229825019836426, "train/clean_loss_weight": 1.2459650039672852, "train/combined_loss": 1.2264968156814575, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.014454976303317536, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.022080659866333 }, { "epoch": 0.14691943127962084, "step": 62, "train/alpha": 0.2, "train/clean_ce_loss": 0.11572265625, "train/clean_loss_scale": 4.854091644287109, "train/clean_loss_weight": 0.9708183407783508, "train/combined_loss": 0.6740740537643433, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.014691943127962086, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5617283582687378 }, { "epoch": 0.14691943127962084, "step": 62, "train/alpha": 0.2, "train/clean_ce_loss": 0.1162109375, "train/clean_loss_scale": 4.763429641723633, "train/clean_loss_weight": 0.9526859521865845, "train/combined_loss": 0.6642751693725586, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.014691943127962086, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5535626411437988 }, { "epoch": 0.14691943127962084, "step": 62, "train/alpha": 0.2, "train/clean_ce_loss": 0.154296875, "train/clean_loss_scale": 5.7675933837890625, "train/clean_loss_weight": 1.1535186767578125, "train/combined_loss": 1.0679059028625488, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.014691943127962086, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8899216055870056 }, { "epoch": 0.14691943127962084, "step": 62, "train/alpha": 0.2, "train/clean_ce_loss": 0.1494140625, "train/clean_loss_scale": 5.3968000411987305, "train/clean_loss_weight": 1.079360008239746, "train/combined_loss": 0.9676293730735779, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.014691943127962086, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8063578009605408 }, { "epoch": 0.14928909952606634, "step": 63, "train/alpha": 0.2, "train/clean_ce_loss": 0.09912109375, "train/clean_loss_scale": 4.413624286651611, "train/clean_loss_weight": 0.8827248811721802, "train/combined_loss": 0.5249798893928528, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.014928909952606635, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4374832510948181 }, { "epoch": 0.14928909952606634, "step": 63, "train/alpha": 0.2, "train/clean_ce_loss": 0.15234375, "train/clean_loss_scale": 4.7665863037109375, "train/clean_loss_weight": 0.9533172845840454, "train/combined_loss": 0.8713915348052979, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.014928909952606635, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7261596322059631 }, { "epoch": 0.14928909952606634, "step": 63, "train/alpha": 0.2, "train/clean_ce_loss": 0.1337890625, "train/clean_loss_scale": 4.335940837860107, "train/clean_loss_weight": 0.8671881556510925, "train/combined_loss": 0.6961216926574707, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.014928909952606635, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5801014304161072 }, { "epoch": 0.14928909952606634, "step": 63, "train/alpha": 0.2, "train/clean_ce_loss": 0.103515625, "train/clean_loss_scale": 5.33608341217041, "train/clean_loss_weight": 1.0672167539596558, "train/combined_loss": 0.6628416180610657, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.014928909952606635, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5523679852485657 }, { "epoch": 0.15165876777251186, "step": 64, "train/alpha": 0.2, "train/clean_ce_loss": 0.09619140625, "train/clean_loss_scale": 5.6482930183410645, "train/clean_loss_weight": 1.129658579826355, "train/combined_loss": 0.6519806981086731, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.015165876777251185, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5433172583580017 }, { "epoch": 0.15165876777251186, "step": 64, "train/alpha": 0.2, "train/clean_ce_loss": 0.2060546875, "train/clean_loss_scale": 6.002173900604248, "train/clean_loss_weight": 1.2004348039627075, "train/combined_loss": 1.4841313362121582, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.015165876777251185, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2367761135101318 }, { "epoch": 0.15165876777251186, "step": 64, "train/alpha": 0.2, "train/clean_ce_loss": 0.08544921875, "train/clean_loss_scale": 5.514604568481445, "train/clean_loss_weight": 1.1029208898544312, "train/combined_loss": 0.5654623508453369, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.015165876777251185, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.47121864557266235 }, { "epoch": 0.15165876777251186, "step": 64, "train/alpha": 0.2, "train/clean_ce_loss": 0.17578125, "train/clean_loss_scale": 5.785099983215332, "train/clean_loss_weight": 1.1570199728012085, "train/combined_loss": 1.22029447555542, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.015165876777251185, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0169121026992798 }, { "epoch": 0.15402843601895735, "grad_norm": 17.33542251586914, "learning_rate": 1.0078740157480315e-06, "loss": 4.226, "step": 65 }, { "epoch": 0.15402843601895735, "step": 65, "train/alpha": 0.2, "train/clean_ce_loss": 0.1298828125, "train/clean_loss_scale": 4.2227783203125, "train/clean_loss_weight": 0.844555675983429, "train/combined_loss": 0.658159613609314, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.015402843601895734, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5484663248062134 }, { "epoch": 0.15402843601895735, "step": 65, "train/alpha": 0.2, "train/clean_ce_loss": 0.10205078125, "train/clean_loss_scale": 6.131288051605225, "train/clean_loss_weight": 1.2262576818466187, "train/combined_loss": 0.7508432865142822, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.015402843601895734, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6257027387619019 }, { "epoch": 0.15402843601895735, "step": 65, "train/alpha": 0.2, "train/clean_ce_loss": 0.212890625, "train/clean_loss_scale": 4.591989040374756, "train/clean_loss_weight": 0.918397843837738, "train/combined_loss": 1.1731096506118774, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.015402843601895734, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9775913953781128 }, { "epoch": 0.15402843601895735, "step": 65, "train/alpha": 0.2, "train/clean_ce_loss": 0.11962890625, "train/clean_loss_scale": 4.7018280029296875, "train/clean_loss_weight": 0.9403656125068665, "train/combined_loss": 0.6749694347381592, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.015402843601895734, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5624745488166809 }, { "epoch": 0.15639810426540285, "step": 66, "train/alpha": 0.2, "train/clean_ce_loss": 0.2021484375, "train/clean_loss_scale": 6.544087886810303, "train/clean_loss_weight": 1.3088176250457764, "train/combined_loss": 1.5874526500701904, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.015639810426540283, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3228771686553955 }, { "epoch": 0.15639810426540285, "step": 66, "train/alpha": 0.2, "train/clean_ce_loss": 0.1259765625, "train/clean_loss_scale": 5.5141754150390625, "train/clean_loss_weight": 1.1028350591659546, "train/combined_loss": 0.8335882425308228, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.015639810426540283, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6946568489074707 }, { "epoch": 0.15639810426540285, "step": 66, "train/alpha": 0.2, "train/clean_ce_loss": 0.10498046875, "train/clean_loss_scale": 5.165822982788086, "train/clean_loss_weight": 1.033164620399475, "train/combined_loss": 0.6507726311683655, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.015639810426540283, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5423105359077454 }, { "epoch": 0.15639810426540285, "step": 66, "train/alpha": 0.2, "train/clean_ce_loss": 0.1904296875, "train/clean_loss_scale": 5.722105979919434, "train/clean_loss_weight": 1.1444212198257446, "train/combined_loss": 1.3075906038284302, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.015639810426540283, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0896588563919067 }, { "epoch": 0.15876777251184834, "step": 67, "train/alpha": 0.2, "train/clean_ce_loss": 0.1318359375, "train/clean_loss_scale": 5.4797258377075195, "train/clean_loss_weight": 1.0959452390670776, "train/combined_loss": 0.8669097423553467, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.015876777251184835, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7224248051643372 }, { "epoch": 0.15876777251184834, "step": 67, "train/alpha": 0.2, "train/clean_ce_loss": 0.1240234375, "train/clean_loss_scale": 5.113492488861084, "train/clean_loss_weight": 1.0226985216140747, "train/combined_loss": 0.7610315084457397, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.015876777251184835, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.634192943572998 }, { "epoch": 0.15876777251184834, "step": 67, "train/alpha": 0.2, "train/clean_ce_loss": 0.1826171875, "train/clean_loss_scale": 5.721022129058838, "train/clean_loss_weight": 1.1442044973373413, "train/combined_loss": 1.2537083625793457, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.015876777251184835, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0447570085525513 }, { "epoch": 0.15876777251184834, "step": 67, "train/alpha": 0.2, "train/clean_ce_loss": 0.10107421875, "train/clean_loss_scale": 5.522477626800537, "train/clean_loss_weight": 1.1044955253601074, "train/combined_loss": 0.6698161363601685, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.015876777251184835, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5581800937652588 }, { "epoch": 0.16113744075829384, "step": 68, "train/alpha": 0.2, "train/clean_ce_loss": 0.126953125, "train/clean_loss_scale": 6.366908550262451, "train/clean_loss_weight": 1.2733817100524902, "train/combined_loss": 0.9699587225914001, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.016113744075829384, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8082989454269409 }, { "epoch": 0.16113744075829384, "step": 68, "train/alpha": 0.2, "train/clean_ce_loss": 0.099609375, "train/clean_loss_scale": 5.866055488586426, "train/clean_loss_weight": 1.1732110977172852, "train/combined_loss": 0.7011769413948059, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.016113744075829384, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5843141078948975 }, { "epoch": 0.16113744075829384, "step": 68, "train/alpha": 0.2, "train/clean_ce_loss": 0.14453125, "train/clean_loss_scale": 4.986661911010742, "train/clean_loss_weight": 0.9973323941230774, "train/combined_loss": 0.8648741245269775, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.016113744075829384, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7207284569740295 }, { "epoch": 0.16113744075829384, "step": 68, "train/alpha": 0.2, "train/clean_ce_loss": 0.1806640625, "train/clean_loss_scale": 6.598851680755615, "train/clean_loss_weight": 1.319770336151123, "train/combined_loss": 1.4306104183197021, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.016113744075829384, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1921753883361816 }, { "epoch": 0.16350710900473933, "step": 69, "train/alpha": 0.2, "train/clean_ce_loss": 0.1376953125, "train/clean_loss_scale": 5.1713175773620605, "train/clean_loss_weight": 1.0342634916305542, "train/combined_loss": 0.8544794321060181, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.016350710900473932, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7120661735534668 }, { "epoch": 0.16350710900473933, "step": 69, "train/alpha": 0.2, "train/clean_ce_loss": 0.1123046875, "train/clean_loss_scale": 5.712642669677734, "train/clean_loss_weight": 1.1425285339355469, "train/combined_loss": 0.7698678970336914, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.016350710900473932, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6415565609931946 }, { "epoch": 0.16350710900473933, "step": 69, "train/alpha": 0.2, "train/clean_ce_loss": 0.09716796875, "train/clean_loss_scale": 5.141294002532959, "train/clean_loss_weight": 1.0282588005065918, "train/combined_loss": 0.5994828939437866, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.016350710900473932, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4995690882205963 }, { "epoch": 0.16350710900473933, "step": 69, "train/alpha": 0.2, "train/clean_ce_loss": 0.103515625, "train/clean_loss_scale": 5.837416648864746, "train/clean_loss_weight": 1.1674833297729492, "train/combined_loss": 0.7251166105270386, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.016350710900473932, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6042638421058655 }, { "epoch": 0.16587677725118483, "grad_norm": 13.40555477142334, "learning_rate": 1.0866141732283465e-06, "loss": 3.9934, "step": 70 }, { "epoch": 0.16587677725118483, "step": 70, "train/alpha": 0.2, "train/clean_ce_loss": 0.1884765625, "train/clean_loss_scale": 6.3351216316223145, "train/clean_loss_weight": 1.2670243978500366, "train/combined_loss": 1.432826280593872, "train/pos_loss/pos0": 0.7438402771949768, "train/pos_loss/pos1": 1.3535047769546509, "train/pos_loss/pos2": 1.7074153423309326, "train/pos_loss/pos3": 2.414970636367798, "train/pos_loss/pos4": 3.0356814861297607, "train/pos_loss/pos5": 3.390486717224121, "train/pos_loss/pos6": 3.5970287322998047, "train/pos_loss/weight0": 3.401078224182129, "train/pos_loss/weight1": 2.071284055709839, "train/pos_loss/weight2": 1.0145217180252075, "train/pos_loss/weight3": 0.3789410889148712, "train/pos_loss/weight4": 0.10623718053102493, "train/pos_loss/weight5": 0.022955218330025673, "train/pos_loss/weight6": 0.004982042126357555, "train/pos_weight_progress": 0.016587677725118485, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1940219402313232 }, { "epoch": 0.16587677725118483, "step": 70, "train/alpha": 0.2, "train/clean_ce_loss": 0.130859375, "train/clean_loss_scale": 4.857799530029297, "train/clean_loss_weight": 0.9715599417686462, "train/combined_loss": 0.7628263235092163, "train/pos_loss/pos0": 0.39934295415878296, "train/pos_loss/pos1": 0.6995024681091309, "train/pos_loss/pos2": 0.9285302758216858, "train/pos_loss/pos3": 1.3071283102035522, "train/pos_loss/pos4": 1.8833670616149902, "train/pos_loss/pos5": 2.257512331008911, "train/pos_loss/pos6": 2.4968297481536865, "train/pos_loss/weight0": 3.440950632095337, "train/pos_loss/weight1": 2.123257637023926, "train/pos_loss/weight2": 0.9487939476966858, "train/pos_loss/weight3": 0.35958346724510193, "train/pos_loss/weight4": 0.10364548861980438, "train/pos_loss/weight5": 0.020016543567180634, "train/pos_loss/weight6": 0.00375236663967371, "train/pos_weight_progress": 0.016587677725118485, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6356886029243469 }, { "epoch": 0.16587677725118483, "step": 70, "train/alpha": 0.2, "train/clean_ce_loss": 0.1806640625, "train/clean_loss_scale": 8.173971176147461, "train/clean_loss_weight": 1.6347942352294922, "train/combined_loss": 1.7720913887023926, "train/pos_loss/pos0": 0.7184296250343323, "train/pos_loss/pos1": 1.3558365106582642, "train/pos_loss/pos2": 1.9762728214263916, "train/pos_loss/pos3": 2.783705711364746, "train/pos_loss/pos4": 3.338247060775757, "train/pos_loss/pos5": 3.641779899597168, "train/pos_loss/pos6": 4.109023571014404, "train/pos_loss/weight0": 2.7971835136413574, "train/pos_loss/weight1": 1.9242241382598877, "train/pos_loss/weight2": 1.171030044555664, "train/pos_loss/weight3": 0.6369890570640564, "train/pos_loss/weight4": 0.2989225685596466, "train/pos_loss/weight5": 0.1204109713435173, "train/pos_loss/weight6": 0.051239609718322754, "train/pos_weight_progress": 0.016587677725118485, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4767428636550903 }, { "epoch": 0.16587677725118483, "step": 70, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 5.672337055206299, "train/clean_loss_weight": 1.1344674825668335, "train/combined_loss": 0.5417525172233582, "train/pos_loss/pos0": 0.24049736559391022, "train/pos_loss/pos1": 0.4868246614933014, "train/pos_loss/pos2": 0.7101596593856812, "train/pos_loss/pos3": 1.117451786994934, "train/pos_loss/pos4": 1.6806985139846802, "train/pos_loss/pos5": 2.117004632949829, "train/pos_loss/pos6": 2.694551706314087, "train/pos_loss/weight0": 3.4254331588745117, "train/pos_loss/weight1": 2.093625068664551, "train/pos_loss/weight2": 1.0128357410430908, "train/pos_loss/weight3": 0.3505195379257202, "train/pos_loss/weight4": 0.09543388336896896, "train/pos_loss/weight5": 0.018570996820926666, "train/pos_loss/weight6": 0.003581451717764139, "train/pos_weight_progress": 0.016587677725118485, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45146042108535767 }, { "epoch": 0.16824644549763032, "step": 71, "train/alpha": 0.2, "train/clean_ce_loss": 0.1845703125, "train/clean_loss_scale": 6.25060510635376, "train/clean_loss_weight": 1.250120997428894, "train/combined_loss": 1.384411334991455, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.016824644549763033, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.153676152229309 }, { "epoch": 0.16824644549763032, "step": 71, "train/alpha": 0.2, "train/clean_ce_loss": 0.0869140625, "train/clean_loss_scale": 6.981712818145752, "train/clean_loss_weight": 1.3963426351547241, "train/combined_loss": 0.7281708121299744, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.016824644549763033, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6068090200424194 }, { "epoch": 0.16824644549763032, "step": 71, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 8.557573318481445, "train/clean_loss_weight": 1.7115147113800049, "train/combined_loss": 0.8975423574447632, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.016824644549763033, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7479519844055176 }, { "epoch": 0.16824644549763032, "step": 71, "train/alpha": 0.2, "train/clean_ce_loss": 0.1396484375, "train/clean_loss_scale": 6.022249698638916, "train/clean_loss_weight": 1.204450011253357, "train/combined_loss": 1.0091973543167114, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.016824644549763033, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8409977555274963 }, { "epoch": 0.17061611374407584, "step": 72, "train/alpha": 0.2, "train/clean_ce_loss": 0.181640625, "train/clean_loss_scale": 6.887914657592773, "train/clean_loss_weight": 1.3775829076766968, "train/combined_loss": 1.5013501644134521, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.017061611374407582, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2511250972747803 }, { "epoch": 0.17061611374407584, "step": 72, "train/alpha": 0.2, "train/clean_ce_loss": 0.10107421875, "train/clean_loss_scale": 8.041158676147461, "train/clean_loss_weight": 1.608231782913208, "train/combined_loss": 0.9753046035766602, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.017061611374407582, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8127537965774536 }, { "epoch": 0.17061611374407584, "step": 72, "train/alpha": 0.2, "train/clean_ce_loss": 0.08935546875, "train/clean_loss_scale": 8.014313697814941, "train/clean_loss_weight": 1.6028627157211304, "train/combined_loss": 0.8593472838401794, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.017061611374407582, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7161227464675903 }, { "epoch": 0.17061611374407584, "step": 72, "train/alpha": 0.2, "train/clean_ce_loss": 0.1298828125, "train/clean_loss_scale": 6.65120267868042, "train/clean_loss_weight": 1.3302406072616577, "train/combined_loss": 1.0366523265838623, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.017061611374407582, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8638769388198853 }, { "epoch": 0.17298578199052134, "step": 73, "train/alpha": 0.2, "train/clean_ce_loss": 0.1669921875, "train/clean_loss_scale": 7.746452808380127, "train/clean_loss_weight": 1.5492905378341675, "train/combined_loss": 1.5523165464401245, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.017298578199052134, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2935971021652222 }, { "epoch": 0.17298578199052134, "step": 73, "train/alpha": 0.2, "train/clean_ce_loss": 0.11376953125, "train/clean_loss_scale": 8.062734603881836, "train/clean_loss_weight": 1.6125469207763672, "train/combined_loss": 1.1007522344589233, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.017298578199052134, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9172934889793396 }, { "epoch": 0.17298578199052134, "step": 73, "train/alpha": 0.2, "train/clean_ce_loss": 0.1162109375, "train/clean_loss_scale": 7.392768859863281, "train/clean_loss_weight": 1.4785537719726562, "train/combined_loss": 1.0309447050094604, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.017298578199052134, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8591206073760986 }, { "epoch": 0.17298578199052134, "step": 73, "train/alpha": 0.2, "train/clean_ce_loss": 0.0908203125, "train/clean_loss_scale": 7.675000190734863, "train/clean_loss_weight": 1.5350000858306885, "train/combined_loss": 0.8364551067352295, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.017298578199052134, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6970459222793579 }, { "epoch": 0.17535545023696683, "step": 74, "train/alpha": 0.2, "train/clean_ce_loss": 0.09228515625, "train/clean_loss_scale": 8.335082054138184, "train/clean_loss_weight": 1.6670163869857788, "train/combined_loss": 0.92304527759552, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.017535545023696683, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7692043781280518 }, { "epoch": 0.17535545023696683, "step": 74, "train/alpha": 0.2, "train/clean_ce_loss": 0.07470703125, "train/clean_loss_scale": 9.277131080627441, "train/clean_loss_weight": 1.8554261922836304, "train/combined_loss": 0.8316803574562073, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.017535545023696683, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6930669546127319 }, { "epoch": 0.17535545023696683, "step": 74, "train/alpha": 0.2, "train/clean_ce_loss": 0.14453125, "train/clean_loss_scale": 7.289440155029297, "train/clean_loss_weight": 1.4578880071640015, "train/combined_loss": 1.264262318611145, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.017535545023696683, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0535519123077393 }, { "epoch": 0.17535545023696683, "step": 74, "train/alpha": 0.2, "train/clean_ce_loss": 0.1259765625, "train/clean_loss_scale": 6.602758884429932, "train/clean_loss_weight": 1.3205517530441284, "train/combined_loss": 0.9981514811515808, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.017535545023696683, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8317928910255432 }, { "epoch": 0.17772511848341233, "grad_norm": 15.932074546813965, "learning_rate": 1.1653543307086612e-06, "loss": 4.0569, "step": 75 }, { "epoch": 0.17772511848341233, "step": 75, "train/alpha": 0.2, "train/clean_ce_loss": 0.12109375, "train/clean_loss_scale": 7.355785846710205, "train/clean_loss_weight": 1.471157193183899, "train/combined_loss": 1.0688875913619995, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.017772511848341232, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8907396793365479 }, { "epoch": 0.17772511848341233, "step": 75, "train/alpha": 0.2, "train/clean_ce_loss": 0.11572265625, "train/clean_loss_scale": 6.605693817138672, "train/clean_loss_weight": 1.3211387395858765, "train/combined_loss": 0.9173141121864319, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.017772511848341232, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.764428436756134 }, { "epoch": 0.17772511848341233, "step": 75, "train/alpha": 0.2, "train/clean_ce_loss": 0.10302734375, "train/clean_loss_scale": 8.088851928710938, "train/clean_loss_weight": 1.6177704334259033, "train/combined_loss": 1.0000475645065308, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.017772511848341232, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.833372950553894 }, { "epoch": 0.17772511848341233, "step": 75, "train/alpha": 0.2, "train/clean_ce_loss": 0.1201171875, "train/clean_loss_scale": 7.8877949714660645, "train/clean_loss_weight": 1.577558994293213, "train/combined_loss": 1.1369516849517822, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.017772511848341232, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9474597573280334 }, { "epoch": 0.18009478672985782, "step": 76, "train/alpha": 0.2, "train/clean_ce_loss": 0.0908203125, "train/clean_loss_scale": 8.182280540466309, "train/clean_loss_weight": 1.6364561319351196, "train/combined_loss": 0.8917407393455505, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.01800947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7431172728538513 }, { "epoch": 0.18009478672985782, "step": 76, "train/alpha": 0.2, "train/clean_ce_loss": 0.12890625, "train/clean_loss_scale": 11.775530815124512, "train/clean_loss_weight": 2.3551061153411865, "train/combined_loss": 1.8215274810791016, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.01800947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.517939567565918 }, { "epoch": 0.18009478672985782, "step": 76, "train/alpha": 0.2, "train/clean_ce_loss": 0.171875, "train/clean_loss_scale": 7.666956901550293, "train/clean_loss_weight": 1.5333913564682007, "train/combined_loss": 1.5813097953796387, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.01800947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3177582025527954 }, { "epoch": 0.18009478672985782, "step": 76, "train/alpha": 0.2, "train/clean_ce_loss": 0.107421875, "train/clean_loss_scale": 6.444835662841797, "train/clean_loss_weight": 1.2889671325683594, "train/combined_loss": 0.8307796120643616, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.01800947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6923163533210754 }, { "epoch": 0.18246445497630331, "step": 77, "train/alpha": 0.2, "train/clean_ce_loss": 0.09619140625, "train/clean_loss_scale": 7.504505634307861, "train/clean_loss_weight": 1.5009011030197144, "train/combined_loss": 0.8662427067756653, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.018246445497630333, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7218689322471619 }, { "epoch": 0.18246445497630331, "step": 77, "train/alpha": 0.2, "train/clean_ce_loss": 0.126953125, "train/clean_loss_scale": 8.313920021057129, "train/clean_loss_weight": 1.6627839803695679, "train/combined_loss": 1.2665736675262451, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.018246445497630333, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0554780960083008 }, { "epoch": 0.18246445497630331, "step": 77, "train/alpha": 0.2, "train/clean_ce_loss": 0.1708984375, "train/clean_loss_scale": 8.492825508117676, "train/clean_loss_weight": 1.698565125465393, "train/combined_loss": 1.7416926622390747, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.018246445497630333, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4514105319976807 }, { "epoch": 0.18246445497630331, "step": 77, "train/alpha": 0.2, "train/clean_ce_loss": 0.1240234375, "train/clean_loss_scale": 7.74900484085083, "train/clean_loss_weight": 1.549800992012024, "train/combined_loss": 1.15326988697052, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.018246445497630333, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9610581994056702 }, { "epoch": 0.1848341232227488, "step": 78, "train/alpha": 0.2, "train/clean_ce_loss": 0.0751953125, "train/clean_loss_scale": 8.241544723510742, "train/clean_loss_weight": 1.6483089923858643, "train/combined_loss": 0.743670642375946, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.01848341232227488, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6197255253791809 }, { "epoch": 0.1848341232227488, "step": 78, "train/alpha": 0.2, "train/clean_ce_loss": 0.111328125, "train/clean_loss_scale": 7.4233574867248535, "train/clean_loss_weight": 1.4846714735031128, "train/combined_loss": 0.9917141795158386, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.01848341232227488, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8264284729957581 }, { "epoch": 0.1848341232227488, "step": 78, "train/alpha": 0.2, "train/clean_ce_loss": 0.08642578125, "train/clean_loss_scale": 8.484142303466797, "train/clean_loss_weight": 1.6968284845352173, "train/combined_loss": 0.8798983693122864, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.01848341232227488, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7332486510276794 }, { "epoch": 0.1848341232227488, "step": 78, "train/alpha": 0.2, "train/clean_ce_loss": 0.16015625, "train/clean_loss_scale": 8.014455795288086, "train/clean_loss_weight": 1.602891206741333, "train/combined_loss": 1.5402781963348389, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.01848341232227488, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2835651636123657 }, { "epoch": 0.1872037914691943, "step": 79, "train/alpha": 0.2, "train/clean_ce_loss": 0.189453125, "train/clean_loss_scale": 7.999196529388428, "train/clean_loss_weight": 1.5998393297195435, "train/combined_loss": 1.8185672760009766, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.01872037914691943, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5154727697372437 }, { "epoch": 0.1872037914691943, "step": 79, "train/alpha": 0.2, "train/clean_ce_loss": 0.0849609375, "train/clean_loss_scale": 6.992909908294678, "train/clean_loss_weight": 1.3985819816589355, "train/combined_loss": 0.7129490375518799, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.01872037914691943, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5941241979598999 }, { "epoch": 0.1872037914691943, "step": 79, "train/alpha": 0.2, "train/clean_ce_loss": 0.09228515625, "train/clean_loss_scale": 8.027215003967285, "train/clean_loss_weight": 1.605443000793457, "train/combined_loss": 0.8889513611793518, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.01872037914691943, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7407928109169006 }, { "epoch": 0.1872037914691943, "step": 79, "train/alpha": 0.2, "train/clean_ce_loss": 0.0859375, "train/clean_loss_scale": 8.379802703857422, "train/clean_loss_weight": 1.6759605407714844, "train/combined_loss": 0.8641672134399414, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.01872037914691943, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7201393246650696 }, { "epoch": 0.1895734597156398, "grad_norm": 14.902511596679688, "learning_rate": 1.2440944881889764e-06, "loss": 4.1383, "step": 80 }, { "epoch": 0.1895734597156398, "step": 80, "train/alpha": 0.2, "train/clean_ce_loss": 0.1328125, "train/clean_loss_scale": 9.980281829833984, "train/clean_loss_weight": 1.9960564374923706, "train/combined_loss": 1.5906074047088623, "train/pos_loss/pos0": 0.6090853810310364, "train/pos_loss/pos1": 1.0862692594528198, "train/pos_loss/pos2": 1.7085233926773071, "train/pos_loss/pos3": 2.360884428024292, "train/pos_loss/pos4": 2.8591461181640625, "train/pos_loss/pos5": 3.2703652381896973, "train/pos_loss/pos6": 3.4950530529022217, "train/pos_loss/weight0": 2.5662026405334473, "train/pos_loss/weight1": 1.870182752609253, "train/pos_loss/weight2": 1.2143046855926514, "train/pos_loss/weight3": 0.7291612029075623, "train/pos_loss/weight4": 0.3749952018260956, "train/pos_loss/weight5": 0.1679297387599945, "train/pos_loss/weight6": 0.07722317427396774, "train/pos_weight_progress": 0.018957345971563982, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3255062103271484 }, { "epoch": 0.1895734597156398, "step": 80, "train/alpha": 0.2, "train/clean_ce_loss": 0.1279296875, "train/clean_loss_scale": 5.404772758483887, "train/clean_loss_weight": 1.0809545516967773, "train/combined_loss": 0.8297170400619507, "train/pos_loss/pos0": 0.3498210310935974, "train/pos_loss/pos1": 0.7266690731048584, "train/pos_loss/pos2": 1.0262947082519531, "train/pos_loss/pos3": 1.480296015739441, "train/pos_loss/pos4": 2.008744239807129, "train/pos_loss/pos5": 2.6040773391723633, "train/pos_loss/pos6": 3.122682809829712, "train/pos_loss/weight0": 3.230250835418701, "train/pos_loss/weight1": 2.0063655376434326, "train/pos_loss/weight2": 1.0871285200119019, "train/pos_loss/weight3": 0.47893035411834717, "train/pos_loss/weight4": 0.1521487534046173, "train/pos_loss/weight5": 0.03710799664258957, "train/pos_loss/weight6": 0.008068263530731201, "train/pos_weight_progress": 0.018957345971563982, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6914308667182922 }, { "epoch": 0.1895734597156398, "step": 80, "train/alpha": 0.2, "train/clean_ce_loss": 0.1416015625, "train/clean_loss_scale": 6.173111915588379, "train/clean_loss_weight": 1.2346223592758179, "train/combined_loss": 1.048946738243103, "train/pos_loss/pos0": 0.4126040041446686, "train/pos_loss/pos1": 0.7146595120429993, "train/pos_loss/pos2": 1.3065600395202637, "train/pos_loss/pos3": 1.8333133459091187, "train/pos_loss/pos4": 2.3859620094299316, "train/pos_loss/pos5": 2.5974342823028564, "train/pos_loss/pos6": 3.029127836227417, "train/pos_loss/weight0": 2.8944203853607178, "train/pos_loss/weight1": 1.9985671043395996, "train/pos_loss/weight2": 1.1482703685760498, "train/pos_loss/weight3": 0.588324785232544, "train/pos_loss/weight4": 0.25352704524993896, "train/pos_loss/weight5": 0.08635151386260986, "train/pos_loss/weight6": 0.030538465827703476, "train/pos_weight_progress": 0.018957345971563982, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8741223216056824 }, { "epoch": 0.1895734597156398, "step": 80, "train/alpha": 0.2, "train/clean_ce_loss": 0.06884765625, "train/clean_loss_scale": 6.114880561828613, "train/clean_loss_weight": 1.2229760885238647, "train/combined_loss": 0.5051942467689514, "train/pos_loss/pos0": 0.16797006130218506, "train/pos_loss/pos1": 0.4165171682834625, "train/pos_loss/pos2": 0.5998693108558655, "train/pos_loss/pos3": 0.7921848297119141, "train/pos_loss/pos4": 1.175559401512146, "train/pos_loss/pos5": 1.46115243434906, "train/pos_loss/pos6": 1.8357841968536377, "train/pos_loss/weight0": 2.9234626293182373, "train/pos_loss/weight1": 1.9600260257720947, "train/pos_loss/weight2": 1.1647725105285645, "train/pos_loss/weight3": 0.5845261216163635, "train/pos_loss/weight4": 0.24561917781829834, "train/pos_loss/weight5": 0.08776795864105225, "train/pos_loss/weight6": 0.03382578492164612, "train/pos_weight_progress": 0.018957345971563982, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.42099520564079285 }, { "epoch": 0.19194312796208532, "step": 81, "train/alpha": 0.2, "train/clean_ce_loss": 0.1044921875, "train/clean_loss_scale": 8.949816703796387, "train/clean_loss_weight": 1.7899633646011353, "train/combined_loss": 1.122223138809204, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.01919431279620853, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9351859092712402 }, { "epoch": 0.19194312796208532, "step": 81, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 10.11292839050293, "train/clean_loss_weight": 2.02258563041687, "train/combined_loss": 0.9895658493041992, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.01919431279620853, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8246381878852844 }, { "epoch": 0.19194312796208532, "step": 81, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 11.377742767333984, "train/clean_loss_weight": 2.2755486965179443, "train/combined_loss": 1.1066632270812988, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.01919431279620853, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9222193956375122 }, { "epoch": 0.19194312796208532, "step": 81, "train/alpha": 0.2, "train/clean_ce_loss": 0.11767578125, "train/clean_loss_scale": 8.541797637939453, "train/clean_loss_weight": 1.7083595991134644, "train/combined_loss": 1.2061952352523804, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.01919431279620853, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0051627159118652 }, { "epoch": 0.1943127962085308, "step": 82, "train/alpha": 0.2, "train/clean_ce_loss": 0.1416015625, "train/clean_loss_scale": 9.148083686828613, "train/clean_loss_weight": 1.8296167850494385, "train/combined_loss": 1.554459571838379, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.01943127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2953829765319824 }, { "epoch": 0.1943127962085308, "step": 82, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 9.738308906555176, "train/clean_loss_weight": 1.9476617574691772, "train/combined_loss": 0.9472026824951172, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.01943127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7893355488777161 }, { "epoch": 0.1943127962085308, "step": 82, "train/alpha": 0.2, "train/clean_ce_loss": 0.083984375, "train/clean_loss_scale": 10.242426872253418, "train/clean_loss_weight": 2.048485517501831, "train/combined_loss": 1.0322445631027222, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.01943127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8602038025856018 }, { "epoch": 0.1943127962085308, "step": 82, "train/alpha": 0.2, "train/clean_ce_loss": 0.11572265625, "train/clean_loss_scale": 9.11135482788086, "train/clean_loss_weight": 1.8222709894180298, "train/combined_loss": 1.2652682065963745, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.01943127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0543901920318604 }, { "epoch": 0.1966824644549763, "step": 83, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 7.885278701782227, "train/clean_loss_weight": 1.577055811882019, "train/combined_loss": 0.6191175580024719, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.01966824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5159313082695007 }, { "epoch": 0.1966824644549763, "step": 83, "train/alpha": 0.2, "train/clean_ce_loss": 0.1484375, "train/clean_loss_scale": 9.683797836303711, "train/clean_loss_weight": 1.9367595911026, "train/combined_loss": 1.724926471710205, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.01966824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.437438726425171 }, { "epoch": 0.1966824644549763, "step": 83, "train/alpha": 0.2, "train/clean_ce_loss": 0.11474609375, "train/clean_loss_scale": 8.44767951965332, "train/clean_loss_weight": 1.6895359754562378, "train/combined_loss": 1.163205862045288, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.01966824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9693382382392883 }, { "epoch": 0.1966824644549763, "step": 83, "train/alpha": 0.2, "train/clean_ce_loss": 0.123046875, "train/clean_loss_scale": 9.314826011657715, "train/clean_loss_weight": 1.8629652261734009, "train/combined_loss": 1.3753923177719116, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.01966824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1461602449417114 }, { "epoch": 0.1990521327014218, "step": 84, "train/alpha": 0.2, "train/clean_ce_loss": 0.107421875, "train/clean_loss_scale": 10.597816467285156, "train/clean_loss_weight": 2.119563341140747, "train/combined_loss": 1.3661247491836548, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.01990521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.138437271118164 }, { "epoch": 0.1990521327014218, "step": 84, "train/alpha": 0.2, "train/clean_ce_loss": 0.177734375, "train/clean_loss_scale": 8.527219772338867, "train/clean_loss_weight": 1.7054439783096313, "train/combined_loss": 1.8186960220336914, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.01990521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5155800580978394 }, { "epoch": 0.1990521327014218, "step": 84, "train/alpha": 0.2, "train/clean_ce_loss": 0.0986328125, "train/clean_loss_scale": 9.19470500946045, "train/clean_loss_weight": 1.838940978050232, "train/combined_loss": 1.0882794857025146, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.01990521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.906899631023407 }, { "epoch": 0.1990521327014218, "step": 84, "train/alpha": 0.2, "train/clean_ce_loss": 0.0859375, "train/clean_loss_scale": 10.47469425201416, "train/clean_loss_weight": 2.0949389934539795, "train/combined_loss": 1.0802029371261597, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.01990521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9001690745353699 }, { "epoch": 0.2014218009478673, "grad_norm": 15.050902366638184, "learning_rate": 1.3228346456692914e-06, "loss": 4.1584, "step": 85 }, { "epoch": 0.2014218009478673, "step": 85, "train/alpha": 0.2, "train/clean_ce_loss": 0.06884765625, "train/clean_loss_scale": 11.28862476348877, "train/clean_loss_weight": 2.2577250003814697, "train/combined_loss": 0.9326344132423401, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.02014218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7771953344345093 }, { "epoch": 0.2014218009478673, "step": 85, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 11.327113151550293, "train/clean_loss_weight": 2.2654225826263428, "train/combined_loss": 1.1349236965179443, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.02014218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.945769727230072 }, { "epoch": 0.2014218009478673, "step": 85, "train/alpha": 0.2, "train/clean_ce_loss": 0.10546875, "train/clean_loss_scale": 8.776082992553711, "train/clean_loss_weight": 1.7552165985107422, "train/combined_loss": 1.1107230186462402, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.02014218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9256024956703186 }, { "epoch": 0.2014218009478673, "step": 85, "train/alpha": 0.2, "train/clean_ce_loss": 0.1416015625, "train/clean_loss_scale": 7.927305221557617, "train/clean_loss_weight": 1.5854610204696655, "train/combined_loss": 1.347022533416748, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.02014218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.12251877784729 }, { "epoch": 0.2037914691943128, "step": 86, "train/alpha": 0.2, "train/clean_ce_loss": 0.039794921875, "train/clean_loss_scale": 25.336671829223633, "train/clean_loss_weight": 5.0673346519470215, "train/combined_loss": 1.2099250555038452, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.02037914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0082708597183228 }, { "epoch": 0.2037914691943128, "step": 86, "train/alpha": 0.2, "train/clean_ce_loss": 0.11328125, "train/clean_loss_scale": 8.959795951843262, "train/clean_loss_weight": 1.7919591665267944, "train/combined_loss": 1.2179722785949707, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.02037914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0149768590927124 }, { "epoch": 0.2037914691943128, "step": 86, "train/alpha": 0.2, "train/clean_ce_loss": 0.052978515625, "train/clean_loss_scale": 10.93168830871582, "train/clean_loss_weight": 2.18633770942688, "train/combined_loss": 0.6949735283851624, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.02037914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5791445970535278 }, { "epoch": 0.2037914691943128, "step": 86, "train/alpha": 0.2, "train/clean_ce_loss": 0.11669921875, "train/clean_loss_scale": 11.239558219909668, "train/clean_loss_weight": 2.2479116916656494, "train/combined_loss": 1.5739772319793701, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.02037914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.311647653579712 }, { "epoch": 0.20616113744075829, "step": 87, "train/alpha": 0.2, "train/clean_ce_loss": 0.1044921875, "train/clean_loss_scale": 14.282509803771973, "train/clean_loss_weight": 2.856502056121826, "train/combined_loss": 1.7908928394317627, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.02061611374407583, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.492410659790039 }, { "epoch": 0.20616113744075829, "step": 87, "train/alpha": 0.2, "train/clean_ce_loss": 0.12255859375, "train/clean_loss_scale": 10.249510765075684, "train/clean_loss_weight": 2.0499022006988525, "train/combined_loss": 1.5073987245559692, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.02061611374407583, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.256165623664856 }, { "epoch": 0.20616113744075829, "step": 87, "train/alpha": 0.2, "train/clean_ce_loss": 0.0703125, "train/clean_loss_scale": 9.601322174072266, "train/clean_loss_weight": 1.920264482498169, "train/combined_loss": 0.8101115226745605, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.02061611374407583, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6750929355621338 }, { "epoch": 0.20616113744075829, "step": 87, "train/alpha": 0.2, "train/clean_ce_loss": 0.11572265625, "train/clean_loss_scale": 8.11049747467041, "train/clean_loss_weight": 1.62209951877594, "train/combined_loss": 1.126281976699829, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.02061611374407583, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9385683536529541 }, { "epoch": 0.20853080568720378, "step": 88, "train/alpha": 0.2, "train/clean_ce_loss": 0.1240234375, "train/clean_loss_scale": 9.688337326049805, "train/clean_loss_weight": 1.9376674890518188, "train/combined_loss": 1.4418970346450806, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.02085308056872038, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2015808820724487 }, { "epoch": 0.20853080568720378, "step": 88, "train/alpha": 0.2, "train/clean_ce_loss": 0.09033203125, "train/clean_loss_scale": 11.663347244262695, "train/clean_loss_weight": 2.332669496536255, "train/combined_loss": 1.2642886638641357, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.02085308056872038, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0535738468170166 }, { "epoch": 0.20853080568720378, "step": 88, "train/alpha": 0.2, "train/clean_ce_loss": 0.1142578125, "train/clean_loss_scale": 11.941779136657715, "train/clean_loss_weight": 2.3883559703826904, "train/combined_loss": 1.6373298168182373, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.02085308056872038, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3644415140151978 }, { "epoch": 0.20853080568720378, "step": 88, "train/alpha": 0.2, "train/clean_ce_loss": 0.12890625, "train/clean_loss_scale": 13.791642189025879, "train/clean_loss_weight": 2.758328437805176, "train/combined_loss": 2.133394718170166, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.02085308056872038, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.7778289318084717 }, { "epoch": 0.2109004739336493, "step": 89, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 11.251127243041992, "train/clean_loss_weight": 2.25022554397583, "train/combined_loss": 0.8767968416213989, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.021090047393364928, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7306640148162842 }, { "epoch": 0.2109004739336493, "step": 89, "train/alpha": 0.2, "train/clean_ce_loss": 0.09375, "train/clean_loss_scale": 9.425854682922363, "train/clean_loss_weight": 1.8851709365844727, "train/combined_loss": 1.0604087114334106, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.021090047393364928, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.883673906326294 }, { "epoch": 0.2109004739336493, "step": 89, "train/alpha": 0.2, "train/clean_ce_loss": 0.06298828125, "train/clean_loss_scale": 15.470813751220703, "train/clean_loss_weight": 3.094162702560425, "train/combined_loss": 1.169376015663147, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.021090047393364928, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9744799733161926 }, { "epoch": 0.2109004739336493, "step": 89, "train/alpha": 0.2, "train/clean_ce_loss": 0.1435546875, "train/clean_loss_scale": 9.861950874328613, "train/clean_loss_weight": 1.9723901748657227, "train/combined_loss": 1.6988751888275146, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.021090047393364928, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.415729284286499 }, { "epoch": 0.2132701421800948, "grad_norm": 16.6069278717041, "learning_rate": 1.4015748031496062e-06, "loss": 4.2287, "step": 90 }, { "epoch": 0.2132701421800948, "step": 90, "train/alpha": 0.2, "train/clean_ce_loss": 0.1689453125, "train/clean_loss_scale": 9.36862564086914, "train/clean_loss_weight": 1.873725175857544, "train/combined_loss": 1.89934241771698, "train/pos_loss/pos0": 0.6682514548301697, "train/pos_loss/pos1": 1.2602189779281616, "train/pos_loss/pos2": 1.7150862216949463, "train/pos_loss/pos3": 2.5988152027130127, "train/pos_loss/pos4": 3.0589568614959717, "train/pos_loss/pos5": 3.3034791946411133, "train/pos_loss/pos6": 3.73219633102417, "train/pos_loss/weight0": 2.2476725578308105, "train/pos_loss/weight1": 1.7322512865066528, "train/pos_loss/weight2": 1.2246112823486328, "train/pos_loss/weight3": 0.8112246990203857, "train/pos_loss/weight4": 0.5061615705490112, "train/pos_loss/weight5": 0.29913610219955444, "train/pos_loss/weight6": 0.1789426952600479, "train/pos_weight_progress": 0.02132701421800948, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5827853679656982 }, { "epoch": 0.2132701421800948, "step": 90, "train/alpha": 0.2, "train/clean_ce_loss": 0.1591796875, "train/clean_loss_scale": 7.311239719390869, "train/clean_loss_weight": 1.4622479677200317, "train/combined_loss": 1.3965610265731812, "train/pos_loss/pos0": 0.6828587651252747, "train/pos_loss/pos1": 1.2699092626571655, "train/pos_loss/pos2": 1.818293809890747, "train/pos_loss/pos3": 2.4195215702056885, "train/pos_loss/pos4": 2.8192126750946045, "train/pos_loss/pos5": 3.4089927673339844, "train/pos_loss/pos6": 3.96476674079895, "train/pos_loss/weight0": 3.376246213912964, "train/pos_loss/weight1": 2.078134775161743, "train/pos_loss/weight2": 1.0015428066253662, "train/pos_loss/weight3": 0.4017121195793152, "train/pos_loss/weight4": 0.11157167702913284, "train/pos_loss/weight5": 0.02496238797903061, "train/pos_loss/weight6": 0.005829824600368738, "train/pos_weight_progress": 0.02132701421800948, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.163800835609436 }, { "epoch": 0.2132701421800948, "step": 90, "train/alpha": 0.2, "train/clean_ce_loss": 0.051025390625, "train/clean_loss_scale": 5.2425923347473145, "train/clean_loss_weight": 1.0485185384750366, "train/combined_loss": 0.3210063874721527, "train/pos_loss/pos0": 0.10172738879919052, "train/pos_loss/pos1": 0.2656137943267822, "train/pos_loss/pos2": 0.4875662624835968, "train/pos_loss/pos3": 0.9092140793800354, "train/pos_loss/pos4": 1.183670997619629, "train/pos_loss/pos5": 1.4902219772338867, "train/pos_loss/pos6": 1.7525466680526733, "train/pos_loss/weight0": 3.445960760116577, "train/pos_loss/weight1": 2.090618848800659, "train/pos_loss/weight2": 0.9648615717887878, "train/pos_loss/weight3": 0.3713124692440033, "train/pos_loss/weight4": 0.10154716670513153, "train/pos_loss/weight5": 0.02134627476334572, "train/pos_loss/weight6": 0.004353215917944908, "train/pos_weight_progress": 0.02132701421800948, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.26750531792640686 }, { "epoch": 0.2132701421800948, "step": 90, "train/alpha": 0.2, "train/clean_ce_loss": 0.09423828125, "train/clean_loss_scale": 9.197702407836914, "train/clean_loss_weight": 1.8395404815673828, "train/combined_loss": 1.040130853652954, "train/pos_loss/pos0": 0.14444851875305176, "train/pos_loss/pos1": 0.40011274814605713, "train/pos_loss/pos2": 0.8269469141960144, "train/pos_loss/pos3": 1.407821536064148, "train/pos_loss/pos4": 1.9935126304626465, "train/pos_loss/pos5": 2.728031635284424, "train/pos_loss/pos6": 3.2227303981781006, "train/pos_loss/weight0": 2.092559337615967, "train/pos_loss/weight1": 1.6511316299438477, "train/pos_loss/weight2": 1.2194342613220215, "train/pos_loss/weight3": 0.8759660124778748, "train/pos_loss/weight4": 0.5651042461395264, "train/pos_loss/weight5": 0.35969558358192444, "train/pos_loss/weight6": 0.23610961437225342, "train/pos_weight_progress": 0.02132701421800948, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8667756915092468 }, { "epoch": 0.2156398104265403, "step": 91, "train/alpha": 0.2, "train/clean_ce_loss": 0.185546875, "train/clean_loss_scale": 6.754693984985352, "train/clean_loss_weight": 1.3509387969970703, "train/combined_loss": 1.5039747953414917, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.02156398104265403, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.253312349319458 }, { "epoch": 0.2156398104265403, "step": 91, "train/alpha": 0.2, "train/clean_ce_loss": 0.12255859375, "train/clean_loss_scale": 8.63094425201416, "train/clean_loss_weight": 1.7261888980865479, "train/combined_loss": 1.2693556547164917, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.02156398104265403, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0577963590621948 }, { "epoch": 0.2156398104265403, "step": 91, "train/alpha": 0.2, "train/clean_ce_loss": 0.0888671875, "train/clean_loss_scale": 9.249427795410156, "train/clean_loss_weight": 1.8498855829238892, "train/combined_loss": 0.9863647818565369, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.02156398104265403, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8219706416130066 }, { "epoch": 0.2156398104265403, "step": 91, "train/alpha": 0.2, "train/clean_ce_loss": 0.056640625, "train/clean_loss_scale": 8.472511291503906, "train/clean_loss_weight": 1.6945022344589233, "train/combined_loss": 0.5758659839630127, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.02156398104265403, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.47988831996917725 }, { "epoch": 0.21800947867298578, "step": 92, "train/alpha": 0.2, "train/clean_ce_loss": 0.11328125, "train/clean_loss_scale": 10.683588027954102, "train/clean_loss_weight": 2.1367175579071045, "train/combined_loss": 1.4523003101348877, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.021800947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2102502584457397 }, { "epoch": 0.21800947867298578, "step": 92, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 7.092349529266357, "train/clean_loss_weight": 1.4184699058532715, "train/combined_loss": 0.5610159039497375, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.021800947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4675132632255554 }, { "epoch": 0.21800947867298578, "step": 92, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 8.56579875946045, "train/clean_loss_weight": 1.7131597995758057, "train/combined_loss": 0.813081681728363, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.021800947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6775680780410767 }, { "epoch": 0.21800947867298578, "step": 92, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 9.837249755859375, "train/clean_loss_weight": 1.9674500226974487, "train/combined_loss": 0.9395341873168945, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.021800947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7829451560974121 }, { "epoch": 0.22037914691943128, "step": 93, "train/alpha": 0.2, "train/clean_ce_loss": 0.09130859375, "train/clean_loss_scale": 8.397583961486816, "train/clean_loss_weight": 1.6795167922973633, "train/combined_loss": 0.9201258420944214, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.022037914691943127, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7667715549468994 }, { "epoch": 0.22037914691943128, "step": 93, "train/alpha": 0.2, "train/clean_ce_loss": 0.061279296875, "train/clean_loss_scale": 12.132441520690918, "train/clean_loss_weight": 2.4264883995056152, "train/combined_loss": 0.8921610116958618, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.022037914691943127, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7434675097465515 }, { "epoch": 0.22037914691943128, "step": 93, "train/alpha": 0.2, "train/clean_ce_loss": 0.10791015625, "train/clean_loss_scale": 7.890707969665527, "train/clean_loss_weight": 1.5781415700912476, "train/combined_loss": 1.021785020828247, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.022037914691943127, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8514875173568726 }, { "epoch": 0.22037914691943128, "step": 93, "train/alpha": 0.2, "train/clean_ce_loss": 0.12255859375, "train/clean_loss_scale": 7.635255813598633, "train/clean_loss_weight": 1.5270512104034424, "train/combined_loss": 1.1229194402694702, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.022037914691943127, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9357662200927734 }, { "epoch": 0.22274881516587677, "step": 94, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 9.926922798156738, "train/clean_loss_weight": 1.9853845834732056, "train/combined_loss": 0.9422821402549744, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.02227488151658768, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7852351069450378 }, { "epoch": 0.22274881516587677, "step": 94, "train/alpha": 0.2, "train/clean_ce_loss": 0.08203125, "train/clean_loss_scale": 6.207318305969238, "train/clean_loss_weight": 1.2414636611938477, "train/combined_loss": 0.6110329031944275, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.02227488151658768, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5091940760612488 }, { "epoch": 0.22274881516587677, "step": 94, "train/alpha": 0.2, "train/clean_ce_loss": 0.125, "train/clean_loss_scale": 6.466464519500732, "train/clean_loss_weight": 1.2932928800582886, "train/combined_loss": 0.9699696898460388, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.02227488151658768, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8083080649375916 }, { "epoch": 0.22274881516587677, "step": 94, "train/alpha": 0.2, "train/clean_ce_loss": 0.177734375, "train/clean_loss_scale": 8.947588920593262, "train/clean_loss_weight": 1.7895177602767944, "train/combined_loss": 1.9083529710769653, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.02227488151658768, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5902941226959229 }, { "epoch": 0.22511848341232227, "grad_norm": 14.591339111328125, "learning_rate": 1.4803149606299211e-06, "loss": 4.137, "step": 95 }, { "epoch": 0.22511848341232227, "step": 95, "train/alpha": 0.2, "train/clean_ce_loss": 0.09814453125, "train/clean_loss_scale": 7.9493865966796875, "train/clean_loss_weight": 1.5898773670196533, "train/combined_loss": 0.9362265467643738, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.022511848341232227, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.780188798904419 }, { "epoch": 0.22511848341232227, "step": 95, "train/alpha": 0.2, "train/clean_ce_loss": 0.057861328125, "train/clean_loss_scale": 9.159346580505371, "train/clean_loss_weight": 1.83186936378479, "train/combined_loss": 0.6359663605690002, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.022511848341232227, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5299719572067261 }, { "epoch": 0.22511848341232227, "step": 95, "train/alpha": 0.2, "train/clean_ce_loss": 0.1279296875, "train/clean_loss_scale": 9.256431579589844, "train/clean_loss_weight": 1.8512862920761108, "train/combined_loss": 1.4210069179534912, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.022511848341232227, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1841723918914795 }, { "epoch": 0.22511848341232227, "step": 95, "train/alpha": 0.2, "train/clean_ce_loss": 0.09130859375, "train/clean_loss_scale": 6.2376017570495605, "train/clean_loss_weight": 1.2475203275680542, "train/combined_loss": 0.6834559440612793, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.022511848341232227, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.569546639919281 }, { "epoch": 0.22748815165876776, "step": 96, "train/alpha": 0.2, "train/clean_ce_loss": 0.11083984375, "train/clean_loss_scale": 8.644017219543457, "train/clean_loss_weight": 1.7288035154342651, "train/combined_loss": 1.1497218608856201, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.022748815165876776, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9581015110015869 }, { "epoch": 0.22748815165876776, "step": 96, "train/alpha": 0.2, "train/clean_ce_loss": 0.11572265625, "train/clean_loss_scale": 6.883800506591797, "train/clean_loss_weight": 1.3767601251602173, "train/combined_loss": 0.9559339880943298, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.022748815165876776, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7966116666793823 }, { "epoch": 0.22748815165876776, "step": 96, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 8.20400619506836, "train/clean_loss_weight": 1.6408013105392456, "train/combined_loss": 0.6489497423171997, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.022748815165876776, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5407914519309998 }, { "epoch": 0.22748815165876776, "step": 96, "train/alpha": 0.2, "train/clean_ce_loss": 0.150390625, "train/clean_loss_scale": 7.8117194175720215, "train/clean_loss_weight": 1.562343955039978, "train/combined_loss": 1.409771203994751, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.022748815165876776, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1748093366622925 }, { "epoch": 0.22985781990521326, "step": 97, "train/alpha": 0.2, "train/clean_ce_loss": 0.12353515625, "train/clean_loss_scale": 7.347040176391602, "train/clean_loss_weight": 1.4694080352783203, "train/combined_loss": 1.0891412496566772, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.02298578199052133, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9076177477836609 }, { "epoch": 0.22985781990521326, "step": 97, "train/alpha": 0.2, "train/clean_ce_loss": 0.1005859375, "train/clean_loss_scale": 8.403892517089844, "train/clean_loss_weight": 1.6807785034179688, "train/combined_loss": 1.014376163482666, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.02298578199052133, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8453134298324585 }, { "epoch": 0.22985781990521326, "step": 97, "train/alpha": 0.2, "train/clean_ce_loss": 0.14453125, "train/clean_loss_scale": 9.458650588989258, "train/clean_loss_weight": 1.8917301893234253, "train/combined_loss": 1.6404846906661987, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.02298578199052133, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3670705556869507 }, { "epoch": 0.22985781990521326, "step": 97, "train/alpha": 0.2, "train/clean_ce_loss": 0.08544921875, "train/clean_loss_scale": 9.11845874786377, "train/clean_loss_weight": 1.823691725730896, "train/combined_loss": 0.9349982738494873, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.02298578199052133, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7791652083396912 }, { "epoch": 0.23222748815165878, "step": 98, "train/alpha": 0.2, "train/clean_ce_loss": 0.078125, "train/clean_loss_scale": 9.215999603271484, "train/clean_loss_weight": 1.8431999683380127, "train/combined_loss": 0.8639999628067017, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.023222748815165877, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7199999690055847 }, { "epoch": 0.23222748815165878, "step": 98, "train/alpha": 0.2, "train/clean_ce_loss": 0.05908203125, "train/clean_loss_scale": 10.831622123718262, "train/clean_loss_weight": 2.1663243770599365, "train/combined_loss": 0.7679450511932373, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.023222748815165877, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6399542093276978 }, { "epoch": 0.23222748815165878, "step": 98, "train/alpha": 0.2, "train/clean_ce_loss": 0.1787109375, "train/clean_loss_scale": 6.792318344116211, "train/clean_loss_weight": 1.3584636449813843, "train/combined_loss": 1.4566339254379272, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.023222748815165877, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2138615846633911 }, { "epoch": 0.23222748815165878, "step": 98, "train/alpha": 0.2, "train/clean_ce_loss": 0.1005859375, "train/clean_loss_scale": 7.842259883880615, "train/clean_loss_weight": 1.568452000617981, "train/combined_loss": 0.9465852379798889, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.023222748815165877, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7888210415840149 }, { "epoch": 0.23459715639810427, "step": 99, "train/alpha": 0.2, "train/clean_ce_loss": 0.11767578125, "train/clean_loss_scale": 7.60653018951416, "train/clean_loss_weight": 1.521306037902832, "train/combined_loss": 1.0741252899169922, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.023459715639810426, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8951044082641602 }, { "epoch": 0.23459715639810427, "step": 99, "train/alpha": 0.2, "train/clean_ce_loss": 0.072265625, "train/clean_loss_scale": 8.370773315429688, "train/clean_loss_weight": 1.6741546392440796, "train/combined_loss": 0.7259030342102051, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.023459715639810426, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6049191951751709 }, { "epoch": 0.23459715639810427, "step": 99, "train/alpha": 0.2, "train/clean_ce_loss": 0.08251953125, "train/clean_loss_scale": 8.412933349609375, "train/clean_loss_weight": 1.682586669921875, "train/combined_loss": 0.833077609539032, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.023459715639810426, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6942313313484192 }, { "epoch": 0.23459715639810427, "step": 99, "train/alpha": 0.2, "train/clean_ce_loss": 0.10302734375, "train/clean_loss_scale": 7.838932514190674, "train/clean_loss_weight": 1.5677865743637085, "train/combined_loss": 0.9691492915153503, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.023459715639810426, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8076243996620178 }, { "epoch": 0.23696682464454977, "grad_norm": 14.4638032913208, "learning_rate": 1.5590551181102361e-06, "loss": 4.203, "step": 100 }, { "epoch": 0.23696682464454977, "step": 100, "train/alpha": 0.2, "train/clean_ce_loss": 0.07763671875, "train/clean_loss_scale": 7.4226250648498535, "train/clean_loss_weight": 1.4845250844955444, "train/combined_loss": 0.6915218830108643, "train/pos_loss/pos0": 0.19209738075733185, "train/pos_loss/pos1": 0.4300755262374878, "train/pos_loss/pos2": 0.753320038318634, "train/pos_loss/pos3": 1.2175244092941284, "train/pos_loss/pos4": 1.6757245063781738, "train/pos_loss/pos5": 2.092836618423462, "train/pos_loss/pos6": 2.5155463218688965, "train/pos_loss/weight0": 2.648097515106201, "train/pos_loss/weight1": 1.9321414232254028, "train/pos_loss/weight2": 1.231208324432373, "train/pos_loss/weight3": 0.6898121237754822, "train/pos_loss/weight4": 0.3196031451225281, "train/pos_loss/weight5": 0.12782977521419525, "train/pos_loss/weight6": 0.05130789428949356, "train/pos_weight_progress": 0.023696682464454975, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5762682557106018 }, { "epoch": 0.23696682464454977, "step": 100, "train/alpha": 0.2, "train/clean_ce_loss": 0.1279296875, "train/clean_loss_scale": 7.897039890289307, "train/clean_loss_weight": 1.579408049583435, "train/combined_loss": 1.212319016456604, "train/pos_loss/pos0": 0.4633263945579529, "train/pos_loss/pos1": 0.6956936717033386, "train/pos_loss/pos2": 1.2472858428955078, "train/pos_loss/pos3": 1.7742102146148682, "train/pos_loss/pos4": 2.4646999835968018, "train/pos_loss/pos5": 2.9064295291900635, "train/pos_loss/pos6": 3.112025260925293, "train/pos_loss/weight0": 2.515756130218506, "train/pos_loss/weight1": 1.853196382522583, "train/pos_loss/weight2": 1.2260655164718628, "train/pos_loss/weight3": 0.7422369122505188, "train/pos_loss/weight4": 0.3898093104362488, "train/pos_loss/weight5": 0.1791944056749344, "train/pos_loss/weight6": 0.093741774559021, "train/pos_weight_progress": 0.023696682464454975, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.010265827178955 }, { "epoch": 0.23696682464454977, "step": 100, "train/alpha": 0.2, "train/clean_ce_loss": 0.1328125, "train/clean_loss_scale": 9.233397483825684, "train/clean_loss_weight": 1.8466795682907104, "train/combined_loss": 1.471572756767273, "train/pos_loss/pos0": 0.6814654469490051, "train/pos_loss/pos1": 1.1876016855239868, "train/pos_loss/pos2": 1.8399708271026611, "train/pos_loss/pos3": 2.3136932849884033, "train/pos_loss/pos4": 3.043595314025879, "train/pos_loss/pos5": 3.4922471046447754, "train/pos_loss/pos6": 3.909515142440796, "train/pos_loss/weight0": 3.0975472927093506, "train/pos_loss/weight1": 1.9651620388031006, "train/pos_loss/weight2": 1.1525077819824219, "train/pos_loss/weight3": 0.5219614505767822, "train/pos_loss/weight4": 0.19594524800777435, "train/pos_loss/weight5": 0.05256598815321922, "train/pos_loss/weight6": 0.014309978112578392, "train/pos_weight_progress": 0.023696682464454975, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2263106107711792 }, { "epoch": 0.23696682464454977, "step": 100, "train/alpha": 0.2, "train/clean_ce_loss": 0.09521484375, "train/clean_loss_scale": 7.282999515533447, "train/clean_loss_weight": 1.4565999507904053, "train/combined_loss": 0.8321396112442017, "train/pos_loss/pos0": 0.4252159595489502, "train/pos_loss/pos1": 0.7021850943565369, "train/pos_loss/pos2": 1.0785844326019287, "train/pos_loss/pos3": 1.5251505374908447, "train/pos_loss/pos4": 2.0075323581695557, "train/pos_loss/pos5": 2.5441291332244873, "train/pos_loss/pos6": 3.1059176921844482, "train/pos_loss/weight0": 3.411540985107422, "train/pos_loss/weight1": 2.0504462718963623, "train/pos_loss/weight2": 1.0332159996032715, "train/pos_loss/weight3": 0.36851173639297485, "train/pos_loss/weight4": 0.11074701696634293, "train/pos_loss/weight5": 0.02138259820640087, "train/pos_loss/weight6": 0.004155035596340895, "train/pos_weight_progress": 0.023696682464454975, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6934496760368347 }, { "epoch": 0.23933649289099526, "step": 101, "train/alpha": 0.2, "train/clean_ce_loss": 0.140625, "train/clean_loss_scale": 7.17156457901001, "train/clean_loss_weight": 1.4343129396438599, "train/combined_loss": 1.2102015018463135, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.023933649289099527, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0085012912750244 }, { "epoch": 0.23933649289099526, "step": 101, "train/alpha": 0.2, "train/clean_ce_loss": 0.08544921875, "train/clean_loss_scale": 10.02873420715332, "train/clean_loss_weight": 2.005746841430664, "train/combined_loss": 1.028337001800537, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.023933649289099527, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8569475412368774 }, { "epoch": 0.23933649289099526, "step": 101, "train/alpha": 0.2, "train/clean_ce_loss": 0.0673828125, "train/clean_loss_scale": 9.813617706298828, "train/clean_loss_weight": 1.9627236127853394, "train/combined_loss": 0.7935230135917664, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.023933649289099527, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6612691879272461 }, { "epoch": 0.23933649289099526, "step": 101, "train/alpha": 0.2, "train/clean_ce_loss": 0.08984375, "train/clean_loss_scale": 8.380097389221191, "train/clean_loss_weight": 1.676019549369812, "train/combined_loss": 0.9034792184829712, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.023933649289099527, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7528993487358093 }, { "epoch": 0.24170616113744076, "step": 102, "train/alpha": 0.2, "train/clean_ce_loss": 0.1298828125, "train/clean_loss_scale": 7.940077304840088, "train/clean_loss_weight": 1.5880154371261597, "train/combined_loss": 1.2375354766845703, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.024170616113744076, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0312795639038086 }, { "epoch": 0.24170616113744076, "step": 102, "train/alpha": 0.2, "train/clean_ce_loss": 0.08056640625, "train/clean_loss_scale": 7.994319438934326, "train/clean_loss_weight": 1.598863959312439, "train/combined_loss": 0.7728883028030396, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.024170616113744076, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6440736055374146 }, { "epoch": 0.24170616113744076, "step": 102, "train/alpha": 0.2, "train/clean_ce_loss": 0.10595703125, "train/clean_loss_scale": 6.6424479484558105, "train/clean_loss_weight": 1.3284896612167358, "train/combined_loss": 0.844576895236969, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.024170616113744076, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7038140892982483 }, { "epoch": 0.24170616113744076, "step": 102, "train/alpha": 0.2, "train/clean_ce_loss": 0.1572265625, "train/clean_loss_scale": 10.336153984069824, "train/clean_loss_weight": 2.0672309398651123, "train/combined_loss": 1.9501416683197021, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.024170616113744076, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.6251180171966553 }, { "epoch": 0.24407582938388625, "step": 103, "train/alpha": 0.2, "train/clean_ce_loss": 0.1552734375, "train/clean_loss_scale": 9.515609741210938, "train/clean_loss_weight": 1.9031219482421875, "train/combined_loss": 1.7730257511138916, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.024407582938388624, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4775214195251465 }, { "epoch": 0.24407582938388625, "step": 103, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 6.602130889892578, "train/clean_loss_weight": 1.3204262256622314, "train/combined_loss": 0.6305550932884216, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.024407582938388624, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5254625678062439 }, { "epoch": 0.24407582938388625, "step": 103, "train/alpha": 0.2, "train/clean_ce_loss": 0.1064453125, "train/clean_loss_scale": 8.247400283813477, "train/clean_loss_weight": 1.6494801044464111, "train/combined_loss": 1.0534765720367432, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.024407582938388624, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8778970837593079 }, { "epoch": 0.24407582938388625, "step": 103, "train/alpha": 0.2, "train/clean_ce_loss": 0.1162109375, "train/clean_loss_scale": 6.8337812423706055, "train/clean_loss_weight": 1.3667563199996948, "train/combined_loss": 0.9529921412467957, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.024407582938388624, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7941601276397705 }, { "epoch": 0.24644549763033174, "step": 104, "train/alpha": 0.2, "train/clean_ce_loss": 0.056640625, "train/clean_loss_scale": 11.657665252685547, "train/clean_loss_weight": 2.331533193588257, "train/combined_loss": 0.7923569679260254, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.024644549763033177, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6602974534034729 }, { "epoch": 0.24644549763033174, "step": 104, "train/alpha": 0.2, "train/clean_ce_loss": 0.130859375, "train/clean_loss_scale": 10.829360008239746, "train/clean_loss_weight": 2.165872097015381, "train/combined_loss": 1.7005479335784912, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.024644549763033177, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4171233177185059 }, { "epoch": 0.24644549763033174, "step": 104, "train/alpha": 0.2, "train/clean_ce_loss": 0.05029296875, "train/clean_loss_scale": 11.785097122192383, "train/clean_loss_weight": 2.3570194244384766, "train/combined_loss": 0.7112489938735962, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.024644549763033177, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5927075147628784 }, { "epoch": 0.24644549763033174, "step": 104, "train/alpha": 0.2, "train/clean_ce_loss": 0.130859375, "train/clean_loss_scale": 8.374800682067871, "train/clean_loss_weight": 1.6749601364135742, "train/combined_loss": 1.3151054382324219, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.024644549763033177, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0959211587905884 }, { "epoch": 0.24881516587677724, "grad_norm": 14.431876182556152, "learning_rate": 1.637795275590551e-06, "loss": 4.2002, "step": 105 }, { "epoch": 0.24881516587677724, "step": 105, "train/alpha": 0.2, "train/clean_ce_loss": 0.1357421875, "train/clean_loss_scale": 7.108262062072754, "train/clean_loss_weight": 1.4216524362564087, "train/combined_loss": 1.1578692197799683, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.024881516587677725, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9648910164833069 }, { "epoch": 0.24881516587677724, "step": 105, "train/alpha": 0.2, "train/clean_ce_loss": 0.10986328125, "train/clean_loss_scale": 11.597970962524414, "train/clean_loss_weight": 2.319594144821167, "train/combined_loss": 1.529029369354248, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.024881516587677725, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.27419114112854 }, { "epoch": 0.24881516587677724, "step": 105, "train/alpha": 0.2, "train/clean_ce_loss": 0.119140625, "train/clean_loss_scale": 6.869548797607422, "train/clean_loss_weight": 1.373909831047058, "train/combined_loss": 0.9821308255195618, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.024881516587677725, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8184423446655273 }, { "epoch": 0.24881516587677724, "step": 105, "train/alpha": 0.2, "train/clean_ce_loss": 0.11572265625, "train/clean_loss_scale": 7.385566234588623, "train/clean_loss_weight": 1.4771132469177246, "train/combined_loss": 1.0256128311157227, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.024881516587677725, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8546773195266724 }, { "epoch": 0.25118483412322273, "step": 106, "train/alpha": 0.2, "train/clean_ce_loss": 0.08642578125, "train/clean_loss_scale": 10.617032051086426, "train/clean_loss_weight": 2.123406410217285, "train/combined_loss": 1.101102352142334, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.025118483412322274, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9175853133201599 }, { "epoch": 0.25118483412322273, "step": 106, "train/alpha": 0.2, "train/clean_ce_loss": 0.0498046875, "train/clean_loss_scale": 9.464179992675781, "train/clean_loss_weight": 1.8928359746932983, "train/combined_loss": 0.5656326413154602, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.025118483412322274, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.47136053442955017 }, { "epoch": 0.25118483412322273, "step": 106, "train/alpha": 0.2, "train/clean_ce_loss": 0.0986328125, "train/clean_loss_scale": 7.875470161437988, "train/clean_loss_weight": 1.5750941038131714, "train/combined_loss": 0.9321357011795044, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.025118483412322274, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7767797708511353 }, { "epoch": 0.25118483412322273, "step": 106, "train/alpha": 0.2, "train/clean_ce_loss": 0.0849609375, "train/clean_loss_scale": 9.184134483337402, "train/clean_loss_weight": 1.8368269205093384, "train/combined_loss": 0.9363512396812439, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.025118483412322274, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7802926898002625 }, { "epoch": 0.25355450236966826, "step": 107, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 8.86658763885498, "train/clean_loss_weight": 1.773317575454712, "train/combined_loss": 0.7533135414123535, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.025355450236966826, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6277613043785095 }, { "epoch": 0.25355450236966826, "step": 107, "train/alpha": 0.2, "train/clean_ce_loss": 0.05908203125, "train/clean_loss_scale": 11.112299919128418, "train/clean_loss_weight": 2.2224600315093994, "train/combined_loss": 0.7878446578979492, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.025355450236966826, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6565372347831726 }, { "epoch": 0.25355450236966826, "step": 107, "train/alpha": 0.2, "train/clean_ce_loss": 0.1572265625, "train/clean_loss_scale": 6.077043533325195, "train/clean_loss_weight": 1.2154086828231812, "train/combined_loss": 1.1465672254562378, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.025355450236966826, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9554726481437683 }, { "epoch": 0.25355450236966826, "step": 107, "train/alpha": 0.2, "train/clean_ce_loss": 0.09375, "train/clean_loss_scale": 9.376908302307129, "train/clean_loss_weight": 1.8753817081451416, "train/combined_loss": 1.054902195930481, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.025355450236966826, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.879085123538971 }, { "epoch": 0.2559241706161137, "step": 108, "train/alpha": 0.2, "train/clean_ce_loss": 0.09228515625, "train/clean_loss_scale": 7.356634616851807, "train/clean_loss_weight": 1.4713269472122192, "train/combined_loss": 0.8146898150444031, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.025592417061611375, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6789081692695618 }, { "epoch": 0.2559241706161137, "step": 108, "train/alpha": 0.2, "train/clean_ce_loss": 0.1728515625, "train/clean_loss_scale": 7.365312099456787, "train/clean_loss_weight": 1.4730623960494995, "train/combined_loss": 1.5277268886566162, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.025592417061611375, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2731057405471802 }, { "epoch": 0.2559241706161137, "step": 108, "train/alpha": 0.2, "train/clean_ce_loss": 0.09033203125, "train/clean_loss_scale": 7.252279758453369, "train/clean_loss_weight": 1.4504560232162476, "train/combined_loss": 0.7861357927322388, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.025592417061611375, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.655113160610199 }, { "epoch": 0.2559241706161137, "step": 108, "train/alpha": 0.2, "train/clean_ce_loss": 0.1181640625, "train/clean_loss_scale": 6.189124584197998, "train/clean_loss_weight": 1.2378249168395996, "train/combined_loss": 0.8775985240936279, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.025592417061611375, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7313321232795715 }, { "epoch": 0.25829383886255924, "step": 109, "train/alpha": 0.2, "train/clean_ce_loss": 0.08544921875, "train/clean_loss_scale": 9.702882766723633, "train/clean_loss_weight": 1.9405765533447266, "train/combined_loss": 0.9949245452880859, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.025829383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8291037678718567 }, { "epoch": 0.25829383886255924, "step": 109, "train/alpha": 0.2, "train/clean_ce_loss": 0.078125, "train/clean_loss_scale": 8.116722106933594, "train/clean_loss_weight": 1.6233444213867188, "train/combined_loss": 0.7609426975250244, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.025829383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.634118914604187 }, { "epoch": 0.25829383886255924, "step": 109, "train/alpha": 0.2, "train/clean_ce_loss": 0.08056640625, "train/clean_loss_scale": 8.359472274780273, "train/clean_loss_weight": 1.6718944311141968, "train/combined_loss": 0.808191180229187, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.025829383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6734926700592041 }, { "epoch": 0.25829383886255924, "step": 109, "train/alpha": 0.2, "train/clean_ce_loss": 0.125, "train/clean_loss_scale": 7.351858139038086, "train/clean_loss_weight": 1.4703716039657593, "train/combined_loss": 1.102778673171997, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.025829383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9189822673797607 }, { "epoch": 0.26066350710900477, "grad_norm": 15.194916725158691, "learning_rate": 1.716535433070866e-06, "loss": 4.237, "step": 110 }, { "epoch": 0.26066350710900477, "step": 110, "train/alpha": 0.2, "train/clean_ce_loss": 0.15234375, "train/clean_loss_scale": 8.142471313476562, "train/clean_loss_weight": 1.6284942626953125, "train/combined_loss": 1.4885456562042236, "train/pos_loss/pos0": 0.5069084167480469, "train/pos_loss/pos1": 1.1668260097503662, "train/pos_loss/pos2": 1.6532068252563477, "train/pos_loss/pos3": 2.342851400375366, "train/pos_loss/pos4": 3.1770265102386475, "train/pos_loss/pos5": 3.6835477352142334, "train/pos_loss/pos6": 3.731297254562378, "train/pos_loss/weight0": 2.7871556282043457, "train/pos_loss/weight1": 1.9112030267715454, "train/pos_loss/weight2": 1.1877648830413818, "train/pos_loss/weight3": 0.645238995552063, "train/pos_loss/weight4": 0.3008890748023987, "train/pos_loss/weight5": 0.11801828444004059, "train/pos_loss/weight6": 0.049730367958545685, "train/pos_weight_progress": 0.026066350710900472, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2404546737670898 }, { "epoch": 0.26066350710900477, "step": 110, "train/alpha": 0.2, "train/clean_ce_loss": 0.1396484375, "train/clean_loss_scale": 7.552839279174805, "train/clean_loss_weight": 1.5105679035186768, "train/combined_loss": 1.2656906843185425, "train/pos_loss/pos0": 0.5072816610336304, "train/pos_loss/pos1": 1.117059588432312, "train/pos_loss/pos2": 1.833931803703308, "train/pos_loss/pos3": 2.4065020084381104, "train/pos_loss/pos4": 3.019176959991455, "train/pos_loss/pos5": 3.6467630863189697, "train/pos_loss/pos6": 3.906099557876587, "train/pos_loss/weight0": 3.316241502761841, "train/pos_loss/weight1": 2.071937322616577, "train/pos_loss/weight2": 1.0464540719985962, "train/pos_loss/weight3": 0.4157644808292389, "train/pos_loss/weight4": 0.11897974461317062, "train/pos_loss/weight5": 0.025068366900086403, "train/pos_loss/weight6": 0.005554498173296452, "train/pos_weight_progress": 0.026066350710900472, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0547422170639038 }, { "epoch": 0.26066350710900477, "step": 110, "train/alpha": 0.2, "train/clean_ce_loss": 0.07666015625, "train/clean_loss_scale": 6.901741981506348, "train/clean_loss_weight": 1.3803484439849854, "train/combined_loss": 0.6349063515663147, "train/pos_loss/pos0": 0.2567697763442993, "train/pos_loss/pos1": 0.5524238348007202, "train/pos_loss/pos2": 0.8758106827735901, "train/pos_loss/pos3": 1.2417241334915161, "train/pos_loss/pos4": 1.5528841018676758, "train/pos_loss/pos5": 1.9507821798324585, "train/pos_loss/pos6": 2.372224807739258, "train/pos_loss/weight0": 3.3096961975097656, "train/pos_loss/weight1": 2.0489554405212402, "train/pos_loss/weight2": 1.0533018112182617, "train/pos_loss/weight3": 0.42117810249328613, "train/pos_loss/weight4": 0.13036763668060303, "train/pos_loss/weight5": 0.029868649318814278, "train/pos_loss/weight6": 0.006632143631577492, "train/pos_weight_progress": 0.026066350710900472, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5290886163711548 }, { "epoch": 0.26066350710900477, "step": 110, "train/alpha": 0.2, "train/clean_ce_loss": 0.134765625, "train/clean_loss_scale": 11.553943634033203, "train/clean_loss_weight": 2.310788869857788, "train/combined_loss": 1.8684892654418945, "train/pos_loss/pos0": 0.7059276103973389, "train/pos_loss/pos1": 1.2600511312484741, "train/pos_loss/pos2": 2.086329936981201, "train/pos_loss/pos3": 2.5372724533081055, "train/pos_loss/pos4": 3.1322622299194336, "train/pos_loss/pos5": 3.554290294647217, "train/pos_loss/pos6": 4.011155605316162, "train/pos_loss/weight0": 2.4895904064178467, "train/pos_loss/weight1": 1.824030876159668, "train/pos_loss/weight2": 1.2252401113510132, "train/pos_loss/weight3": 0.7483718991279602, "train/pos_loss/weight4": 0.40326225757598877, "train/pos_loss/weight5": 0.20177137851715088, "train/pos_loss/weight6": 0.10773330181837082, "train/pos_weight_progress": 0.026066350710900472, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5570744276046753 }, { "epoch": 0.26303317535545023, "step": 111, "train/alpha": 0.2, "train/clean_ce_loss": 0.125, "train/clean_loss_scale": 8.706076622009277, "train/clean_loss_weight": 1.7412153482437134, "train/combined_loss": 1.3059115409851074, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.026303317535545025, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0882595777511597 }, { "epoch": 0.26303317535545023, "step": 111, "train/alpha": 0.2, "train/clean_ce_loss": 0.08642578125, "train/clean_loss_scale": 8.130783081054688, "train/clean_loss_weight": 1.6261566877365112, "train/combined_loss": 0.8432511687278748, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.026303317535545025, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7027093172073364 }, { "epoch": 0.26303317535545023, "step": 111, "train/alpha": 0.2, "train/clean_ce_loss": 0.0888671875, "train/clean_loss_scale": 6.2856268882751465, "train/clean_loss_weight": 1.2571253776550293, "train/combined_loss": 0.670303225517273, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.026303317535545025, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5585860013961792 }, { "epoch": 0.26303317535545023, "step": 111, "train/alpha": 0.2, "train/clean_ce_loss": 0.055419921875, "train/clean_loss_scale": 7.1249775886535645, "train/clean_loss_weight": 1.4249955415725708, "train/combined_loss": 0.47383883595466614, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.026303317535545025, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3948656916618347 }, { "epoch": 0.26540284360189575, "step": 112, "train/alpha": 0.2, "train/clean_ce_loss": 0.1953125, "train/clean_loss_scale": 5.756765842437744, "train/clean_loss_weight": 1.1513532400131226, "train/combined_loss": 1.3492419719696045, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.026540284360189573, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1243683099746704 }, { "epoch": 0.26540284360189575, "step": 112, "train/alpha": 0.2, "train/clean_ce_loss": 0.07666015625, "train/clean_loss_scale": 6.306121826171875, "train/clean_loss_weight": 1.261224389076233, "train/combined_loss": 0.58011394739151, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.026540284360189573, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.48342829942703247 }, { "epoch": 0.26540284360189575, "step": 112, "train/alpha": 0.2, "train/clean_ce_loss": 0.10498046875, "train/clean_loss_scale": 7.4629740715026855, "train/clean_loss_weight": 1.492594838142395, "train/combined_loss": 0.940159797668457, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.026540284360189573, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7834665179252625 }, { "epoch": 0.26540284360189575, "step": 112, "train/alpha": 0.2, "train/clean_ce_loss": 0.0303955078125, "train/clean_loss_scale": 10.439717292785645, "train/clean_loss_weight": 2.0879435539245605, "train/combined_loss": 0.3807846009731293, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.026540284360189573, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.31732049584388733 }, { "epoch": 0.2677725118483412, "step": 113, "train/alpha": 0.2, "train/clean_ce_loss": 0.1748046875, "train/clean_loss_scale": 7.383641242980957, "train/clean_loss_weight": 1.4767283201217651, "train/combined_loss": 1.5488340854644775, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.026777251184834122, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.290695071220398 }, { "epoch": 0.2677725118483412, "step": 113, "train/alpha": 0.2, "train/clean_ce_loss": 0.138671875, "train/clean_loss_scale": 6.490065574645996, "train/clean_loss_weight": 1.2980130910873413, "train/combined_loss": 1.0799874067306519, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.026777251184834122, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8999895453453064 }, { "epoch": 0.2677725118483412, "step": 113, "train/alpha": 0.2, "train/clean_ce_loss": 0.11328125, "train/clean_loss_scale": 10.166272163391113, "train/clean_loss_weight": 2.033254384994507, "train/combined_loss": 1.3819776773452759, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.026777251184834122, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1516480445861816 }, { "epoch": 0.2677725118483412, "step": 113, "train/alpha": 0.2, "train/clean_ce_loss": 0.072265625, "train/clean_loss_scale": 6.989177703857422, "train/clean_loss_weight": 1.397835612297058, "train/combined_loss": 0.6060927510261536, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.026777251184834122, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5050773024559021 }, { "epoch": 0.27014218009478674, "step": 114, "train/alpha": 0.2, "train/clean_ce_loss": 0.044189453125, "train/clean_loss_scale": 5.730712413787842, "train/clean_loss_weight": 1.1461424827575684, "train/combined_loss": 0.30388444662094116, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.027014218009478674, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2532370388507843 }, { "epoch": 0.27014218009478674, "step": 114, "train/alpha": 0.2, "train/clean_ce_loss": 0.059814453125, "train/clean_loss_scale": 5.860137462615967, "train/clean_loss_weight": 1.1720274686813354, "train/combined_loss": 0.42062509059906006, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.027014218009478674, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.35052090883255005 }, { "epoch": 0.27014218009478674, "step": 114, "train/alpha": 0.2, "train/clean_ce_loss": 0.10595703125, "train/clean_loss_scale": 6.582294464111328, "train/clean_loss_weight": 1.3164589405059814, "train/combined_loss": 0.8369284868240356, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.027014218009478674, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6974403858184814 }, { "epoch": 0.27014218009478674, "step": 114, "train/alpha": 0.2, "train/clean_ce_loss": 0.169921875, "train/clean_loss_scale": 5.040882587432861, "train/clean_loss_weight": 1.008176565170288, "train/combined_loss": 1.0278675556182861, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.027014218009478674, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.856556236743927 }, { "epoch": 0.2725118483412322, "grad_norm": 12.62503433227539, "learning_rate": 1.795275590551181e-06, "loss": 3.8074, "step": 115 }, { "epoch": 0.2725118483412322, "step": 115, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 5.959814548492432, "train/clean_loss_weight": 1.1919629573822021, "train/combined_loss": 0.47143062949180603, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.027251184834123223, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3928588628768921 }, { "epoch": 0.2725118483412322, "step": 115, "train/alpha": 0.2, "train/clean_ce_loss": 0.09814453125, "train/clean_loss_scale": 5.394107818603516, "train/clean_loss_weight": 1.0788215398788452, "train/combined_loss": 0.6352826356887817, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.027251184834123223, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5294021964073181 }, { "epoch": 0.2725118483412322, "step": 115, "train/alpha": 0.2, "train/clean_ce_loss": 0.08935546875, "train/clean_loss_scale": 6.052790641784668, "train/clean_loss_weight": 1.2105581760406494, "train/combined_loss": 0.6490198969841003, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.027251184834123223, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5408499240875244 }, { "epoch": 0.2725118483412322, "step": 115, "train/alpha": 0.2, "train/clean_ce_loss": 0.130859375, "train/clean_loss_scale": 7.428195476531982, "train/clean_loss_weight": 1.4856390953063965, "train/combined_loss": 1.1664588451385498, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.027251184834123223, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9720489978790283 }, { "epoch": 0.27488151658767773, "step": 116, "train/alpha": 0.2, "train/clean_ce_loss": 0.08837890625, "train/clean_loss_scale": 6.6766791343688965, "train/clean_loss_weight": 1.3353358507156372, "train/combined_loss": 0.7080931067466736, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.027488151658767772, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5900775790214539 }, { "epoch": 0.27488151658767773, "step": 116, "train/alpha": 0.2, "train/clean_ce_loss": 0.146484375, "train/clean_loss_scale": 6.467016220092773, "train/clean_loss_weight": 1.2934032678604126, "train/combined_loss": 1.1367801427841187, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.027488151658767772, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9473168253898621 }, { "epoch": 0.27488151658767773, "step": 116, "train/alpha": 0.2, "train/clean_ce_loss": 0.134765625, "train/clean_loss_scale": 4.567587375640869, "train/clean_loss_weight": 0.9135174751281738, "train/combined_loss": 0.7386645674705505, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.027488151658767772, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6155537962913513 }, { "epoch": 0.27488151658767773, "step": 116, "train/alpha": 0.2, "train/clean_ce_loss": 0.11279296875, "train/clean_loss_scale": 5.788697242736816, "train/clean_loss_weight": 1.157739520072937, "train/combined_loss": 0.7835092544555664, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.027488151658767772, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6529243588447571 }, { "epoch": 0.2772511848341232, "step": 117, "train/alpha": 0.2, "train/clean_ce_loss": 0.10888671875, "train/clean_loss_scale": 6.577661037445068, "train/clean_loss_weight": 1.3155322074890137, "train/combined_loss": 0.8594638705253601, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.02772511848341232, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7162199020385742 }, { "epoch": 0.2772511848341232, "step": 117, "train/alpha": 0.2, "train/clean_ce_loss": 0.1220703125, "train/clean_loss_scale": 9.661580085754395, "train/clean_loss_weight": 1.9323160648345947, "train/combined_loss": 1.4152705669403076, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.02772511848341232, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1793920993804932 }, { "epoch": 0.2772511848341232, "step": 117, "train/alpha": 0.2, "train/clean_ce_loss": 0.14453125, "train/clean_loss_scale": 4.705601692199707, "train/clean_loss_weight": 0.9411203265190125, "train/combined_loss": 0.8161278367042542, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.02772511848341232, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.680106520652771 }, { "epoch": 0.2772511848341232, "step": 117, "train/alpha": 0.2, "train/clean_ce_loss": 0.140625, "train/clean_loss_scale": 8.802048683166504, "train/clean_loss_weight": 1.7604097127914429, "train/combined_loss": 1.485345721244812, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.02772511848341232, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2377880811691284 }, { "epoch": 0.2796208530805687, "step": 118, "train/alpha": 0.2, "train/clean_ce_loss": 0.1259765625, "train/clean_loss_scale": 7.577296257019043, "train/clean_loss_weight": 1.5154592990875244, "train/combined_loss": 1.1454740762710571, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.027962085308056873, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.954561710357666 }, { "epoch": 0.2796208530805687, "step": 118, "train/alpha": 0.2, "train/clean_ce_loss": 0.1318359375, "train/clean_loss_scale": 5.8784098625183105, "train/clean_loss_weight": 1.1756819486618042, "train/combined_loss": 0.9299827814102173, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.027962085308056873, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.774985671043396 }, { "epoch": 0.2796208530805687, "step": 118, "train/alpha": 0.2, "train/clean_ce_loss": 0.09326171875, "train/clean_loss_scale": 6.012860298156738, "train/clean_loss_weight": 1.2025721073150635, "train/combined_loss": 0.6729236245155334, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.027962085308056873, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5607696771621704 }, { "epoch": 0.2796208530805687, "step": 118, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 5.069005966186523, "train/clean_loss_weight": 1.0138012170791626, "train/combined_loss": 0.4336376190185547, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.027962085308056873, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3613646924495697 }, { "epoch": 0.28199052132701424, "step": 119, "train/alpha": 0.2, "train/clean_ce_loss": 0.06787109375, "train/clean_loss_scale": 6.674619197845459, "train/clean_loss_weight": 1.3349238634109497, "train/combined_loss": 0.5436164736747742, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.02819905213270142, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45301371812820435 }, { "epoch": 0.28199052132701424, "step": 119, "train/alpha": 0.2, "train/clean_ce_loss": 0.11572265625, "train/clean_loss_scale": 6.80821418762207, "train/clean_loss_weight": 1.361642837524414, "train/combined_loss": 0.9454375505447388, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.02819905213270142, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.787864625453949 }, { "epoch": 0.28199052132701424, "step": 119, "train/alpha": 0.2, "train/clean_ce_loss": 0.1572265625, "train/clean_loss_scale": 8.221719741821289, "train/clean_loss_weight": 1.6443439722061157, "train/combined_loss": 1.5512073040008545, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.02819905213270142, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2926727533340454 }, { "epoch": 0.28199052132701424, "step": 119, "train/alpha": 0.2, "train/clean_ce_loss": 0.052978515625, "train/clean_loss_scale": 7.152757167816162, "train/clean_loss_weight": 1.4305514097213745, "train/combined_loss": 0.45473095774650574, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.02819905213270142, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.37894245982170105 }, { "epoch": 0.2843601895734597, "grad_norm": 11.404234886169434, "learning_rate": 1.874015748031496e-06, "loss": 3.9023, "step": 120 }, { "epoch": 0.2843601895734597, "step": 120, "train/alpha": 0.2, "train/clean_ce_loss": 0.0986328125, "train/clean_loss_scale": 9.11103343963623, "train/clean_loss_weight": 1.822206735610962, "train/combined_loss": 1.0783761739730835, "train/pos_loss/pos0": 0.4346798360347748, "train/pos_loss/pos1": 1.0675705671310425, "train/pos_loss/pos2": 1.61229407787323, "train/pos_loss/pos3": 1.8963373899459839, "train/pos_loss/pos4": 2.8322856426239014, "train/pos_loss/pos5": 3.0138425827026367, "train/pos_loss/pos6": 3.8016395568847656, "train/pos_loss/weight0": 3.4823694229125977, "train/pos_loss/weight1": 2.0479679107666016, "train/pos_loss/weight2": 1.0044759511947632, "train/pos_loss/weight3": 0.3541938066482544, "train/pos_loss/weight4": 0.09083253890275955, "train/pos_loss/weight5": 0.016970612108707428, "train/pos_loss/weight6": 0.0031895197462290525, "train/pos_weight_progress": 0.02843601895734597, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8986468315124512 }, { "epoch": 0.2843601895734597, "step": 120, "train/alpha": 0.2, "train/clean_ce_loss": 0.0888671875, "train/clean_loss_scale": 7.818353176116943, "train/clean_loss_weight": 1.5636706352233887, "train/combined_loss": 0.8337540626525879, "train/pos_loss/pos0": 0.31508496403694153, "train/pos_loss/pos1": 0.7291222810745239, "train/pos_loss/pos2": 0.9605226516723633, "train/pos_loss/pos3": 1.522178053855896, "train/pos_loss/pos4": 2.0938825607299805, "train/pos_loss/pos5": 2.4802346229553223, "train/pos_loss/pos6": 3.2378041744232178, "train/pos_loss/weight0": 3.0827083587646484, "train/pos_loss/weight1": 2.0452394485473633, "train/pos_loss/weight2": 1.1175090074539185, "train/pos_loss/weight3": 0.5002025961875916, "train/pos_loss/weight4": 0.18619412183761597, "train/pos_loss/weight5": 0.05341967195272446, "train/pos_loss/weight6": 0.014726865105330944, "train/pos_weight_progress": 0.02843601895734597, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6947950720787048 }, { "epoch": 0.2843601895734597, "step": 120, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 7.097861289978027, "train/clean_loss_weight": 1.4195722341537476, "train/combined_loss": 0.6196765303611755, "train/pos_loss/pos0": 0.23665566742420197, "train/pos_loss/pos1": 0.3890765607357025, "train/pos_loss/pos2": 0.7233636975288391, "train/pos_loss/pos3": 1.0904639959335327, "train/pos_loss/pos4": 1.5695220232009888, "train/pos_loss/pos5": 2.027913808822632, "train/pos_loss/pos6": 2.5300393104553223, "train/pos_loss/weight0": 2.8588452339172363, "train/pos_loss/weight1": 1.9836992025375366, "train/pos_loss/weight2": 1.1655898094177246, "train/pos_loss/weight3": 0.6256840825080872, "train/pos_loss/weight4": 0.2509002089500427, "train/pos_loss/weight5": 0.08410586416721344, "train/pos_loss/weight6": 0.031175825744867325, "train/pos_weight_progress": 0.02843601895734597, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5163971185684204 }, { "epoch": 0.2843601895734597, "step": 120, "train/alpha": 0.2, "train/clean_ce_loss": 0.1640625, "train/clean_loss_scale": 8.383946418762207, "train/clean_loss_weight": 1.6767892837524414, "train/combined_loss": 1.6505894660949707, "train/pos_loss/pos0": 0.7292377352714539, "train/pos_loss/pos1": 1.115533471107483, "train/pos_loss/pos2": 1.6023237705230713, "train/pos_loss/pos3": 2.3405747413635254, "train/pos_loss/pos4": 2.634976863861084, "train/pos_loss/pos5": 3.2268104553222656, "train/pos_loss/pos6": 3.8279290199279785, "train/pos_loss/weight0": 2.504676342010498, "train/pos_loss/weight1": 1.8175691366195679, "train/pos_loss/weight2": 1.2300715446472168, "train/pos_loss/weight3": 0.7517133951187134, "train/pos_loss/weight4": 0.4038807153701782, "train/pos_loss/weight5": 0.19527536630630493, "train/pos_loss/weight6": 0.09681318700313568, "train/pos_weight_progress": 0.02843601895734597, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3754912614822388 }, { "epoch": 0.28672985781990523, "step": 121, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 8.426804542541504, "train/clean_loss_weight": 1.6853609085083008, "train/combined_loss": 0.8048257231712341, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.028672985781990522, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6706880927085876 }, { "epoch": 0.28672985781990523, "step": 121, "train/alpha": 0.2, "train/clean_ce_loss": 0.1748046875, "train/clean_loss_scale": 7.150811195373535, "train/clean_loss_weight": 1.4301623106002808, "train/combined_loss": 1.4999943971633911, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.028672985781990522, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2499953508377075 }, { "epoch": 0.28672985781990523, "step": 121, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 6.084627628326416, "train/clean_loss_weight": 1.2169255018234253, "train/combined_loss": 0.6096511483192444, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.028672985781990522, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5080426335334778 }, { "epoch": 0.28672985781990523, "step": 121, "train/alpha": 0.2, "train/clean_ce_loss": 0.11767578125, "train/clean_loss_scale": 9.370162963867188, "train/clean_loss_weight": 1.8740326166152954, "train/combined_loss": 1.323169469833374, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.028672985781990522, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.102641224861145 }, { "epoch": 0.2890995260663507, "step": 122, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 8.217121124267578, "train/clean_loss_weight": 1.6434242725372314, "train/combined_loss": 0.9147966504096985, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.02890995260663507, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7623305320739746 }, { "epoch": 0.2890995260663507, "step": 122, "train/alpha": 0.2, "train/clean_ce_loss": 0.06396484375, "train/clean_loss_scale": 7.840730667114258, "train/clean_loss_weight": 1.5681461095809937, "train/combined_loss": 0.6018373370170593, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.02890995260663507, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5015311241149902 }, { "epoch": 0.2890995260663507, "step": 122, "train/alpha": 0.2, "train/clean_ce_loss": 0.1591796875, "train/clean_loss_scale": 7.365092754364014, "train/clean_loss_weight": 1.4730185270309448, "train/combined_loss": 1.4068478345870972, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.02890995260663507, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1723731756210327 }, { "epoch": 0.2890995260663507, "step": 122, "train/alpha": 0.2, "train/clean_ce_loss": 0.171875, "train/clean_loss_scale": 6.281715393066406, "train/clean_loss_weight": 1.256343126296997, "train/combined_loss": 1.2956037521362305, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.02890995260663507, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0796698331832886 }, { "epoch": 0.2914691943127962, "step": 123, "train/alpha": 0.2, "train/clean_ce_loss": 0.1064453125, "train/clean_loss_scale": 6.627328395843506, "train/clean_loss_weight": 1.3254656791687012, "train/combined_loss": 0.8465376496315002, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.02914691943127962, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7054480314254761 }, { "epoch": 0.2914691943127962, "step": 123, "train/alpha": 0.2, "train/clean_ce_loss": 0.06884765625, "train/clean_loss_scale": 5.1046648025512695, "train/clean_loss_weight": 1.0209330320358276, "train/combined_loss": 0.4217330515384674, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.02914691943127962, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.35144421458244324 }, { "epoch": 0.2914691943127962, "step": 123, "train/alpha": 0.2, "train/clean_ce_loss": 0.169921875, "train/clean_loss_scale": 6.980672359466553, "train/clean_loss_weight": 1.3961344957351685, "train/combined_loss": 1.4234026670455933, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.02914691943127962, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.186168909072876 }, { "epoch": 0.2914691943127962, "step": 123, "train/alpha": 0.2, "train/clean_ce_loss": 0.20703125, "train/clean_loss_scale": 5.60741662979126, "train/clean_loss_weight": 1.121483325958252, "train/combined_loss": 1.3930926322937012, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.02914691943127962, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1609104871749878 }, { "epoch": 0.2938388625592417, "step": 124, "train/alpha": 0.2, "train/clean_ce_loss": 0.11083984375, "train/clean_loss_scale": 5.779308319091797, "train/clean_loss_weight": 1.155861735343933, "train/combined_loss": 0.7686931490898132, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.029383886255924172, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6405776143074036 }, { "epoch": 0.2938388625592417, "step": 124, "train/alpha": 0.2, "train/clean_ce_loss": 0.0869140625, "train/clean_loss_scale": 9.089435577392578, "train/clean_loss_weight": 1.8178871870040894, "train/combined_loss": 0.9479997158050537, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.029383886255924172, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.789999783039093 }, { "epoch": 0.2938388625592417, "step": 124, "train/alpha": 0.2, "train/clean_ce_loss": 0.0087890625, "train/clean_loss_scale": 43.58010482788086, "train/clean_loss_weight": 8.716021537780762, "train/combined_loss": 0.4596339166164398, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.029383886255924172, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3830282688140869 }, { "epoch": 0.2938388625592417, "step": 124, "train/alpha": 0.2, "train/clean_ce_loss": 0.10009765625, "train/clean_loss_scale": 6.784634113311768, "train/clean_loss_weight": 1.3569267988204956, "train/combined_loss": 0.8149511814117432, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.029383886255924172, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.679125964641571 }, { "epoch": 0.2962085308056872, "grad_norm": 13.74691390991211, "learning_rate": 1.952755905511811e-06, "loss": 4.0445, "step": 125 }, { "epoch": 0.2962085308056872, "step": 125, "train/alpha": 0.2, "train/clean_ce_loss": 0.1142578125, "train/clean_loss_scale": 5.110098361968994, "train/clean_loss_weight": 1.0220197439193726, "train/combined_loss": 0.7006424069404602, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.02962085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.583868682384491 }, { "epoch": 0.2962085308056872, "step": 125, "train/alpha": 0.2, "train/clean_ce_loss": 0.185546875, "train/clean_loss_scale": 7.20167350769043, "train/clean_loss_weight": 1.440334677696228, "train/combined_loss": 1.6034976243972778, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.02962085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3362480401992798 }, { "epoch": 0.2962085308056872, "step": 125, "train/alpha": 0.2, "train/clean_ce_loss": 0.09130859375, "train/clean_loss_scale": 6.024261474609375, "train/clean_loss_weight": 1.2048523426055908, "train/combined_loss": 0.6600801944732666, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.02962085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5500668287277222 }, { "epoch": 0.2962085308056872, "step": 125, "train/alpha": 0.2, "train/clean_ce_loss": 0.1689453125, "train/clean_loss_scale": 6.545890808105469, "train/clean_loss_weight": 1.3091782331466675, "train/combined_loss": 1.327077031135559, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.02962085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1058975458145142 }, { "epoch": 0.2985781990521327, "step": 126, "train/alpha": 0.2, "train/clean_ce_loss": 0.09228515625, "train/clean_loss_scale": 5.868474960327148, "train/clean_loss_weight": 1.1736949682235718, "train/combined_loss": 0.64988774061203, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.02985781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5415731072425842 }, { "epoch": 0.2985781990521327, "step": 126, "train/alpha": 0.2, "train/clean_ce_loss": 0.087890625, "train/clean_loss_scale": 7.0406413078308105, "train/clean_loss_weight": 1.408128261566162, "train/combined_loss": 0.7425676584243774, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.02985781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6188063621520996 }, { "epoch": 0.2985781990521327, "step": 126, "train/alpha": 0.2, "train/clean_ce_loss": 0.150390625, "train/clean_loss_scale": 7.790990829467773, "train/clean_loss_weight": 1.5581982135772705, "train/combined_loss": 1.4060304164886475, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.02985781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1716920137405396 }, { "epoch": 0.2985781990521327, "step": 126, "train/alpha": 0.2, "train/clean_ce_loss": 0.1376953125, "train/clean_loss_scale": 9.896080017089844, "train/clean_loss_weight": 1.9792159795761108, "train/combined_loss": 1.6351726055145264, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.02985781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.362643837928772 }, { "epoch": 0.3009478672985782, "step": 127, "train/alpha": 0.2, "train/clean_ce_loss": 0.0220947265625, "train/clean_loss_scale": 10.116719245910645, "train/clean_loss_weight": 2.023343801498413, "train/combined_loss": 0.2682313919067383, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.03009478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.22352614998817444 }, { "epoch": 0.3009478672985782, "step": 127, "train/alpha": 0.2, "train/clean_ce_loss": 0.1171875, "train/clean_loss_scale": 5.389208793640137, "train/clean_loss_weight": 1.0778417587280273, "train/combined_loss": 0.7578575015068054, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.03009478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6315479278564453 }, { "epoch": 0.3009478672985782, "step": 127, "train/alpha": 0.2, "train/clean_ce_loss": 0.1220703125, "train/clean_loss_scale": 4.913789749145508, "train/clean_loss_weight": 0.9827579855918884, "train/combined_loss": 0.7197933793067932, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.03009478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5998278260231018 }, { "epoch": 0.3009478672985782, "step": 127, "train/alpha": 0.2, "train/clean_ce_loss": 0.0908203125, "train/clean_loss_scale": 6.498845100402832, "train/clean_loss_weight": 1.2997690439224243, "train/combined_loss": 0.7082725763320923, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.03009478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5902271270751953 }, { "epoch": 0.3033175355450237, "step": 128, "train/alpha": 0.2, "train/clean_ce_loss": 0.181640625, "train/clean_loss_scale": 6.9780168533325195, "train/clean_loss_weight": 1.3956034183502197, "train/combined_loss": 1.5209896564483643, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.03033175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.267491340637207 }, { "epoch": 0.3033175355450237, "step": 128, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 5.975621700286865, "train/clean_loss_weight": 1.1951243877410889, "train/combined_loss": 0.5111957788467407, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.03033175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.42599648237228394 }, { "epoch": 0.3033175355450237, "step": 128, "train/alpha": 0.2, "train/clean_ce_loss": 0.1357421875, "train/clean_loss_scale": 6.668565273284912, "train/clean_loss_weight": 1.3337130546569824, "train/combined_loss": 1.0862467288970947, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.03033175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9052056670188904 }, { "epoch": 0.3033175355450237, "step": 128, "train/alpha": 0.2, "train/clean_ce_loss": 0.012451171875, "train/clean_loss_scale": 18.23672866821289, "train/clean_loss_weight": 3.647345781326294, "train/combined_loss": 0.27248236536979675, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.03033175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.22706864774227142 }, { "epoch": 0.3056872037914692, "step": 129, "train/alpha": 0.2, "train/clean_ce_loss": 0.0830078125, "train/clean_loss_scale": 6.536713123321533, "train/clean_loss_weight": 1.3073426485061646, "train/combined_loss": 0.6511179208755493, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.03056872037914692, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5425982475280762 }, { "epoch": 0.3056872037914692, "step": 129, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 5.096177577972412, "train/clean_loss_weight": 1.0192354917526245, "train/combined_loss": 0.43297600746154785, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.03056872037914692, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.36081334948539734 }, { "epoch": 0.3056872037914692, "step": 129, "train/alpha": 0.2, "train/clean_ce_loss": 0.06884765625, "train/clean_loss_scale": 6.462002277374268, "train/clean_loss_weight": 1.2924004793167114, "train/combined_loss": 0.5338724851608276, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.03056872037914692, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4448937177658081 }, { "epoch": 0.3056872037914692, "step": 129, "train/alpha": 0.2, "train/clean_ce_loss": 0.09814453125, "train/clean_loss_scale": 6.120573043823242, "train/clean_loss_weight": 1.2241146564483643, "train/combined_loss": 0.7208409309387207, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.03056872037914692, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6007007956504822 }, { "epoch": 0.3080568720379147, "grad_norm": 13.877212524414062, "learning_rate": 2e-06, "loss": 3.7897, "step": 130 }, { "epoch": 0.3080568720379147, "step": 130, "train/alpha": 0.2, "train/clean_ce_loss": 0.0252685546875, "train/clean_loss_scale": 12.397001266479492, "train/clean_loss_weight": 2.479400396347046, "train/combined_loss": 0.3759051561355591, "train/pos_loss/pos0": 0.034956011921167374, "train/pos_loss/pos1": 0.18415053188800812, "train/pos_loss/pos2": 0.7135148048400879, "train/pos_loss/pos3": 1.329639196395874, "train/pos_loss/pos4": 1.986741542816162, "train/pos_loss/pos5": 2.7488601207733154, "train/pos_loss/pos6": 3.2034037113189697, "train/pos_loss/weight0": 3.3460581302642822, "train/pos_loss/weight1": 2.0376274585723877, "train/pos_loss/weight2": 0.9860528707504272, "train/pos_loss/weight3": 0.44665753841400146, "train/pos_loss/weight4": 0.1403484344482422, "train/pos_loss/weight5": 0.035125549882650375, "train/pos_loss/weight6": 0.008130389265716076, "train/pos_weight_progress": 0.030805687203791468, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.31325429677963257 }, { "epoch": 0.3080568720379147, "step": 130, "train/alpha": 0.2, "train/clean_ce_loss": 0.09423828125, "train/clean_loss_scale": 7.805016994476318, "train/clean_loss_weight": 1.5610034465789795, "train/combined_loss": 0.8826376795768738, "train/pos_loss/pos0": 0.36358556151390076, "train/pos_loss/pos1": 0.6238425374031067, "train/pos_loss/pos2": 0.975184440612793, "train/pos_loss/pos3": 1.2518304586410522, "train/pos_loss/pos4": 1.5534602403640747, "train/pos_loss/pos5": 2.0970354080200195, "train/pos_loss/pos6": 2.634410858154297, "train/pos_loss/weight0": 2.5965771675109863, "train/pos_loss/weight1": 1.9051944017410278, "train/pos_loss/weight2": 1.274388074874878, "train/pos_loss/weight3": 0.7541605234146118, "train/pos_loss/weight4": 0.318652868270874, "train/pos_loss/weight5": 0.11219462007284164, "train/pos_loss/weight6": 0.03883270174264908, "train/pos_weight_progress": 0.030805687203791468, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7355313897132874 }, { "epoch": 0.3080568720379147, "step": 130, "train/alpha": 0.2, "train/clean_ce_loss": 0.10400390625, "train/clean_loss_scale": 8.772217750549316, "train/clean_loss_weight": 1.7544435262680054, "train/combined_loss": 1.0948138236999512, "train/pos_loss/pos0": 0.366897851228714, "train/pos_loss/pos1": 0.7796083092689514, "train/pos_loss/pos2": 1.1753809452056885, "train/pos_loss/pos3": 1.6572062969207764, "train/pos_loss/pos4": 2.097313642501831, "train/pos_loss/pos5": 2.582216739654541, "train/pos_loss/pos6": 3.0340490341186523, "train/pos_loss/weight0": 2.642962694168091, "train/pos_loss/weight1": 1.8756039142608643, "train/pos_loss/weight2": 1.1950485706329346, "train/pos_loss/weight3": 0.6871593594551086, "train/pos_loss/weight4": 0.35301417112350464, "train/pos_loss/weight5": 0.16553586721420288, "train/pos_loss/weight6": 0.0806756317615509, "train/pos_weight_progress": 0.030805687203791468, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9123448729515076 }, { "epoch": 0.3080568720379147, "step": 130, "train/alpha": 0.2, "train/clean_ce_loss": 0.0908203125, "train/clean_loss_scale": 6.6709442138671875, "train/clean_loss_weight": 1.3341888189315796, "train/combined_loss": 0.7270287275314331, "train/pos_loss/pos0": 0.2745899260044098, "train/pos_loss/pos1": 0.5705921649932861, "train/pos_loss/pos2": 0.8438339829444885, "train/pos_loss/pos3": 1.1149044036865234, "train/pos_loss/pos4": 1.5894829034805298, "train/pos_loss/pos5": 1.8868050575256348, "train/pos_loss/pos6": 2.248396396636963, "train/pos_loss/weight0": 2.8706886768341064, "train/pos_loss/weight1": 1.955948829650879, "train/pos_loss/weight2": 1.1768202781677246, "train/pos_loss/weight3": 0.610862135887146, "train/pos_loss/weight4": 0.2573484480381012, "train/pos_loss/weight5": 0.09304771572351456, "train/pos_loss/weight6": 0.03528452664613724, "train/pos_weight_progress": 0.030805687203791468, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.605857253074646 }, { "epoch": 0.3104265402843602, "step": 131, "train/alpha": 0.2, "train/clean_ce_loss": 0.07861328125, "train/clean_loss_scale": 9.440916061401367, "train/clean_loss_weight": 1.8881832361221313, "train/combined_loss": 0.8906176090240479, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03104265402843602, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7421813607215881 }, { "epoch": 0.3104265402843602, "step": 131, "train/alpha": 0.2, "train/clean_ce_loss": 0.045654296875, "train/clean_loss_scale": 12.655265808105469, "train/clean_loss_weight": 2.531053304672241, "train/combined_loss": 0.6933206915855408, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03104265402843602, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5777672529220581 }, { "epoch": 0.3104265402843602, "step": 131, "train/alpha": 0.2, "train/clean_ce_loss": 0.140625, "train/clean_loss_scale": 9.945202827453613, "train/clean_loss_weight": 1.9890406131744385, "train/combined_loss": 1.6782530546188354, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03104265402843602, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.398544192314148 }, { "epoch": 0.3104265402843602, "step": 131, "train/alpha": 0.2, "train/clean_ce_loss": 0.1748046875, "train/clean_loss_scale": 9.091034889221191, "train/clean_loss_weight": 1.818207025527954, "train/combined_loss": 1.9069867134094238, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03104265402843602, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5891555547714233 }, { "epoch": 0.3127962085308057, "step": 132, "train/alpha": 0.2, "train/clean_ce_loss": 0.10205078125, "train/clean_loss_scale": 6.860504150390625, "train/clean_loss_weight": 1.372100830078125, "train/combined_loss": 0.8401437401771545, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.031279620853080566, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7001197934150696 }, { "epoch": 0.3127962085308057, "step": 132, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 11.231801986694336, "train/clean_loss_weight": 2.2463605403900146, "train/combined_loss": 1.1780229806900024, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.031279620853080566, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9816858172416687 }, { "epoch": 0.3127962085308057, "step": 132, "train/alpha": 0.2, "train/clean_ce_loss": 0.1865234375, "train/clean_loss_scale": 7.879689693450928, "train/clean_loss_weight": 1.5759379863739014, "train/combined_loss": 1.7636961936950684, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.031279620853080566, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4697468280792236 }, { "epoch": 0.3127962085308057, "step": 132, "train/alpha": 0.2, "train/clean_ce_loss": 0.1640625, "train/clean_loss_scale": 9.949628829956055, "train/clean_loss_weight": 1.989925742149353, "train/combined_loss": 1.9588332176208496, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.031279620853080566, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.6323610544204712 }, { "epoch": 0.31516587677725116, "step": 133, "train/alpha": 0.2, "train/clean_ce_loss": 0.09765625, "train/clean_loss_scale": 7.315150737762451, "train/clean_loss_weight": 1.463030219078064, "train/combined_loss": 0.8572442531585693, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03151658767772512, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7143701910972595 }, { "epoch": 0.31516587677725116, "step": 133, "train/alpha": 0.2, "train/clean_ce_loss": 0.1611328125, "train/clean_loss_scale": 10.234657287597656, "train/clean_loss_weight": 2.046931505203247, "train/combined_loss": 1.9789669513702393, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03151658767772512, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.649139165878296 }, { "epoch": 0.31516587677725116, "step": 133, "train/alpha": 0.2, "train/clean_ce_loss": 0.061279296875, "train/clean_loss_scale": 9.387069702148438, "train/clean_loss_weight": 1.8774139881134033, "train/combined_loss": 0.6902796626091003, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03151658767772512, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5752330422401428 }, { "epoch": 0.31516587677725116, "step": 133, "train/alpha": 0.2, "train/clean_ce_loss": 0.054931640625, "train/clean_loss_scale": 13.60422134399414, "train/clean_loss_weight": 2.720844268798828, "train/combined_loss": 0.8967626094818115, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03151658767772512, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7473021745681763 }, { "epoch": 0.3175355450236967, "step": 134, "train/alpha": 0.2, "train/clean_ce_loss": 0.126953125, "train/clean_loss_scale": 8.095634460449219, "train/clean_loss_weight": 1.6191269159317017, "train/combined_loss": 1.2333192825317383, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03175355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0277661085128784 }, { "epoch": 0.3175355450236967, "step": 134, "train/alpha": 0.2, "train/clean_ce_loss": 0.17578125, "train/clean_loss_scale": 8.016087532043457, "train/clean_loss_weight": 1.6032174825668335, "train/combined_loss": 1.6908934116363525, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03175355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4090778827667236 }, { "epoch": 0.3175355450236967, "step": 134, "train/alpha": 0.2, "train/clean_ce_loss": 0.076171875, "train/clean_loss_scale": 9.108377456665039, "train/clean_loss_weight": 1.8216755390167236, "train/combined_loss": 0.8325626254081726, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03175355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6938021779060364 }, { "epoch": 0.3175355450236967, "step": 134, "train/alpha": 0.2, "train/clean_ce_loss": 0.109375, "train/clean_loss_scale": 11.7114839553833, "train/clean_loss_weight": 2.342296838760376, "train/combined_loss": 1.5371322631835938, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03175355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.280943512916565 }, { "epoch": 0.31990521327014215, "grad_norm": 20.862064361572266, "learning_rate": 2e-06, "loss": 4.4567, "step": 135 }, { "epoch": 0.31990521327014215, "step": 135, "train/alpha": 0.2, "train/clean_ce_loss": 0.11767578125, "train/clean_loss_scale": 9.301199913024902, "train/clean_loss_weight": 1.8602399826049805, "train/combined_loss": 1.3134311437606812, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.031990521327014215, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.094525933265686 }, { "epoch": 0.31990521327014215, "step": 135, "train/alpha": 0.2, "train/clean_ce_loss": 0.123046875, "train/clean_loss_scale": 10.281200408935547, "train/clean_loss_weight": 2.0562400817871094, "train/combined_loss": 1.5180835723876953, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.031990521327014215, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.265069603919983 }, { "epoch": 0.31990521327014215, "step": 135, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 7.857872486114502, "train/clean_loss_weight": 1.5715745687484741, "train/combined_loss": 0.7873219847679138, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.031990521327014215, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.656101644039154 }, { "epoch": 0.31990521327014215, "step": 135, "train/alpha": 0.2, "train/clean_ce_loss": 0.059326171875, "train/clean_loss_scale": 9.240106582641602, "train/clean_loss_weight": 1.848021388053894, "train/combined_loss": 0.6578161716461182, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.031990521327014215, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5481801629066467 }, { "epoch": 0.3222748815165877, "step": 136, "train/alpha": 0.2, "train/clean_ce_loss": 0.08935546875, "train/clean_loss_scale": 7.977595806121826, "train/clean_loss_weight": 1.5955191850662231, "train/combined_loss": 0.8554101586341858, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03222748815165877, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.712841808795929 }, { "epoch": 0.3222748815165877, "step": 136, "train/alpha": 0.2, "train/clean_ce_loss": 0.1005859375, "train/clean_loss_scale": 8.929595947265625, "train/clean_loss_weight": 1.785919189453125, "train/combined_loss": 1.077830195426941, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03222748815165877, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8981918096542358 }, { "epoch": 0.3222748815165877, "step": 136, "train/alpha": 0.2, "train/clean_ce_loss": 0.048583984375, "train/clean_loss_scale": 17.130521774291992, "train/clean_loss_weight": 3.4261043071746826, "train/combined_loss": 0.9987227916717529, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03222748815165877, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.832269012928009 }, { "epoch": 0.3222748815165877, "step": 136, "train/alpha": 0.2, "train/clean_ce_loss": 0.06396484375, "train/clean_loss_scale": 10.75713062286377, "train/clean_loss_weight": 2.151426076889038, "train/combined_loss": 0.8256937861442566, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03222748815165877, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6880781650543213 }, { "epoch": 0.3246445497630332, "step": 137, "train/alpha": 0.2, "train/clean_ce_loss": 0.091796875, "train/clean_loss_scale": 7.910001754760742, "train/clean_loss_weight": 1.5820003747940063, "train/combined_loss": 0.8713361024856567, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03246445497630332, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7261134386062622 }, { "epoch": 0.3246445497630332, "step": 137, "train/alpha": 0.2, "train/clean_ce_loss": 0.1357421875, "train/clean_loss_scale": 9.397405624389648, "train/clean_loss_weight": 1.8794811964035034, "train/combined_loss": 1.5307493209838867, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03246445497630332, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.275624394416809 }, { "epoch": 0.3246445497630332, "step": 137, "train/alpha": 0.2, "train/clean_ce_loss": 0.14453125, "train/clean_loss_scale": 8.413229942321777, "train/clean_loss_weight": 1.6826460361480713, "train/combined_loss": 1.459169626235962, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03246445497630332, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2159746885299683 }, { "epoch": 0.3246445497630332, "step": 137, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 10.227089881896973, "train/clean_loss_weight": 2.0454180240631104, "train/combined_loss": 1.072645902633667, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03246445497630332, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8938716053962708 }, { "epoch": 0.32701421800947866, "step": 138, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 8.621953964233398, "train/clean_loss_weight": 1.7243908643722534, "train/combined_loss": 0.8234639763832092, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.032701421800947865, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6862199902534485 }, { "epoch": 0.32701421800947866, "step": 138, "train/alpha": 0.2, "train/clean_ce_loss": 0.1669921875, "train/clean_loss_scale": 7.814615249633789, "train/clean_loss_weight": 1.5629230737686157, "train/combined_loss": 1.5659756660461426, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.032701421800947865, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.304979681968689 }, { "epoch": 0.32701421800947866, "step": 138, "train/alpha": 0.2, "train/clean_ce_loss": 0.1015625, "train/clean_loss_scale": 8.544044494628906, "train/clean_loss_weight": 1.7088088989257812, "train/combined_loss": 1.041305422782898, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.032701421800947865, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8677545189857483 }, { "epoch": 0.32701421800947866, "step": 138, "train/alpha": 0.2, "train/clean_ce_loss": 0.06884765625, "train/clean_loss_scale": 8.668540954589844, "train/clean_loss_weight": 1.7337082624435425, "train/combined_loss": 0.7161704897880554, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.032701421800947865, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5968087315559387 }, { "epoch": 0.3293838862559242, "step": 139, "train/alpha": 0.2, "train/clean_ce_loss": 0.1787109375, "train/clean_loss_scale": 7.324519634246826, "train/clean_loss_weight": 1.4649039506912231, "train/combined_loss": 1.5707660913467407, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03293838862559242, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3089717626571655 }, { "epoch": 0.3293838862559242, "step": 139, "train/alpha": 0.2, "train/clean_ce_loss": 0.05224609375, "train/clean_loss_scale": 9.080355644226074, "train/clean_loss_weight": 1.8160711526870728, "train/combined_loss": 0.5692957043647766, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03293838862559242, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.47441309690475464 }, { "epoch": 0.3293838862559242, "step": 139, "train/alpha": 0.2, "train/clean_ce_loss": 0.09326171875, "train/clean_loss_scale": 7.5196380615234375, "train/clean_loss_weight": 1.5039275884628296, "train/combined_loss": 0.8415532112121582, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03293838862559242, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7012943625450134 }, { "epoch": 0.3293838862559242, "step": 139, "train/alpha": 0.2, "train/clean_ce_loss": 0.09912109375, "train/clean_loss_scale": 8.204788208007812, "train/clean_loss_weight": 1.6409577131271362, "train/combined_loss": 0.975921094417572, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03293838862559242, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8132675886154175 }, { "epoch": 0.33175355450236965, "grad_norm": 16.264009475708008, "learning_rate": 2e-06, "loss": 4.2704, "step": 140 }, { "epoch": 0.33175355450236965, "step": 140, "train/alpha": 0.2, "train/clean_ce_loss": 0.10693359375, "train/clean_loss_scale": 10.606654167175293, "train/clean_loss_weight": 2.121330976486206, "train/combined_loss": 1.3610491752624512, "train/pos_loss/pos0": 0.44627922773361206, "train/pos_loss/pos1": 0.6452670693397522, "train/pos_loss/pos2": 1.4306389093399048, "train/pos_loss/pos3": 2.1327733993530273, "train/pos_loss/pos4": 2.983022928237915, "train/pos_loss/pos5": 3.2984492778778076, "train/pos_loss/pos6": 3.5519912242889404, "train/pos_loss/weight0": 2.4775571823120117, "train/pos_loss/weight1": 1.8145971298217773, "train/pos_loss/weight2": 1.2255542278289795, "train/pos_loss/weight3": 0.743371307849884, "train/pos_loss/weight4": 0.4116758406162262, "train/pos_loss/weight5": 0.21068842709064484, "train/pos_loss/weight6": 0.11655652523040771, "train/pos_weight_progress": 0.03317535545023697, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1342076063156128 }, { "epoch": 0.33175355450236965, "step": 140, "train/alpha": 0.2, "train/clean_ce_loss": 0.051025390625, "train/clean_loss_scale": 6.9341840744018555, "train/clean_loss_weight": 1.3868368864059448, "train/combined_loss": 0.4245833456516266, "train/pos_loss/pos0": 0.15918777883052826, "train/pos_loss/pos1": 0.33746370673179626, "train/pos_loss/pos2": 0.4749865233898163, "train/pos_loss/pos3": 0.8030915856361389, "train/pos_loss/pos4": 1.0973198413848877, "train/pos_loss/pos5": 1.4276273250579834, "train/pos_loss/pos6": 1.7454208135604858, "train/pos_loss/weight0": 3.0327250957489014, "train/pos_loss/weight1": 1.9653756618499756, "train/pos_loss/weight2": 1.1910978555679321, "train/pos_loss/weight3": 0.5324010252952576, "train/pos_loss/weight4": 0.19962672889232635, "train/pos_loss/weight5": 0.06005223095417023, "train/pos_loss/weight6": 0.018721092492341995, "train/pos_weight_progress": 0.03317535545023697, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.35381945967674255 }, { "epoch": 0.33175355450236965, "step": 140, "train/alpha": 0.2, "train/clean_ce_loss": 0.1201171875, "train/clean_loss_scale": 13.47390079498291, "train/clean_loss_weight": 2.694780111312866, "train/combined_loss": 1.942136526107788, "train/pos_loss/pos0": 0.5860076546669006, "train/pos_loss/pos1": 1.1029430627822876, "train/pos_loss/pos2": 1.815817952156067, "train/pos_loss/pos3": 2.578089714050293, "train/pos_loss/pos4": 3.021615505218506, "train/pos_loss/pos5": 3.641974687576294, "train/pos_loss/pos6": 3.9413158893585205, "train/pos_loss/weight0": 2.1404619216918945, "train/pos_loss/weight1": 1.6627099514007568, "train/pos_loss/weight2": 1.2067720890045166, "train/pos_loss/weight3": 0.8540095686912537, "train/pos_loss/weight4": 0.5544860363006592, "train/pos_loss/weight5": 0.3521932065486908, "train/pos_loss/weight6": 0.22936765849590302, "train/pos_weight_progress": 0.03317535545023697, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.6184470653533936 }, { "epoch": 0.33175355450236965, "step": 140, "train/alpha": 0.2, "train/clean_ce_loss": 0.09130859375, "train/clean_loss_scale": 5.150632381439209, "train/clean_loss_weight": 1.0301264524459839, "train/combined_loss": 0.5643563866615295, "train/pos_loss/pos0": 0.20782877504825592, "train/pos_loss/pos1": 0.48897629976272583, "train/pos_loss/pos2": 0.7823203802108765, "train/pos_loss/pos3": 1.277838945388794, "train/pos_loss/pos4": 1.9229984283447266, "train/pos_loss/pos5": 2.315368175506592, "train/pos_loss/pos6": 2.8274855613708496, "train/pos_loss/weight0": 3.4012293815612793, "train/pos_loss/weight1": 2.031139612197876, "train/pos_loss/weight2": 1.0369751453399658, "train/pos_loss/weight3": 0.3924306631088257, "train/pos_loss/weight4": 0.11028224229812622, "train/pos_loss/weight5": 0.022823208943009377, "train/pos_loss/weight6": 0.005119708366692066, "train/pos_weight_progress": 0.03317535545023697, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4702969789505005 }, { "epoch": 0.3341232227488152, "step": 141, "train/alpha": 0.2, "train/clean_ce_loss": 0.1357421875, "train/clean_loss_scale": 9.315899848937988, "train/clean_loss_weight": 1.8631800413131714, "train/combined_loss": 1.517472743988037, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.033412322274881515, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.264560580253601 }, { "epoch": 0.3341232227488152, "step": 141, "train/alpha": 0.2, "train/clean_ce_loss": 0.1328125, "train/clean_loss_scale": 8.946117401123047, "train/clean_loss_weight": 1.789223551750183, "train/combined_loss": 1.4257874488830566, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.033412322274881515, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.188156247138977 }, { "epoch": 0.3341232227488152, "step": 141, "train/alpha": 0.2, "train/clean_ce_loss": 0.1083984375, "train/clean_loss_scale": 8.368470191955566, "train/clean_loss_weight": 1.6736940145492554, "train/combined_loss": 1.088554859161377, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.033412322274881515, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9071291089057922 }, { "epoch": 0.3341232227488152, "step": 141, "train/alpha": 0.2, "train/clean_ce_loss": 0.177734375, "train/clean_loss_scale": 8.654280662536621, "train/clean_loss_weight": 1.73085618019104, "train/combined_loss": 1.8457958698272705, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.033412322274881515, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.538163185119629 }, { "epoch": 0.33649289099526064, "step": 142, "train/alpha": 0.2, "train/clean_ce_loss": 0.1240234375, "train/clean_loss_scale": 7.171302795410156, "train/clean_loss_weight": 1.434260606765747, "train/combined_loss": 1.067291498184204, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.03364928909952607, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.889409601688385 }, { "epoch": 0.33649289099526064, "step": 142, "train/alpha": 0.2, "train/clean_ce_loss": 0.07373046875, "train/clean_loss_scale": 7.721415996551514, "train/clean_loss_weight": 1.5442832708358765, "train/combined_loss": 0.6831643581390381, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.03364928909952607, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5693036317825317 }, { "epoch": 0.33649289099526064, "step": 142, "train/alpha": 0.2, "train/clean_ce_loss": 0.091796875, "train/clean_loss_scale": 9.996770858764648, "train/clean_loss_weight": 1.9993542432785034, "train/combined_loss": 1.1012067794799805, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.03364928909952607, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.917672336101532 }, { "epoch": 0.33649289099526064, "step": 142, "train/alpha": 0.2, "train/clean_ce_loss": 0.056396484375, "train/clean_loss_scale": 9.168231010437012, "train/clean_loss_weight": 1.8336461782455444, "train/combined_loss": 0.6204671859741211, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.03364928909952607, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5170559883117676 }, { "epoch": 0.33886255924170616, "step": 143, "train/alpha": 0.2, "train/clean_ce_loss": 0.09814453125, "train/clean_loss_scale": 7.910330772399902, "train/clean_loss_weight": 1.5820661783218384, "train/combined_loss": 0.9316267967224121, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.03388625592417062, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7763556838035583 }, { "epoch": 0.33886255924170616, "step": 143, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 7.79207181930542, "train/clean_loss_weight": 1.558414340019226, "train/combined_loss": 0.602668046951294, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.03388625592417062, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5022233724594116 }, { "epoch": 0.33886255924170616, "step": 143, "train/alpha": 0.2, "train/clean_ce_loss": 0.119140625, "train/clean_loss_scale": 7.4528489112854, "train/clean_loss_weight": 1.490569829940796, "train/combined_loss": 1.0655244588851929, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.03388625592417062, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.887937068939209 }, { "epoch": 0.33886255924170616, "step": 143, "train/alpha": 0.2, "train/clean_ce_loss": 0.06396484375, "train/clean_loss_scale": 9.682659149169922, "train/clean_loss_weight": 1.936531901359558, "train/combined_loss": 0.7432197332382202, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.03388625592417062, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6193497776985168 }, { "epoch": 0.3412322274881517, "step": 144, "train/alpha": 0.2, "train/clean_ce_loss": 0.051513671875, "train/clean_loss_scale": 14.085210800170898, "train/clean_loss_weight": 2.817042112350464, "train/combined_loss": 0.8706971406936646, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.034123222748815164, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7255809307098389 }, { "epoch": 0.3412322274881517, "step": 144, "train/alpha": 0.2, "train/clean_ce_loss": 0.078125, "train/clean_loss_scale": 7.437294960021973, "train/clean_loss_weight": 1.4874590635299683, "train/combined_loss": 0.6972463726997375, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.034123222748815164, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5810386538505554 }, { "epoch": 0.3412322274881517, "step": 144, "train/alpha": 0.2, "train/clean_ce_loss": 0.095703125, "train/clean_loss_scale": 8.325933456420898, "train/clean_loss_weight": 1.6651867628097534, "train/combined_loss": 0.9561814069747925, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.034123222748815164, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7968178391456604 }, { "epoch": 0.3412322274881517, "step": 144, "train/alpha": 0.2, "train/clean_ce_loss": 0.1494140625, "train/clean_loss_scale": 10.112822532653809, "train/clean_loss_weight": 2.022564649581909, "train/combined_loss": 1.8131974935531616, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.034123222748815164, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5109978914260864 }, { "epoch": 0.34360189573459715, "grad_norm": 13.466164588928223, "learning_rate": 2e-06, "loss": 3.8727, "step": 145 }, { "epoch": 0.34360189573459715, "step": 145, "train/alpha": 0.2, "train/clean_ce_loss": 0.09521484375, "train/clean_loss_scale": 8.578706741333008, "train/clean_loss_weight": 1.7157413959503174, "train/combined_loss": 0.9801843166351318, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.034360189573459717, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8168202638626099 }, { "epoch": 0.34360189573459715, "step": 145, "train/alpha": 0.2, "train/clean_ce_loss": 0.07470703125, "train/clean_loss_scale": 9.918815612792969, "train/clean_loss_weight": 1.9837630987167358, "train/combined_loss": 0.8892063498497009, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.034360189573459717, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7410053014755249 }, { "epoch": 0.34360189573459715, "step": 145, "train/alpha": 0.2, "train/clean_ce_loss": 0.061279296875, "train/clean_loss_scale": 9.255725860595703, "train/clean_loss_weight": 1.8511451482772827, "train/combined_loss": 0.6806212663650513, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.034360189573459717, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5671843886375427 }, { "epoch": 0.34360189573459715, "step": 145, "train/alpha": 0.2, "train/clean_ce_loss": 0.09375, "train/clean_loss_scale": 12.55984115600586, "train/clean_loss_weight": 2.5119683742523193, "train/combined_loss": 1.4129821062088013, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.034360189573459717, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1774851083755493 }, { "epoch": 0.3459715639810427, "step": 146, "train/alpha": 0.2, "train/clean_ce_loss": 0.1162109375, "train/clean_loss_scale": 7.577042579650879, "train/clean_loss_weight": 1.5154085159301758, "train/combined_loss": 1.0566422939300537, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.03459715639810427, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8805352449417114 }, { "epoch": 0.3459715639810427, "step": 146, "train/alpha": 0.2, "train/clean_ce_loss": 0.060791015625, "train/clean_loss_scale": 7.489643096923828, "train/clean_loss_weight": 1.4979286193847656, "train/combined_loss": 0.5463635921478271, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.03459715639810427, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45530301332473755 }, { "epoch": 0.3459715639810427, "step": 146, "train/alpha": 0.2, "train/clean_ce_loss": 0.0576171875, "train/clean_loss_scale": 10.241867065429688, "train/clean_loss_weight": 2.0483734607696533, "train/combined_loss": 0.7081290483474731, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.03459715639810427, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5901075601577759 }, { "epoch": 0.3459715639810427, "step": 146, "train/alpha": 0.2, "train/clean_ce_loss": 0.0751953125, "train/clean_loss_scale": 9.779064178466797, "train/clean_loss_weight": 1.9558128118515015, "train/combined_loss": 0.8824077844619751, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.03459715639810427, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7353398203849792 }, { "epoch": 0.34834123222748814, "step": 147, "train/alpha": 0.2, "train/clean_ce_loss": 0.169921875, "train/clean_loss_scale": 6.2783427238464355, "train/clean_loss_weight": 1.2556685209274292, "train/combined_loss": 1.2801933288574219, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.034834123222748814, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0668277740478516 }, { "epoch": 0.34834123222748814, "step": 147, "train/alpha": 0.2, "train/clean_ce_loss": 0.09423828125, "train/clean_loss_scale": 8.312453269958496, "train/clean_loss_weight": 1.662490725517273, "train/combined_loss": 0.9400215744972229, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.034834123222748814, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7833513021469116 }, { "epoch": 0.34834123222748814, "step": 147, "train/alpha": 0.2, "train/clean_ce_loss": 0.040283203125, "train/clean_loss_scale": 20.3404598236084, "train/clean_loss_weight": 4.068091869354248, "train/combined_loss": 0.9832546710968018, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.034834123222748814, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8193789124488831 }, { "epoch": 0.34834123222748814, "step": 147, "train/alpha": 0.2, "train/clean_ce_loss": 0.10693359375, "train/clean_loss_scale": 7.638242721557617, "train/clean_loss_weight": 1.5276485681533813, "train/combined_loss": 0.9801416993141174, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.034834123222748814, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8167847394943237 }, { "epoch": 0.35071090047393366, "step": 148, "train/alpha": 0.2, "train/clean_ce_loss": 0.1572265625, "train/clean_loss_scale": 8.213349342346191, "train/clean_loss_weight": 1.642669916152954, "train/combined_loss": 1.5496280193328857, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.035071090047393366, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2913566827774048 }, { "epoch": 0.35071090047393366, "step": 148, "train/alpha": 0.2, "train/clean_ce_loss": 0.16796875, "train/clean_loss_scale": 9.169893264770508, "train/clean_loss_weight": 1.8339786529541016, "train/combined_loss": 1.848306655883789, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.035071090047393366, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5402555465698242 }, { "epoch": 0.35071090047393366, "step": 148, "train/alpha": 0.2, "train/clean_ce_loss": 0.05908203125, "train/clean_loss_scale": 11.359183311462402, "train/clean_loss_weight": 2.271836757659912, "train/combined_loss": 0.8053483366966248, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.035071090047393366, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6711236238479614 }, { "epoch": 0.35071090047393366, "step": 148, "train/alpha": 0.2, "train/clean_ce_loss": 0.08984375, "train/clean_loss_scale": 8.438328742980957, "train/clean_loss_weight": 1.6876658201217651, "train/combined_loss": 0.9097573161125183, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.035071090047393366, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7581310868263245 }, { "epoch": 0.35308056872037913, "step": 149, "train/alpha": 0.2, "train/clean_ce_loss": 0.1357421875, "train/clean_loss_scale": 8.162874221801758, "train/clean_loss_weight": 1.6325749158859253, "train/combined_loss": 1.329655647277832, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.03530805687203791, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1080464124679565 }, { "epoch": 0.35308056872037913, "step": 149, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 7.558798789978027, "train/clean_loss_weight": 1.5117597579956055, "train/combined_loss": 0.5890548229217529, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.03530805687203791, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.49087902903556824 }, { "epoch": 0.35308056872037913, "step": 149, "train/alpha": 0.2, "train/clean_ce_loss": 0.126953125, "train/clean_loss_scale": 9.848289489746094, "train/clean_loss_weight": 1.9696578979492188, "train/combined_loss": 1.500325322151184, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.03530805687203791, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2502710819244385 }, { "epoch": 0.35308056872037913, "step": 149, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 7.527474880218506, "train/clean_loss_weight": 1.5054949522018433, "train/combined_loss": 0.789502739906311, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.03530805687203791, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6579189300537109 }, { "epoch": 0.35545023696682465, "grad_norm": 13.601014137268066, "learning_rate": 2e-06, "loss": 3.8446, "step": 150 }, { "epoch": 0.35545023696682465, "step": 150, "train/alpha": 0.2, "train/clean_ce_loss": 0.10009765625, "train/clean_loss_scale": 7.61848783493042, "train/clean_loss_weight": 1.5236976146697998, "train/combined_loss": 0.9151113629341125, "train/pos_loss/pos0": 0.27050095796585083, "train/pos_loss/pos1": 0.6033226847648621, "train/pos_loss/pos2": 0.9666163325309753, "train/pos_loss/pos3": 1.5361974239349365, "train/pos_loss/pos4": 1.9372174739837646, "train/pos_loss/pos5": 2.333197832107544, "train/pos_loss/pos6": 2.628159761428833, "train/pos_loss/weight0": 2.5988268852233887, "train/pos_loss/weight1": 1.9007154703140259, "train/pos_loss/weight2": 1.2131625413894653, "train/pos_loss/weight3": 0.7093943357467651, "train/pos_loss/weight4": 0.35385334491729736, "train/pos_loss/weight5": 0.15233147144317627, "train/pos_loss/weight6": 0.07171618193387985, "train/pos_weight_progress": 0.035545023696682464, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7625927925109863 }, { "epoch": 0.35545023696682465, "step": 150, "train/alpha": 0.2, "train/clean_ce_loss": 0.15234375, "train/clean_loss_scale": 7.202974796295166, "train/clean_loss_weight": 1.440595030784607, "train/combined_loss": 1.3167937994003296, "train/pos_loss/pos0": 0.45591405034065247, "train/pos_loss/pos1": 0.9680526852607727, "train/pos_loss/pos2": 1.4397096633911133, "train/pos_loss/pos3": 2.2202308177948, "train/pos_loss/pos4": 2.7513110637664795, "train/pos_loss/pos5": 3.313688039779663, "train/pos_loss/pos6": 3.629721164703369, "train/pos_loss/weight0": 2.7193641662597656, "train/pos_loss/weight1": 1.969496250152588, "train/pos_loss/weight2": 1.1997833251953125, "train/pos_loss/weight3": 0.6566296815872192, "train/pos_loss/weight4": 0.294038861989975, "train/pos_loss/weight5": 0.11428213864564896, "train/pos_loss/weight6": 0.04640538617968559, "train/pos_weight_progress": 0.035545023696682464, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0973281860351562 }, { "epoch": 0.35545023696682465, "step": 150, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 6.5739850997924805, "train/clean_loss_weight": 1.314797043800354, "train/combined_loss": 0.6317188739776611, "train/pos_loss/pos0": 0.2390361726284027, "train/pos_loss/pos1": 0.5551248788833618, "train/pos_loss/pos2": 0.8540620803833008, "train/pos_loss/pos3": 1.270583987236023, "train/pos_loss/pos4": 1.6515620946884155, "train/pos_loss/pos5": 2.057804584503174, "train/pos_loss/pos6": 2.5473270416259766, "train/pos_loss/weight0": 3.264051914215088, "train/pos_loss/weight1": 2.070547342300415, "train/pos_loss/weight2": 1.0577081441879272, "train/pos_loss/weight3": 0.44122135639190674, "train/pos_loss/weight4": 0.13129501044750214, "train/pos_loss/weight5": 0.028756065294146538, "train/pos_loss/weight6": 0.006420147605240345, "train/pos_weight_progress": 0.035545023696682464, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5264323949813843 }, { "epoch": 0.35545023696682465, "step": 150, "train/alpha": 0.2, "train/clean_ce_loss": 0.138671875, "train/clean_loss_scale": 8.485146522521973, "train/clean_loss_weight": 1.6970293521881104, "train/combined_loss": 1.4119813442230225, "train/pos_loss/pos0": 0.39350733160972595, "train/pos_loss/pos1": 0.9125859141349792, "train/pos_loss/pos2": 1.567943811416626, "train/pos_loss/pos3": 2.157409191131592, "train/pos_loss/pos4": 2.729592800140381, "train/pos_loss/pos5": 3.4432332515716553, "train/pos_loss/pos6": 3.6086764335632324, "train/pos_loss/weight0": 2.5122737884521484, "train/pos_loss/weight1": 1.8495516777038574, "train/pos_loss/weight2": 1.2261273860931396, "train/pos_loss/weight3": 0.739139199256897, "train/pos_loss/weight4": 0.39019614458084106, "train/pos_loss/weight5": 0.1862480342388153, "train/pos_loss/weight6": 0.09646444767713547, "train/pos_weight_progress": 0.035545023696682464, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.176651120185852 }, { "epoch": 0.3578199052132701, "step": 151, "train/alpha": 0.2, "train/clean_ce_loss": 0.05322265625, "train/clean_loss_scale": 7.0953688621521, "train/clean_loss_weight": 1.4190738201141357, "train/combined_loss": 0.45316123962402344, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.035781990521327016, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3776343762874603 }, { "epoch": 0.3578199052132701, "step": 151, "train/alpha": 0.2, "train/clean_ce_loss": 0.1533203125, "train/clean_loss_scale": 8.144731521606445, "train/clean_loss_weight": 1.628946304321289, "train/combined_loss": 1.4985032081604004, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.035781990521327016, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2487527132034302 }, { "epoch": 0.3578199052132701, "step": 151, "train/alpha": 0.2, "train/clean_ce_loss": 0.10546875, "train/clean_loss_scale": 7.2882080078125, "train/clean_loss_weight": 1.4576416015625, "train/combined_loss": 0.9224138259887695, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.035781990521327016, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7686781883239746 }, { "epoch": 0.3578199052132701, "step": 151, "train/alpha": 0.2, "train/clean_ce_loss": 0.09814453125, "train/clean_loss_scale": 6.498228549957275, "train/clean_loss_weight": 1.2996457815170288, "train/combined_loss": 0.7653186917304993, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.035781990521327016, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6377655863761902 }, { "epoch": 0.36018957345971564, "step": 152, "train/alpha": 0.2, "train/clean_ce_loss": 0.1328125, "train/clean_loss_scale": 8.556018829345703, "train/clean_loss_weight": 1.7112038135528564, "train/combined_loss": 1.3636155128479004, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.03601895734597156, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1363462209701538 }, { "epoch": 0.36018957345971564, "step": 152, "train/alpha": 0.2, "train/clean_ce_loss": 0.0732421875, "train/clean_loss_scale": 7.612486839294434, "train/clean_loss_weight": 1.5224974155426025, "train/combined_loss": 0.6690662503242493, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.03601895734597156, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5575551986694336 }, { "epoch": 0.36018957345971564, "step": 152, "train/alpha": 0.2, "train/clean_ce_loss": 0.0849609375, "train/clean_loss_scale": 6.306519031524658, "train/clean_loss_weight": 1.2613037824630737, "train/combined_loss": 0.6429693698883057, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.03601895734597156, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5358077883720398 }, { "epoch": 0.36018957345971564, "step": 152, "train/alpha": 0.2, "train/clean_ce_loss": 0.1337890625, "train/clean_loss_scale": 8.375715255737305, "train/clean_loss_weight": 1.6751431226730347, "train/combined_loss": 1.344694972038269, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.03601895734597156, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1205791234970093 }, { "epoch": 0.36255924170616116, "step": 153, "train/alpha": 0.2, "train/clean_ce_loss": 0.08251953125, "train/clean_loss_scale": 9.384481430053711, "train/clean_loss_weight": 1.8768962621688843, "train/combined_loss": 0.929283618927002, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.03625592417061611, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7744030356407166 }, { "epoch": 0.36255924170616116, "step": 153, "train/alpha": 0.2, "train/clean_ce_loss": 0.12451171875, "train/clean_loss_scale": 6.371980667114258, "train/clean_loss_weight": 1.2743961811065674, "train/combined_loss": 0.9520635604858398, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.03625592417061611, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7933862805366516 }, { "epoch": 0.36255924170616116, "step": 153, "train/alpha": 0.2, "train/clean_ce_loss": 0.123046875, "train/clean_loss_scale": 9.595477104187012, "train/clean_loss_weight": 1.9190953969955444, "train/combined_loss": 1.4168322086334229, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.03625592417061611, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.180693507194519 }, { "epoch": 0.36255924170616116, "step": 153, "train/alpha": 0.2, "train/clean_ce_loss": 0.1650390625, "train/clean_loss_scale": 6.028199195861816, "train/clean_loss_weight": 1.2056398391723633, "train/combined_loss": 1.1938660144805908, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.03625592417061611, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9948883056640625 }, { "epoch": 0.36492890995260663, "step": 154, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 6.7786455154418945, "train/clean_loss_weight": 1.355729103088379, "train/combined_loss": 0.5401732921600342, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.036492890995260666, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4501444399356842 }, { "epoch": 0.36492890995260663, "step": 154, "train/alpha": 0.2, "train/clean_ce_loss": 0.10498046875, "train/clean_loss_scale": 6.563847541809082, "train/clean_loss_weight": 1.3127695322036743, "train/combined_loss": 0.8268909454345703, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.036492890995260666, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.689075767993927 }, { "epoch": 0.36492890995260663, "step": 154, "train/alpha": 0.2, "train/clean_ce_loss": 0.140625, "train/clean_loss_scale": 6.710843563079834, "train/clean_loss_weight": 1.3421686887741089, "train/combined_loss": 1.1324548721313477, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.036492890995260666, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9437123537063599 }, { "epoch": 0.36492890995260663, "step": 154, "train/alpha": 0.2, "train/clean_ce_loss": 0.134765625, "train/clean_loss_scale": 9.113224029541016, "train/clean_loss_weight": 1.822644829750061, "train/combined_loss": 1.4737792015075684, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.036492890995260666, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2281492948532104 }, { "epoch": 0.36729857819905215, "grad_norm": 14.113810539245605, "learning_rate": 2e-06, "loss": 3.8837, "step": 155 }, { "epoch": 0.36729857819905215, "step": 155, "train/alpha": 0.2, "train/clean_ce_loss": 0.099609375, "train/clean_loss_scale": 6.928337574005127, "train/clean_loss_weight": 1.3856675624847412, "train/combined_loss": 0.8281528353691101, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.03672985781990521, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6901273727416992 }, { "epoch": 0.36729857819905215, "step": 155, "train/alpha": 0.2, "train/clean_ce_loss": 0.06982421875, "train/clean_loss_scale": 7.057004451751709, "train/clean_loss_weight": 1.4114009141921997, "train/combined_loss": 0.5912997722625732, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.03672985781990521, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.49274981021881104 }, { "epoch": 0.36729857819905215, "step": 155, "train/alpha": 0.2, "train/clean_ce_loss": 0.134765625, "train/clean_loss_scale": 5.667751312255859, "train/clean_loss_weight": 1.1335502862930298, "train/combined_loss": 0.9165816307067871, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.03672985781990521, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7638180255889893 }, { "epoch": 0.36729857819905215, "step": 155, "train/alpha": 0.2, "train/clean_ce_loss": 0.1591796875, "train/clean_loss_scale": 5.660552501678467, "train/clean_loss_weight": 1.1321104764938354, "train/combined_loss": 1.081254005432129, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.03672985781990521, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9010449647903442 }, { "epoch": 0.3696682464454976, "step": 156, "train/alpha": 0.2, "train/clean_ce_loss": 0.1123046875, "train/clean_loss_scale": 8.483402252197266, "train/clean_loss_weight": 1.6966804265975952, "train/combined_loss": 1.1432709693908691, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.03696682464454976, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9527258276939392 }, { "epoch": 0.3696682464454976, "step": 156, "train/alpha": 0.2, "train/clean_ce_loss": 0.126953125, "train/clean_loss_scale": 8.06838321685791, "train/clean_loss_weight": 1.61367666721344, "train/combined_loss": 1.2291676998138428, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.03696682464454976, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0243064165115356 }, { "epoch": 0.3696682464454976, "step": 156, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 6.04306697845459, "train/clean_loss_weight": 1.208613395690918, "train/combined_loss": 0.47093433141708374, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.03696682464454976, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.39244526624679565 }, { "epoch": 0.3696682464454976, "step": 156, "train/alpha": 0.2, "train/clean_ce_loss": 0.09716796875, "train/clean_loss_scale": 7.411225318908691, "train/clean_loss_weight": 1.4822450876235962, "train/combined_loss": 0.8641604781150818, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.03696682464454976, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7201337218284607 }, { "epoch": 0.37203791469194314, "step": 157, "train/alpha": 0.2, "train/clean_ce_loss": 0.0849609375, "train/clean_loss_scale": 7.061115741729736, "train/clean_loss_weight": 1.412223219871521, "train/combined_loss": 0.7199028134346008, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.037203791469194315, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5999190211296082 }, { "epoch": 0.37203791469194314, "step": 157, "train/alpha": 0.2, "train/clean_ce_loss": 0.1630859375, "train/clean_loss_scale": 6.025681495666504, "train/clean_loss_weight": 1.2051362991333008, "train/combined_loss": 1.1792447566986084, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.037203791469194315, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9827039241790771 }, { "epoch": 0.37203791469194314, "step": 157, "train/alpha": 0.2, "train/clean_ce_loss": 0.06982421875, "train/clean_loss_scale": 6.637152194976807, "train/clean_loss_weight": 1.3274304866790771, "train/combined_loss": 0.556120753288269, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.037203791469194315, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46343398094177246 }, { "epoch": 0.37203791469194314, "step": 157, "train/alpha": 0.2, "train/clean_ce_loss": 0.1640625, "train/clean_loss_scale": 6.811182498931885, "train/clean_loss_weight": 1.362236499786377, "train/combined_loss": 1.3409515619277954, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.037203791469194315, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1174596548080444 }, { "epoch": 0.3744075829383886, "step": 158, "train/alpha": 0.2, "train/clean_ce_loss": 0.055908203125, "train/clean_loss_scale": 6.548968315124512, "train/clean_loss_weight": 1.3097937107086182, "train/combined_loss": 0.439369261264801, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.03744075829383886, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.36614105105400085 }, { "epoch": 0.3744075829383886, "step": 158, "train/alpha": 0.2, "train/clean_ce_loss": 0.0830078125, "train/clean_loss_scale": 6.377126693725586, "train/clean_loss_weight": 1.2754253149032593, "train/combined_loss": 0.6352216005325317, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.03744075829383886, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5293513536453247 }, { "epoch": 0.3744075829383886, "step": 158, "train/alpha": 0.2, "train/clean_ce_loss": 0.0869140625, "train/clean_loss_scale": 5.828486919403076, "train/clean_loss_weight": 1.165697455406189, "train/combined_loss": 0.6078929901123047, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.03744075829383886, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5065774917602539 }, { "epoch": 0.3744075829383886, "step": 158, "train/alpha": 0.2, "train/clean_ce_loss": 0.1748046875, "train/clean_loss_scale": 6.345023155212402, "train/clean_loss_weight": 1.2690047025680542, "train/combined_loss": 1.3309677839279175, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.03744075829383886, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1091398000717163 }, { "epoch": 0.3767772511848341, "step": 159, "train/alpha": 0.2, "train/clean_ce_loss": 0.1748046875, "train/clean_loss_scale": 5.651362895965576, "train/clean_loss_weight": 1.130272626876831, "train/combined_loss": 1.1854617595672607, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.03767772511848341, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9878847599029541 }, { "epoch": 0.3767772511848341, "step": 159, "train/alpha": 0.2, "train/clean_ce_loss": 0.1279296875, "train/clean_loss_scale": 5.471645832061768, "train/clean_loss_weight": 1.0943292379379272, "train/combined_loss": 0.8399831056594849, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.03767772511848341, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.699985921382904 }, { "epoch": 0.3767772511848341, "step": 159, "train/alpha": 0.2, "train/clean_ce_loss": 0.0771484375, "train/clean_loss_scale": 6.627222537994385, "train/clean_loss_weight": 1.3254445791244507, "train/combined_loss": 0.6135358810424805, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.03767772511848341, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5112798810005188 }, { "epoch": 0.3767772511848341, "step": 159, "train/alpha": 0.2, "train/clean_ce_loss": 0.1376953125, "train/clean_loss_scale": 4.986250877380371, "train/clean_loss_weight": 0.9972501993179321, "train/combined_loss": 0.8239001035690308, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.03767772511848341, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.686583399772644 }, { "epoch": 0.3791469194312796, "grad_norm": 10.901315689086914, "learning_rate": 2e-06, "loss": 3.7629, "step": 160 }, { "epoch": 0.3791469194312796, "step": 160, "train/alpha": 0.2, "train/clean_ce_loss": 0.1220703125, "train/clean_loss_scale": 9.009074211120605, "train/clean_loss_weight": 1.8018149137496948, "train/combined_loss": 1.3196885585784912, "train/pos_loss/pos0": 0.6181595921516418, "train/pos_loss/pos1": 1.0459622144699097, "train/pos_loss/pos2": 1.715496301651001, "train/pos_loss/pos3": 2.3450675010681152, "train/pos_loss/pos4": 2.9441707134246826, "train/pos_loss/pos5": 3.1286420822143555, "train/pos_loss/pos6": 3.5586068630218506, "train/pos_loss/weight0": 3.1567883491516113, "train/pos_loss/weight1": 2.0525219440460205, "train/pos_loss/weight2": 1.1386514902114868, "train/pos_loss/weight3": 0.46345415711402893, "train/pos_loss/weight4": 0.14746417105197906, "train/pos_loss/weight5": 0.033805813640356064, "train/pos_loss/weight6": 0.007313992828130722, "train/pos_weight_progress": 0.037914691943127965, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0997405052185059 }, { "epoch": 0.3791469194312796, "step": 160, "train/alpha": 0.2, "train/clean_ce_loss": 0.15625, "train/clean_loss_scale": 7.6157121658325195, "train/clean_loss_weight": 1.5231424570083618, "train/combined_loss": 1.4279459714889526, "train/pos_loss/pos0": 0.7637249231338501, "train/pos_loss/pos1": 1.144392728805542, "train/pos_loss/pos2": 1.8144288063049316, "train/pos_loss/pos3": 2.386403799057007, "train/pos_loss/pos4": 3.0074100494384766, "train/pos_loss/pos5": 3.39290452003479, "train/pos_loss/pos6": 3.596510887145996, "train/pos_loss/weight0": 3.286459445953369, "train/pos_loss/weight1": 2.051893949508667, "train/pos_loss/weight2": 1.062001347541809, "train/pos_loss/weight3": 0.42908117175102234, "train/pos_loss/weight4": 0.13345350325107574, "train/pos_loss/weight5": 0.030137278139591217, "train/pos_loss/weight6": 0.006973105017095804, "train/pos_weight_progress": 0.037914691943127965, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1899549961090088 }, { "epoch": 0.3791469194312796, "step": 160, "train/alpha": 0.2, "train/clean_ce_loss": 0.068359375, "train/clean_loss_scale": 6.044639587402344, "train/clean_loss_weight": 1.2089279890060425, "train/combined_loss": 0.4958493709564209, "train/pos_loss/pos0": 0.19671860337257385, "train/pos_loss/pos1": 0.4347118139266968, "train/pos_loss/pos2": 0.656646728515625, "train/pos_loss/pos3": 0.9204391241073608, "train/pos_loss/pos4": 1.3718469142913818, "train/pos_loss/pos5": 1.6927043199539185, "train/pos_loss/pos6": 2.1672933101654053, "train/pos_loss/weight0": 3.3331844806671143, "train/pos_loss/weight1": 1.9822107553482056, "train/pos_loss/weight2": 1.0636850595474243, "train/pos_loss/weight3": 0.4282400608062744, "train/pos_loss/weight4": 0.14507576823234558, "train/pos_loss/weight5": 0.037343524396419525, "train/pos_loss/weight6": 0.010260578244924545, "train/pos_weight_progress": 0.037914691943127965, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4132077991962433 }, { "epoch": 0.3791469194312796, "step": 160, "train/alpha": 0.2, "train/clean_ce_loss": 0.09912109375, "train/clean_loss_scale": 10.73791790008545, "train/clean_loss_weight": 2.1475837230682373, "train/combined_loss": 1.2772250175476074, "train/pos_loss/pos0": 0.4035336673259735, "train/pos_loss/pos1": 0.8344869613647461, "train/pos_loss/pos2": 1.3023138046264648, "train/pos_loss/pos3": 1.7133647203445435, "train/pos_loss/pos4": 2.1694843769073486, "train/pos_loss/pos5": 2.5863585472106934, "train/pos_loss/pos6": 2.8976290225982666, "train/pos_loss/weight0": 2.3509113788604736, "train/pos_loss/weight1": 1.775519609451294, "train/pos_loss/weight2": 1.2275376319885254, "train/pos_loss/weight3": 0.7859495282173157, "train/pos_loss/weight4": 0.4627910256385803, "train/pos_loss/weight5": 0.2506747841835022, "train/pos_loss/weight6": 0.146615669131279, "train/pos_weight_progress": 0.037914691943127965, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0643541812896729 }, { "epoch": 0.3815165876777251, "step": 161, "train/alpha": 0.2, "train/clean_ce_loss": 0.12109375, "train/clean_loss_scale": 11.163617134094238, "train/clean_loss_weight": 2.2327234745025635, "train/combined_loss": 1.6222131252288818, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03815165876777251, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.351844310760498 }, { "epoch": 0.3815165876777251, "step": 161, "train/alpha": 0.2, "train/clean_ce_loss": 0.12158203125, "train/clean_loss_scale": 6.189572811126709, "train/clean_loss_weight": 1.2379145622253418, "train/combined_loss": 0.9030489921569824, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03815165876777251, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7525408267974854 }, { "epoch": 0.3815165876777251, "step": 161, "train/alpha": 0.2, "train/clean_ce_loss": 0.0947265625, "train/clean_loss_scale": 6.010976791381836, "train/clean_loss_weight": 1.202195405960083, "train/combined_loss": 0.6832790374755859, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03815165876777251, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5693991780281067 }, { "epoch": 0.3815165876777251, "step": 161, "train/alpha": 0.2, "train/clean_ce_loss": 0.1083984375, "train/clean_loss_scale": 7.088913440704346, "train/clean_loss_weight": 1.4177826642990112, "train/combined_loss": 0.9221125841140747, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03815165876777251, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7684271335601807 }, { "epoch": 0.38388625592417064, "step": 162, "train/alpha": 0.2, "train/clean_ce_loss": 0.11181640625, "train/clean_loss_scale": 6.686613082885742, "train/clean_loss_weight": 1.3373225927352905, "train/combined_loss": 0.8972076177597046, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03838862559241706, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7476730346679688 }, { "epoch": 0.38388625592417064, "step": 162, "train/alpha": 0.2, "train/clean_ce_loss": 0.11181640625, "train/clean_loss_scale": 9.388303756713867, "train/clean_loss_weight": 1.8776607513427734, "train/combined_loss": 1.259719729423523, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03838862559241706, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0497664213180542 }, { "epoch": 0.38388625592417064, "step": 162, "train/alpha": 0.2, "train/clean_ce_loss": 0.10400390625, "train/clean_loss_scale": 7.463063716888428, "train/clean_loss_weight": 1.4926127195358276, "train/combined_loss": 0.9314253330230713, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03838862559241706, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7761877775192261 }, { "epoch": 0.38388625592417064, "step": 162, "train/alpha": 0.2, "train/clean_ce_loss": 0.1240234375, "train/clean_loss_scale": 11.034285545349121, "train/clean_loss_weight": 2.206857204437256, "train/combined_loss": 1.6422120332717896, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03838862559241706, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3685100078582764 }, { "epoch": 0.3862559241706161, "step": 163, "train/alpha": 0.2, "train/clean_ce_loss": 0.0908203125, "train/clean_loss_scale": 8.596819877624512, "train/clean_loss_weight": 1.719364047050476, "train/combined_loss": 0.936919093132019, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.038625592417061615, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7807658910751343 }, { "epoch": 0.3862559241706161, "step": 163, "train/alpha": 0.2, "train/clean_ce_loss": 0.07666015625, "train/clean_loss_scale": 8.149809837341309, "train/clean_loss_weight": 1.6299619674682617, "train/combined_loss": 0.7497188448905945, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.038625592417061615, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6247656941413879 }, { "epoch": 0.3862559241706161, "step": 163, "train/alpha": 0.2, "train/clean_ce_loss": 0.1708984375, "train/clean_loss_scale": 8.081687927246094, "train/clean_loss_weight": 1.6163376569747925, "train/combined_loss": 1.6573774814605713, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.038625592417061615, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.381147861480713 }, { "epoch": 0.3862559241706161, "step": 163, "train/alpha": 0.2, "train/clean_ce_loss": 0.0888671875, "train/clean_loss_scale": 7.414450168609619, "train/clean_loss_weight": 1.482890009880066, "train/combined_loss": 0.7906816005706787, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.038625592417061615, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6589013338088989 }, { "epoch": 0.3886255924170616, "step": 164, "train/alpha": 0.2, "train/clean_ce_loss": 0.09326171875, "train/clean_loss_scale": 6.262844085693359, "train/clean_loss_weight": 1.2525688409805298, "train/combined_loss": 0.7009003162384033, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03886255924170616, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.584083616733551 }, { "epoch": 0.3886255924170616, "step": 164, "train/alpha": 0.2, "train/clean_ce_loss": 0.0458984375, "train/clean_loss_scale": 14.888927459716797, "train/clean_loss_weight": 2.977785587310791, "train/combined_loss": 0.8200542330741882, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03886255924170616, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6833785176277161 }, { "epoch": 0.3886255924170616, "step": 164, "train/alpha": 0.2, "train/clean_ce_loss": 0.0703125, "train/clean_loss_scale": 8.465179443359375, "train/clean_loss_weight": 1.6930359601974487, "train/combined_loss": 0.7142494916915894, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03886255924170616, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.595207929611206 }, { "epoch": 0.3886255924170616, "step": 164, "train/alpha": 0.2, "train/clean_ce_loss": 0.022216796875, "train/clean_loss_scale": 11.936967849731445, "train/clean_loss_weight": 2.3873937129974365, "train/combined_loss": 0.3182414174079895, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03886255924170616, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2652011811733246 }, { "epoch": 0.3909952606635071, "grad_norm": 14.233325958251953, "learning_rate": 2e-06, "loss": 3.6518, "step": 165 }, { "epoch": 0.3909952606635071, "step": 165, "train/alpha": 0.2, "train/clean_ce_loss": 0.04833984375, "train/clean_loss_scale": 6.3120646476745605, "train/clean_loss_weight": 1.2624129056930542, "train/combined_loss": 0.36614906787872314, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03909952606635071, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3051242232322693 }, { "epoch": 0.3909952606635071, "step": 165, "train/alpha": 0.2, "train/clean_ce_loss": 0.12451171875, "train/clean_loss_scale": 9.830306053161621, "train/clean_loss_weight": 1.9660612344741821, "train/combined_loss": 1.4687860012054443, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03909952606635071, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2239882946014404 }, { "epoch": 0.3909952606635071, "step": 165, "train/alpha": 0.2, "train/clean_ce_loss": 0.1455078125, "train/clean_loss_scale": 8.435701370239258, "train/clean_loss_weight": 1.6871403455734253, "train/combined_loss": 1.4729524850845337, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03909952606635071, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2274603843688965 }, { "epoch": 0.3909952606635071, "step": 165, "train/alpha": 0.2, "train/clean_ce_loss": 0.0830078125, "train/clean_loss_scale": 8.520414352416992, "train/clean_loss_weight": 1.7040828466415405, "train/combined_loss": 0.8487131595611572, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03909952606635071, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7072609663009644 }, { "epoch": 0.3933649289099526, "step": 166, "train/alpha": 0.2, "train/clean_ce_loss": 0.142578125, "train/clean_loss_scale": 6.979774475097656, "train/clean_loss_weight": 1.395954966545105, "train/combined_loss": 1.1941957473754883, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03933649289099526, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9951631426811218 }, { "epoch": 0.3933649289099526, "step": 166, "train/alpha": 0.2, "train/clean_ce_loss": 0.1171875, "train/clean_loss_scale": 5.765474319458008, "train/clean_loss_weight": 1.1530948877334595, "train/combined_loss": 0.8107698559761047, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03933649289099526, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6756415367126465 }, { "epoch": 0.3933649289099526, "step": 166, "train/alpha": 0.2, "train/clean_ce_loss": 0.06689453125, "train/clean_loss_scale": 10.54568099975586, "train/clean_loss_weight": 2.1091363430023193, "train/combined_loss": 0.8465380668640137, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03933649289099526, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7054483890533447 }, { "epoch": 0.3933649289099526, "step": 166, "train/alpha": 0.2, "train/clean_ce_loss": 0.09619140625, "train/clean_loss_scale": 8.310202598571777, "train/clean_loss_weight": 1.6620405912399292, "train/combined_loss": 0.9592441320419312, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03933649289099526, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7993701100349426 }, { "epoch": 0.3957345971563981, "step": 167, "train/alpha": 0.2, "train/clean_ce_loss": 0.0869140625, "train/clean_loss_scale": 5.977802276611328, "train/clean_loss_weight": 1.1955604553222656, "train/combined_loss": 0.6234661340713501, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03957345971563981, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5195550918579102 }, { "epoch": 0.3957345971563981, "step": 167, "train/alpha": 0.2, "train/clean_ce_loss": 0.10986328125, "train/clean_loss_scale": 5.95251989364624, "train/clean_loss_weight": 1.1905039548873901, "train/combined_loss": 0.784756064414978, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03957345971563981, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6539633870124817 }, { "epoch": 0.3957345971563981, "step": 167, "train/alpha": 0.2, "train/clean_ce_loss": 0.12255859375, "train/clean_loss_scale": 9.476874351501465, "train/clean_loss_weight": 1.8953748941421509, "train/combined_loss": 1.3937668800354004, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03957345971563981, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1614724397659302 }, { "epoch": 0.3957345971563981, "step": 167, "train/alpha": 0.2, "train/clean_ce_loss": 0.11279296875, "train/clean_loss_scale": 8.447020530700684, "train/clean_loss_weight": 1.6894041299819946, "train/combined_loss": 1.143317461013794, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03957345971563981, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9527645707130432 }, { "epoch": 0.3981042654028436, "step": 168, "train/alpha": 0.2, "train/clean_ce_loss": 0.1298828125, "train/clean_loss_scale": 10.151069641113281, "train/clean_loss_weight": 2.0302140712738037, "train/combined_loss": 1.5821393728256226, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03981042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3184494972229004 }, { "epoch": 0.3981042654028436, "step": 168, "train/alpha": 0.2, "train/clean_ce_loss": 0.1484375, "train/clean_loss_scale": 8.360089302062988, "train/clean_loss_weight": 1.6720179319381714, "train/combined_loss": 1.4891408681869507, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03981042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2409507036209106 }, { "epoch": 0.3981042654028436, "step": 168, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 7.244144439697266, "train/clean_loss_weight": 1.448828935623169, "train/combined_loss": 0.7046062350273132, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03981042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5871718525886536 }, { "epoch": 0.3981042654028436, "step": 168, "train/alpha": 0.2, "train/clean_ce_loss": 0.1474609375, "train/clean_loss_scale": 8.052123069763184, "train/clean_loss_weight": 1.6104246377944946, "train/combined_loss": 1.4248483180999756, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.03981042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1873736381530762 }, { "epoch": 0.4004739336492891, "step": 169, "train/alpha": 0.2, "train/clean_ce_loss": 0.09814453125, "train/clean_loss_scale": 7.057665824890137, "train/clean_loss_weight": 1.411533236503601, "train/combined_loss": 0.8312055468559265, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.04004739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6926712989807129 }, { "epoch": 0.4004739336492891, "step": 169, "train/alpha": 0.2, "train/clean_ce_loss": 0.1015625, "train/clean_loss_scale": 10.24777889251709, "train/clean_loss_weight": 2.049555778503418, "train/combined_loss": 1.248948097229004, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.04004739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.04079008102417 }, { "epoch": 0.4004739336492891, "step": 169, "train/alpha": 0.2, "train/clean_ce_loss": 0.08203125, "train/clean_loss_scale": 9.007696151733398, "train/clean_loss_weight": 1.8015393018722534, "train/combined_loss": 0.8866950869560242, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.04004739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7389125823974609 }, { "epoch": 0.4004739336492891, "step": 169, "train/alpha": 0.2, "train/clean_ce_loss": 0.0498046875, "train/clean_loss_scale": 9.69593620300293, "train/clean_loss_weight": 1.9391872882843018, "train/combined_loss": 0.579483687877655, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.04004739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.48290306329727173 }, { "epoch": 0.4028436018957346, "grad_norm": 13.909271240234375, "learning_rate": 2e-06, "loss": 3.5119, "step": 170 }, { "epoch": 0.4028436018957346, "step": 170, "train/alpha": 0.2, "train/clean_ce_loss": 0.1123046875, "train/clean_loss_scale": 7.7696027755737305, "train/clean_loss_weight": 1.5539206266403198, "train/combined_loss": 1.0470753908157349, "train/pos_loss/pos0": 0.4320808947086334, "train/pos_loss/pos1": 0.7760789394378662, "train/pos_loss/pos2": 1.2301450967788696, "train/pos_loss/pos3": 1.6888724565505981, "train/pos_loss/pos4": 2.1242496967315674, "train/pos_loss/pos5": 2.5522067546844482, "train/pos_loss/pos6": 2.8013689517974854, "train/pos_loss/weight0": 2.880424976348877, "train/pos_loss/weight1": 1.9590702056884766, "train/pos_loss/weight2": 1.1696763038635254, "train/pos_loss/weight3": 0.6034475564956665, "train/pos_loss/weight4": 0.25626546144485474, "train/pos_loss/weight5": 0.09525129199028015, "train/pos_loss/weight6": 0.035864684730768204, "train/pos_weight_progress": 0.04028436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.872562825679779 }, { "epoch": 0.4028436018957346, "step": 170, "train/alpha": 0.2, "train/clean_ce_loss": 0.09423828125, "train/clean_loss_scale": 7.341911792755127, "train/clean_loss_weight": 1.4683823585510254, "train/combined_loss": 0.8302670121192932, "train/pos_loss/pos0": 0.31303274631500244, "train/pos_loss/pos1": 0.5965776443481445, "train/pos_loss/pos2": 0.9454675912857056, "train/pos_loss/pos3": 1.3429348468780518, "train/pos_loss/pos4": 1.9436591863632202, "train/pos_loss/pos5": 2.350531816482544, "train/pos_loss/pos6": 2.540580987930298, "train/pos_loss/weight0": 2.8703839778900146, "train/pos_loss/weight1": 1.9362461566925049, "train/pos_loss/weight2": 1.170748233795166, "train/pos_loss/weight3": 0.6151950359344482, "train/pos_loss/weight4": 0.26557692885398865, "train/pos_loss/weight5": 0.10272763669490814, "train/pos_loss/weight6": 0.03912203013896942, "train/pos_weight_progress": 0.04028436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6918891668319702 }, { "epoch": 0.4028436018957346, "step": 170, "train/alpha": 0.2, "train/clean_ce_loss": 0.12451171875, "train/clean_loss_scale": 7.536777496337891, "train/clean_loss_weight": 1.5073555707931519, "train/combined_loss": 1.1261005401611328, "train/pos_loss/pos0": 0.35437220335006714, "train/pos_loss/pos1": 0.7742924094200134, "train/pos_loss/pos2": 1.2310445308685303, "train/pos_loss/pos3": 1.7402085065841675, "train/pos_loss/pos4": 2.26389479637146, "train/pos_loss/pos5": 2.6848793029785156, "train/pos_loss/pos6": 3.0557429790496826, "train/pos_loss/weight0": 2.625192642211914, "train/pos_loss/weight1": 1.866754412651062, "train/pos_loss/weight2": 1.2039932012557983, "train/pos_loss/weight3": 0.7020902037620544, "train/pos_loss/weight4": 0.362740695476532, "train/pos_loss/weight5": 0.16232100129127502, "train/pos_loss/weight6": 0.07690819352865219, "train/pos_weight_progress": 0.04028436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9384171366691589 }, { "epoch": 0.4028436018957346, "step": 170, "train/alpha": 0.2, "train/clean_ce_loss": 0.053955078125, "train/clean_loss_scale": 8.58091926574707, "train/clean_loss_weight": 1.7161839008331299, "train/combined_loss": 0.5555809736251831, "train/pos_loss/pos0": 0.2137903869152069, "train/pos_loss/pos1": 0.41864004731178284, "train/pos_loss/pos2": 0.642509400844574, "train/pos_loss/pos3": 0.9942941665649414, "train/pos_loss/pos4": 1.2815443277359009, "train/pos_loss/pos5": 1.7985073328018188, "train/pos_loss/pos6": 2.184751272201538, "train/pos_loss/weight0": 2.960240602493286, "train/pos_loss/weight1": 1.9995319843292236, "train/pos_loss/weight2": 1.1525734663009644, "train/pos_loss/weight3": 0.5695019960403442, "train/pos_loss/weight4": 0.22295376658439636, "train/pos_loss/weight5": 0.07105579972267151, "train/pos_loss/weight6": 0.024142442271113396, "train/pos_weight_progress": 0.04028436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4629841446876526 }, { "epoch": 0.4052132701421801, "step": 171, "train/alpha": 0.2, "train/clean_ce_loss": 0.076171875, "train/clean_loss_scale": 15.391865730285645, "train/clean_loss_weight": 3.0783731937408447, "train/combined_loss": 1.4069128036499023, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04052132701421801, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1724272966384888 }, { "epoch": 0.4052132701421801, "step": 171, "train/alpha": 0.2, "train/clean_ce_loss": 0.03662109375, "train/clean_loss_scale": 13.58178997039795, "train/clean_loss_weight": 2.716357946395874, "train/combined_loss": 0.5968559980392456, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04052132701421801, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.49737998843193054 }, { "epoch": 0.4052132701421801, "step": 171, "train/alpha": 0.2, "train/clean_ce_loss": 0.0966796875, "train/clean_loss_scale": 8.74929141998291, "train/clean_loss_weight": 1.7498582601547241, "train/combined_loss": 1.01505446434021, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04052132701421801, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8458787798881531 }, { "epoch": 0.4052132701421801, "step": 171, "train/alpha": 0.2, "train/clean_ce_loss": 0.07177734375, "train/clean_loss_scale": 10.634531021118164, "train/clean_loss_weight": 2.126906156539917, "train/combined_loss": 0.9159820079803467, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04052132701421801, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7633183598518372 }, { "epoch": 0.4075829383886256, "step": 172, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 8.844223022460938, "train/clean_loss_weight": 1.7688446044921875, "train/combined_loss": 0.8861497044563293, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04075829383886256, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7384580969810486 }, { "epoch": 0.4075829383886256, "step": 172, "train/alpha": 0.2, "train/clean_ce_loss": 0.193359375, "train/clean_loss_scale": 7.530999660491943, "train/clean_loss_weight": 1.5061999559402466, "train/combined_loss": 1.747427225112915, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04075829383886256, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4561893939971924 }, { "epoch": 0.4075829383886256, "step": 172, "train/alpha": 0.2, "train/clean_ce_loss": 0.14453125, "train/clean_loss_scale": 8.671247482299805, "train/clean_loss_weight": 1.734249472618103, "train/combined_loss": 1.5039194822311401, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04075829383886256, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2532662153244019 }, { "epoch": 0.4075829383886256, "step": 172, "train/alpha": 0.2, "train/clean_ce_loss": 0.1142578125, "train/clean_loss_scale": 11.691010475158691, "train/clean_loss_weight": 2.3382022380828857, "train/combined_loss": 1.6029472351074219, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04075829383886256, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3357893228530884 }, { "epoch": 0.4099526066350711, "step": 173, "train/alpha": 0.2, "train/clean_ce_loss": 0.08935546875, "train/clean_loss_scale": 10.21124267578125, "train/clean_loss_weight": 2.042248487472534, "train/combined_loss": 1.0949164628982544, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04099526066350711, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9124304056167603 }, { "epoch": 0.4099526066350711, "step": 173, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 10.728900909423828, "train/clean_loss_weight": 2.145780324935913, "train/combined_loss": 1.0435532331466675, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04099526066350711, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8696277141571045 }, { "epoch": 0.4099526066350711, "step": 173, "train/alpha": 0.2, "train/clean_ce_loss": 0.08837890625, "train/clean_loss_scale": 8.916570663452148, "train/clean_loss_weight": 1.7833141088485718, "train/combined_loss": 0.9456441402435303, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04099526066350711, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7880367636680603 }, { "epoch": 0.4099526066350711, "step": 173, "train/alpha": 0.2, "train/clean_ce_loss": 0.1435546875, "train/clean_loss_scale": 8.817378044128418, "train/clean_loss_weight": 1.7634756565093994, "train/combined_loss": 1.5189311504364014, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04099526066350711, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2657759189605713 }, { "epoch": 0.41232227488151657, "step": 174, "train/alpha": 0.2, "train/clean_ce_loss": 0.07470703125, "train/clean_loss_scale": 11.051148414611816, "train/clean_loss_weight": 2.2102296352386475, "train/combined_loss": 0.9907181859016418, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04123222748815166, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8255984783172607 }, { "epoch": 0.41232227488151657, "step": 174, "train/alpha": 0.2, "train/clean_ce_loss": 0.13671875, "train/clean_loss_scale": 8.457074165344238, "train/clean_loss_weight": 1.6914148330688477, "train/combined_loss": 1.3874887228012085, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04123222748815166, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1562405824661255 }, { "epoch": 0.41232227488151657, "step": 174, "train/alpha": 0.2, "train/clean_ce_loss": 0.12890625, "train/clean_loss_scale": 8.91515827178955, "train/clean_loss_weight": 1.783031702041626, "train/combined_loss": 1.379063606262207, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04123222748815166, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1492196321487427 }, { "epoch": 0.41232227488151657, "step": 174, "train/alpha": 0.2, "train/clean_ce_loss": 0.10791015625, "train/clean_loss_scale": 7.8925018310546875, "train/clean_loss_weight": 1.5785003900527954, "train/combined_loss": 1.0220173597335815, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04123222748815166, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.851681113243103 }, { "epoch": 0.4146919431279621, "grad_norm": 13.043194770812988, "learning_rate": 2e-06, "loss": 3.8902, "step": 175 }, { "epoch": 0.4146919431279621, "step": 175, "train/alpha": 0.2, "train/clean_ce_loss": 0.134765625, "train/clean_loss_scale": 8.604049682617188, "train/clean_loss_weight": 1.7208099365234375, "train/combined_loss": 1.391436219215393, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.041469194312796206, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1595301628112793 }, { "epoch": 0.4146919431279621, "step": 175, "train/alpha": 0.2, "train/clean_ce_loss": 0.10986328125, "train/clean_loss_scale": 7.43794584274292, "train/clean_loss_weight": 1.4875892400741577, "train/combined_loss": 0.9805885553359985, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.041469194312796206, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8171571493148804 }, { "epoch": 0.4146919431279621, "step": 175, "train/alpha": 0.2, "train/clean_ce_loss": 0.1396484375, "train/clean_loss_scale": 9.95272159576416, "train/clean_loss_weight": 1.990544319152832, "train/combined_loss": 1.667858362197876, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.041469194312796206, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.38988196849823 }, { "epoch": 0.4146919431279621, "step": 175, "train/alpha": 0.2, "train/clean_ce_loss": 0.12158203125, "train/clean_loss_scale": 6.897220611572266, "train/clean_loss_weight": 1.3794441223144531, "train/combined_loss": 1.006293773651123, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.041469194312796206, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8385781049728394 }, { "epoch": 0.41706161137440756, "step": 176, "train/alpha": 0.2, "train/clean_ce_loss": 0.06298828125, "train/clean_loss_scale": 10.590978622436523, "train/clean_loss_weight": 2.1181957721710205, "train/combined_loss": 0.8005290031433105, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04170616113744076, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6671075224876404 }, { "epoch": 0.41706161137440756, "step": 176, "train/alpha": 0.2, "train/clean_ce_loss": 0.0947265625, "train/clean_loss_scale": 7.970959663391113, "train/clean_loss_weight": 1.5941919088363647, "train/combined_loss": 0.9060739278793335, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04170616113744076, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7550616264343262 }, { "epoch": 0.41706161137440756, "step": 176, "train/alpha": 0.2, "train/clean_ce_loss": 0.08447265625, "train/clean_loss_scale": 9.263724327087402, "train/clean_loss_weight": 1.8527449369430542, "train/combined_loss": 0.9390376806259155, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04170616113744076, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7825313806533813 }, { "epoch": 0.41706161137440756, "step": 176, "train/alpha": 0.2, "train/clean_ce_loss": 0.1201171875, "train/clean_loss_scale": 10.37468433380127, "train/clean_loss_weight": 2.074936866760254, "train/combined_loss": 1.4954135417938232, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04170616113744076, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2461779117584229 }, { "epoch": 0.4194312796208531, "step": 177, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 9.101728439331055, "train/clean_loss_weight": 1.8203457593917847, "train/combined_loss": 0.7146278619766235, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04194312796208531, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5955232381820679 }, { "epoch": 0.4194312796208531, "step": 177, "train/alpha": 0.2, "train/clean_ce_loss": 0.1875, "train/clean_loss_scale": 8.210005760192871, "train/clean_loss_weight": 1.6420011520385742, "train/combined_loss": 1.8472511768341064, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04194312796208531, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5393760204315186 }, { "epoch": 0.4194312796208531, "step": 177, "train/alpha": 0.2, "train/clean_ce_loss": 0.16015625, "train/clean_loss_scale": 9.206491470336914, "train/clean_loss_weight": 1.8412983417510986, "train/combined_loss": 1.769372582435608, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04194312796208531, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4744771718978882 }, { "epoch": 0.4194312796208531, "step": 177, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 8.986889839172363, "train/clean_loss_weight": 1.7973779439926147, "train/combined_loss": 0.7688003778457642, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04194312796208531, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6406669616699219 }, { "epoch": 0.4218009478672986, "step": 178, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 10.109603881835938, "train/clean_loss_weight": 2.021920919418335, "train/combined_loss": 0.9892405271530151, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.042180094786729856, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.824367105960846 }, { "epoch": 0.4218009478672986, "step": 178, "train/alpha": 0.2, "train/clean_ce_loss": 0.1123046875, "train/clean_loss_scale": 11.137593269348145, "train/clean_loss_weight": 2.2275187969207764, "train/combined_loss": 1.5009647607803345, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.042180094786729856, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2508039474487305 }, { "epoch": 0.4218009478672986, "step": 178, "train/alpha": 0.2, "train/clean_ce_loss": 0.056396484375, "train/clean_loss_scale": 10.93632984161377, "train/clean_loss_weight": 2.1872661113739014, "train/combined_loss": 0.7401247024536133, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.042180094786729856, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6167705655097961 }, { "epoch": 0.4218009478672986, "step": 178, "train/alpha": 0.2, "train/clean_ce_loss": 0.047119140625, "train/clean_loss_scale": 14.683008193969727, "train/clean_loss_weight": 2.9366016387939453, "train/combined_loss": 0.8302208781242371, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.042180094786729856, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6918507218360901 }, { "epoch": 0.42417061611374407, "step": 179, "train/alpha": 0.2, "train/clean_ce_loss": 0.12109375, "train/clean_loss_scale": 10.129132270812988, "train/clean_loss_weight": 2.0258264541625977, "train/combined_loss": 1.471889615058899, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04241706161137441, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2265746593475342 }, { "epoch": 0.42417061611374407, "step": 179, "train/alpha": 0.2, "train/clean_ce_loss": 0.11962890625, "train/clean_loss_scale": 10.81546688079834, "train/clean_loss_weight": 2.163093328475952, "train/combined_loss": 1.5526108741760254, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04241706161137441, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2938424348831177 }, { "epoch": 0.42417061611374407, "step": 179, "train/alpha": 0.2, "train/clean_ce_loss": 0.1923828125, "train/clean_loss_scale": 7.593131065368652, "train/clean_loss_weight": 1.5186262130737305, "train/combined_loss": 1.7529454231262207, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04241706161137441, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4607878923416138 }, { "epoch": 0.42417061611374407, "step": 179, "train/alpha": 0.2, "train/clean_ce_loss": 0.0751953125, "train/clean_loss_scale": 8.634552001953125, "train/clean_loss_weight": 1.7269104719161987, "train/combined_loss": 0.7791334390640259, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04241706161137441, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6492778658866882 }, { "epoch": 0.4265402843601896, "grad_norm": 15.747260093688965, "learning_rate": 2e-06, "loss": 3.7602, "step": 180 }, { "epoch": 0.4265402843601896, "step": 180, "train/alpha": 0.2, "train/clean_ce_loss": 0.0771484375, "train/clean_loss_scale": 11.640790939331055, "train/clean_loss_weight": 2.328158140182495, "train/combined_loss": 1.077682614326477, "train/pos_loss/pos0": 0.3309166431427002, "train/pos_loss/pos1": 0.5551695823669434, "train/pos_loss/pos2": 1.0335874557495117, "train/pos_loss/pos3": 1.5547714233398438, "train/pos_loss/pos4": 2.1051981449127197, "train/pos_loss/pos5": 2.5698957443237305, "train/pos_loss/pos6": 3.0420150756835938, "train/pos_loss/weight0": 2.3657209873199463, "train/pos_loss/weight1": 1.7851368188858032, "train/pos_loss/weight2": 1.2156771421432495, "train/pos_loss/weight3": 0.7853690385818481, "train/pos_loss/weight4": 0.4521196186542511, "train/pos_loss/weight5": 0.25212857127189636, "train/pos_loss/weight6": 0.1438482105731964, "train/pos_weight_progress": 0.04265402843601896, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8980688452720642 }, { "epoch": 0.4265402843601896, "step": 180, "train/alpha": 0.2, "train/clean_ce_loss": 0.1103515625, "train/clean_loss_scale": 8.339515686035156, "train/clean_loss_weight": 1.667903184890747, "train/combined_loss": 1.1043343544006348, "train/pos_loss/pos0": 0.4004599452018738, "train/pos_loss/pos1": 0.870613157749176, "train/pos_loss/pos2": 1.2763376235961914, "train/pos_loss/pos3": 1.7882862091064453, "train/pos_loss/pos4": 2.124983549118042, "train/pos_loss/pos5": 2.5418174266815186, "train/pos_loss/pos6": 3.0316293239593506, "train/pos_loss/weight0": 2.8125786781311035, "train/pos_loss/weight1": 1.9284992218017578, "train/pos_loss/weight2": 1.2031382322311401, "train/pos_loss/weight3": 0.642058789730072, "train/pos_loss/weight4": 0.274713933467865, "train/pos_loss/weight5": 0.1011071503162384, "train/pos_loss/weight6": 0.037904031574726105, "train/pos_weight_progress": 0.04265402843601896, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9202786087989807 }, { "epoch": 0.4265402843601896, "step": 180, "train/alpha": 0.2, "train/clean_ce_loss": 0.09423828125, "train/clean_loss_scale": 7.995582580566406, "train/clean_loss_weight": 1.599116563796997, "train/combined_loss": 0.9041879773139954, "train/pos_loss/pos0": 0.33084017038345337, "train/pos_loss/pos1": 0.6518003940582275, "train/pos_loss/pos2": 1.0974560976028442, "train/pos_loss/pos3": 1.602905035018921, "train/pos_loss/pos4": 1.970689058303833, "train/pos_loss/pos5": 2.503274440765381, "train/pos_loss/pos6": 2.826869249343872, "train/pos_loss/weight0": 2.886122941970825, "train/pos_loss/weight1": 1.989662528038025, "train/pos_loss/weight2": 1.1659600734710693, "train/pos_loss/weight3": 0.5905900001525879, "train/pos_loss/weight4": 0.2470017820596695, "train/pos_loss/weight5": 0.0876876637339592, "train/pos_loss/weight6": 0.03297531232237816, "train/pos_weight_progress": 0.04265402843601896, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7534899711608887 }, { "epoch": 0.4265402843601896, "step": 180, "train/alpha": 0.2, "train/clean_ce_loss": 0.169921875, "train/clean_loss_scale": 7.4651336669921875, "train/clean_loss_weight": 1.4930267333984375, "train/combined_loss": 1.522187352180481, "train/pos_loss/pos0": 0.569693922996521, "train/pos_loss/pos1": 1.1509040594100952, "train/pos_loss/pos2": 1.730931043624878, "train/pos_loss/pos3": 2.267362356185913, "train/pos_loss/pos4": 2.799929141998291, "train/pos_loss/pos5": 3.2476348876953125, "train/pos_loss/pos6": 3.722909688949585, "train/pos_loss/weight0": 2.7124886512756348, "train/pos_loss/weight1": 1.9034637212753296, "train/pos_loss/weight2": 1.2038317918777466, "train/pos_loss/weight3": 0.6694272756576538, "train/pos_loss/weight4": 0.3165385127067566, "train/pos_loss/weight5": 0.13257871568202972, "train/pos_loss/weight6": 0.061670832335948944, "train/pos_weight_progress": 0.04265402843601896, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2684894800186157 }, { "epoch": 0.42890995260663506, "step": 181, "train/alpha": 0.2, "train/clean_ce_loss": 0.154296875, "train/clean_loss_scale": 7.234823226928711, "train/clean_loss_weight": 1.4469646215438843, "train/combined_loss": 1.3395726680755615, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.042890995260663506, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1163105964660645 }, { "epoch": 0.42890995260663506, "step": 181, "train/alpha": 0.2, "train/clean_ce_loss": 0.1025390625, "train/clean_loss_scale": 6.619039058685303, "train/clean_loss_weight": 1.3238078355789185, "train/combined_loss": 0.814452052116394, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.042890995260663506, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6787100434303284 }, { "epoch": 0.42890995260663506, "step": 181, "train/alpha": 0.2, "train/clean_ce_loss": 0.0673828125, "train/clean_loss_scale": 6.42011022567749, "train/clean_loss_weight": 1.2840220928192139, "train/combined_loss": 0.5191261172294617, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.042890995260663506, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4326050877571106 }, { "epoch": 0.42890995260663506, "step": 181, "train/alpha": 0.2, "train/clean_ce_loss": 0.1142578125, "train/clean_loss_scale": 6.574915885925293, "train/clean_loss_weight": 1.3149832487106323, "train/combined_loss": 0.9014825820922852, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.042890995260663506, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7512354850769043 }, { "epoch": 0.4312796208530806, "step": 182, "train/alpha": 0.2, "train/clean_ce_loss": 0.04345703125, "train/clean_loss_scale": 11.124088287353516, "train/clean_loss_weight": 2.2248177528381348, "train/combined_loss": 0.5801038146018982, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.04312796208530806, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.48341983556747437 }, { "epoch": 0.4312796208530806, "step": 182, "train/alpha": 0.2, "train/clean_ce_loss": 0.08984375, "train/clean_loss_scale": 11.633635520935059, "train/clean_loss_weight": 2.3267271518707275, "train/combined_loss": 1.25425124168396, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.04312796208530806, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0452094078063965 }, { "epoch": 0.4312796208530806, "step": 182, "train/alpha": 0.2, "train/clean_ce_loss": 0.0947265625, "train/clean_loss_scale": 10.19446849822998, "train/clean_loss_weight": 2.038893699645996, "train/combined_loss": 1.1588243246078491, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.04312796208530806, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9656869173049927 }, { "epoch": 0.4312796208530806, "step": 182, "train/alpha": 0.2, "train/clean_ce_loss": 0.1318359375, "train/clean_loss_scale": 6.437130928039551, "train/clean_loss_weight": 1.287426233291626, "train/combined_loss": 1.0183742046356201, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.04312796208530806, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8486452102661133 }, { "epoch": 0.43364928909952605, "step": 183, "train/alpha": 0.2, "train/clean_ce_loss": 0.09326171875, "train/clean_loss_scale": 5.500178813934326, "train/clean_loss_weight": 1.1000357866287231, "train/combined_loss": 0.6155473589897156, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.0433649289099526, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5129561424255371 }, { "epoch": 0.43364928909952605, "step": 183, "train/alpha": 0.2, "train/clean_ce_loss": 0.0947265625, "train/clean_loss_scale": 7.022294998168945, "train/clean_loss_weight": 1.404458999633789, "train/combined_loss": 0.7982374429702759, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.0433649289099526, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6651978492736816 }, { "epoch": 0.43364928909952605, "step": 183, "train/alpha": 0.2, "train/clean_ce_loss": 0.1708984375, "train/clean_loss_scale": 6.811635971069336, "train/clean_loss_weight": 1.362327218055725, "train/combined_loss": 1.396917462348938, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.0433649289099526, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1640979051589966 }, { "epoch": 0.43364928909952605, "step": 183, "train/alpha": 0.2, "train/clean_ce_loss": 0.02490234375, "train/clean_loss_scale": 12.323223114013672, "train/clean_loss_weight": 2.46464467048645, "train/combined_loss": 0.36825257539749146, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.0433649289099526, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.30687713623046875 }, { "epoch": 0.43601895734597157, "step": 184, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 7.727094650268555, "train/clean_loss_weight": 1.5454189777374268, "train/combined_loss": 0.6565011739730835, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.043601895734597156, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5470843315124512 }, { "epoch": 0.43601895734597157, "step": 184, "train/alpha": 0.2, "train/clean_ce_loss": 0.032958984375, "train/clean_loss_scale": 9.130937576293945, "train/clean_loss_weight": 1.8261874914169312, "train/combined_loss": 0.36113569140434265, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.043601895734597156, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3009464144706726 }, { "epoch": 0.43601895734597157, "step": 184, "train/alpha": 0.2, "train/clean_ce_loss": 0.123046875, "train/clean_loss_scale": 8.381021499633789, "train/clean_loss_weight": 1.6762043237686157, "train/combined_loss": 1.2375102043151855, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.043601895734597156, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.031258463859558 }, { "epoch": 0.43601895734597157, "step": 184, "train/alpha": 0.2, "train/clean_ce_loss": 0.0693359375, "train/clean_loss_scale": 7.66654109954834, "train/clean_loss_weight": 1.5333082675933838, "train/combined_loss": 0.6378801465034485, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.043601895734597156, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5315667986869812 }, { "epoch": 0.43838862559241704, "grad_norm": 15.357961654663086, "learning_rate": 2e-06, "loss": 3.8281, "step": 185 }, { "epoch": 0.43838862559241704, "step": 185, "train/alpha": 0.2, "train/clean_ce_loss": 0.1396484375, "train/clean_loss_scale": 6.084261894226074, "train/clean_loss_weight": 1.2168524265289307, "train/combined_loss": 1.0195891857147217, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.04383886255924171, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8496576547622681 }, { "epoch": 0.43838862559241704, "step": 185, "train/alpha": 0.2, "train/clean_ce_loss": 0.10498046875, "train/clean_loss_scale": 6.852015972137451, "train/clean_loss_weight": 1.3704031705856323, "train/combined_loss": 0.8631934523582458, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.04383886255924171, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7193278670310974 }, { "epoch": 0.43838862559241704, "step": 185, "train/alpha": 0.2, "train/clean_ce_loss": 0.03076171875, "train/clean_loss_scale": 9.429668426513672, "train/clean_loss_weight": 1.885933756828308, "train/combined_loss": 0.3480873703956604, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.04383886255924171, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.29007279872894287 }, { "epoch": 0.43838862559241704, "step": 185, "train/alpha": 0.2, "train/clean_ce_loss": 0.0849609375, "train/clean_loss_scale": 5.9496612548828125, "train/clean_loss_weight": 1.1899322271347046, "train/combined_loss": 0.6065865755081177, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.04383886255924171, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5054888129234314 }, { "epoch": 0.44075829383886256, "step": 186, "train/alpha": 0.2, "train/clean_ce_loss": 0.059814453125, "train/clean_loss_scale": 7.150488376617432, "train/clean_loss_weight": 1.4300976991653442, "train/combined_loss": 0.5132430791854858, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.04407582938388625, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.42770254611968994 }, { "epoch": 0.44075829383886256, "step": 186, "train/alpha": 0.2, "train/clean_ce_loss": 0.08544921875, "train/clean_loss_scale": 6.3753767013549805, "train/clean_loss_weight": 1.2750753164291382, "train/combined_loss": 0.6537251472473145, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.04407582938388625, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5447709560394287 }, { "epoch": 0.44075829383886256, "step": 186, "train/alpha": 0.2, "train/clean_ce_loss": 0.07568359375, "train/clean_loss_scale": 9.160566329956055, "train/clean_loss_weight": 1.832113265991211, "train/combined_loss": 0.8319655060768127, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.04407582938388625, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6933045983314514 }, { "epoch": 0.44075829383886256, "step": 186, "train/alpha": 0.2, "train/clean_ce_loss": 0.09765625, "train/clean_loss_scale": 5.898181915283203, "train/clean_loss_weight": 1.1796363592147827, "train/combined_loss": 0.691193163394928, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.04407582938388625, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5759943127632141 }, { "epoch": 0.4431279620853081, "step": 187, "train/alpha": 0.2, "train/clean_ce_loss": 0.11328125, "train/clean_loss_scale": 4.910879611968994, "train/clean_loss_weight": 0.9821759462356567, "train/combined_loss": 0.6675727367401123, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.044312796208530805, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5563105940818787 }, { "epoch": 0.4431279620853081, "step": 187, "train/alpha": 0.2, "train/clean_ce_loss": 0.1025390625, "train/clean_loss_scale": 5.224650859832764, "train/clean_loss_weight": 1.0449302196502686, "train/combined_loss": 0.642876923084259, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.044312796208530805, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.53573077917099 }, { "epoch": 0.4431279620853081, "step": 187, "train/alpha": 0.2, "train/clean_ce_loss": 0.09033203125, "train/clean_loss_scale": 7.314042091369629, "train/clean_loss_weight": 1.4628084897994995, "train/combined_loss": 0.7928307056427002, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.044312796208530805, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6606922745704651 }, { "epoch": 0.4431279620853081, "step": 187, "train/alpha": 0.2, "train/clean_ce_loss": 0.083984375, "train/clean_loss_scale": 6.245538711547852, "train/clean_loss_weight": 1.2491077184677124, "train/combined_loss": 0.6294332146644592, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.044312796208530805, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5245276689529419 }, { "epoch": 0.44549763033175355, "step": 188, "train/alpha": 0.2, "train/clean_ce_loss": 0.060791015625, "train/clean_loss_scale": 6.869311332702637, "train/clean_loss_weight": 1.3738622665405273, "train/combined_loss": 0.5011109113693237, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.04454976303317536, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4175924062728882 }, { "epoch": 0.44549763033175355, "step": 188, "train/alpha": 0.2, "train/clean_ce_loss": 0.07373046875, "train/clean_loss_scale": 6.045029640197754, "train/clean_loss_weight": 1.2090059518814087, "train/combined_loss": 0.5348434448242188, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.04454976303317536, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4457028806209564 }, { "epoch": 0.44549763033175355, "step": 188, "train/alpha": 0.2, "train/clean_ce_loss": 0.06884765625, "train/clean_loss_scale": 5.638636589050293, "train/clean_loss_weight": 1.1277273893356323, "train/combined_loss": 0.46584829688072205, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.04454976303317536, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3882068991661072 }, { "epoch": 0.44549763033175355, "step": 188, "train/alpha": 0.2, "train/clean_ce_loss": 0.08203125, "train/clean_loss_scale": 7.34982967376709, "train/clean_loss_weight": 1.469965934753418, "train/combined_loss": 0.7234988212585449, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.04454976303317536, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6029157042503357 }, { "epoch": 0.44786729857819907, "step": 189, "train/alpha": 0.2, "train/clean_ce_loss": 0.06103515625, "train/clean_loss_scale": 7.172458648681641, "train/clean_loss_weight": 1.434491753578186, "train/combined_loss": 0.5253265500068665, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.0447867298578199, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43777212500572205 }, { "epoch": 0.44786729857819907, "step": 189, "train/alpha": 0.2, "train/clean_ce_loss": 0.16015625, "train/clean_loss_scale": 5.818526268005371, "train/clean_loss_weight": 1.1637052297592163, "train/combined_loss": 1.1182479858398438, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.0447867298578199, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9318733811378479 }, { "epoch": 0.44786729857819907, "step": 189, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 5.932364463806152, "train/clean_loss_weight": 1.1864728927612305, "train/combined_loss": 0.5804911255836487, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.0447867298578199, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4837426245212555 }, { "epoch": 0.44786729857819907, "step": 189, "train/alpha": 0.2, "train/clean_ce_loss": 0.07177734375, "train/clean_loss_scale": 6.56997013092041, "train/clean_loss_weight": 1.31399405002594, "train/combined_loss": 0.5658900141716003, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.0447867298578199, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.471574991941452 }, { "epoch": 0.45023696682464454, "grad_norm": 12.475541114807129, "learning_rate": 2e-06, "loss": 3.7967, "step": 190 }, { "epoch": 0.45023696682464454, "step": 190, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 5.385260581970215, "train/clean_loss_weight": 1.077052116394043, "train/combined_loss": 0.47962477803230286, "train/pos_loss/pos0": 0.17852140963077545, "train/pos_loss/pos1": 0.36245784163475037, "train/pos_loss/pos2": 0.7202925682067871, "train/pos_loss/pos3": 1.181327223777771, "train/pos_loss/pos4": 1.6831415891647339, "train/pos_loss/pos5": 2.0390803813934326, "train/pos_loss/pos6": 2.3992652893066406, "train/pos_loss/weight0": 3.385481595993042, "train/pos_loss/weight1": 2.0653209686279297, "train/pos_loss/weight2": 1.0162214040756226, "train/pos_loss/weight3": 0.3881327211856842, "train/pos_loss/weight4": 0.11493474245071411, "train/pos_loss/weight5": 0.025057367980480194, "train/pos_loss/weight6": 0.0048513952642679214, "train/pos_weight_progress": 0.045023696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3996873199939728 }, { "epoch": 0.45023696682464454, "step": 190, "train/alpha": 0.2, "train/clean_ce_loss": 0.072265625, "train/clean_loss_scale": 9.16804027557373, "train/clean_loss_weight": 1.8336080312728882, "train/combined_loss": 0.795041024684906, "train/pos_loss/pos0": 0.29622822999954224, "train/pos_loss/pos1": 0.5046590566635132, "train/pos_loss/pos2": 0.7186654210090637, "train/pos_loss/pos3": 1.1253803968429565, "train/pos_loss/pos4": 1.5199998617172241, "train/pos_loss/pos5": 1.9132684469223022, "train/pos_loss/pos6": 2.442094087600708, "train/pos_loss/weight0": 2.484070062637329, "train/pos_loss/weight1": 1.8280720710754395, "train/pos_loss/weight2": 1.224328637123108, "train/pos_loss/weight3": 0.7469531893730164, "train/pos_loss/weight4": 0.4178473949432373, "train/pos_loss/weight5": 0.19655385613441467, "train/pos_loss/weight6": 0.10217445343732834, "train/pos_weight_progress": 0.045023696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6625341773033142 }, { "epoch": 0.45023696682464454, "step": 190, "train/alpha": 0.2, "train/clean_ce_loss": 0.0341796875, "train/clean_loss_scale": 21.096038818359375, "train/clean_loss_weight": 4.219207763671875, "train/combined_loss": 0.8652671575546265, "train/pos_loss/pos0": 0.1303320676088333, "train/pos_loss/pos1": 0.445541113615036, "train/pos_loss/pos2": 0.9489859938621521, "train/pos_loss/pos3": 1.349128246307373, "train/pos_loss/pos4": 1.857822060585022, "train/pos_loss/pos5": 2.159803867340088, "train/pos_loss/pos6": 2.9242231845855713, "train/pos_loss/weight0": 2.3918023109436035, "train/pos_loss/weight1": 1.8030697107315063, "train/pos_loss/weight2": 1.206918478012085, "train/pos_loss/weight3": 0.7754082083702087, "train/pos_loss/weight4": 0.4499475061893463, "train/pos_loss/weight5": 0.23712456226348877, "train/pos_loss/weight6": 0.1357293725013733, "train/pos_weight_progress": 0.045023696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7210559844970703 }, { "epoch": 0.45023696682464454, "step": 190, "train/alpha": 0.2, "train/clean_ce_loss": 0.1767578125, "train/clean_loss_scale": 8.659956932067871, "train/clean_loss_weight": 1.7319914102554321, "train/combined_loss": 1.8368580341339111, "train/pos_loss/pos0": 0.7668980360031128, "train/pos_loss/pos1": 1.1984856128692627, "train/pos_loss/pos2": 2.002176523208618, "train/pos_loss/pos3": 2.583139657974243, "train/pos_loss/pos4": 3.2059919834136963, "train/pos_loss/pos5": 3.7226858139038086, "train/pos_loss/pos6": 4.13360071182251, "train/pos_loss/weight0": 2.5346531867980957, "train/pos_loss/weight1": 1.8587232828140259, "train/pos_loss/weight2": 1.2503777742385864, "train/pos_loss/weight3": 0.7410361170768738, "train/pos_loss/weight4": 0.3685176372528076, "train/pos_loss/weight5": 0.16728347539901733, "train/pos_loss/weight6": 0.07940856367349625, "train/pos_weight_progress": 0.045023696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.530714988708496 }, { "epoch": 0.45260663507109006, "step": 191, "train/alpha": 0.2, "train/clean_ce_loss": 0.12451171875, "train/clean_loss_scale": 8.003622055053711, "train/clean_loss_weight": 1.600724458694458, "train/combined_loss": 1.1958537101745605, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.04526066350710901, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9965447783470154 }, { "epoch": 0.45260663507109006, "step": 191, "train/alpha": 0.2, "train/clean_ce_loss": 0.12890625, "train/clean_loss_scale": 7.4211015701293945, "train/clean_loss_weight": 1.4842203855514526, "train/combined_loss": 1.147951602935791, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.04526066350710901, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9566263556480408 }, { "epoch": 0.45260663507109006, "step": 191, "train/alpha": 0.2, "train/clean_ce_loss": 0.11279296875, "train/clean_loss_scale": 6.056275844573975, "train/clean_loss_weight": 1.2112551927566528, "train/combined_loss": 0.8197264075279236, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.04526066350710901, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6831053495407104 }, { "epoch": 0.45260663507109006, "step": 191, "train/alpha": 0.2, "train/clean_ce_loss": 0.07177734375, "train/clean_loss_scale": 7.10051965713501, "train/clean_loss_weight": 1.420103907585144, "train/combined_loss": 0.6115877032279968, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.04526066350710901, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5096564292907715 }, { "epoch": 0.4549763033175355, "step": 192, "train/alpha": 0.2, "train/clean_ce_loss": 0.0859375, "train/clean_loss_scale": 6.856506824493408, "train/clean_loss_weight": 1.3713014125823975, "train/combined_loss": 0.7070772647857666, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.04549763033175355, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5892310738563538 }, { "epoch": 0.4549763033175355, "step": 192, "train/alpha": 0.2, "train/clean_ce_loss": 0.1728515625, "train/clean_loss_scale": 5.973702430725098, "train/clean_loss_weight": 1.1947405338287354, "train/combined_loss": 1.2390766143798828, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.04549763033175355, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0325638055801392 }, { "epoch": 0.4549763033175355, "step": 192, "train/alpha": 0.2, "train/clean_ce_loss": 0.1640625, "train/clean_loss_scale": 6.646373271942139, "train/clean_loss_weight": 1.3292746543884277, "train/combined_loss": 1.3085047006607056, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.04549763033175355, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0904206037521362 }, { "epoch": 0.4549763033175355, "step": 192, "train/alpha": 0.2, "train/clean_ce_loss": 0.1533203125, "train/clean_loss_scale": 4.9441118240356445, "train/clean_loss_weight": 0.9888224005699158, "train/combined_loss": 0.9096393585205078, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.04549763033175355, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7580327987670898 }, { "epoch": 0.45734597156398105, "step": 193, "train/alpha": 0.2, "train/clean_ce_loss": 0.1083984375, "train/clean_loss_scale": 5.884084701538086, "train/clean_loss_weight": 1.1768169403076172, "train/combined_loss": 0.7653907537460327, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.045734597156398105, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.637825608253479 }, { "epoch": 0.45734597156398105, "step": 193, "train/alpha": 0.2, "train/clean_ce_loss": 0.12451171875, "train/clean_loss_scale": 6.042792797088623, "train/clean_loss_weight": 1.2085585594177246, "train/combined_loss": 0.9028781652450562, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.045734597156398105, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7523984909057617 }, { "epoch": 0.45734597156398105, "step": 193, "train/alpha": 0.2, "train/clean_ce_loss": 0.1669921875, "train/clean_loss_scale": 7.288321495056152, "train/clean_loss_weight": 1.4576643705368042, "train/combined_loss": 1.460511326789856, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.045734597156398105, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.217092752456665 }, { "epoch": 0.45734597156398105, "step": 193, "train/alpha": 0.2, "train/clean_ce_loss": 0.0693359375, "train/clean_loss_scale": 6.963620185852051, "train/clean_loss_weight": 1.3927240371704102, "train/combined_loss": 0.5793949365615845, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.045734597156398105, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.48282912373542786 }, { "epoch": 0.4597156398104265, "step": 194, "train/alpha": 0.2, "train/clean_ce_loss": 0.146484375, "train/clean_loss_scale": 8.425036430358887, "train/clean_loss_weight": 1.6850073337554932, "train/combined_loss": 1.4809634685516357, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.04597156398104266, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2341362237930298 }, { "epoch": 0.4597156398104265, "step": 194, "train/alpha": 0.2, "train/clean_ce_loss": 0.1279296875, "train/clean_loss_scale": 5.97426176071167, "train/clean_loss_weight": 1.194852352142334, "train/combined_loss": 0.9171425104141235, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.04597156398104266, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7642854452133179 }, { "epoch": 0.4597156398104265, "step": 194, "train/alpha": 0.2, "train/clean_ce_loss": 0.13671875, "train/clean_loss_scale": 10.1478910446167, "train/clean_loss_weight": 2.02957820892334, "train/combined_loss": 1.6648883819580078, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.04597156398104266, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3874069452285767 }, { "epoch": 0.4597156398104265, "step": 194, "train/alpha": 0.2, "train/clean_ce_loss": 0.034912109375, "train/clean_loss_scale": 6.796826362609863, "train/clean_loss_weight": 1.3593653440475464, "train/combined_loss": 0.2847498655319214, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.04597156398104266, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.23729154467582703 }, { "epoch": 0.46208530805687204, "grad_norm": 14.667984962463379, "learning_rate": 2e-06, "loss": 4.0583, "step": 195 }, { "epoch": 0.46208530805687204, "step": 195, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 7.123274326324463, "train/clean_loss_weight": 1.4246548414230347, "train/combined_loss": 0.5676358938217163, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.0462085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4730299413204193 }, { "epoch": 0.46208530805687204, "step": 195, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 6.785248756408691, "train/clean_loss_weight": 1.357049822807312, "train/combined_loss": 0.536723792552948, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.0462085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.44726982712745667 }, { "epoch": 0.46208530805687204, "step": 195, "train/alpha": 0.2, "train/clean_ce_loss": 0.0888671875, "train/clean_loss_scale": 5.8196821212768555, "train/clean_loss_weight": 1.1639364957809448, "train/combined_loss": 0.6206145286560059, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.0462085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5171787738800049 }, { "epoch": 0.46208530805687204, "step": 195, "train/alpha": 0.2, "train/clean_ce_loss": 0.12060546875, "train/clean_loss_scale": 6.546350002288818, "train/clean_loss_weight": 1.3092700242996216, "train/combined_loss": 0.9474307298660278, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.0462085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7895256280899048 }, { "epoch": 0.46445497630331756, "step": 196, "train/alpha": 0.2, "train/clean_ce_loss": 0.09521484375, "train/clean_loss_scale": 6.0500264167785645, "train/clean_loss_weight": 1.210005283355713, "train/combined_loss": 0.6912627816200256, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.046445497630331754, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5760523080825806 }, { "epoch": 0.46445497630331756, "step": 196, "train/alpha": 0.2, "train/clean_ce_loss": 0.078125, "train/clean_loss_scale": 5.949760437011719, "train/clean_loss_weight": 1.1899521350860596, "train/combined_loss": 0.5577900409698486, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.046445497630331754, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4648250341415405 }, { "epoch": 0.46445497630331756, "step": 196, "train/alpha": 0.2, "train/clean_ce_loss": 0.01312255859375, "train/clean_loss_scale": 15.795988082885742, "train/clean_loss_weight": 3.1591975688934326, "train/combined_loss": 0.2487405389547348, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.046445497630331754, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.20728377997875214 }, { "epoch": 0.46445497630331756, "step": 196, "train/alpha": 0.2, "train/clean_ce_loss": 0.10009765625, "train/clean_loss_scale": 6.463438510894775, "train/clean_loss_weight": 1.2926877737045288, "train/combined_loss": 0.7763700485229492, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.046445497630331754, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.646975040435791 }, { "epoch": 0.466824644549763, "step": 197, "train/alpha": 0.2, "train/clean_ce_loss": 0.07568359375, "train/clean_loss_scale": 6.15471887588501, "train/clean_loss_weight": 1.2309437990188599, "train/combined_loss": 0.558973491191864, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.04668246445497631, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46581125259399414 }, { "epoch": 0.466824644549763, "step": 197, "train/alpha": 0.2, "train/clean_ce_loss": 0.0751953125, "train/clean_loss_scale": 5.626549243927002, "train/clean_loss_weight": 1.1253098249435425, "train/combined_loss": 0.5077081322669983, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.04668246445497631, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4230901300907135 }, { "epoch": 0.466824644549763, "step": 197, "train/alpha": 0.2, "train/clean_ce_loss": 0.1572265625, "train/clean_loss_scale": 6.905981063842773, "train/clean_loss_weight": 1.3811962604522705, "train/combined_loss": 1.3029643297195435, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.04668246445497631, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0858036279678345 }, { "epoch": 0.466824644549763, "step": 197, "train/alpha": 0.2, "train/clean_ce_loss": 0.10400390625, "train/clean_loss_scale": 5.394187927246094, "train/clean_loss_weight": 1.0788376331329346, "train/combined_loss": 0.6732199192047119, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.04668246445497631, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5610166192054749 }, { "epoch": 0.46919431279620855, "step": 198, "train/alpha": 0.2, "train/clean_ce_loss": 0.07177734375, "train/clean_loss_scale": 5.634237289428711, "train/clean_loss_weight": 1.126847505569458, "train/combined_loss": 0.4852926731109619, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.04691943127962085, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.40441057085990906 }, { "epoch": 0.46919431279620855, "step": 198, "train/alpha": 0.2, "train/clean_ce_loss": 0.10791015625, "train/clean_loss_scale": 7.981311798095703, "train/clean_loss_weight": 1.5962623357772827, "train/combined_loss": 1.0335174798965454, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.04691943127962085, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8612645864486694 }, { "epoch": 0.46919431279620855, "step": 198, "train/alpha": 0.2, "train/clean_ce_loss": 0.09765625, "train/clean_loss_scale": 6.02084493637085, "train/clean_loss_weight": 1.2041690349578857, "train/combined_loss": 0.7055677175521851, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.04691943127962085, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5879731178283691 }, { "epoch": 0.46919431279620855, "step": 198, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 5.712878704071045, "train/clean_loss_weight": 1.142575740814209, "train/combined_loss": 0.548971951007843, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.04691943127962085, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4574766159057617 }, { "epoch": 0.471563981042654, "step": 199, "train/alpha": 0.2, "train/clean_ce_loss": 0.1630859375, "train/clean_loss_scale": 6.239516735076904, "train/clean_loss_weight": 1.2479033470153809, "train/combined_loss": 1.221092939376831, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.047156398104265404, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0175774097442627 }, { "epoch": 0.471563981042654, "step": 199, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 7.307918548583984, "train/clean_loss_weight": 1.4615837335586548, "train/combined_loss": 0.6208876371383667, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.047156398104265404, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5174063444137573 }, { "epoch": 0.471563981042654, "step": 199, "train/alpha": 0.2, "train/clean_ce_loss": 0.134765625, "train/clean_loss_scale": 7.599166393280029, "train/clean_loss_weight": 1.5198333263397217, "train/combined_loss": 1.2289276123046875, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.047156398104265404, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0241063833236694 }, { "epoch": 0.471563981042654, "step": 199, "train/alpha": 0.2, "train/clean_ce_loss": 0.1767578125, "train/clean_loss_scale": 5.436890125274658, "train/clean_loss_weight": 1.0873780250549316, "train/combined_loss": 1.1532154083251953, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.047156398104265404, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9610128402709961 }, { "epoch": 0.47393364928909953, "grad_norm": 14.797430992126465, "learning_rate": 2e-06, "loss": 3.7623, "step": 200 }, { "epoch": 0.47393364928909953, "step": 200, "train/alpha": 0.2, "train/clean_ce_loss": 0.08642578125, "train/clean_loss_scale": 6.588406562805176, "train/clean_loss_weight": 1.3176813125610352, "train/combined_loss": 0.6832898259162903, "train/pos_loss/pos0": 0.2528619170188904, "train/pos_loss/pos1": 0.5330188274383545, "train/pos_loss/pos2": 1.1252355575561523, "train/pos_loss/pos3": 1.6301549673080444, "train/pos_loss/pos4": 2.180319309234619, "train/pos_loss/pos5": 2.6050145626068115, "train/pos_loss/pos6": 2.9954469203948975, "train/pos_loss/weight0": 3.4480319023132324, "train/pos_loss/weight1": 2.0594687461853027, "train/pos_loss/weight2": 0.993809700012207, "train/pos_loss/weight3": 0.36566662788391113, "train/pos_loss/weight4": 0.10685522854328156, "train/pos_loss/weight5": 0.021987074986100197, "train/pos_loss/weight6": 0.004180411342531443, "train/pos_weight_progress": 0.04739336492890995, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5694081783294678 }, { "epoch": 0.47393364928909953, "step": 200, "train/alpha": 0.2, "train/clean_ce_loss": 0.08984375, "train/clean_loss_scale": 8.242921829223633, "train/clean_loss_weight": 1.6485843658447266, "train/combined_loss": 0.8886899948120117, "train/pos_loss/pos0": 0.26663270592689514, "train/pos_loss/pos1": 0.6140275001525879, "train/pos_loss/pos2": 1.067431926727295, "train/pos_loss/pos3": 1.5715351104736328, "train/pos_loss/pos4": 2.039973497390747, "train/pos_loss/pos5": 2.496840238571167, "train/pos_loss/pos6": 2.8746230602264404, "train/pos_loss/weight0": 2.770775318145752, "train/pos_loss/weight1": 1.9354162216186523, "train/pos_loss/weight2": 1.2357620000839233, "train/pos_loss/weight3": 0.6406232118606567, "train/pos_loss/weight4": 0.2735355496406555, "train/pos_loss/weight5": 0.10224025696516037, "train/pos_loss/weight6": 0.04164751619100571, "train/pos_weight_progress": 0.04739336492890995, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7405750155448914 }, { "epoch": 0.47393364928909953, "step": 200, "train/alpha": 0.2, "train/clean_ce_loss": 0.10693359375, "train/clean_loss_scale": 6.864932060241699, "train/clean_loss_weight": 1.3729864358901978, "train/combined_loss": 0.8809102773666382, "train/pos_loss/pos0": 0.32378336787223816, "train/pos_loss/pos1": 0.6138749122619629, "train/pos_loss/pos2": 1.0536329746246338, "train/pos_loss/pos3": 1.4144184589385986, "train/pos_loss/pos4": 1.7742558717727661, "train/pos_loss/pos5": 2.15073299407959, "train/pos_loss/pos6": 2.587698221206665, "train/pos_loss/weight0": 2.7659199237823486, "train/pos_loss/weight1": 1.930415153503418, "train/pos_loss/weight2": 1.214881420135498, "train/pos_loss/weight3": 0.6423957347869873, "train/pos_loss/weight4": 0.28734177350997925, "train/pos_loss/weight5": 0.11314152926206589, "train/pos_loss/weight6": 0.04590450972318649, "train/pos_weight_progress": 0.04739336492890995, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7340918779373169 }, { "epoch": 0.47393364928909953, "step": 200, "train/alpha": 0.2, "train/clean_ce_loss": 0.0634765625, "train/clean_loss_scale": 9.997105598449707, "train/clean_loss_weight": 1.9994211196899414, "train/combined_loss": 0.7614983320236206, "train/pos_loss/pos0": 0.2630878984928131, "train/pos_loss/pos1": 0.467913419008255, "train/pos_loss/pos2": 0.7984517216682434, "train/pos_loss/pos3": 1.2985866069793701, "train/pos_loss/pos4": 1.5886973142623901, "train/pos_loss/pos5": 2.0075442790985107, "train/pos_loss/pos6": 2.2285239696502686, "train/pos_loss/weight0": 2.6676528453826904, "train/pos_loss/weight1": 1.8597646951675415, "train/pos_loss/weight2": 1.223476767539978, "train/pos_loss/weight3": 0.6864650249481201, "train/pos_loss/weight4": 0.33833521604537964, "train/pos_loss/weight5": 0.15198209881782532, "train/pos_loss/weight6": 0.07232353836297989, "train/pos_weight_progress": 0.04739336492890995, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6345819234848022 }, { "epoch": 0.476303317535545, "step": 201, "train/alpha": 0.2, "train/clean_ce_loss": 0.1708984375, "train/clean_loss_scale": 7.375005722045898, "train/clean_loss_weight": 1.4750012159347534, "train/combined_loss": 1.5124523639678955, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.0476303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2603769302368164 }, { "epoch": 0.476303317535545, "step": 201, "train/alpha": 0.2, "train/clean_ce_loss": 0.1201171875, "train/clean_loss_scale": 8.37582778930664, "train/clean_loss_weight": 1.6751655340194702, "train/combined_loss": 1.2072970867156982, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.0476303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0060808658599854 }, { "epoch": 0.476303317535545, "step": 201, "train/alpha": 0.2, "train/clean_ce_loss": 0.051025390625, "train/clean_loss_scale": 9.996248245239258, "train/clean_loss_weight": 1.9992496967315674, "train/combined_loss": 0.6120749711990356, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.0476303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5100624561309814 }, { "epoch": 0.476303317535545, "step": 201, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 7.413558483123779, "train/clean_loss_weight": 1.482711672782898, "train/combined_loss": 0.7428038120269775, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.0476303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.619003176689148 }, { "epoch": 0.4786729857819905, "step": 202, "train/alpha": 0.2, "train/clean_ce_loss": 0.2041015625, "train/clean_loss_scale": 8.937633514404297, "train/clean_loss_weight": 1.7875267267227173, "train/combined_loss": 2.1890220642089844, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.047867298578199054, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.8241850137710571 }, { "epoch": 0.4786729857819905, "step": 202, "train/alpha": 0.2, "train/clean_ce_loss": 0.12353515625, "train/clean_loss_scale": 7.5887298583984375, "train/clean_loss_weight": 1.5177459716796875, "train/combined_loss": 1.1249699592590332, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.047867298578199054, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9374749064445496 }, { "epoch": 0.4786729857819905, "step": 202, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 9.635653495788574, "train/clean_loss_weight": 1.9271306991577148, "train/combined_loss": 1.0106143951416016, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.047867298578199054, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8421787023544312 }, { "epoch": 0.4786729857819905, "step": 202, "train/alpha": 0.2, "train/clean_ce_loss": 0.07177734375, "train/clean_loss_scale": 7.688226699829102, "train/clean_loss_weight": 1.5376453399658203, "train/combined_loss": 0.6622085571289062, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.047867298578199054, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5518404841423035 }, { "epoch": 0.48104265402843605, "step": 203, "train/alpha": 0.2, "train/clean_ce_loss": 0.09326171875, "train/clean_loss_scale": 7.568664073944092, "train/clean_loss_weight": 1.5137327909469604, "train/combined_loss": 0.8470399379730225, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.0481042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7058666348457336 }, { "epoch": 0.48104265402843605, "step": 203, "train/alpha": 0.2, "train/clean_ce_loss": 0.050048828125, "train/clean_loss_scale": 13.168956756591797, "train/clean_loss_weight": 2.633791446685791, "train/combined_loss": 0.7909090518951416, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.0481042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6590908765792847 }, { "epoch": 0.48104265402843605, "step": 203, "train/alpha": 0.2, "train/clean_ce_loss": 0.126953125, "train/clean_loss_scale": 9.957561492919922, "train/clean_loss_weight": 1.9915122985839844, "train/combined_loss": 1.516972303390503, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.0481042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2641435861587524 }, { "epoch": 0.48104265402843605, "step": 203, "train/alpha": 0.2, "train/clean_ce_loss": 0.0242919921875, "train/clean_loss_scale": 29.06707763671875, "train/clean_loss_weight": 5.81341552734375, "train/combined_loss": 0.8473166823387146, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.0481042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7060972452163696 }, { "epoch": 0.4834123222748815, "step": 204, "train/alpha": 0.2, "train/clean_ce_loss": 0.0478515625, "train/clean_loss_scale": 12.950230598449707, "train/clean_loss_weight": 2.5900461673736572, "train/combined_loss": 0.7436264753341675, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.04834123222748815, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6196887493133545 }, { "epoch": 0.4834123222748815, "step": 204, "train/alpha": 0.2, "train/clean_ce_loss": 0.040283203125, "train/clean_loss_scale": 19.801572799682617, "train/clean_loss_weight": 3.9603145122528076, "train/combined_loss": 0.9572049379348755, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.04834123222748815, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7976707816123962 }, { "epoch": 0.4834123222748815, "step": 204, "train/alpha": 0.2, "train/clean_ce_loss": 0.11376953125, "train/clean_loss_scale": 7.154356002807617, "train/clean_loss_weight": 1.4308712482452393, "train/combined_loss": 0.9767372608184814, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.04834123222748815, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8139477372169495 }, { "epoch": 0.4834123222748815, "step": 204, "train/alpha": 0.2, "train/clean_ce_loss": 0.06298828125, "train/clean_loss_scale": 9.972396850585938, "train/clean_loss_weight": 1.9944794178009033, "train/combined_loss": 0.7537729740142822, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.04834123222748815, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6281441450119019 }, { "epoch": 0.48578199052132703, "grad_norm": 14.589866638183594, "learning_rate": 2e-06, "loss": 3.8026, "step": 205 }, { "epoch": 0.48578199052132703, "step": 205, "train/alpha": 0.2, "train/clean_ce_loss": 0.09228515625, "train/clean_loss_scale": 8.559441566467285, "train/clean_loss_weight": 1.711888313293457, "train/combined_loss": 0.9478912949562073, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.0485781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7899094223976135 }, { "epoch": 0.48578199052132703, "step": 205, "train/alpha": 0.2, "train/clean_ce_loss": 0.0634765625, "train/clean_loss_scale": 11.112586975097656, "train/clean_loss_weight": 2.222517490386963, "train/combined_loss": 0.846466600894928, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.0485781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7053888440132141 }, { "epoch": 0.48578199052132703, "step": 205, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 7.78777551651001, "train/clean_loss_weight": 1.557555079460144, "train/combined_loss": 0.8168038129806519, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.0485781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6806698441505432 }, { "epoch": 0.48578199052132703, "step": 205, "train/alpha": 0.2, "train/clean_ce_loss": 0.046875, "train/clean_loss_scale": 13.046587944030762, "train/clean_loss_weight": 2.6093175411224365, "train/combined_loss": 0.7338705658912659, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.0485781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6115587949752808 }, { "epoch": 0.4881516587677725, "step": 206, "train/alpha": 0.2, "train/clean_ce_loss": 0.11181640625, "train/clean_loss_scale": 7.8549652099609375, "train/clean_loss_weight": 1.5709930658340454, "train/combined_loss": 1.0539767742156982, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.04881516587677725, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8783139586448669 }, { "epoch": 0.4881516587677725, "step": 206, "train/alpha": 0.2, "train/clean_ce_loss": 0.1328125, "train/clean_loss_scale": 8.999344825744629, "train/clean_loss_weight": 1.7998689413070679, "train/combined_loss": 1.4342706203460693, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.04881516587677725, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.195225477218628 }, { "epoch": 0.4881516587677725, "step": 206, "train/alpha": 0.2, "train/clean_ce_loss": 0.1181640625, "train/clean_loss_scale": 7.173387050628662, "train/clean_loss_weight": 1.4346774816513062, "train/combined_loss": 1.017163872718811, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.04881516587677725, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8476365804672241 }, { "epoch": 0.4881516587677725, "step": 206, "train/alpha": 0.2, "train/clean_ce_loss": 0.146484375, "train/clean_loss_scale": 9.303884506225586, "train/clean_loss_weight": 1.8607769012451172, "train/combined_loss": 1.6354484558105469, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.04881516587677725, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3628736734390259 }, { "epoch": 0.490521327014218, "step": 207, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 9.162888526916504, "train/clean_loss_weight": 1.8325777053833008, "train/combined_loss": 0.896602988243103, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.0490521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7471691370010376 }, { "epoch": 0.490521327014218, "step": 207, "train/alpha": 0.2, "train/clean_ce_loss": 0.15625, "train/clean_loss_scale": 9.920507431030273, "train/clean_loss_weight": 1.9841015338897705, "train/combined_loss": 1.8600950241088867, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.0490521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5500792264938354 }, { "epoch": 0.490521327014218, "step": 207, "train/alpha": 0.2, "train/clean_ce_loss": 0.18359375, "train/clean_loss_scale": 8.38624095916748, "train/clean_loss_weight": 1.677248239517212, "train/combined_loss": 1.8475936651229858, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.0490521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5396614074707031 }, { "epoch": 0.490521327014218, "step": 207, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 7.624726295471191, "train/clean_loss_weight": 1.5249452590942383, "train/combined_loss": 0.7237533330917358, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.0490521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6031277775764465 }, { "epoch": 0.4928909952606635, "step": 208, "train/alpha": 0.2, "train/clean_ce_loss": 0.08642578125, "train/clean_loss_scale": 9.017127990722656, "train/clean_loss_weight": 1.803425669670105, "train/combined_loss": 0.935174822807312, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.04928909952606635, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7793123722076416 }, { "epoch": 0.4928909952606635, "step": 208, "train/alpha": 0.2, "train/clean_ce_loss": 0.1123046875, "train/clean_loss_scale": 7.249225616455078, "train/clean_loss_weight": 1.4498451948165894, "train/combined_loss": 0.9769464135169983, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.04928909952606635, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8141220211982727 }, { "epoch": 0.4928909952606635, "step": 208, "train/alpha": 0.2, "train/clean_ce_loss": 0.09375, "train/clean_loss_scale": 8.559901237487793, "train/clean_loss_weight": 1.7119802236557007, "train/combined_loss": 0.9629889130592346, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.04928909952606635, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.802490770816803 }, { "epoch": 0.4928909952606635, "step": 208, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 10.74881362915039, "train/clean_loss_weight": 2.1497628688812256, "train/combined_loss": 1.0328937768936157, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.04928909952606635, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.860744833946228 }, { "epoch": 0.495260663507109, "step": 209, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 10.399478912353516, "train/clean_loss_weight": 2.0798957347869873, "train/combined_loss": 1.0176053047180176, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.0495260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8480044007301331 }, { "epoch": 0.495260663507109, "step": 209, "train/alpha": 0.2, "train/clean_ce_loss": 0.04150390625, "train/clean_loss_scale": 13.281071662902832, "train/clean_loss_weight": 2.656214475631714, "train/combined_loss": 0.6614596247673035, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.0495260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5512163639068604 }, { "epoch": 0.495260663507109, "step": 209, "train/alpha": 0.2, "train/clean_ce_loss": 0.03076171875, "train/clean_loss_scale": 20.139413833618164, "train/clean_loss_weight": 4.027883052825928, "train/combined_loss": 0.743427574634552, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.0495260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6195229887962341 }, { "epoch": 0.495260663507109, "step": 209, "train/alpha": 0.2, "train/clean_ce_loss": 0.12451171875, "train/clean_loss_scale": 7.708837985992432, "train/clean_loss_weight": 1.5417675971984863, "train/combined_loss": 1.151808738708496, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.0495260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9598406553268433 }, { "epoch": 0.4976303317535545, "grad_norm": 12.709029197692871, "learning_rate": 2e-06, "loss": 3.9757, "step": 210 }, { "epoch": 0.4976303317535545, "step": 210, "train/alpha": 0.2, "train/clean_ce_loss": 0.125, "train/clean_loss_scale": 10.453930854797363, "train/clean_loss_weight": 2.0907862186431885, "train/combined_loss": 1.5680896043777466, "train/pos_loss/pos0": 0.4994902014732361, "train/pos_loss/pos1": 1.0905861854553223, "train/pos_loss/pos2": 1.5635597705841064, "train/pos_loss/pos3": 2.3562793731689453, "train/pos_loss/pos4": 2.7169463634490967, "train/pos_loss/pos5": 3.273723840713501, "train/pos_loss/pos6": 3.5228161811828613, "train/pos_loss/weight0": 2.463439702987671, "train/pos_loss/weight1": 1.8031272888183594, "train/pos_loss/weight2": 1.2217791080474854, "train/pos_loss/weight3": 0.7567427158355713, "train/pos_loss/weight4": 0.4219622015953064, "train/pos_loss/weight5": 0.21695518493652344, "train/pos_loss/weight6": 0.11599373072385788, "train/pos_weight_progress": 0.04976303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3067413568496704 }, { "epoch": 0.4976303317535545, "step": 210, "train/alpha": 0.2, "train/clean_ce_loss": 0.0859375, "train/clean_loss_scale": 6.667750835418701, "train/clean_loss_weight": 1.333550214767456, "train/combined_loss": 0.6876118183135986, "train/pos_loss/pos0": 0.2739873230457306, "train/pos_loss/pos1": 0.5546033978462219, "train/pos_loss/pos2": 0.9199235439300537, "train/pos_loss/pos3": 1.3718957901000977, "train/pos_loss/pos4": 1.7780588865280151, "train/pos_loss/pos5": 2.2527525424957275, "train/pos_loss/pos6": 2.6442503929138184, "train/pos_loss/weight0": 3.237712860107422, "train/pos_loss/weight1": 2.0417847633361816, "train/pos_loss/weight2": 1.0538339614868164, "train/pos_loss/weight3": 0.4655037522315979, "train/pos_loss/weight4": 0.15158165991306305, "train/pos_loss/weight5": 0.03895903751254082, "train/pos_loss/weight6": 0.010624291375279427, "train/pos_weight_progress": 0.04976303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5730098485946655 }, { "epoch": 0.4976303317535545, "step": 210, "train/alpha": 0.2, "train/clean_ce_loss": 0.0986328125, "train/clean_loss_scale": 7.797677516937256, "train/clean_loss_weight": 1.5595355033874512, "train/combined_loss": 0.9229282140731812, "train/pos_loss/pos0": 0.38190513849258423, "train/pos_loss/pos1": 0.5686119198799133, "train/pos_loss/pos2": 0.9171980619430542, "train/pos_loss/pos3": 1.4238373041152954, "train/pos_loss/pos4": 1.918570637702942, "train/pos_loss/pos5": 2.2231459617614746, "train/pos_loss/pos6": 2.6199991703033447, "train/pos_loss/weight0": 2.630204677581787, "train/pos_loss/weight1": 1.87398362159729, "train/pos_loss/weight2": 1.22019362449646, "train/pos_loss/weight3": 0.7039976119995117, "train/pos_loss/weight4": 0.34789544343948364, "train/pos_loss/weight5": 0.15238820016384125, "train/pos_loss/weight6": 0.07133691012859344, "train/pos_weight_progress": 0.04976303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7691068649291992 }, { "epoch": 0.4976303317535545, "step": 210, "train/alpha": 0.2, "train/clean_ce_loss": 0.1884765625, "train/clean_loss_scale": 8.567410469055176, "train/clean_loss_weight": 1.713482141494751, "train/combined_loss": 1.9377073049545288, "train/pos_loss/pos0": 0.8931020498275757, "train/pos_loss/pos1": 1.4347556829452515, "train/pos_loss/pos2": 2.172637462615967, "train/pos_loss/pos3": 2.6931610107421875, "train/pos_loss/pos4": 3.078456401824951, "train/pos_loss/pos5": 3.8161587715148926, "train/pos_loss/pos6": 4.571468830108643, "train/pos_loss/weight0": 2.741170883178711, "train/pos_loss/weight1": 1.8960055112838745, "train/pos_loss/weight2": 1.1766003370285034, "train/pos_loss/weight3": 0.6748368740081787, "train/pos_loss/weight4": 0.3172513544559479, "train/pos_loss/weight5": 0.13255532085895538, "train/pos_loss/weight6": 0.0615793839097023, "train/pos_weight_progress": 0.04976303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.6147561073303223 }, { "epoch": 0.5, "step": 211, "train/alpha": 0.2, "train/clean_ce_loss": 0.0247802734375, "train/clean_loss_scale": 11.396142959594727, "train/clean_loss_weight": 2.279228687286377, "train/combined_loss": 0.33887943625450134, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.05, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2823995351791382 }, { "epoch": 0.5, "step": 211, "train/alpha": 0.2, "train/clean_ce_loss": 0.0693359375, "train/clean_loss_scale": 6.147999286651611, "train/clean_loss_weight": 1.2295998334884644, "train/combined_loss": 0.5115327835083008, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.05, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4262773096561432 }, { "epoch": 0.5, "step": 211, "train/alpha": 0.2, "train/clean_ce_loss": 0.1865234375, "train/clean_loss_scale": 5.853665828704834, "train/clean_loss_weight": 1.1707332134246826, "train/combined_loss": 1.3102149963378906, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.05, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0918458700180054 }, { "epoch": 0.5, "step": 211, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 5.931197166442871, "train/clean_loss_weight": 1.18623948097229, "train/combined_loss": 0.5073953866958618, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.05, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4228294789791107 }, { "epoch": 0.5023696682464455, "step": 212, "train/alpha": 0.2, "train/clean_ce_loss": 0.06884765625, "train/clean_loss_scale": 7.207103252410889, "train/clean_loss_weight": 1.4414206743240356, "train/combined_loss": 0.5954306125640869, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.05023696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.49619215726852417 }, { "epoch": 0.5023696682464455, "step": 212, "train/alpha": 0.2, "train/clean_ce_loss": 0.1494140625, "train/clean_loss_scale": 5.016376495361328, "train/clean_loss_weight": 1.0032752752304077, "train/combined_loss": 0.8994206190109253, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.05023696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7495172023773193 }, { "epoch": 0.5023696682464455, "step": 212, "train/alpha": 0.2, "train/clean_ce_loss": 0.134765625, "train/clean_loss_scale": 8.355627059936523, "train/clean_loss_weight": 1.6711254119873047, "train/combined_loss": 1.3512616157531738, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.05023696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1260513067245483 }, { "epoch": 0.5023696682464455, "step": 212, "train/alpha": 0.2, "train/clean_ce_loss": 0.0771484375, "train/clean_loss_scale": 6.591864109039307, "train/clean_loss_weight": 1.3183728456497192, "train/combined_loss": 0.610262393951416, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.05023696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5085520148277283 }, { "epoch": 0.504739336492891, "step": 213, "train/alpha": 0.2, "train/clean_ce_loss": 0.10498046875, "train/clean_loss_scale": 7.216818809509277, "train/clean_loss_weight": 1.4433637857437134, "train/combined_loss": 0.9091500639915466, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.0504739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7576250433921814 }, { "epoch": 0.504739336492891, "step": 213, "train/alpha": 0.2, "train/clean_ce_loss": 0.044921875, "train/clean_loss_scale": 8.17012882232666, "train/clean_loss_weight": 1.6340258121490479, "train/combined_loss": 0.4404210150241852, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.0504739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3670175075531006 }, { "epoch": 0.504739336492891, "step": 213, "train/alpha": 0.2, "train/clean_ce_loss": 0.04345703125, "train/clean_loss_scale": 7.9884033203125, "train/clean_loss_weight": 1.597680687904358, "train/combined_loss": 0.41658276319503784, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.0504739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3471522927284241 }, { "epoch": 0.504739336492891, "step": 213, "train/alpha": 0.2, "train/clean_ce_loss": 0.10205078125, "train/clean_loss_scale": 9.587237358093262, "train/clean_loss_weight": 1.9174474477767944, "train/combined_loss": 1.1740621328353882, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.0504739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9783850908279419 }, { "epoch": 0.5071090047393365, "step": 214, "train/alpha": 0.2, "train/clean_ce_loss": 0.10546875, "train/clean_loss_scale": 6.073468208312988, "train/clean_loss_weight": 1.2146936655044556, "train/combined_loss": 0.768673300743103, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.05071090047393365, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6405611038208008 }, { "epoch": 0.5071090047393365, "step": 214, "train/alpha": 0.2, "train/clean_ce_loss": 0.1083984375, "train/clean_loss_scale": 6.391565799713135, "train/clean_loss_weight": 1.278313159942627, "train/combined_loss": 0.8314028978347778, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.05071090047393365, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6928357481956482 }, { "epoch": 0.5071090047393365, "step": 214, "train/alpha": 0.2, "train/clean_ce_loss": 0.042724609375, "train/clean_loss_scale": 7.715219020843506, "train/clean_loss_weight": 1.543043851852417, "train/combined_loss": 0.39555567502975464, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.05071090047393365, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.32962971925735474 }, { "epoch": 0.5071090047393365, "step": 214, "train/alpha": 0.2, "train/clean_ce_loss": 0.1357421875, "train/clean_loss_scale": 5.340425491333008, "train/clean_loss_weight": 1.0680850744247437, "train/combined_loss": 0.8699052333831787, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.05071090047393365, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7249210476875305 }, { "epoch": 0.509478672985782, "grad_norm": 17.213176727294922, "learning_rate": 2e-06, "loss": 3.801, "step": 215 }, { "epoch": 0.509478672985782, "step": 215, "train/alpha": 0.2, "train/clean_ce_loss": 0.0966796875, "train/clean_loss_scale": 6.8307647705078125, "train/clean_loss_weight": 1.3661530017852783, "train/combined_loss": 0.7924754619598389, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.0509478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6603962182998657 }, { "epoch": 0.509478672985782, "step": 215, "train/alpha": 0.2, "train/clean_ce_loss": 0.035888671875, "train/clean_loss_scale": 11.421487808227539, "train/clean_loss_weight": 2.2842977046966553, "train/combined_loss": 0.49188244342803955, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.0509478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.40990203619003296 }, { "epoch": 0.509478672985782, "step": 215, "train/alpha": 0.2, "train/clean_ce_loss": 0.0703125, "train/clean_loss_scale": 8.37515640258789, "train/clean_loss_weight": 1.675031304359436, "train/combined_loss": 0.7066538333892822, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.0509478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5888782143592834 }, { "epoch": 0.509478672985782, "step": 215, "train/alpha": 0.2, "train/clean_ce_loss": 0.06201171875, "train/clean_loss_scale": 6.090289115905762, "train/clean_loss_weight": 1.2180578708648682, "train/combined_loss": 0.4532031714916229, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.0509478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3776693046092987 }, { "epoch": 0.5118483412322274, "step": 216, "train/alpha": 0.2, "train/clean_ce_loss": 0.140625, "train/clean_loss_scale": 7.1443047523498535, "train/clean_loss_weight": 1.4288610219955444, "train/combined_loss": 1.205601453781128, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.05118483412322275, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.00466787815094 }, { "epoch": 0.5118483412322274, "step": 216, "train/alpha": 0.2, "train/clean_ce_loss": 0.12060546875, "train/clean_loss_scale": 7.261280059814453, "train/clean_loss_weight": 1.4522560834884644, "train/combined_loss": 1.0509001016616821, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.05118483412322275, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8757500648498535 }, { "epoch": 0.5118483412322274, "step": 216, "train/alpha": 0.2, "train/clean_ce_loss": 0.1708984375, "train/clean_loss_scale": 5.18678092956543, "train/clean_loss_weight": 1.0373562574386597, "train/combined_loss": 1.0636953115463257, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.05118483412322275, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8864127397537231 }, { "epoch": 0.5118483412322274, "step": 216, "train/alpha": 0.2, "train/clean_ce_loss": 0.130859375, "train/clean_loss_scale": 8.314449310302734, "train/clean_loss_weight": 1.662889838218689, "train/combined_loss": 1.3056284189224243, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.05118483412322275, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0880236625671387 }, { "epoch": 0.514218009478673, "step": 217, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 5.796388149261475, "train/clean_loss_weight": 1.1592776775360107, "train/combined_loss": 0.45171070098876953, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.051421800947867295, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3764255940914154 }, { "epoch": 0.514218009478673, "step": 217, "train/alpha": 0.2, "train/clean_ce_loss": 0.181640625, "train/clean_loss_scale": 6.363606929779053, "train/clean_loss_weight": 1.2727214097976685, "train/combined_loss": 1.387067437171936, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.051421800947867295, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1558895111083984 }, { "epoch": 0.514218009478673, "step": 217, "train/alpha": 0.2, "train/clean_ce_loss": 0.16796875, "train/clean_loss_scale": 7.742941379547119, "train/clean_loss_weight": 1.5485882759094238, "train/combined_loss": 1.5606865882873535, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.051421800947867295, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.300572156906128 }, { "epoch": 0.514218009478673, "step": 217, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 6.31360387802124, "train/clean_loss_weight": 1.2627208232879639, "train/combined_loss": 0.6029984951019287, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.051421800947867295, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5024987459182739 }, { "epoch": 0.5165876777251185, "step": 218, "train/alpha": 0.2, "train/clean_ce_loss": 0.0439453125, "train/clean_loss_scale": 8.001266479492188, "train/clean_loss_weight": 1.6002533435821533, "train/combined_loss": 0.4219418168067932, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.05165876777251185, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3516181707382202 }, { "epoch": 0.5165876777251185, "step": 218, "train/alpha": 0.2, "train/clean_ce_loss": 0.125, "train/clean_loss_scale": 6.591228485107422, "train/clean_loss_weight": 1.318245768547058, "train/combined_loss": 0.9886842966079712, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.05165876777251185, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8239035606384277 }, { "epoch": 0.5165876777251185, "step": 218, "train/alpha": 0.2, "train/clean_ce_loss": 0.171875, "train/clean_loss_scale": 6.173497200012207, "train/clean_loss_weight": 1.2346994876861572, "train/combined_loss": 1.273283839225769, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.05165876777251185, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0610698461532593 }, { "epoch": 0.5165876777251185, "step": 218, "train/alpha": 0.2, "train/clean_ce_loss": 0.1103515625, "train/clean_loss_scale": 5.143661022186279, "train/clean_loss_weight": 1.028732180595398, "train/combined_loss": 0.6811332702636719, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.05165876777251185, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.567611038684845 }, { "epoch": 0.518957345971564, "step": 219, "train/alpha": 0.2, "train/clean_ce_loss": 0.0625, "train/clean_loss_scale": 6.605418682098389, "train/clean_loss_weight": 1.3210837841033936, "train/combined_loss": 0.4954063892364502, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.0518957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4128386676311493 }, { "epoch": 0.518957345971564, "step": 219, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 10.334855079650879, "train/clean_loss_weight": 2.0669710636138916, "train/combined_loss": 0.9022812843322754, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.0518957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7519010901451111 }, { "epoch": 0.518957345971564, "step": 219, "train/alpha": 0.2, "train/clean_ce_loss": 0.1103515625, "train/clean_loss_scale": 5.259997844696045, "train/clean_loss_weight": 1.051999568939209, "train/combined_loss": 0.6965388059616089, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.0518957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5804489850997925 }, { "epoch": 0.518957345971564, "step": 219, "train/alpha": 0.2, "train/clean_ce_loss": 0.107421875, "train/clean_loss_scale": 6.465440273284912, "train/clean_loss_weight": 1.2930880784988403, "train/combined_loss": 0.8334356546401978, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.0518957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6945297122001648 }, { "epoch": 0.5213270142180095, "grad_norm": 16.12134552001953, "learning_rate": 2e-06, "loss": 3.9085, "step": 220 }, { "epoch": 0.5213270142180095, "step": 220, "train/alpha": 0.2, "train/clean_ce_loss": 0.140625, "train/clean_loss_scale": 7.75838041305542, "train/clean_loss_weight": 1.5516761541366577, "train/combined_loss": 1.3092267513275146, "train/pos_loss/pos0": 0.5601005554199219, "train/pos_loss/pos1": 1.3067355155944824, "train/pos_loss/pos2": 1.9379897117614746, "train/pos_loss/pos3": 2.4149932861328125, "train/pos_loss/pos4": 3.192722797393799, "train/pos_loss/pos5": 4.290172576904297, "train/pos_loss/pos6": 3.910400390625, "train/pos_loss/weight0": 3.587512969970703, "train/pos_loss/weight1": 1.9963321685791016, "train/pos_loss/weight2": 0.968815267086029, "train/pos_loss/weight3": 0.3341616094112396, "train/pos_loss/weight4": 0.09152774512767792, "train/pos_loss/weight5": 0.018108829855918884, "train/pos_loss/weight6": 0.0035416411701589823, "train/pos_weight_progress": 0.052132701421800945, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.091022253036499 }, { "epoch": 0.5213270142180095, "step": 220, "train/alpha": 0.2, "train/clean_ce_loss": 0.0216064453125, "train/clean_loss_scale": 17.975976943969727, "train/clean_loss_weight": 3.5951955318450928, "train/combined_loss": 0.4660763740539551, "train/pos_loss/pos0": 0.11207545548677444, "train/pos_loss/pos1": 0.36703842878341675, "train/pos_loss/pos2": 0.8011788129806519, "train/pos_loss/pos3": 1.406773328781128, "train/pos_loss/pos4": 1.8695573806762695, "train/pos_loss/pos5": 2.2831201553344727, "train/pos_loss/pos6": 2.83567476272583, "train/pos_loss/weight0": 3.3786678314208984, "train/pos_loss/weight1": 2.1214652061462402, "train/pos_loss/weight2": 1.0171239376068115, "train/pos_loss/weight3": 0.35702764987945557, "train/pos_loss/weight4": 0.10289552807807922, "train/pos_loss/weight5": 0.019085358828306198, "train/pos_loss/weight6": 0.0037349762860685587, "train/pos_weight_progress": 0.052132701421800945, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3883969783782959 }, { "epoch": 0.5213270142180095, "step": 220, "train/alpha": 0.2, "train/clean_ce_loss": 0.0576171875, "train/clean_loss_scale": 8.98673152923584, "train/clean_loss_weight": 1.7973463535308838, "train/combined_loss": 0.6213482618331909, "train/pos_loss/pos0": 0.20602408051490784, "train/pos_loss/pos1": 0.41515445709228516, "train/pos_loss/pos2": 0.6383290886878967, "train/pos_loss/pos3": 0.9982593059539795, "train/pos_loss/pos4": 1.305989146232605, "train/pos_loss/pos5": 1.626061201095581, "train/pos_loss/pos6": 1.909238338470459, "train/pos_loss/weight0": 2.5804805755615234, "train/pos_loss/weight1": 1.8946259021759033, "train/pos_loss/weight2": 1.2673920392990112, "train/pos_loss/weight3": 0.7244990468025208, "train/pos_loss/weight4": 0.3385845422744751, "train/pos_loss/weight5": 0.1355334371328354, "train/pos_loss/weight6": 0.058884382247924805, "train/pos_weight_progress": 0.052132701421800945, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5177901983261108 }, { "epoch": 0.5213270142180095, "step": 220, "train/alpha": 0.2, "train/clean_ce_loss": 0.068359375, "train/clean_loss_scale": 12.546395301818848, "train/clean_loss_weight": 2.5092790126800537, "train/combined_loss": 1.0291965007781982, "train/pos_loss/pos0": 0.26476046442985535, "train/pos_loss/pos1": 0.5561507344245911, "train/pos_loss/pos2": 0.9024195671081543, "train/pos_loss/pos3": 1.3627727031707764, "train/pos_loss/pos4": 1.9326887130737305, "train/pos_loss/pos5": 2.2845544815063477, "train/pos_loss/pos6": 2.826809883117676, "train/pos_loss/weight0": 2.230333089828491, "train/pos_loss/weight1": 1.7080857753753662, "train/pos_loss/weight2": 1.2230942249298096, "train/pos_loss/weight3": 0.8247315883636475, "train/pos_loss/weight4": 0.5149797797203064, "train/pos_loss/weight5": 0.3056573271751404, "train/pos_loss/weight6": 0.19311855733394623, "train/pos_weight_progress": 0.052132701421800945, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8576637506484985 }, { "epoch": 0.523696682464455, "step": 221, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 7.93311071395874, "train/clean_loss_weight": 1.5866221189498901, "train/combined_loss": 0.7623223662376404, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.0523696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6352686285972595 }, { "epoch": 0.523696682464455, "step": 221, "train/alpha": 0.2, "train/clean_ce_loss": 0.06787109375, "train/clean_loss_scale": 7.915754795074463, "train/clean_loss_weight": 1.5831509828567505, "train/combined_loss": 0.6447011232376099, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.0523696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5372509360313416 }, { "epoch": 0.523696682464455, "step": 221, "train/alpha": 0.2, "train/clean_ce_loss": 0.1748046875, "train/clean_loss_scale": 8.721209526062012, "train/clean_loss_weight": 1.7442419528961182, "train/combined_loss": 1.8294100761413574, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.0523696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5245083570480347 }, { "epoch": 0.523696682464455, "step": 221, "train/alpha": 0.2, "train/clean_ce_loss": 0.12353515625, "train/clean_loss_scale": 8.78810977935791, "train/clean_loss_weight": 1.7576220035552979, "train/combined_loss": 1.3027687072753906, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.0523696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.085640549659729 }, { "epoch": 0.5260663507109005, "step": 222, "train/alpha": 0.2, "train/clean_ce_loss": 0.1328125, "train/clean_loss_scale": 8.391151428222656, "train/clean_loss_weight": 1.6782302856445312, "train/combined_loss": 1.3373396396636963, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.05260663507109005, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1144497394561768 }, { "epoch": 0.5260663507109005, "step": 222, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 9.568218231201172, "train/clean_loss_weight": 1.913643717765808, "train/combined_loss": 1.0652118921279907, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.05260663507109005, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8876765370368958 }, { "epoch": 0.5260663507109005, "step": 222, "train/alpha": 0.2, "train/clean_ce_loss": 0.072265625, "train/clean_loss_scale": 8.643835067749023, "train/clean_loss_weight": 1.7287670373916626, "train/combined_loss": 0.7495825886726379, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.05260663507109005, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6246521472930908 }, { "epoch": 0.5260663507109005, "step": 222, "train/alpha": 0.2, "train/clean_ce_loss": 0.09423828125, "train/clean_loss_scale": 9.523444175720215, "train/clean_loss_weight": 1.904688835144043, "train/combined_loss": 1.0769675970077515, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.05260663507109005, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8974729776382446 }, { "epoch": 0.5284360189573459, "step": 223, "train/alpha": 0.2, "train/clean_ce_loss": 0.0311279296875, "train/clean_loss_scale": 18.855947494506836, "train/clean_loss_weight": 3.7711894512176514, "train/combined_loss": 0.7043359279632568, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.052843601895734595, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5869466066360474 }, { "epoch": 0.5284360189573459, "step": 223, "train/alpha": 0.2, "train/clean_ce_loss": 0.154296875, "train/clean_loss_scale": 9.35616397857666, "train/clean_loss_weight": 1.8712328672409058, "train/combined_loss": 1.7323522567749023, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.052843601895734595, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.443626880645752 }, { "epoch": 0.5284360189573459, "step": 223, "train/alpha": 0.2, "train/clean_ce_loss": 0.08203125, "train/clean_loss_scale": 8.329733848571777, "train/clean_loss_weight": 1.6659468412399292, "train/combined_loss": 0.8199581503868103, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.052843601895734595, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6832984685897827 }, { "epoch": 0.5284360189573459, "step": 223, "train/alpha": 0.2, "train/clean_ce_loss": 0.06005859375, "train/clean_loss_scale": 8.824848175048828, "train/clean_loss_weight": 1.7649697065353394, "train/combined_loss": 0.6360095739364624, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.052843601895734595, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5300079584121704 }, { "epoch": 0.5308056872037915, "step": 224, "train/alpha": 0.2, "train/clean_ce_loss": 0.0732421875, "train/clean_loss_scale": 8.42330265045166, "train/clean_loss_weight": 1.68466055393219, "train/combined_loss": 0.7403293251991272, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.05308056872037915, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6169410943984985 }, { "epoch": 0.5308056872037915, "step": 224, "train/alpha": 0.2, "train/clean_ce_loss": 0.041259765625, "train/clean_loss_scale": 13.234442710876465, "train/clean_loss_weight": 2.646888494491577, "train/combined_loss": 0.6552600264549255, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.05308056872037915, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5460500121116638 }, { "epoch": 0.5308056872037915, "step": 224, "train/alpha": 0.2, "train/clean_ce_loss": 0.09130859375, "train/clean_loss_scale": 7.189395904541016, "train/clean_loss_weight": 1.437879204750061, "train/combined_loss": 0.7877443432807922, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.05308056872037915, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6564536094665527 }, { "epoch": 0.5308056872037915, "step": 224, "train/alpha": 0.2, "train/clean_ce_loss": 0.08544921875, "train/clean_loss_scale": 7.525470733642578, "train/clean_loss_weight": 1.5050941705703735, "train/combined_loss": 0.7716547250747681, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.05308056872037915, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6430456042289734 }, { "epoch": 0.533175355450237, "grad_norm": 12.944866180419922, "learning_rate": 2e-06, "loss": 3.8421, "step": 225 }, { "epoch": 0.533175355450237, "step": 225, "train/alpha": 0.2, "train/clean_ce_loss": 0.07763671875, "train/clean_loss_scale": 11.792513847351074, "train/clean_loss_weight": 2.3585028648376465, "train/combined_loss": 1.0986385345458984, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.0533175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9155320525169373 }, { "epoch": 0.533175355450237, "step": 225, "train/alpha": 0.2, "train/clean_ce_loss": 0.1669921875, "train/clean_loss_scale": 8.965363502502441, "train/clean_loss_weight": 1.7930727005004883, "train/combined_loss": 1.7965748310089111, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.0533175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.497145652770996 }, { "epoch": 0.533175355450237, "step": 225, "train/alpha": 0.2, "train/clean_ce_loss": 0.0986328125, "train/clean_loss_scale": 7.720211982727051, "train/clean_loss_weight": 1.5440424680709839, "train/combined_loss": 0.9137594699859619, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.0533175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7614662051200867 }, { "epoch": 0.533175355450237, "step": 225, "train/alpha": 0.2, "train/clean_ce_loss": 0.068359375, "train/clean_loss_scale": 10.072397232055664, "train/clean_loss_weight": 2.014479398727417, "train/combined_loss": 0.8262513279914856, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.0533175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6885427832603455 }, { "epoch": 0.5355450236966824, "step": 226, "train/alpha": 0.2, "train/clean_ce_loss": 0.060546875, "train/clean_loss_scale": 9.112818717956543, "train/clean_loss_weight": 1.8225637674331665, "train/combined_loss": 0.6621032357215881, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.053554502369668244, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5517526865005493 }, { "epoch": 0.5355450236966824, "step": 226, "train/alpha": 0.2, "train/clean_ce_loss": 0.08447265625, "train/clean_loss_scale": 7.432928562164307, "train/clean_loss_weight": 1.4865857362747192, "train/combined_loss": 0.7534550428390503, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.053554502369668244, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6278792023658752 }, { "epoch": 0.5355450236966824, "step": 226, "train/alpha": 0.2, "train/clean_ce_loss": 0.1845703125, "train/clean_loss_scale": 6.394505500793457, "train/clean_loss_weight": 1.2789011001586914, "train/combined_loss": 1.4162830114364624, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.053554502369668244, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1802358627319336 }, { "epoch": 0.5355450236966824, "step": 226, "train/alpha": 0.2, "train/clean_ce_loss": 0.05224609375, "train/clean_loss_scale": 10.81184196472168, "train/clean_loss_weight": 2.1623685359954834, "train/combined_loss": 0.6778517961502075, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.053554502369668244, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5648764967918396 }, { "epoch": 0.5379146919431279, "step": 227, "train/alpha": 0.2, "train/clean_ce_loss": 0.1015625, "train/clean_loss_scale": 8.458852767944336, "train/clean_loss_weight": 1.6917705535888672, "train/combined_loss": 1.0309226512908936, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.053791469194312796, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.859102189540863 }, { "epoch": 0.5379146919431279, "step": 227, "train/alpha": 0.2, "train/clean_ce_loss": 0.11767578125, "train/clean_loss_scale": 5.580258846282959, "train/clean_loss_weight": 1.1160517930984497, "train/combined_loss": 0.7879936099052429, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.053791469194312796, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.656661331653595 }, { "epoch": 0.5379146919431279, "step": 227, "train/alpha": 0.2, "train/clean_ce_loss": 0.062255859375, "train/clean_loss_scale": 10.047605514526367, "train/clean_loss_weight": 2.009521245956421, "train/combined_loss": 0.750626802444458, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.053791469194312796, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6255223155021667 }, { "epoch": 0.5379146919431279, "step": 227, "train/alpha": 0.2, "train/clean_ce_loss": 0.1005859375, "train/clean_loss_scale": 10.14763069152832, "train/clean_loss_weight": 2.0295262336730957, "train/combined_loss": 1.2248506546020508, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.053791469194312796, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0207089185714722 }, { "epoch": 0.5402843601895735, "step": 228, "train/alpha": 0.2, "train/clean_ce_loss": 0.057861328125, "train/clean_loss_scale": 7.855093955993652, "train/clean_loss_weight": 1.5710188150405884, "train/combined_loss": 0.5454074144363403, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.05402843601895735, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45450615882873535 }, { "epoch": 0.5402843601895735, "step": 228, "train/alpha": 0.2, "train/clean_ce_loss": 0.08642578125, "train/clean_loss_scale": 7.8790082931518555, "train/clean_loss_weight": 1.5758017301559448, "train/combined_loss": 0.8171393275260925, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.05402843601895735, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6809494495391846 }, { "epoch": 0.5402843601895735, "step": 228, "train/alpha": 0.2, "train/clean_ce_loss": 0.11328125, "train/clean_loss_scale": 8.387303352355957, "train/clean_loss_weight": 1.6774606704711914, "train/combined_loss": 1.1401489973068237, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.05402843601895735, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.950124204158783 }, { "epoch": 0.5402843601895735, "step": 228, "train/alpha": 0.2, "train/clean_ce_loss": 0.03564453125, "train/clean_loss_scale": 14.21950912475586, "train/clean_loss_weight": 2.8439018726348877, "train/combined_loss": 0.6082172989845276, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.05402843601895735, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5068477392196655 }, { "epoch": 0.542654028436019, "step": 229, "train/alpha": 0.2, "train/clean_ce_loss": 0.1376953125, "train/clean_loss_scale": 10.983872413635254, "train/clean_loss_weight": 2.196774482727051, "train/combined_loss": 1.8149133920669556, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.054265402843601894, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.512427806854248 }, { "epoch": 0.542654028436019, "step": 229, "train/alpha": 0.2, "train/clean_ce_loss": 0.154296875, "train/clean_loss_scale": 6.819482803344727, "train/clean_loss_weight": 1.3638966083526611, "train/combined_loss": 1.2626698017120361, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.054265402843601894, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.05222487449646 }, { "epoch": 0.542654028436019, "step": 229, "train/alpha": 0.2, "train/clean_ce_loss": 0.052001953125, "train/clean_loss_scale": 11.577744483947754, "train/clean_loss_weight": 2.315548896789551, "train/combined_loss": 0.7224783897399902, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.054265402843601894, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6020653247833252 }, { "epoch": 0.542654028436019, "step": 229, "train/alpha": 0.2, "train/clean_ce_loss": 0.08544921875, "train/clean_loss_scale": 8.013375282287598, "train/clean_loss_weight": 1.6026750802993774, "train/combined_loss": 0.8216840028762817, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.054265402843601894, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6847366690635681 }, { "epoch": 0.5450236966824644, "grad_norm": 16.054414749145508, "learning_rate": 2e-06, "loss": 3.7283, "step": 230 }, { "epoch": 0.5450236966824644, "step": 230, "train/alpha": 0.2, "train/clean_ce_loss": 0.0859375, "train/clean_loss_scale": 6.668150424957275, "train/clean_loss_weight": 1.333630084991455, "train/combined_loss": 0.6876530051231384, "train/pos_loss/pos0": 0.26465871930122375, "train/pos_loss/pos1": 0.5026355981826782, "train/pos_loss/pos2": 0.6610429883003235, "train/pos_loss/pos3": 1.0523288249969482, "train/pos_loss/pos4": 1.4169105291366577, "train/pos_loss/pos5": 1.758895993232727, "train/pos_loss/pos6": 2.010045051574707, "train/pos_loss/weight0": 2.6682586669921875, "train/pos_loss/weight1": 1.9138931035995483, "train/pos_loss/weight2": 1.2079036235809326, "train/pos_loss/weight3": 0.6864076852798462, "train/pos_loss/weight4": 0.32869675755500793, "train/pos_loss/weight5": 0.13338828086853027, "train/pos_loss/weight6": 0.061451733112335205, "train/pos_weight_progress": 0.054502369668246446, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5730441808700562 }, { "epoch": 0.5450236966824644, "step": 230, "train/alpha": 0.2, "train/clean_ce_loss": 0.10693359375, "train/clean_loss_scale": 11.095453262329102, "train/clean_loss_weight": 2.219090700149536, "train/combined_loss": 1.4237720966339111, "train/pos_loss/pos0": 0.466829776763916, "train/pos_loss/pos1": 0.7580975294113159, "train/pos_loss/pos2": 1.536517858505249, "train/pos_loss/pos3": 2.0802268981933594, "train/pos_loss/pos4": 2.6150004863739014, "train/pos_loss/pos5": 2.9535536766052246, "train/pos_loss/pos6": 3.321345567703247, "train/pos_loss/weight0": 2.385092258453369, "train/pos_loss/weight1": 1.7922965288162231, "train/pos_loss/weight2": 1.209419846534729, "train/pos_loss/weight3": 0.772910475730896, "train/pos_loss/weight4": 0.4502907395362854, "train/pos_loss/weight5": 0.24868682026863098, "train/pos_loss/weight6": 0.1413034200668335, "train/pos_weight_progress": 0.054502369668246446, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.186476707458496 }, { "epoch": 0.5450236966824644, "step": 230, "train/alpha": 0.2, "train/clean_ce_loss": 0.0224609375, "train/clean_loss_scale": 14.219740867614746, "train/clean_loss_weight": 2.8439481258392334, "train/combined_loss": 0.3832664489746094, "train/pos_loss/pos0": 0.038764990866184235, "train/pos_loss/pos1": 0.18022440373897552, "train/pos_loss/pos2": 0.5920419692993164, "train/pos_loss/pos3": 1.279750108718872, "train/pos_loss/pos4": 1.571769118309021, "train/pos_loss/pos5": 2.1936564445495605, "train/pos_loss/pos6": 2.542698860168457, "train/pos_loss/weight0": 3.11513352394104, "train/pos_loss/weight1": 1.9742145538330078, "train/pos_loss/weight2": 1.1640541553497314, "train/pos_loss/weight3": 0.5116783380508423, "train/pos_loss/weight4": 0.17051956057548523, "train/pos_loss/weight5": 0.05044352635741234, "train/pos_loss/weight6": 0.013955870643258095, "train/pos_weight_progress": 0.054502369668246446, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3193887174129486 }, { "epoch": 0.5450236966824644, "step": 230, "train/alpha": 0.2, "train/clean_ce_loss": 0.125, "train/clean_loss_scale": 10.193143844604492, "train/clean_loss_weight": 2.0386288166046143, "train/combined_loss": 1.528971552848816, "train/pos_loss/pos0": 0.4933638274669647, "train/pos_loss/pos1": 1.0857889652252197, "train/pos_loss/pos2": 1.637057900428772, "train/pos_loss/pos3": 2.1294922828674316, "train/pos_loss/pos4": 2.563351631164551, "train/pos_loss/pos5": 2.9974403381347656, "train/pos_loss/pos6": 3.5489914417266846, "train/pos_loss/weight0": 2.415250778198242, "train/pos_loss/weight1": 1.8094099760055542, "train/pos_loss/weight2": 1.2739256620407104, "train/pos_loss/weight3": 0.7996955513954163, "train/pos_loss/weight4": 0.41325971484184265, "train/pos_loss/weight5": 0.19152796268463135, "train/pos_loss/weight6": 0.09693026542663574, "train/pos_weight_progress": 0.054502369668246446, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2741429805755615 }, { "epoch": 0.54739336492891, "step": 231, "train/alpha": 0.2, "train/clean_ce_loss": 0.130859375, "train/clean_loss_scale": 6.685166835784912, "train/clean_loss_weight": 1.3370333909988403, "train/combined_loss": 1.0497801303863525, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.054739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8748167753219604 }, { "epoch": 0.54739336492891, "step": 231, "train/alpha": 0.2, "train/clean_ce_loss": 0.0732421875, "train/clean_loss_scale": 8.10543155670166, "train/clean_loss_weight": 1.6210863590240479, "train/combined_loss": 0.7123914361000061, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.054739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.593659520149231 }, { "epoch": 0.54739336492891, "step": 231, "train/alpha": 0.2, "train/clean_ce_loss": 0.09375, "train/clean_loss_scale": 6.6598219871521, "train/clean_loss_weight": 1.331964373588562, "train/combined_loss": 0.7492299675941467, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.054739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6243582963943481 }, { "epoch": 0.54739336492891, "step": 231, "train/alpha": 0.2, "train/clean_ce_loss": 0.06982421875, "train/clean_loss_scale": 7.869899272918701, "train/clean_loss_weight": 1.5739798545837402, "train/combined_loss": 0.6594114899635315, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.054739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.549509584903717 }, { "epoch": 0.5497630331753555, "step": 232, "train/alpha": 0.2, "train/clean_ce_loss": 0.1640625, "train/clean_loss_scale": 6.952772617340088, "train/clean_loss_weight": 1.3905545473098755, "train/combined_loss": 1.3688271045684814, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.054976303317535544, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1406892538070679 }, { "epoch": 0.5497630331753555, "step": 232, "train/alpha": 0.2, "train/clean_ce_loss": 0.169921875, "train/clean_loss_scale": 5.833526611328125, "train/clean_loss_weight": 1.1667053699493408, "train/combined_loss": 1.1894925832748413, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.054976303317535544, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9912437796592712 }, { "epoch": 0.5497630331753555, "step": 232, "train/alpha": 0.2, "train/clean_ce_loss": 0.0274658203125, "train/clean_loss_scale": 12.922382354736328, "train/clean_loss_weight": 2.5844764709472656, "train/combined_loss": 0.425908625125885, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.054976303317535544, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3549238443374634 }, { "epoch": 0.5497630331753555, "step": 232, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 8.408891677856445, "train/clean_loss_weight": 1.6817783117294312, "train/combined_loss": 0.6602293848991394, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.054976303317535544, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5501911640167236 }, { "epoch": 0.5521327014218009, "step": 233, "train/alpha": 0.2, "train/clean_ce_loss": 0.0771484375, "train/clean_loss_scale": 6.356504440307617, "train/clean_loss_weight": 1.2713009119033813, "train/combined_loss": 0.5884732604026794, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.055213270142180096, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.49039438366889954 }, { "epoch": 0.5521327014218009, "step": 233, "train/alpha": 0.2, "train/clean_ce_loss": 0.1513671875, "train/clean_loss_scale": 6.077821731567383, "train/clean_loss_weight": 1.2155643701553345, "train/combined_loss": 1.1039793491363525, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.055213270142180096, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9199827909469604 }, { "epoch": 0.5521327014218009, "step": 233, "train/alpha": 0.2, "train/clean_ce_loss": 0.11376953125, "train/clean_loss_scale": 7.488616943359375, "train/clean_loss_weight": 1.4977234601974487, "train/combined_loss": 1.022371768951416, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.055213270142180096, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8519764542579651 }, { "epoch": 0.5521327014218009, "step": 233, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 6.191376686096191, "train/clean_loss_weight": 1.238275408744812, "train/combined_loss": 0.5405362248420715, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.055213270142180096, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45044684410095215 }, { "epoch": 0.5545023696682464, "step": 234, "train/alpha": 0.2, "train/clean_ce_loss": 0.09814453125, "train/clean_loss_scale": 6.227477550506592, "train/clean_loss_weight": 1.2454955577850342, "train/combined_loss": 0.7334314584732056, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.05545023696682464, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6111928820610046 }, { "epoch": 0.5545023696682464, "step": 234, "train/alpha": 0.2, "train/clean_ce_loss": 0.12353515625, "train/clean_loss_scale": 5.864379405975342, "train/clean_loss_weight": 1.1728758811950684, "train/combined_loss": 0.869348406791687, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.05545023696682464, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7244570255279541 }, { "epoch": 0.5545023696682464, "step": 234, "train/alpha": 0.2, "train/clean_ce_loss": 0.0015716552734375, "train/clean_loss_scale": 150.96658325195312, "train/clean_loss_weight": 30.193317413330078, "train/combined_loss": 0.28472089767456055, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.05545023696682464, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2372674196958542 }, { "epoch": 0.5545023696682464, "step": 234, "train/alpha": 0.2, "train/clean_ce_loss": 0.107421875, "train/clean_loss_scale": 6.367303848266602, "train/clean_loss_weight": 1.2734607458114624, "train/combined_loss": 0.8207852840423584, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.05545023696682464, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6839877367019653 }, { "epoch": 0.556872037914692, "grad_norm": 16.225847244262695, "learning_rate": 2e-06, "loss": 3.7922, "step": 235 }, { "epoch": 0.556872037914692, "step": 235, "train/alpha": 0.2, "train/clean_ce_loss": 0.06103515625, "train/clean_loss_scale": 9.205181121826172, "train/clean_loss_weight": 1.8410362005233765, "train/combined_loss": 0.6742075681686401, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.05568720379146919, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5618396401405334 }, { "epoch": 0.556872037914692, "step": 235, "train/alpha": 0.2, "train/clean_ce_loss": 0.035400390625, "train/clean_loss_scale": 10.312493324279785, "train/clean_loss_weight": 2.0624988079071045, "train/combined_loss": 0.4380795359611511, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.05568720379146919, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3650662899017334 }, { "epoch": 0.556872037914692, "step": 235, "train/alpha": 0.2, "train/clean_ce_loss": 0.150390625, "train/clean_loss_scale": 7.490016460418701, "train/clean_loss_weight": 1.498003363609314, "train/combined_loss": 1.3517138957977295, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.05568720379146919, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.126428246498108 }, { "epoch": 0.556872037914692, "step": 235, "train/alpha": 0.2, "train/clean_ce_loss": 0.14453125, "train/clean_loss_scale": 6.121349811553955, "train/clean_loss_weight": 1.224269986152649, "train/combined_loss": 1.0616716146469116, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.05568720379146919, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.884726345539093 }, { "epoch": 0.5592417061611374, "step": 236, "train/alpha": 0.2, "train/clean_ce_loss": 0.11669921875, "train/clean_loss_scale": 7.664383411407471, "train/clean_loss_weight": 1.53287672996521, "train/combined_loss": 1.0733129978179932, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.055924170616113746, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8944275379180908 }, { "epoch": 0.5592417061611374, "step": 236, "train/alpha": 0.2, "train/clean_ce_loss": 0.059814453125, "train/clean_loss_scale": 6.189406394958496, "train/clean_loss_weight": 1.2378813028335571, "train/combined_loss": 0.4442591369152069, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.055924170616113746, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3702159523963928 }, { "epoch": 0.5592417061611374, "step": 236, "train/alpha": 0.2, "train/clean_ce_loss": 0.099609375, "train/clean_loss_scale": 6.077897548675537, "train/clean_loss_weight": 1.2155795097351074, "train/combined_loss": 0.7264987230300903, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.055924170616113746, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6054155826568604 }, { "epoch": 0.5592417061611374, "step": 236, "train/alpha": 0.2, "train/clean_ce_loss": 0.078125, "train/clean_loss_scale": 4.718809127807617, "train/clean_loss_weight": 0.9437618255615234, "train/combined_loss": 0.4423883557319641, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.055924170616113746, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3686569631099701 }, { "epoch": 0.5616113744075829, "step": 237, "train/alpha": 0.2, "train/clean_ce_loss": 0.037353515625, "train/clean_loss_scale": 12.09244441986084, "train/clean_loss_weight": 2.4184889793395996, "train/combined_loss": 0.542034387588501, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.05616113744075829, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4516953229904175 }, { "epoch": 0.5616113744075829, "step": 237, "train/alpha": 0.2, "train/clean_ce_loss": 0.1533203125, "train/clean_loss_scale": 5.961300373077393, "train/clean_loss_weight": 1.1922601461410522, "train/combined_loss": 1.0967861413955688, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.05616113744075829, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9139884114265442 }, { "epoch": 0.5616113744075829, "step": 237, "train/alpha": 0.2, "train/clean_ce_loss": 0.08251953125, "train/clean_loss_scale": 6.688277244567871, "train/clean_loss_weight": 1.3376554250717163, "train/combined_loss": 0.6622961759567261, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.05616113744075829, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5519134998321533 }, { "epoch": 0.5616113744075829, "step": 237, "train/alpha": 0.2, "train/clean_ce_loss": 0.1474609375, "train/clean_loss_scale": 6.13253927230835, "train/clean_loss_weight": 1.2265079021453857, "train/combined_loss": 1.0851719379425049, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.05616113744075829, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9043099880218506 }, { "epoch": 0.5639810426540285, "step": 238, "train/alpha": 0.2, "train/clean_ce_loss": 0.103515625, "train/clean_loss_scale": 6.1924848556518555, "train/clean_loss_weight": 1.238497018814087, "train/combined_loss": 0.7692227363586426, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.05639810426540284, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6410189270973206 }, { "epoch": 0.5639810426540285, "step": 238, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 6.570163249969482, "train/clean_loss_weight": 1.3140326738357544, "train/combined_loss": 0.5620569586753845, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.05639810426540284, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46838077902793884 }, { "epoch": 0.5639810426540285, "step": 238, "train/alpha": 0.2, "train/clean_ce_loss": 0.171875, "train/clean_loss_scale": 9.334918975830078, "train/clean_loss_weight": 1.8669837713241577, "train/combined_loss": 1.925326943397522, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.05639810426540284, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.60443913936615 }, { "epoch": 0.5639810426540285, "step": 238, "train/alpha": 0.2, "train/clean_ce_loss": 0.1240234375, "train/clean_loss_scale": 7.122486591339111, "train/clean_loss_weight": 1.424497365951538, "train/combined_loss": 1.0600262880325317, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.05639810426540284, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8833552598953247 }, { "epoch": 0.566350710900474, "step": 239, "train/alpha": 0.2, "train/clean_ce_loss": 0.18359375, "train/clean_loss_scale": 6.748507022857666, "train/clean_loss_weight": 1.3497014045715332, "train/combined_loss": 1.4867805242538452, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.056635071090047395, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2389837503433228 }, { "epoch": 0.566350710900474, "step": 239, "train/alpha": 0.2, "train/clean_ce_loss": 0.06103515625, "train/clean_loss_scale": 8.375096321105957, "train/clean_loss_weight": 1.6750192642211914, "train/combined_loss": 0.6134104132652283, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.056635071090047395, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5111753344535828 }, { "epoch": 0.566350710900474, "step": 239, "train/alpha": 0.2, "train/clean_ce_loss": 0.048095703125, "train/clean_loss_scale": 7.673814296722412, "train/clean_loss_weight": 1.5347628593444824, "train/combined_loss": 0.44289299845695496, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.056635071090047395, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.36907750368118286 }, { "epoch": 0.566350710900474, "step": 239, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 7.9507737159729, "train/clean_loss_weight": 1.590154767036438, "train/combined_loss": 0.6755051612854004, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.056635071090047395, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5629209876060486 }, { "epoch": 0.5687203791469194, "grad_norm": 15.722723960876465, "learning_rate": 2e-06, "loss": 3.6119, "step": 240 }, { "epoch": 0.5687203791469194, "step": 240, "train/alpha": 0.2, "train/clean_ce_loss": 0.19921875, "train/clean_loss_scale": 7.244471549987793, "train/clean_loss_weight": 1.4488943815231323, "train/combined_loss": 1.7318814992904663, "train/pos_loss/pos0": 0.8273919820785522, "train/pos_loss/pos1": 1.5562494993209839, "train/pos_loss/pos2": 2.125011444091797, "train/pos_loss/pos3": 2.7012534141540527, "train/pos_loss/pos4": 3.1612961292266846, "train/pos_loss/pos5": 3.7592546939849854, "train/pos_loss/pos6": 4.1257781982421875, "train/pos_loss/weight0": 3.1640961170196533, "train/pos_loss/weight1": 2.062664747238159, "train/pos_loss/weight2": 1.0893328189849854, "train/pos_loss/weight3": 0.47362592816352844, "train/pos_loss/weight4": 0.15772223472595215, "train/pos_loss/weight5": 0.041522275656461716, "train/pos_loss/weight6": 0.01103561744093895, "train/pos_weight_progress": 0.05687203791469194, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4432345628738403 }, { "epoch": 0.5687203791469194, "step": 240, "train/alpha": 0.2, "train/clean_ce_loss": 0.12158203125, "train/clean_loss_scale": 4.305046081542969, "train/clean_loss_weight": 0.8610092401504517, "train/combined_loss": 0.6280994415283203, "train/pos_loss/pos0": 0.29117029905319214, "train/pos_loss/pos1": 0.5892701745033264, "train/pos_loss/pos2": 0.900415301322937, "train/pos_loss/pos3": 1.3555972576141357, "train/pos_loss/pos4": 1.7534327507019043, "train/pos_loss/pos5": 2.054899215698242, "train/pos_loss/pos6": 2.348036050796509, "train/pos_loss/weight0": 3.5912258625030518, "train/pos_loss/weight1": 2.105942726135254, "train/pos_loss/weight2": 0.9261673092842102, "train/pos_loss/weight3": 0.2989353835582733, "train/pos_loss/weight4": 0.06495921313762665, "train/pos_loss/weight5": 0.011088031344115734, "train/pos_loss/weight6": 0.0016813190886750817, "train/pos_weight_progress": 0.05687203791469194, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5234162211418152 }, { "epoch": 0.5687203791469194, "step": 240, "train/alpha": 0.2, "train/clean_ce_loss": 0.10302734375, "train/clean_loss_scale": 7.349550724029541, "train/clean_loss_weight": 1.4699101448059082, "train/combined_loss": 0.9086456298828125, "train/pos_loss/pos0": 0.30140578746795654, "train/pos_loss/pos1": 0.6284923553466797, "train/pos_loss/pos2": 0.9329119920730591, "train/pos_loss/pos3": 1.4202593564987183, "train/pos_loss/pos4": 1.9117913246154785, "train/pos_loss/pos5": 2.449587821960449, "train/pos_loss/pos6": 2.9313735961914062, "train/pos_loss/weight0": 2.6670427322387695, "train/pos_loss/weight1": 1.898268222808838, "train/pos_loss/weight2": 1.1931995153427124, "train/pos_loss/weight3": 0.6806488633155823, "train/pos_loss/weight4": 0.3385753035545349, "train/pos_loss/weight5": 0.1517968773841858, "train/pos_loss/weight6": 0.07046815752983093, "train/pos_weight_progress": 0.05687203791469194, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7572047114372253 }, { "epoch": 0.5687203791469194, "step": 240, "train/alpha": 0.2, "train/clean_ce_loss": 0.0859375, "train/clean_loss_scale": 8.16096305847168, "train/clean_loss_weight": 1.632192611694336, "train/combined_loss": 0.8415992856025696, "train/pos_loss/pos0": 0.3070613443851471, "train/pos_loss/pos1": 0.5995963215827942, "train/pos_loss/pos2": 0.9492274522781372, "train/pos_loss/pos3": 1.243073582649231, "train/pos_loss/pos4": 1.7753379344940186, "train/pos_loss/pos5": 2.161067247390747, "train/pos_loss/pos6": 2.569929838180542, "train/pos_loss/weight0": 2.7164087295532227, "train/pos_loss/weight1": 1.9304991960525513, "train/pos_loss/weight2": 1.2072346210479736, "train/pos_loss/weight3": 0.6598313450813293, "train/pos_loss/weight4": 0.30905723571777344, "train/pos_loss/weight5": 0.12330427765846252, "train/pos_loss/weight6": 0.05366448312997818, "train/pos_weight_progress": 0.05687203791469194, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7013327479362488 }, { "epoch": 0.5710900473933649, "step": 241, "train/alpha": 0.2, "train/clean_ce_loss": 0.126953125, "train/clean_loss_scale": 8.548887252807617, "train/clean_loss_weight": 1.7097774744033813, "train/combined_loss": 1.3023695945739746, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05710900473933649, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0853079557418823 }, { "epoch": 0.5710900473933649, "step": 241, "train/alpha": 0.2, "train/clean_ce_loss": 0.060546875, "train/clean_loss_scale": 10.41580581665039, "train/clean_loss_weight": 2.0831611156463623, "train/combined_loss": 0.7567734122276306, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05710900473933649, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6306445002555847 }, { "epoch": 0.5710900473933649, "step": 241, "train/alpha": 0.2, "train/clean_ce_loss": 0.09521484375, "train/clean_loss_scale": 7.04004430770874, "train/clean_loss_weight": 1.4080089330673218, "train/combined_loss": 0.8043800592422485, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05710900473933649, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6703166961669922 }, { "epoch": 0.5710900473933649, "step": 241, "train/alpha": 0.2, "train/clean_ce_loss": 0.017333984375, "train/clean_loss_scale": 22.480798721313477, "train/clean_loss_weight": 4.49616003036499, "train/combined_loss": 0.4676181674003601, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05710900473933649, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3896818161010742 }, { "epoch": 0.5734597156398105, "step": 242, "train/alpha": 0.2, "train/clean_ce_loss": 0.07177734375, "train/clean_loss_scale": 9.16287899017334, "train/clean_loss_weight": 1.832575798034668, "train/combined_loss": 0.7892245650291443, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.057345971563981045, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6576871275901794 }, { "epoch": 0.5734597156398105, "step": 242, "train/alpha": 0.2, "train/clean_ce_loss": 0.11669921875, "train/clean_loss_scale": 6.575939178466797, "train/clean_loss_weight": 1.3151878118515015, "train/combined_loss": 0.9208883047103882, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.057345971563981045, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7674069404602051 }, { "epoch": 0.5734597156398105, "step": 242, "train/alpha": 0.2, "train/clean_ce_loss": 0.1240234375, "train/clean_loss_scale": 7.462796211242676, "train/clean_loss_weight": 1.4925593137741089, "train/combined_loss": 1.1106740236282349, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.057345971563981045, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9255616664886475 }, { "epoch": 0.5734597156398105, "step": 242, "train/alpha": 0.2, "train/clean_ce_loss": 0.0224609375, "train/clean_loss_scale": 22.2127628326416, "train/clean_loss_weight": 4.44255256652832, "train/combined_loss": 0.5987033843994141, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.057345971563981045, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.49891945719718933 }, { "epoch": 0.5758293838862559, "step": 243, "train/alpha": 0.2, "train/clean_ce_loss": 0.1396484375, "train/clean_loss_scale": 7.279393196105957, "train/clean_loss_weight": 1.4558786153793335, "train/combined_loss": 1.2198671102523804, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05758293838862559, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.016555905342102 }, { "epoch": 0.5758293838862559, "step": 243, "train/alpha": 0.2, "train/clean_ce_loss": 0.033203125, "train/clean_loss_scale": 13.62583065032959, "train/clean_loss_weight": 2.725166082382202, "train/combined_loss": 0.5429041981697083, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05758293838862559, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4524201452732086 }, { "epoch": 0.5758293838862559, "step": 243, "train/alpha": 0.2, "train/clean_ce_loss": 0.1943359375, "train/clean_loss_scale": 7.529552459716797, "train/clean_loss_weight": 1.5059105157852173, "train/combined_loss": 1.7559151649475098, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05758293838862559, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.463262677192688 }, { "epoch": 0.5758293838862559, "step": 243, "train/alpha": 0.2, "train/clean_ce_loss": 0.01611328125, "train/clean_loss_scale": 24.312381744384766, "train/clean_loss_weight": 4.862476348876953, "train/combined_loss": 0.4701026976108551, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05758293838862559, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3917522430419922 }, { "epoch": 0.5781990521327014, "step": 244, "train/alpha": 0.2, "train/clean_ce_loss": 0.00482177734375, "train/clean_loss_scale": 66.58419799804688, "train/clean_loss_weight": 13.316840171813965, "train/combined_loss": 0.3852649927139282, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05781990521327014, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3210541605949402 }, { "epoch": 0.5781990521327014, "step": 244, "train/alpha": 0.2, "train/clean_ce_loss": 0.07763671875, "train/clean_loss_scale": 8.362424850463867, "train/clean_loss_weight": 1.6724849939346313, "train/combined_loss": 0.7790775299072266, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05781990521327014, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6492312550544739 }, { "epoch": 0.5781990521327014, "step": 244, "train/alpha": 0.2, "train/clean_ce_loss": 0.0869140625, "train/clean_loss_scale": 8.284476280212402, "train/clean_loss_weight": 1.6568952798843384, "train/combined_loss": 0.8640449643135071, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05781990521327014, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7200374603271484 }, { "epoch": 0.5781990521327014, "step": 244, "train/alpha": 0.2, "train/clean_ce_loss": 0.1142578125, "train/clean_loss_scale": 9.394635200500488, "train/clean_loss_weight": 1.8789271116256714, "train/combined_loss": 1.2880926132202148, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05781990521327014, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0734105110168457 }, { "epoch": 0.580568720379147, "grad_norm": 18.823226928710938, "learning_rate": 2e-06, "loss": 3.7785, "step": 245 }, { "epoch": 0.580568720379147, "step": 245, "train/alpha": 0.2, "train/clean_ce_loss": 0.0262451171875, "train/clean_loss_scale": 17.53953742980957, "train/clean_loss_weight": 3.5079076290130615, "train/combined_loss": 0.5523926615715027, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.058056872037914695, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46032723784446716 }, { "epoch": 0.580568720379147, "step": 245, "train/alpha": 0.2, "train/clean_ce_loss": 0.0235595703125, "train/clean_loss_scale": 16.325700759887695, "train/clean_loss_weight": 3.2651402950286865, "train/combined_loss": 0.4615517854690552, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.058056872037914695, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.38462647795677185 }, { "epoch": 0.580568720379147, "step": 245, "train/alpha": 0.2, "train/clean_ce_loss": 0.1181640625, "train/clean_loss_scale": 8.573409080505371, "train/clean_loss_weight": 1.71468186378479, "train/combined_loss": 1.2156825065612793, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.058056872037914695, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0130687952041626 }, { "epoch": 0.580568720379147, "step": 245, "train/alpha": 0.2, "train/clean_ce_loss": 0.1748046875, "train/clean_loss_scale": 7.977416515350342, "train/clean_loss_weight": 1.5954833030700684, "train/combined_loss": 1.6733877658843994, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.058056872037914695, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3944897651672363 }, { "epoch": 0.5829383886255924, "step": 246, "train/alpha": 0.2, "train/clean_ce_loss": 0.041015625, "train/clean_loss_scale": 11.230388641357422, "train/clean_loss_weight": 2.2460777759552, "train/combined_loss": 0.5527456998825073, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05829383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4606214165687561 }, { "epoch": 0.5829383886255924, "step": 246, "train/alpha": 0.2, "train/clean_ce_loss": 0.11083984375, "train/clean_loss_scale": 8.466714859008789, "train/clean_loss_weight": 1.6933430433273315, "train/combined_loss": 1.1261392831802368, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05829383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9384493827819824 }, { "epoch": 0.5829383886255924, "step": 246, "train/alpha": 0.2, "train/clean_ce_loss": 0.043701171875, "train/clean_loss_scale": 10.626627922058105, "train/clean_loss_weight": 2.1253256797790527, "train/combined_loss": 0.5572752952575684, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05829383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46439608931541443 }, { "epoch": 0.5829383886255924, "step": 246, "train/alpha": 0.2, "train/clean_ce_loss": 0.1533203125, "train/clean_loss_scale": 6.133711814880371, "train/clean_loss_weight": 1.2267423868179321, "train/combined_loss": 1.128507137298584, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05829383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9404225945472717 }, { "epoch": 0.5853080568720379, "step": 247, "train/alpha": 0.2, "train/clean_ce_loss": 0.12109375, "train/clean_loss_scale": 9.696063041687012, "train/clean_loss_weight": 1.939212679862976, "train/combined_loss": 1.408959150314331, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05853080568720379, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1741325855255127 }, { "epoch": 0.5853080568720379, "step": 247, "train/alpha": 0.2, "train/clean_ce_loss": 0.07666015625, "train/clean_loss_scale": 8.249832153320312, "train/clean_loss_weight": 1.6499664783477783, "train/combined_loss": 0.7589200735092163, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05853080568720379, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6324334144592285 }, { "epoch": 0.5853080568720379, "step": 247, "train/alpha": 0.2, "train/clean_ce_loss": 0.11669921875, "train/clean_loss_scale": 7.419841766357422, "train/clean_loss_weight": 1.4839683771133423, "train/combined_loss": 1.0390677452087402, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05853080568720379, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8658897280693054 }, { "epoch": 0.5853080568720379, "step": 247, "train/alpha": 0.2, "train/clean_ce_loss": 0.0299072265625, "train/clean_loss_scale": 16.22068214416504, "train/clean_loss_weight": 3.2441365718841553, "train/combined_loss": 0.5821387767791748, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05853080568720379, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4851156175136566 }, { "epoch": 0.5876777251184834, "step": 248, "train/alpha": 0.2, "train/clean_ce_loss": 0.1083984375, "train/clean_loss_scale": 8.788582801818848, "train/clean_loss_weight": 1.7577165365219116, "train/combined_loss": 1.143202304840088, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.058767772511848344, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9526686072349548 }, { "epoch": 0.5876777251184834, "step": 248, "train/alpha": 0.2, "train/clean_ce_loss": 0.1318359375, "train/clean_loss_scale": 8.060962677001953, "train/clean_loss_weight": 1.6121925115585327, "train/combined_loss": 1.2752695083618164, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.058767772511848344, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0627245903015137 }, { "epoch": 0.5876777251184834, "step": 248, "train/alpha": 0.2, "train/clean_ce_loss": 0.181640625, "train/clean_loss_scale": 8.705419540405273, "train/clean_loss_weight": 1.7410839796066284, "train/combined_loss": 1.8975093364715576, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.058767772511848344, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5812578201293945 }, { "epoch": 0.5876777251184834, "step": 248, "train/alpha": 0.2, "train/clean_ce_loss": 0.0859375, "train/clean_loss_scale": 9.262337684631348, "train/clean_loss_weight": 1.8524675369262695, "train/combined_loss": 0.9551786184310913, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.058767772511848344, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7959821820259094 }, { "epoch": 0.590047393364929, "step": 249, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 8.410491943359375, "train/clean_loss_weight": 1.682098388671875, "train/combined_loss": 0.8180516958236694, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05900473933649289, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6817097663879395 }, { "epoch": 0.590047393364929, "step": 249, "train/alpha": 0.2, "train/clean_ce_loss": 0.08447265625, "train/clean_loss_scale": 9.236847877502441, "train/clean_loss_weight": 1.8473695516586304, "train/combined_loss": 0.9363133311271667, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05900473933649289, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7802610993385315 }, { "epoch": 0.590047393364929, "step": 249, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 7.351244926452637, "train/clean_loss_weight": 1.470249056816101, "train/combined_loss": 0.7365602850914001, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05900473933649289, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6138002276420593 }, { "epoch": 0.590047393364929, "step": 249, "train/alpha": 0.2, "train/clean_ce_loss": 0.11865234375, "train/clean_loss_scale": 11.085958480834961, "train/clean_loss_weight": 2.217191696166992, "train/combined_loss": 1.5784499645233154, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05900473933649289, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3153749704360962 }, { "epoch": 0.5924170616113744, "grad_norm": 14.83861255645752, "learning_rate": 2e-06, "loss": 3.6676, "step": 250 }, { "epoch": 0.5924170616113744, "step": 250, "train/alpha": 0.2, "train/clean_ce_loss": 0.12353515625, "train/clean_loss_scale": 9.756114959716797, "train/clean_loss_weight": 1.9512230157852173, "train/combined_loss": 1.44626784324646, "train/pos_loss/pos0": 0.46825137734413147, "train/pos_loss/pos1": 0.9177239537239075, "train/pos_loss/pos2": 1.6875672340393066, "train/pos_loss/pos3": 2.54693603515625, "train/pos_loss/pos4": 2.6654837131500244, "train/pos_loss/pos5": 2.98333740234375, "train/pos_loss/pos6": 3.8089406490325928, "train/pos_loss/weight0": 2.6348252296447754, "train/pos_loss/weight1": 1.8871724605560303, "train/pos_loss/weight2": 1.2081756591796875, "train/pos_loss/weight3": 0.6867493391036987, "train/pos_loss/weight4": 0.3558172583580017, "train/pos_loss/weight5": 0.15347185730934143, "train/pos_loss/weight6": 0.07378781586885452, "train/pos_weight_progress": 0.05924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2052232027053833 }, { "epoch": 0.5924170616113744, "step": 250, "train/alpha": 0.2, "train/clean_ce_loss": 0.13671875, "train/clean_loss_scale": 6.56574821472168, "train/clean_loss_weight": 1.3131496906280518, "train/combined_loss": 1.0771931409835815, "train/pos_loss/pos0": 0.43578875064849854, "train/pos_loss/pos1": 0.8810344934463501, "train/pos_loss/pos2": 1.4570765495300293, "train/pos_loss/pos3": 2.213657855987549, "train/pos_loss/pos4": 2.6000101566314697, "train/pos_loss/pos5": 3.1506593227386475, "train/pos_loss/pos6": 3.4766228199005127, "train/pos_loss/weight0": 3.22137451171875, "train/pos_loss/weight1": 2.0265536308288574, "train/pos_loss/weight2": 1.1378682851791382, "train/pos_loss/weight3": 0.4517437815666199, "train/pos_loss/weight4": 0.12724943459033966, "train/pos_loss/weight5": 0.028069939464330673, "train/pos_loss/weight6": 0.007140756119042635, "train/pos_weight_progress": 0.05924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8976609110832214 }, { "epoch": 0.5924170616113744, "step": 250, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 6.885848045349121, "train/clean_loss_weight": 1.3771696090698242, "train/combined_loss": 0.6737909913063049, "train/pos_loss/pos0": 0.20061959326267242, "train/pos_loss/pos1": 0.5605740547180176, "train/pos_loss/pos2": 0.9309505820274353, "train/pos_loss/pos3": 1.3638768196105957, "train/pos_loss/pos4": 1.9308860301971436, "train/pos_loss/pos5": 2.211416006088257, "train/pos_loss/pos6": 2.6303296089172363, "train/pos_loss/weight0": 3.1394147872924805, "train/pos_loss/weight1": 2.020780324935913, "train/pos_loss/weight2": 1.1225979328155518, "train/pos_loss/weight3": 0.49252790212631226, "train/pos_loss/weight4": 0.1659129410982132, "train/pos_loss/weight5": 0.04529912397265434, "train/pos_loss/weight6": 0.013467113487422466, "train/pos_weight_progress": 0.05924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5614925026893616 }, { "epoch": 0.5924170616113744, "step": 250, "train/alpha": 0.2, "train/clean_ce_loss": 0.169921875, "train/clean_loss_scale": 8.389053344726562, "train/clean_loss_weight": 1.6778106689453125, "train/combined_loss": 1.7105804681777954, "train/pos_loss/pos0": 0.6976631879806519, "train/pos_loss/pos1": 1.2446244955062866, "train/pos_loss/pos2": 1.7817691564559937, "train/pos_loss/pos3": 2.428314447402954, "train/pos_loss/pos4": 2.9287803173065186, "train/pos_loss/pos5": 3.3289594650268555, "train/pos_loss/pos6": 3.7824883460998535, "train/pos_loss/weight0": 2.558765411376953, "train/pos_loss/weight1": 1.883005142211914, "train/pos_loss/weight2": 1.2304826974868774, "train/pos_loss/weight3": 0.7236863374710083, "train/pos_loss/weight4": 0.3620311915874481, "train/pos_loss/weight5": 0.16300871968269348, "train/pos_loss/weight6": 0.07902069389820099, "train/pos_weight_progress": 0.05924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4254837036132812 }, { "epoch": 0.5947867298578199, "step": 251, "train/alpha": 0.2, "train/clean_ce_loss": 0.1435546875, "train/clean_loss_scale": 4.626436233520508, "train/clean_loss_weight": 0.9252872467041016, "train/combined_loss": 0.7969759106636047, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.05947867298578199, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6641466021537781 }, { "epoch": 0.5947867298578199, "step": 251, "train/alpha": 0.2, "train/clean_ce_loss": 0.059326171875, "train/clean_loss_scale": 6.596224784851074, "train/clean_loss_weight": 1.3192449808120728, "train/combined_loss": 0.4695945084095001, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.05947867298578199, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.39132875204086304 }, { "epoch": 0.5947867298578199, "step": 251, "train/alpha": 0.2, "train/clean_ce_loss": 0.037109375, "train/clean_loss_scale": 7.9359354972839355, "train/clean_loss_weight": 1.5871871709823608, "train/combined_loss": 0.3533971309661865, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.05947867298578199, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.29449760913848877 }, { "epoch": 0.5947867298578199, "step": 251, "train/alpha": 0.2, "train/clean_ce_loss": 0.10498046875, "train/clean_loss_scale": 8.520176887512207, "train/clean_loss_weight": 1.7040354013442993, "train/combined_loss": 1.0733425617218018, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.05947867298578199, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8944521546363831 }, { "epoch": 0.5971563981042654, "step": 252, "train/alpha": 0.2, "train/clean_ce_loss": 0.1259765625, "train/clean_loss_scale": 6.131660461425781, "train/clean_loss_weight": 1.2263320684432983, "train/combined_loss": 0.9269345998764038, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.05971563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7724454998970032 }, { "epoch": 0.5971563981042654, "step": 252, "train/alpha": 0.2, "train/clean_ce_loss": 0.14453125, "train/clean_loss_scale": 5.454991817474365, "train/clean_loss_weight": 1.0909984111785889, "train/combined_loss": 0.9461001753807068, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.05971563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7884168028831482 }, { "epoch": 0.5971563981042654, "step": 252, "train/alpha": 0.2, "train/clean_ce_loss": 0.027099609375, "train/clean_loss_scale": 10.319408416748047, "train/clean_loss_weight": 2.0638816356658936, "train/combined_loss": 0.33558231592178345, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.05971563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.279651939868927 }, { "epoch": 0.5971563981042654, "step": 252, "train/alpha": 0.2, "train/clean_ce_loss": 0.1259765625, "train/clean_loss_scale": 6.1826653480529785, "train/clean_loss_weight": 1.2365330457687378, "train/combined_loss": 0.9346451163291931, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.05971563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7788709402084351 }, { "epoch": 0.5995260663507109, "step": 253, "train/alpha": 0.2, "train/clean_ce_loss": 0.0859375, "train/clean_loss_scale": 5.296424865722656, "train/clean_loss_weight": 1.059285044670105, "train/combined_loss": 0.5461938381195068, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.05995260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45516151189804077 }, { "epoch": 0.5995260663507109, "step": 253, "train/alpha": 0.2, "train/clean_ce_loss": 0.17578125, "train/clean_loss_scale": 6.139350891113281, "train/clean_loss_weight": 1.227870225906372, "train/combined_loss": 1.295019268989563, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.05995260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.079182744026184 }, { "epoch": 0.5995260663507109, "step": 253, "train/alpha": 0.2, "train/clean_ce_loss": 0.1005859375, "train/clean_loss_scale": 5.785996913909912, "train/clean_loss_weight": 1.1571993827819824, "train/combined_loss": 0.6983879208564758, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.05995260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5819899439811707 }, { "epoch": 0.5995260663507109, "step": 253, "train/alpha": 0.2, "train/clean_ce_loss": 0.0263671875, "train/clean_loss_scale": 22.99778938293457, "train/clean_loss_weight": 4.599557876586914, "train/combined_loss": 0.7276644110679626, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.05995260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6063870191574097 }, { "epoch": 0.6018957345971564, "step": 254, "train/alpha": 0.2, "train/clean_ce_loss": 0.11376953125, "train/clean_loss_scale": 5.745262622833252, "train/clean_loss_weight": 1.1490525007247925, "train/combined_loss": 0.7843630313873291, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.06018957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6536358594894409 }, { "epoch": 0.6018957345971564, "step": 254, "train/alpha": 0.2, "train/clean_ce_loss": 0.08984375, "train/clean_loss_scale": 6.731245517730713, "train/clean_loss_weight": 1.3462491035461426, "train/combined_loss": 0.7257124185562134, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.06018957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6047603487968445 }, { "epoch": 0.6018957345971564, "step": 254, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 6.337459564208984, "train/clean_loss_weight": 1.2674919366836548, "train/combined_loss": 0.5050163269042969, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.06018957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4208469092845917 }, { "epoch": 0.6018957345971564, "step": 254, "train/alpha": 0.2, "train/clean_ce_loss": 0.14453125, "train/clean_loss_scale": 6.350634574890137, "train/clean_loss_weight": 1.2701269388198853, "train/combined_loss": 1.1014381647109985, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.06018957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9178651571273804 }, { "epoch": 0.6042654028436019, "grad_norm": 15.690946578979492, "learning_rate": 2e-06, "loss": 3.6967, "step": 255 }, { "epoch": 0.6042654028436019, "step": 255, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 6.101674556732178, "train/clean_loss_weight": 1.2203348875045776, "train/combined_loss": 0.5219792127609253, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.06042654028436019, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43498265743255615 }, { "epoch": 0.6042654028436019, "step": 255, "train/alpha": 0.2, "train/clean_ce_loss": 0.0263671875, "train/clean_loss_scale": 11.205808639526367, "train/clean_loss_weight": 2.241161823272705, "train/combined_loss": 0.3545587658882141, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.06042654028436019, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2954656481742859 }, { "epoch": 0.6042654028436019, "step": 255, "train/alpha": 0.2, "train/clean_ce_loss": 0.162109375, "train/clean_loss_scale": 6.519273281097412, "train/clean_loss_weight": 1.3038547039031982, "train/combined_loss": 1.268202304840088, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.06042654028436019, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0568352937698364 }, { "epoch": 0.6042654028436019, "step": 255, "train/alpha": 0.2, "train/clean_ce_loss": 0.16015625, "train/clean_loss_scale": 5.541961669921875, "train/clean_loss_weight": 1.108392357826233, "train/combined_loss": 1.0650957822799683, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.06042654028436019, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8875797986984253 }, { "epoch": 0.6066350710900474, "step": 256, "train/alpha": 0.2, "train/clean_ce_loss": 0.05712890625, "train/clean_loss_scale": 5.946281433105469, "train/clean_loss_weight": 1.1892563104629517, "train/combined_loss": 0.40764546394348145, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.06066350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3397045433521271 }, { "epoch": 0.6066350710900474, "step": 256, "train/alpha": 0.2, "train/clean_ce_loss": 0.0274658203125, "train/clean_loss_scale": 9.866019248962402, "train/clean_loss_weight": 1.9732038974761963, "train/combined_loss": 0.3251739740371704, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.06066350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.27097830176353455 }, { "epoch": 0.6066350710900474, "step": 256, "train/alpha": 0.2, "train/clean_ce_loss": 0.1005859375, "train/clean_loss_scale": 5.162665367126465, "train/clean_loss_weight": 1.032533049583435, "train/combined_loss": 0.6231498122215271, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.06066350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5192915201187134 }, { "epoch": 0.6066350710900474, "step": 256, "train/alpha": 0.2, "train/clean_ce_loss": 0.08984375, "train/clean_loss_scale": 5.560895919799805, "train/clean_loss_weight": 1.112179160118103, "train/combined_loss": 0.5995340943336487, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.06066350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4996117353439331 }, { "epoch": 0.6090047393364929, "step": 257, "train/alpha": 0.2, "train/clean_ce_loss": 0.09619140625, "train/clean_loss_scale": 7.078251838684082, "train/clean_loss_weight": 1.4156503677368164, "train/combined_loss": 0.8170404434204102, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.060900473933649286, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6808670163154602 }, { "epoch": 0.6090047393364929, "step": 257, "train/alpha": 0.2, "train/clean_ce_loss": 0.01129150390625, "train/clean_loss_scale": 33.986907958984375, "train/clean_loss_weight": 6.79738187789917, "train/combined_loss": 0.46051594614982605, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.060900473933649286, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.38376328349113464 }, { "epoch": 0.6090047393364929, "step": 257, "train/alpha": 0.2, "train/clean_ce_loss": 0.140625, "train/clean_loss_scale": 8.045077323913574, "train/clean_loss_weight": 1.6090154647827148, "train/combined_loss": 1.3576067686080933, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.060900473933649286, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1313389539718628 }, { "epoch": 0.6090047393364929, "step": 257, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 7.323399066925049, "train/clean_loss_weight": 1.4646798372268677, "train/combined_loss": 0.8153002858161926, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.060900473933649286, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6794168949127197 }, { "epoch": 0.6113744075829384, "step": 258, "train/alpha": 0.2, "train/clean_ce_loss": 0.06884765625, "train/clean_loss_scale": 7.7615180015563965, "train/clean_loss_weight": 1.552303671836853, "train/combined_loss": 0.6412347555160522, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.06113744075829384, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5343623161315918 }, { "epoch": 0.6113744075829384, "step": 258, "train/alpha": 0.2, "train/clean_ce_loss": 0.107421875, "train/clean_loss_scale": 5.851755142211914, "train/clean_loss_weight": 1.1703510284423828, "train/combined_loss": 0.7543277740478516, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.06113744075829384, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6286064982414246 }, { "epoch": 0.6113744075829384, "step": 258, "train/alpha": 0.2, "train/clean_ce_loss": 0.0546875, "train/clean_loss_scale": 5.508718967437744, "train/clean_loss_weight": 1.1017438173294067, "train/combined_loss": 0.36150968074798584, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.06113744075829384, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.30125805735588074 }, { "epoch": 0.6113744075829384, "step": 258, "train/alpha": 0.2, "train/clean_ce_loss": 0.1337890625, "train/clean_loss_scale": 6.02115535736084, "train/clean_loss_weight": 1.2042311429977417, "train/combined_loss": 0.9666776657104492, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.06113744075829384, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8055647015571594 }, { "epoch": 0.6137440758293838, "step": 259, "train/alpha": 0.2, "train/clean_ce_loss": 0.1171875, "train/clean_loss_scale": 7.472597122192383, "train/clean_loss_weight": 1.4945194721221924, "train/combined_loss": 1.0508339405059814, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.06137440758293839, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.875694990158081 }, { "epoch": 0.6137440758293838, "step": 259, "train/alpha": 0.2, "train/clean_ce_loss": 0.03369140625, "train/clean_loss_scale": 7.019266128540039, "train/clean_loss_weight": 1.4038532972335815, "train/combined_loss": 0.28378674387931824, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.06137440758293839, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2364889532327652 }, { "epoch": 0.6137440758293838, "step": 259, "train/alpha": 0.2, "train/clean_ce_loss": 0.07177734375, "train/clean_loss_scale": 6.726517677307129, "train/clean_loss_weight": 1.3453035354614258, "train/combined_loss": 0.5793738961219788, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.06137440758293839, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4828115701675415 }, { "epoch": 0.6137440758293838, "step": 259, "train/alpha": 0.2, "train/clean_ce_loss": 0.01214599609375, "train/clean_loss_scale": 26.809114456176758, "train/clean_loss_weight": 5.361823081970215, "train/combined_loss": 0.3907480835914612, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.06137440758293839, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.32562339305877686 }, { "epoch": 0.6161137440758294, "grad_norm": 12.078834533691406, "learning_rate": 2e-06, "loss": 3.559, "step": 260 }, { "epoch": 0.6161137440758294, "step": 260, "train/alpha": 0.2, "train/clean_ce_loss": 0.095703125, "train/clean_loss_scale": 7.339895248413086, "train/clean_loss_weight": 1.467979073524475, "train/combined_loss": 0.8429411053657532, "train/pos_loss/pos0": 0.3672548532485962, "train/pos_loss/pos1": 0.7604404091835022, "train/pos_loss/pos2": 1.1919242143630981, "train/pos_loss/pos3": 1.6201306581497192, "train/pos_loss/pos4": 2.0652050971984863, "train/pos_loss/pos5": 2.3980133533477783, "train/pos_loss/pos6": 2.897418737411499, "train/pos_loss/weight0": 3.420557975769043, "train/pos_loss/weight1": 2.06095814704895, "train/pos_loss/weight2": 1.005354642868042, "train/pos_loss/weight3": 0.3879770338535309, "train/pos_loss/weight4": 0.09876877814531326, "train/pos_loss/weight5": 0.0221124105155468, "train/pos_loss/weight6": 0.0042710755951702595, "train/pos_weight_progress": 0.061611374407582936, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7024509310722351 }, { "epoch": 0.6161137440758294, "step": 260, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 8.953927993774414, "train/clean_loss_weight": 1.7907856702804565, "train/combined_loss": 0.7135161757469177, "train/pos_loss/pos0": 0.2550848126411438, "train/pos_loss/pos1": 0.5632335543632507, "train/pos_loss/pos2": 0.7927475571632385, "train/pos_loss/pos3": 1.081391453742981, "train/pos_loss/pos4": 1.5433322191238403, "train/pos_loss/pos5": 2.032238721847534, "train/pos_loss/pos6": 2.4645493030548096, "train/pos_loss/weight0": 2.835613965988159, "train/pos_loss/weight1": 1.9319874048233032, "train/pos_loss/weight2": 1.190282940864563, "train/pos_loss/weight3": 0.6229854226112366, "train/pos_loss/weight4": 0.27423787117004395, "train/pos_loss/weight5": 0.10287882387638092, "train/pos_loss/weight6": 0.042014241218566895, "train/pos_weight_progress": 0.061611374407582936, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.594596803188324 }, { "epoch": 0.6161137440758294, "step": 260, "train/alpha": 0.2, "train/clean_ce_loss": 0.0233154296875, "train/clean_loss_scale": 21.435667037963867, "train/clean_loss_weight": 4.287133693695068, "train/combined_loss": 0.5997381210327148, "train/pos_loss/pos0": 0.06279955804347992, "train/pos_loss/pos1": 0.23850540816783905, "train/pos_loss/pos2": 0.5470170974731445, "train/pos_loss/pos3": 1.0842902660369873, "train/pos_loss/pos4": 1.8354926109313965, "train/pos_loss/pos5": 2.2299928665161133, "train/pos_loss/pos6": 2.711510419845581, "train/pos_loss/weight0": 2.4960012435913086, "train/pos_loss/weight1": 1.85382080078125, "train/pos_loss/weight2": 1.2274293899536133, "train/pos_loss/weight3": 0.7283468246459961, "train/pos_loss/weight4": 0.3980807363986969, "train/pos_loss/weight5": 0.1955038607120514, "train/pos_loss/weight6": 0.10081742703914642, "train/pos_weight_progress": 0.061611374407582936, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4997817873954773 }, { "epoch": 0.6161137440758294, "step": 260, "train/alpha": 0.2, "train/clean_ce_loss": 0.07470703125, "train/clean_loss_scale": 9.177098274230957, "train/clean_loss_weight": 1.8354196548461914, "train/combined_loss": 0.8227125406265259, "train/pos_loss/pos0": 0.17825742065906525, "train/pos_loss/pos1": 0.5030475854873657, "train/pos_loss/pos2": 0.8679136633872986, "train/pos_loss/pos3": 1.2818119525909424, "train/pos_loss/pos4": 1.751712679862976, "train/pos_loss/pos5": 2.005361557006836, "train/pos_loss/pos6": 2.398681640625, "train/pos_loss/weight0": 2.418961763381958, "train/pos_loss/weight1": 1.8336868286132812, "train/pos_loss/weight2": 1.2509479522705078, "train/pos_loss/weight3": 0.7899905443191528, "train/pos_loss/weight4": 0.42276567220687866, "train/pos_loss/weight5": 0.18613579869270325, "train/pos_loss/weight6": 0.09751135110855103, "train/pos_weight_progress": 0.061611374407582936, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6855937838554382 }, { "epoch": 0.6184834123222749, "step": 261, "train/alpha": 0.2, "train/clean_ce_loss": 0.1748046875, "train/clean_loss_scale": 9.581496238708496, "train/clean_loss_weight": 1.9162992238998413, "train/combined_loss": 2.0098683834075928, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06184834123222749, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.674890398979187 }, { "epoch": 0.6184834123222749, "step": 261, "train/alpha": 0.2, "train/clean_ce_loss": 0.041259765625, "train/clean_loss_scale": 17.476520538330078, "train/clean_loss_weight": 3.4953041076660156, "train/combined_loss": 0.8652925491333008, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06184834123222749, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7210771441459656 }, { "epoch": 0.6184834123222749, "step": 261, "train/alpha": 0.2, "train/clean_ce_loss": 0.048095703125, "train/clean_loss_scale": 12.528865814208984, "train/clean_loss_weight": 2.5057733058929443, "train/combined_loss": 0.7231014966964722, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06184834123222749, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6025846004486084 }, { "epoch": 0.6184834123222749, "step": 261, "train/alpha": 0.2, "train/clean_ce_loss": 0.1416015625, "train/clean_loss_scale": 8.144561767578125, "train/clean_loss_weight": 1.628912329673767, "train/combined_loss": 1.3839391469955444, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06184834123222749, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.153282642364502 }, { "epoch": 0.6208530805687204, "step": 262, "train/alpha": 0.2, "train/clean_ce_loss": 0.15234375, "train/clean_loss_scale": 8.508097648620605, "train/clean_loss_weight": 1.7016195058822632, "train/combined_loss": 1.5553865432739258, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06208530805687204, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2961554527282715 }, { "epoch": 0.6208530805687204, "step": 262, "train/alpha": 0.2, "train/clean_ce_loss": 0.11669921875, "train/clean_loss_scale": 8.111714363098145, "train/clean_loss_weight": 1.6223429441452026, "train/combined_loss": 1.135956883430481, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06208530805687204, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9466307759284973 }, { "epoch": 0.6208530805687204, "step": 262, "train/alpha": 0.2, "train/clean_ce_loss": 0.173828125, "train/clean_loss_scale": 8.062767028808594, "train/clean_loss_weight": 1.6125534772872925, "train/combined_loss": 1.6818429231643677, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06208530805687204, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4015357494354248 }, { "epoch": 0.6208530805687204, "step": 262, "train/alpha": 0.2, "train/clean_ce_loss": 0.083984375, "train/clean_loss_scale": 7.231998920440674, "train/clean_loss_weight": 1.4463998079299927, "train/combined_loss": 0.7288498878479004, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06208530805687204, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.607374906539917 }, { "epoch": 0.6232227488151659, "step": 263, "train/alpha": 0.2, "train/clean_ce_loss": 0.150390625, "train/clean_loss_scale": 7.849396705627441, "train/clean_loss_weight": 1.569879412651062, "train/combined_loss": 1.4165709018707275, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.062322274881516586, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1804757118225098 }, { "epoch": 0.6232227488151659, "step": 263, "train/alpha": 0.2, "train/clean_ce_loss": 0.1279296875, "train/clean_loss_scale": 12.55223560333252, "train/clean_loss_weight": 2.5104472637176514, "train/combined_loss": 1.926964282989502, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.062322274881516586, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.6058036088943481 }, { "epoch": 0.6232227488151659, "step": 263, "train/alpha": 0.2, "train/clean_ce_loss": 0.0201416015625, "train/clean_loss_scale": 24.48615074157715, "train/clean_loss_weight": 4.89723014831543, "train/combined_loss": 0.5918283462524414, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.062322274881516586, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.49319028854370117 }, { "epoch": 0.6232227488151659, "step": 263, "train/alpha": 0.2, "train/clean_ce_loss": 0.109375, "train/clean_loss_scale": 10.465413093566895, "train/clean_loss_weight": 2.0930826663970947, "train/combined_loss": 1.3735854625701904, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.062322274881516586, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1446545124053955 }, { "epoch": 0.6255924170616114, "step": 264, "train/alpha": 0.2, "train/clean_ce_loss": 0.10693359375, "train/clean_loss_scale": 8.876930236816406, "train/clean_loss_weight": 1.775386095046997, "train/combined_loss": 1.1390904188156128, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06255924170616113, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9492420554161072 }, { "epoch": 0.6255924170616114, "step": 264, "train/alpha": 0.2, "train/clean_ce_loss": 0.06103515625, "train/clean_loss_scale": 9.578239440917969, "train/clean_loss_weight": 1.9156478643417358, "train/combined_loss": 0.701531171798706, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06255924170616113, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.58460932970047 }, { "epoch": 0.6255924170616114, "step": 264, "train/alpha": 0.2, "train/clean_ce_loss": 0.087890625, "train/clean_loss_scale": 8.681466102600098, "train/clean_loss_weight": 1.7362931966781616, "train/combined_loss": 0.9156233072280884, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06255924170616113, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7630194425582886 }, { "epoch": 0.6255924170616114, "step": 264, "train/alpha": 0.2, "train/clean_ce_loss": 0.027587890625, "train/clean_loss_scale": 17.734586715698242, "train/clean_loss_weight": 3.54691743850708, "train/combined_loss": 0.5871118307113647, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06255924170616113, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.48925983905792236 }, { "epoch": 0.6279620853080569, "grad_norm": 17.56761932373047, "learning_rate": 2e-06, "loss": 3.8793, "step": 265 }, { "epoch": 0.6279620853080569, "step": 265, "train/alpha": 0.2, "train/clean_ce_loss": 0.054931640625, "train/clean_loss_scale": 11.616801261901855, "train/clean_loss_weight": 2.3233602046966553, "train/combined_loss": 0.7657559514045715, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06279620853080568, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6381299495697021 }, { "epoch": 0.6279620853080569, "step": 265, "train/alpha": 0.2, "train/clean_ce_loss": 0.08544921875, "train/clean_loss_scale": 7.978315353393555, "train/clean_loss_weight": 1.595663070678711, "train/combined_loss": 0.8180890083312988, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06279620853080568, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6817408204078674 }, { "epoch": 0.6279620853080569, "step": 265, "train/alpha": 0.2, "train/clean_ce_loss": 0.0751953125, "train/clean_loss_scale": 10.041952133178711, "train/clean_loss_weight": 2.008390426635742, "train/combined_loss": 0.9061292409896851, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06279620853080568, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7551077008247375 }, { "epoch": 0.6279620853080569, "step": 265, "train/alpha": 0.2, "train/clean_ce_loss": 0.15234375, "train/clean_loss_scale": 8.415207862854004, "train/clean_loss_weight": 1.6830415725708008, "train/combined_loss": 1.538405179977417, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06279620853080568, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2820043563842773 }, { "epoch": 0.6303317535545023, "step": 266, "train/alpha": 0.2, "train/clean_ce_loss": 0.0233154296875, "train/clean_loss_scale": 21.16068458557129, "train/clean_loss_weight": 4.232137203216553, "train/combined_loss": 0.5920445322990417, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06303317535545024, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4933704435825348 }, { "epoch": 0.6303317535545023, "step": 266, "train/alpha": 0.2, "train/clean_ce_loss": 0.07861328125, "train/clean_loss_scale": 8.579761505126953, "train/clean_loss_weight": 1.7159522771835327, "train/combined_loss": 0.8093798160552979, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06303317535545024, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6744831800460815 }, { "epoch": 0.6303317535545023, "step": 266, "train/alpha": 0.2, "train/clean_ce_loss": 0.1455078125, "train/clean_loss_scale": 8.976165771484375, "train/clean_loss_weight": 1.795233130455017, "train/combined_loss": 1.5673227310180664, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06303317535545024, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3061022758483887 }, { "epoch": 0.6303317535545023, "step": 266, "train/alpha": 0.2, "train/clean_ce_loss": 0.1708984375, "train/clean_loss_scale": 8.157575607299805, "train/clean_loss_weight": 1.6315151453018188, "train/combined_loss": 1.6729402542114258, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06303317535545024, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3941168785095215 }, { "epoch": 0.6327014218009479, "step": 267, "train/alpha": 0.2, "train/clean_ce_loss": 0.1015625, "train/clean_loss_scale": 7.986092567443848, "train/clean_loss_weight": 1.5972185134887695, "train/combined_loss": 0.9733050465583801, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06327014218009479, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8110875487327576 }, { "epoch": 0.6327014218009479, "step": 267, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 8.443881034851074, "train/clean_loss_weight": 1.6887762546539307, "train/combined_loss": 0.8856179714202881, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06327014218009479, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7380149960517883 }, { "epoch": 0.6327014218009479, "step": 267, "train/alpha": 0.2, "train/clean_ce_loss": 0.10205078125, "train/clean_loss_scale": 8.080153465270996, "train/clean_loss_weight": 1.6160306930541992, "train/combined_loss": 0.9895031452178955, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06327014218009479, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8245859742164612 }, { "epoch": 0.6327014218009479, "step": 267, "train/alpha": 0.2, "train/clean_ce_loss": 0.095703125, "train/clean_loss_scale": 8.691757202148438, "train/clean_loss_weight": 1.7383514642715454, "train/combined_loss": 0.9981939792633057, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06327014218009479, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8318282961845398 }, { "epoch": 0.6350710900473934, "step": 268, "train/alpha": 0.2, "train/clean_ce_loss": 0.059814453125, "train/clean_loss_scale": 8.212528228759766, "train/clean_loss_weight": 1.6425056457519531, "train/combined_loss": 0.5894734263420105, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06350710900473934, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4912278652191162 }, { "epoch": 0.6350710900473934, "step": 268, "train/alpha": 0.2, "train/clean_ce_loss": 0.1455078125, "train/clean_loss_scale": 9.637084007263184, "train/clean_loss_weight": 1.9274168014526367, "train/combined_loss": 1.682725191116333, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06350710900473934, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.402271032333374 }, { "epoch": 0.6350710900473934, "step": 268, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 11.30173397064209, "train/clean_loss_weight": 2.2603468894958496, "train/combined_loss": 1.0727818012237549, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06350710900473934, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8939847946166992 }, { "epoch": 0.6350710900473934, "step": 268, "train/alpha": 0.2, "train/clean_ce_loss": 0.00060272216796875, "train/clean_loss_scale": 744.2766723632812, "train/clean_loss_weight": 148.85533142089844, "train/combined_loss": 0.5383104681968689, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06350710900473934, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.44859203696250916 }, { "epoch": 0.6374407582938388, "step": 269, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 9.013225555419922, "train/clean_loss_weight": 1.8026450872421265, "train/combined_loss": 0.8661146759986877, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06374407582938389, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7217622399330139 }, { "epoch": 0.6374407582938388, "step": 269, "train/alpha": 0.2, "train/clean_ce_loss": 0.103515625, "train/clean_loss_scale": 8.471562385559082, "train/clean_loss_weight": 1.6943124532699585, "train/combined_loss": 1.0523269176483154, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06374407582938389, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.876939058303833 }, { "epoch": 0.6374407582938388, "step": 269, "train/alpha": 0.2, "train/clean_ce_loss": 0.09912109375, "train/clean_loss_scale": 8.878281593322754, "train/clean_loss_weight": 1.7756563425064087, "train/combined_loss": 1.056030035018921, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06374407582938389, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8800249695777893 }, { "epoch": 0.6374407582938388, "step": 269, "train/alpha": 0.2, "train/clean_ce_loss": 0.11767578125, "train/clean_loss_scale": 7.2917256355285645, "train/clean_loss_weight": 1.4583451747894287, "train/combined_loss": 1.0296714305877686, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06374407582938389, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8580595254898071 }, { "epoch": 0.6398104265402843, "grad_norm": 14.49075698852539, "learning_rate": 2e-06, "loss": 3.781, "step": 270 }, { "epoch": 0.6398104265402843, "step": 270, "train/alpha": 0.2, "train/clean_ce_loss": 0.107421875, "train/clean_loss_scale": 8.548382759094238, "train/clean_loss_weight": 1.7096766233444214, "train/combined_loss": 1.1019400358200073, "train/pos_loss/pos0": 0.35130825638771057, "train/pos_loss/pos1": 0.669779896736145, "train/pos_loss/pos2": 1.1144983768463135, "train/pos_loss/pos3": 1.6496423482894897, "train/pos_loss/pos4": 2.236340284347534, "train/pos_loss/pos5": 2.5704214572906494, "train/pos_loss/pos6": 3.0321261882781982, "train/pos_loss/weight0": 2.4708213806152344, "train/pos_loss/weight1": 1.8428995609283447, "train/pos_loss/weight2": 1.2391859292984009, "train/pos_loss/weight3": 0.7442918419837952, "train/pos_loss/weight4": 0.40492644906044006, "train/pos_loss/weight5": 0.19413885474205017, "train/pos_loss/weight6": 0.10373599082231522, "train/pos_weight_progress": 0.06398104265402843, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9182833433151245 }, { "epoch": 0.6398104265402843, "step": 270, "train/alpha": 0.2, "train/clean_ce_loss": 0.087890625, "train/clean_loss_scale": 8.911599159240723, "train/clean_loss_weight": 1.7823199033737183, "train/combined_loss": 0.9398952126502991, "train/pos_loss/pos0": 0.324981153011322, "train/pos_loss/pos1": 0.7121604681015015, "train/pos_loss/pos2": 1.132885456085205, "train/pos_loss/pos3": 1.5329314470291138, "train/pos_loss/pos4": 1.9710273742675781, "train/pos_loss/pos5": 2.4433836936950684, "train/pos_loss/pos6": 2.905332565307617, "train/pos_loss/weight0": 2.821455240249634, "train/pos_loss/weight1": 1.9658725261688232, "train/pos_loss/weight2": 1.1912693977355957, "train/pos_loss/weight3": 0.6223946809768677, "train/pos_loss/weight4": 0.270454078912735, "train/pos_loss/weight5": 0.09314597398042679, "train/pos_loss/weight6": 0.03540823981165886, "train/pos_weight_progress": 0.06398104265402843, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7832459807395935 }, { "epoch": 0.6398104265402843, "step": 270, "train/alpha": 0.2, "train/clean_ce_loss": 0.16796875, "train/clean_loss_scale": 8.684305191040039, "train/clean_loss_weight": 1.7368611097335815, "train/combined_loss": 1.7504302263259888, "train/pos_loss/pos0": 0.5289387106895447, "train/pos_loss/pos1": 1.5734809637069702, "train/pos_loss/pos2": 2.024869203567505, "train/pos_loss/pos3": 2.6642966270446777, "train/pos_loss/pos4": 3.222010374069214, "train/pos_loss/pos5": 3.6716907024383545, "train/pos_loss/pos6": 3.813897132873535, "train/pos_loss/weight0": 2.76588773727417, "train/pos_loss/weight1": 1.9321584701538086, "train/pos_loss/weight2": 1.193804144859314, "train/pos_loss/weight3": 0.6450486183166504, "train/pos_loss/weight4": 0.29588767886161804, "train/pos_loss/weight5": 0.11724685877561569, "train/pos_loss/weight6": 0.04996651038527489, "train/pos_weight_progress": 0.06398104265402843, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4586918354034424 }, { "epoch": 0.6398104265402843, "step": 270, "train/alpha": 0.2, "train/clean_ce_loss": 0.083984375, "train/clean_loss_scale": 7.984341621398926, "train/clean_loss_weight": 1.5968683958053589, "train/combined_loss": 0.8046719431877136, "train/pos_loss/pos0": 0.37933671474456787, "train/pos_loss/pos1": 0.725994884967804, "train/pos_loss/pos2": 1.088681697845459, "train/pos_loss/pos3": 1.3615124225616455, "train/pos_loss/pos4": 1.9082322120666504, "train/pos_loss/pos5": 2.288250684738159, "train/pos_loss/pos6": 2.696824073791504, "train/pos_loss/weight0": 3.36197829246521, "train/pos_loss/weight1": 2.113954782485962, "train/pos_loss/weight2": 1.011075735092163, "train/pos_loss/weight3": 0.37897711992263794, "train/pos_loss/weight4": 0.10843221098184586, "train/pos_loss/weight5": 0.02135787345468998, "train/pos_loss/weight6": 0.0042238784953951836, "train/pos_weight_progress": 0.06398104265402843, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6705599427223206 }, { "epoch": 0.6421800947867299, "step": 271, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 6.096076011657715, "train/clean_loss_weight": 1.2192152738571167, "train/combined_loss": 0.5215002298355103, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06421800947867298, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43458354473114014 }, { "epoch": 0.6421800947867299, "step": 271, "train/alpha": 0.2, "train/clean_ce_loss": 0.1640625, "train/clean_loss_scale": 7.419656753540039, "train/clean_loss_weight": 1.4839314222335815, "train/combined_loss": 1.460744857788086, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06421800947867298, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2172874212265015 }, { "epoch": 0.6421800947867299, "step": 271, "train/alpha": 0.2, "train/clean_ce_loss": 0.058837890625, "train/clean_loss_scale": 7.759791374206543, "train/clean_loss_weight": 1.5519583225250244, "train/combined_loss": 0.5478836894035339, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06421800947867298, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45656976103782654 }, { "epoch": 0.6421800947867299, "step": 271, "train/alpha": 0.2, "train/clean_ce_loss": 0.07666015625, "train/clean_loss_scale": 7.217824935913086, "train/clean_loss_weight": 1.443565011024475, "train/combined_loss": 0.6639834642410278, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06421800947867298, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5533195734024048 }, { "epoch": 0.6445497630331753, "step": 272, "train/alpha": 0.2, "train/clean_ce_loss": 0.06396484375, "train/clean_loss_scale": 9.144003868103027, "train/clean_loss_weight": 1.8288007974624634, "train/combined_loss": 0.7018737196922302, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06445497630331753, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5848947763442993 }, { "epoch": 0.6445497630331753, "step": 272, "train/alpha": 0.2, "train/clean_ce_loss": 0.10986328125, "train/clean_loss_scale": 7.9242730140686035, "train/clean_loss_weight": 1.5848546028137207, "train/combined_loss": 1.0447039604187012, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06445497630331753, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.870586633682251 }, { "epoch": 0.6445497630331753, "step": 272, "train/alpha": 0.2, "train/clean_ce_loss": 0.109375, "train/clean_loss_scale": 7.810153007507324, "train/clean_loss_weight": 1.5620306730270386, "train/combined_loss": 1.0250825881958008, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06445497630331753, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8542354702949524 }, { "epoch": 0.6445497630331753, "step": 272, "train/alpha": 0.2, "train/clean_ce_loss": 0.11474609375, "train/clean_loss_scale": 10.639164924621582, "train/clean_loss_weight": 2.127833127975464, "train/combined_loss": 1.4649631977081299, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06445497630331753, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.220802664756775 }, { "epoch": 0.6469194312796208, "step": 273, "train/alpha": 0.2, "train/clean_ce_loss": 0.07763671875, "train/clean_loss_scale": 7.786304950714111, "train/clean_loss_weight": 1.5572609901428223, "train/combined_loss": 0.7254037857055664, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06469194312796209, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6045031547546387 }, { "epoch": 0.6469194312796208, "step": 273, "train/alpha": 0.2, "train/clean_ce_loss": 0.03564453125, "train/clean_loss_scale": 11.120068550109863, "train/clean_loss_weight": 2.2240138053894043, "train/combined_loss": 0.4756435751914978, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06469194312796209, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3963696360588074 }, { "epoch": 0.6469194312796208, "step": 273, "train/alpha": 0.2, "train/clean_ce_loss": 0.004547119140625, "train/clean_loss_scale": 99.47914123535156, "train/clean_loss_weight": 19.895828247070312, "train/combined_loss": 0.5428122282028198, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06469194312796209, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45234352350234985 }, { "epoch": 0.6469194312796208, "step": 273, "train/alpha": 0.2, "train/clean_ce_loss": 0.059326171875, "train/clean_loss_scale": 7.826287269592285, "train/clean_loss_weight": 1.5652574300765991, "train/combined_loss": 0.5571644306182861, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06469194312796209, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4643036723136902 }, { "epoch": 0.6492890995260664, "step": 274, "train/alpha": 0.2, "train/clean_ce_loss": 0.07373046875, "train/clean_loss_scale": 7.868056774139404, "train/clean_loss_weight": 1.5736113786697388, "train/combined_loss": 0.6961386203765869, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06492890995260664, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5801154971122742 }, { "epoch": 0.6492890995260664, "step": 274, "train/alpha": 0.2, "train/clean_ce_loss": 0.1767578125, "train/clean_loss_scale": 7.705462455749512, "train/clean_loss_weight": 1.5410925149917603, "train/combined_loss": 1.6344008445739746, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06492890995260664, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3620007038116455 }, { "epoch": 0.6492890995260664, "step": 274, "train/alpha": 0.2, "train/clean_ce_loss": 0.1044921875, "train/clean_loss_scale": 11.196244239807129, "train/clean_loss_weight": 2.2392489910125732, "train/combined_loss": 1.4039040803909302, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06492890995260664, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1699200868606567 }, { "epoch": 0.6492890995260664, "step": 274, "train/alpha": 0.2, "train/clean_ce_loss": 0.10986328125, "train/clean_loss_scale": 7.495048999786377, "train/clean_loss_weight": 1.4990098476409912, "train/combined_loss": 0.9881168007850647, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06492890995260664, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8234306573867798 }, { "epoch": 0.6516587677725119, "grad_norm": 16.31027603149414, "learning_rate": 2e-06, "loss": 3.8281, "step": 275 }, { "epoch": 0.6516587677725119, "step": 275, "train/alpha": 0.2, "train/clean_ce_loss": 0.041259765625, "train/clean_loss_scale": 16.411584854125977, "train/clean_loss_weight": 3.2823169231414795, "train/combined_loss": 0.812565803527832, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06516587677725119, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6771381497383118 }, { "epoch": 0.6516587677725119, "step": 275, "train/alpha": 0.2, "train/clean_ce_loss": 0.0703125, "train/clean_loss_scale": 7.655806541442871, "train/clean_loss_weight": 1.5311613082885742, "train/combined_loss": 0.645958662033081, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06516587677725119, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5382989048957825 }, { "epoch": 0.6516587677725119, "step": 275, "train/alpha": 0.2, "train/clean_ce_loss": 0.11572265625, "train/clean_loss_scale": 6.765317440032959, "train/clean_loss_weight": 1.3530634641647339, "train/combined_loss": 0.9394806027412415, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06516587677725119, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7829005122184753 }, { "epoch": 0.6516587677725119, "step": 275, "train/alpha": 0.2, "train/clean_ce_loss": 0.1484375, "train/clean_loss_scale": 9.594456672668457, "train/clean_loss_weight": 1.9188913106918335, "train/combined_loss": 1.7090126276016235, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06516587677725119, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4241771697998047 }, { "epoch": 0.6540284360189573, "step": 276, "train/alpha": 0.2, "train/clean_ce_loss": 0.134765625, "train/clean_loss_scale": 9.396793365478516, "train/clean_loss_weight": 1.8793586492538452, "train/combined_loss": 1.5196375846862793, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06540284360189573, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2663646936416626 }, { "epoch": 0.6540284360189573, "step": 276, "train/alpha": 0.2, "train/clean_ce_loss": 0.11328125, "train/clean_loss_scale": 7.836991786956787, "train/clean_loss_weight": 1.5673984289169312, "train/combined_loss": 1.0653411149978638, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06540284360189573, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8877842426300049 }, { "epoch": 0.6540284360189573, "step": 276, "train/alpha": 0.2, "train/clean_ce_loss": 0.10400390625, "train/clean_loss_scale": 7.392886638641357, "train/clean_loss_weight": 1.4785773754119873, "train/combined_loss": 0.9226669073104858, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06540284360189573, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7688890695571899 }, { "epoch": 0.6540284360189573, "step": 276, "train/alpha": 0.2, "train/clean_ce_loss": 0.049560546875, "train/clean_loss_scale": 11.53174114227295, "train/clean_loss_weight": 2.3063483238220215, "train/combined_loss": 0.6858232617378235, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06540284360189573, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5715193748474121 }, { "epoch": 0.6563981042654028, "step": 277, "train/alpha": 0.2, "train/clean_ce_loss": 0.12451171875, "train/clean_loss_scale": 6.4633331298828125, "train/clean_loss_weight": 1.2926666736602783, "train/combined_loss": 0.96571284532547, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06563981042654028, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8047606945037842 }, { "epoch": 0.6563981042654028, "step": 277, "train/alpha": 0.2, "train/clean_ce_loss": 0.0286865234375, "train/clean_loss_scale": 14.123504638671875, "train/clean_loss_weight": 2.8247010707855225, "train/combined_loss": 0.48618510365486145, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06563981042654028, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4051542580127716 }, { "epoch": 0.6563981042654028, "step": 277, "train/alpha": 0.2, "train/clean_ce_loss": 0.1591796875, "train/clean_loss_scale": 7.888865947723389, "train/clean_loss_weight": 1.5777732133865356, "train/combined_loss": 1.5068966150283813, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06563981042654028, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2557471990585327 }, { "epoch": 0.6563981042654028, "step": 277, "train/alpha": 0.2, "train/clean_ce_loss": 0.1708984375, "train/clean_loss_scale": 7.731688022613525, "train/clean_loss_weight": 1.546337604522705, "train/combined_loss": 1.5856001377105713, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06563981042654028, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.321333408355713 }, { "epoch": 0.6587677725118484, "step": 278, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 8.726643562316895, "train/clean_loss_weight": 1.7453287839889526, "train/combined_loss": 0.690291166305542, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06587677725118483, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5752426385879517 }, { "epoch": 0.6587677725118484, "step": 278, "train/alpha": 0.2, "train/clean_ce_loss": 0.1416015625, "train/clean_loss_scale": 10.918972969055176, "train/clean_loss_weight": 2.1837947368621826, "train/combined_loss": 1.855372428894043, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06587677725118483, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.546143651008606 }, { "epoch": 0.6587677725118484, "step": 278, "train/alpha": 0.2, "train/clean_ce_loss": 0.1708984375, "train/clean_loss_scale": 7.45693302154541, "train/clean_loss_weight": 1.4913866519927979, "train/combined_loss": 1.5292538404464722, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06587677725118483, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2743781805038452 }, { "epoch": 0.6587677725118484, "step": 278, "train/alpha": 0.2, "train/clean_ce_loss": 0.1201171875, "train/clean_loss_scale": 10.37606430053711, "train/clean_loss_weight": 2.0752129554748535, "train/combined_loss": 1.495612382888794, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06587677725118483, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2463436126708984 }, { "epoch": 0.6611374407582938, "step": 279, "train/alpha": 0.2, "train/clean_ce_loss": 0.107421875, "train/clean_loss_scale": 8.601070404052734, "train/clean_loss_weight": 1.7202141284942627, "train/combined_loss": 1.108731746673584, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06611374407582939, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9239431023597717 }, { "epoch": 0.6611374407582938, "step": 279, "train/alpha": 0.2, "train/clean_ce_loss": 0.0006103515625, "train/clean_loss_scale": 536.3587646484375, "train/clean_loss_weight": 107.2717514038086, "train/combined_loss": 0.3928409218788147, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06611374407582939, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3273674249649048 }, { "epoch": 0.6611374407582938, "step": 279, "train/alpha": 0.2, "train/clean_ce_loss": 0.10107421875, "train/clean_loss_scale": 7.621261119842529, "train/clean_loss_weight": 1.5242522954940796, "train/combined_loss": 0.9243756532669067, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06611374407582939, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.770313024520874 }, { "epoch": 0.6611374407582938, "step": 279, "train/alpha": 0.2, "train/clean_ce_loss": 0.049560546875, "train/clean_loss_scale": 10.175436019897461, "train/clean_loss_weight": 2.0350873470306396, "train/combined_loss": 0.6051602363586426, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06611374407582939, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5043001770973206 }, { "epoch": 0.6635071090047393, "grad_norm": 18.16552734375, "learning_rate": 2e-06, "loss": 3.7486, "step": 280 }, { "epoch": 0.6635071090047393, "step": 280, "train/alpha": 0.2, "train/clean_ce_loss": 0.166015625, "train/clean_loss_scale": 6.996201515197754, "train/clean_loss_weight": 1.3992403745651245, "train/combined_loss": 1.3937745094299316, "train/pos_loss/pos0": 0.5574227571487427, "train/pos_loss/pos1": 1.0132797956466675, "train/pos_loss/pos2": 1.5152772665023804, "train/pos_loss/pos3": 2.1569275856018066, "train/pos_loss/pos4": 2.521909713745117, "train/pos_loss/pos5": 3.0920701026916504, "train/pos_loss/pos6": 3.422588586807251, "train/pos_loss/weight0": 2.696026563644409, "train/pos_loss/weight1": 1.9230024814605713, "train/pos_loss/weight2": 1.20458984375, "train/pos_loss/weight3": 0.6551782488822937, "train/pos_loss/weight4": 0.32757100462913513, "train/pos_loss/weight5": 0.1331043243408203, "train/pos_loss/weight6": 0.060527484863996506, "train/pos_weight_progress": 0.06635071090047394, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1614787578582764 }, { "epoch": 0.6635071090047393, "step": 280, "train/alpha": 0.2, "train/clean_ce_loss": 0.08251953125, "train/clean_loss_scale": 7.376357078552246, "train/clean_loss_weight": 1.475271463394165, "train/combined_loss": 0.7304322719573975, "train/pos_loss/pos0": 0.2757692337036133, "train/pos_loss/pos1": 0.5668022036552429, "train/pos_loss/pos2": 1.0256773233413696, "train/pos_loss/pos3": 1.457939863204956, "train/pos_loss/pos4": 1.8233084678649902, "train/pos_loss/pos5": 2.207136631011963, "train/pos_loss/pos6": 2.5424532890319824, "train/pos_loss/weight0": 3.2085483074188232, "train/pos_loss/weight1": 2.0181708335876465, "train/pos_loss/weight2": 1.0659993886947632, "train/pos_loss/weight3": 0.47781071066856384, "train/pos_loss/weight4": 0.16863280534744263, "train/pos_loss/weight5": 0.04710513725876808, "train/pos_loss/weight6": 0.013732805848121643, "train/pos_weight_progress": 0.06635071090047394, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.608693540096283 }, { "epoch": 0.6635071090047393, "step": 280, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 9.529024124145508, "train/clean_loss_weight": 1.9058048725128174, "train/combined_loss": 0.9994308352470398, "train/pos_loss/pos0": 0.2338472306728363, "train/pos_loss/pos1": 0.6773439049720764, "train/pos_loss/pos2": 1.136377215385437, "train/pos_loss/pos3": 1.518655776977539, "train/pos_loss/pos4": 2.2756757736206055, "train/pos_loss/pos5": 2.5357062816619873, "train/pos_loss/pos6": 2.8542118072509766, "train/pos_loss/weight0": 2.6018967628479004, "train/pos_loss/weight1": 1.8432800769805908, "train/pos_loss/weight2": 1.2274863719940186, "train/pos_loss/weight3": 0.7057533860206604, "train/pos_loss/weight4": 0.36689817905426025, "train/pos_loss/weight5": 0.1712445616722107, "train/pos_loss/weight6": 0.08344049751758575, "train/pos_weight_progress": 0.06635071090047394, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8328590393066406 }, { "epoch": 0.6635071090047393, "step": 280, "train/alpha": 0.2, "train/clean_ce_loss": 0.1015625, "train/clean_loss_scale": 7.251054763793945, "train/clean_loss_weight": 1.4502109289169312, "train/combined_loss": 0.8837223052978516, "train/pos_loss/pos0": 0.31921225786209106, "train/pos_loss/pos1": 0.6194599866867065, "train/pos_loss/pos2": 1.0743993520736694, "train/pos_loss/pos3": 1.4538174867630005, "train/pos_loss/pos4": 1.8870975971221924, "train/pos_loss/pos5": 2.316525459289551, "train/pos_loss/pos6": 2.5434956550598145, "train/pos_loss/weight0": 2.7711222171783447, "train/pos_loss/weight1": 1.9985493421554565, "train/pos_loss/weight2": 1.1899412870407104, "train/pos_loss/weight3": 0.6317296028137207, "train/pos_loss/weight4": 0.2738140821456909, "train/pos_loss/weight5": 0.09639880806207657, "train/pos_loss/weight6": 0.038444533944129944, "train/pos_weight_progress": 0.06635071090047394, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7364352345466614 }, { "epoch": 0.6658767772511849, "step": 281, "train/alpha": 0.2, "train/clean_ce_loss": 0.0966796875, "train/clean_loss_scale": 8.766283988952637, "train/clean_loss_weight": 1.7532567977905273, "train/combined_loss": 1.0170259475708008, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06658767772511848, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8475216031074524 }, { "epoch": 0.6658767772511849, "step": 281, "train/alpha": 0.2, "train/clean_ce_loss": 0.0279541015625, "train/clean_loss_scale": 13.410923957824707, "train/clean_loss_weight": 2.682184934616089, "train/combined_loss": 0.449868381023407, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06658767772511848, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3748903274536133 }, { "epoch": 0.6658767772511849, "step": 281, "train/alpha": 0.2, "train/clean_ce_loss": 0.13671875, "train/clean_loss_scale": 9.305452346801758, "train/clean_loss_weight": 1.8610905408859253, "train/combined_loss": 1.5266757011413574, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06658767772511848, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.272229790687561 }, { "epoch": 0.6658767772511849, "step": 281, "train/alpha": 0.2, "train/clean_ce_loss": 0.171875, "train/clean_loss_scale": 8.580543518066406, "train/clean_loss_weight": 1.7161086797714233, "train/combined_loss": 1.7697371244430542, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06658767772511848, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4747809171676636 }, { "epoch": 0.6682464454976303, "step": 282, "train/alpha": 0.2, "train/clean_ce_loss": 0.049560546875, "train/clean_loss_scale": 10.975329399108887, "train/clean_loss_weight": 2.195065975189209, "train/combined_loss": 0.6527320146560669, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06682464454976303, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5439433455467224 }, { "epoch": 0.6682464454976303, "step": 282, "train/alpha": 0.2, "train/clean_ce_loss": 0.08984375, "train/clean_loss_scale": 7.246969699859619, "train/clean_loss_weight": 1.4493939876556396, "train/combined_loss": 0.7813138961791992, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06682464454976303, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.651094913482666 }, { "epoch": 0.6682464454976303, "step": 282, "train/alpha": 0.2, "train/clean_ce_loss": 0.053466796875, "train/clean_loss_scale": 9.27786636352539, "train/clean_loss_weight": 1.855573296546936, "train/combined_loss": 0.5952693223953247, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06682464454976303, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.49605777859687805 }, { "epoch": 0.6682464454976303, "step": 282, "train/alpha": 0.2, "train/clean_ce_loss": 0.046630859375, "train/clean_loss_scale": 7.905673503875732, "train/clean_loss_weight": 1.5811346769332886, "train/combined_loss": 0.4423780143260956, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06682464454976303, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.36864835023880005 }, { "epoch": 0.6706161137440758, "step": 283, "train/alpha": 0.2, "train/clean_ce_loss": 0.1357421875, "train/clean_loss_scale": 10.695630073547363, "train/clean_loss_weight": 2.1391260623931885, "train/combined_loss": 1.7422178983688354, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06706161137440758, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4518482685089111 }, { "epoch": 0.6706161137440758, "step": 283, "train/alpha": 0.2, "train/clean_ce_loss": 0.0262451171875, "train/clean_loss_scale": 19.12731170654297, "train/clean_loss_weight": 3.8254623413085938, "train/combined_loss": 0.6023982763290405, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06706161137440758, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5019985437393188 }, { "epoch": 0.6706161137440758, "step": 283, "train/alpha": 0.2, "train/clean_ce_loss": 0.06884765625, "train/clean_loss_scale": 8.2009916305542, "train/clean_loss_weight": 1.6401983499526978, "train/combined_loss": 0.6775428652763367, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06706161137440758, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5646190643310547 }, { "epoch": 0.6706161137440758, "step": 283, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 7.642550468444824, "train/clean_loss_weight": 1.5285100936889648, "train/combined_loss": 0.8508307933807373, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06706161137440758, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7090256810188293 }, { "epoch": 0.6729857819905213, "step": 284, "train/alpha": 0.2, "train/clean_ce_loss": 0.0859375, "train/clean_loss_scale": 8.58749771118164, "train/clean_loss_weight": 1.7174996137619019, "train/combined_loss": 0.8855856657028198, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06729857819905213, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7379880547523499 }, { "epoch": 0.6729857819905213, "step": 284, "train/alpha": 0.2, "train/clean_ce_loss": 0.1083984375, "train/clean_loss_scale": 8.984671592712402, "train/clean_loss_weight": 1.7969343662261963, "train/combined_loss": 1.1687092781066895, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06729857819905213, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9739243388175964 }, { "epoch": 0.6729857819905213, "step": 284, "train/alpha": 0.2, "train/clean_ce_loss": 0.150390625, "train/clean_loss_scale": 6.357377529144287, "train/clean_loss_weight": 1.2714755535125732, "train/combined_loss": 1.1473079919815063, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06729857819905213, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.956089973449707 }, { "epoch": 0.6729857819905213, "step": 284, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 8.663155555725098, "train/clean_loss_weight": 1.7326310873031616, "train/combined_loss": 0.741105854511261, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06729857819905213, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.617588222026825 }, { "epoch": 0.6753554502369669, "grad_norm": 15.349434852600098, "learning_rate": 2e-06, "loss": 4.0565, "step": 285 }, { "epoch": 0.6753554502369669, "step": 285, "train/alpha": 0.2, "train/clean_ce_loss": 0.171875, "train/clean_loss_scale": 6.824174880981445, "train/clean_loss_weight": 1.3648350238800049, "train/combined_loss": 1.407486081123352, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06753554502369669, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1729050874710083 }, { "epoch": 0.6753554502369669, "step": 285, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 7.026210308074951, "train/clean_loss_weight": 1.4052420854568481, "train/combined_loss": 0.6669410467147827, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06753554502369669, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5557842254638672 }, { "epoch": 0.6753554502369669, "step": 285, "train/alpha": 0.2, "train/clean_ce_loss": 0.1689453125, "train/clean_loss_scale": 7.349734783172607, "train/clean_loss_weight": 1.4699469804763794, "train/combined_loss": 1.4900438785552979, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06753554502369669, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2417032718658447 }, { "epoch": 0.6753554502369669, "step": 285, "train/alpha": 0.2, "train/clean_ce_loss": 0.0220947265625, "train/clean_loss_scale": 27.11199951171875, "train/clean_loss_weight": 5.422399997711182, "train/combined_loss": 0.7188386917114258, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06753554502369669, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5990322232246399 }, { "epoch": 0.6777251184834123, "step": 286, "train/alpha": 0.2, "train/clean_ce_loss": 0.06103515625, "train/clean_loss_scale": 9.141253471374512, "train/clean_loss_weight": 1.828250765800476, "train/combined_loss": 0.6695254445075989, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06777251184834124, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5579378604888916 }, { "epoch": 0.6777251184834123, "step": 286, "train/alpha": 0.2, "train/clean_ce_loss": 0.03369140625, "train/clean_loss_scale": 11.906213760375977, "train/clean_loss_weight": 2.3812427520751953, "train/combined_loss": 0.48136448860168457, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06777251184834124, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4011370837688446 }, { "epoch": 0.6777251184834123, "step": 286, "train/alpha": 0.2, "train/clean_ce_loss": 0.154296875, "train/clean_loss_scale": 7.974482536315918, "train/clean_loss_weight": 1.5948965549468994, "train/combined_loss": 1.4765253067016602, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06777251184834124, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2304377555847168 }, { "epoch": 0.6777251184834123, "step": 286, "train/alpha": 0.2, "train/clean_ce_loss": 0.060302734375, "train/clean_loss_scale": 7.368833541870117, "train/clean_loss_weight": 1.4737666845321655, "train/combined_loss": 0.5332329869270325, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06777251184834124, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4443608224391937 }, { "epoch": 0.6800947867298578, "step": 287, "train/alpha": 0.2, "train/clean_ce_loss": 0.1328125, "train/clean_loss_scale": 9.44956111907959, "train/clean_loss_weight": 1.8899122476577759, "train/combined_loss": 1.5060237646102905, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06800947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2550197839736938 }, { "epoch": 0.6800947867298578, "step": 287, "train/alpha": 0.2, "train/clean_ce_loss": 0.0294189453125, "train/clean_loss_scale": 14.437378883361816, "train/clean_loss_weight": 2.8874757289886475, "train/combined_loss": 0.5096789598464966, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06800947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4247324466705322 }, { "epoch": 0.6800947867298578, "step": 287, "train/alpha": 0.2, "train/clean_ce_loss": 0.0751953125, "train/clean_loss_scale": 8.615553855895996, "train/clean_loss_weight": 1.7231107950210571, "train/combined_loss": 0.7774190902709961, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06800947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6478492617607117 }, { "epoch": 0.6800947867298578, "step": 287, "train/alpha": 0.2, "train/clean_ce_loss": 0.0186767578125, "train/clean_loss_scale": 25.455509185791016, "train/clean_loss_weight": 5.091102123260498, "train/combined_loss": 0.5705116391181946, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06800947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4754263758659363 }, { "epoch": 0.6824644549763034, "step": 288, "train/alpha": 0.2, "train/clean_ce_loss": 0.054443359375, "train/clean_loss_scale": 8.842288970947266, "train/clean_loss_weight": 1.7684577703475952, "train/combined_loss": 0.5776847004890442, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06824644549763033, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4814039170742035 }, { "epoch": 0.6824644549763034, "step": 288, "train/alpha": 0.2, "train/clean_ce_loss": 0.0595703125, "train/clean_loss_scale": 7.669949054718018, "train/clean_loss_weight": 1.5339897871017456, "train/combined_loss": 0.5482814908027649, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06824644549763033, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4569012522697449 }, { "epoch": 0.6824644549763034, "step": 288, "train/alpha": 0.2, "train/clean_ce_loss": 0.072265625, "train/clean_loss_scale": 8.256503105163574, "train/clean_loss_weight": 1.6513006687164307, "train/combined_loss": 0.7159935832023621, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06824644549763033, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5966613292694092 }, { "epoch": 0.6824644549763034, "step": 288, "train/alpha": 0.2, "train/clean_ce_loss": 0.0322265625, "train/clean_loss_scale": 12.821887969970703, "train/clean_loss_weight": 2.564377546310425, "train/combined_loss": 0.4958464503288269, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06824644549763033, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4132053852081299 }, { "epoch": 0.6848341232227488, "step": 289, "train/alpha": 0.2, "train/clean_ce_loss": 0.00022792816162109375, "train/clean_loss_scale": 1120.9075927734375, "train/clean_loss_weight": 224.1815185546875, "train/combined_loss": 0.3065836727619171, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06848341232227488, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.255486398935318 }, { "epoch": 0.6848341232227488, "step": 289, "train/alpha": 0.2, "train/clean_ce_loss": 0.09521484375, "train/clean_loss_scale": 7.21531343460083, "train/clean_loss_weight": 1.443062663078308, "train/combined_loss": 0.8244059085845947, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06848341232227488, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6870049238204956 }, { "epoch": 0.6848341232227488, "step": 289, "train/alpha": 0.2, "train/clean_ce_loss": 0.02490234375, "train/clean_loss_scale": 18.857378005981445, "train/clean_loss_weight": 3.7714755535125732, "train/combined_loss": 0.5635114908218384, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06848341232227488, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4695928990840912 }, { "epoch": 0.6848341232227488, "step": 289, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 8.196551322937012, "train/clean_loss_weight": 1.6393102407455444, "train/combined_loss": 0.6531626582145691, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06848341232227488, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.544302225112915 }, { "epoch": 0.6872037914691943, "grad_norm": 13.453370094299316, "learning_rate": 2e-06, "loss": 3.894, "step": 290 }, { "epoch": 0.6872037914691943, "step": 290, "train/alpha": 0.2, "train/clean_ce_loss": 0.134765625, "train/clean_loss_scale": 9.644387245178223, "train/clean_loss_weight": 1.9288774728775024, "train/combined_loss": 1.5596781969070435, "train/pos_loss/pos0": 0.7210587859153748, "train/pos_loss/pos1": 1.3135849237442017, "train/pos_loss/pos2": 1.6436161994934082, "train/pos_loss/pos3": 2.0378804206848145, "train/pos_loss/pos4": 2.723647117614746, "train/pos_loss/pos5": 2.9379308223724365, "train/pos_loss/pos6": 3.5638091564178467, "train/pos_loss/weight0": 2.7880771160125732, "train/pos_loss/weight1": 1.9054954051971436, "train/pos_loss/weight2": 1.1993087530136108, "train/pos_loss/weight3": 0.6426250338554382, "train/pos_loss/weight4": 0.29752978682518005, "train/pos_loss/weight5": 0.11879607290029526, "train/pos_loss/weight6": 0.048167575150728226, "train/pos_weight_progress": 0.06872037914691943, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2997318506240845 }, { "epoch": 0.6872037914691943, "step": 290, "train/alpha": 0.2, "train/clean_ce_loss": 0.09521484375, "train/clean_loss_scale": 9.080339431762695, "train/clean_loss_weight": 1.8160679340362549, "train/combined_loss": 1.0374997854232788, "train/pos_loss/pos0": 0.4733944237232208, "train/pos_loss/pos1": 0.9244102835655212, "train/pos_loss/pos2": 1.3319977521896362, "train/pos_loss/pos3": 1.741553783416748, "train/pos_loss/pos4": 2.1918280124664307, "train/pos_loss/pos5": 2.5499892234802246, "train/pos_loss/pos6": 2.8235273361206055, "train/pos_loss/weight0": 3.2890207767486572, "train/pos_loss/weight1": 2.0121548175811768, "train/pos_loss/weight2": 1.0529354810714722, "train/pos_loss/weight3": 0.4438987374305725, "train/pos_loss/weight4": 0.1518392413854599, "train/pos_loss/weight5": 0.03892918676137924, "train/pos_loss/weight6": 0.011222094297409058, "train/pos_weight_progress": 0.06872037914691943, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8645831346511841 }, { "epoch": 0.6872037914691943, "step": 290, "train/alpha": 0.2, "train/clean_ce_loss": 0.07177734375, "train/clean_loss_scale": 8.636004447937012, "train/clean_loss_weight": 1.7272008657455444, "train/combined_loss": 0.7438433766365051, "train/pos_loss/pos0": 0.17403779923915863, "train/pos_loss/pos1": 0.5288421511650085, "train/pos_loss/pos2": 1.0024008750915527, "train/pos_loss/pos3": 1.5165209770202637, "train/pos_loss/pos4": 2.1754984855651855, "train/pos_loss/pos5": 2.6866228580474854, "train/pos_loss/pos6": 3.0990958213806152, "train/pos_loss/weight0": 2.9754929542541504, "train/pos_loss/weight1": 1.9933114051818848, "train/pos_loss/weight2": 1.1455427408218384, "train/pos_loss/weight3": 0.5585806965827942, "train/pos_loss/weight4": 0.22400397062301636, "train/pos_loss/weight5": 0.07731747627258301, "train/pos_loss/weight6": 0.025750359520316124, "train/pos_weight_progress": 0.06872037914691943, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6198694705963135 }, { "epoch": 0.6872037914691943, "step": 290, "train/alpha": 0.2, "train/clean_ce_loss": 0.05126953125, "train/clean_loss_scale": 12.807204246520996, "train/clean_loss_weight": 2.561440944671631, "train/combined_loss": 0.7879432439804077, "train/pos_loss/pos0": 0.17481857538223267, "train/pos_loss/pos1": 0.4718545377254486, "train/pos_loss/pos2": 0.9579777121543884, "train/pos_loss/pos3": 1.4356220960617065, "train/pos_loss/pos4": 1.9916093349456787, "train/pos_loss/pos5": 2.5757224559783936, "train/pos_loss/pos6": 2.894801616668701, "train/pos_loss/weight0": 2.676961660385132, "train/pos_loss/weight1": 1.9315574169158936, "train/pos_loss/weight2": 1.2353514432907104, "train/pos_loss/weight3": 0.6803271770477295, "train/pos_loss/weight4": 0.31039610505104065, "train/pos_loss/weight5": 0.11622707545757294, "train/pos_loss/weight6": 0.049178846180438995, "train/pos_weight_progress": 0.06872037914691943, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6566193699836731 }, { "epoch": 0.6895734597156398, "step": 291, "train/alpha": 0.2, "train/clean_ce_loss": 0.08544921875, "train/clean_loss_scale": 7.545363426208496, "train/clean_loss_weight": 1.5090726613998413, "train/combined_loss": 0.7736945152282715, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.06895734597156399, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.644745409488678 }, { "epoch": 0.6895734597156398, "step": 291, "train/alpha": 0.2, "train/clean_ce_loss": 0.07470703125, "train/clean_loss_scale": 13.39666748046875, "train/clean_loss_weight": 2.679333448410034, "train/combined_loss": 1.2009903192520142, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.06895734597156399, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.00082528591156 }, { "epoch": 0.6895734597156398, "step": 291, "train/alpha": 0.2, "train/clean_ce_loss": 0.1396484375, "train/clean_loss_scale": 10.3115816116333, "train/clean_loss_weight": 2.062316417694092, "train/combined_loss": 1.72799551486969, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.06895734597156399, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4399962425231934 }, { "epoch": 0.6895734597156398, "step": 291, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 8.847588539123535, "train/clean_loss_weight": 1.7695177793502808, "train/combined_loss": 0.9279599785804749, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.06895734597156399, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7732999920845032 }, { "epoch": 0.6919431279620853, "step": 292, "train/alpha": 0.2, "train/clean_ce_loss": 0.0693359375, "train/clean_loss_scale": 8.178264617919922, "train/clean_loss_weight": 1.6356528997421265, "train/combined_loss": 0.6804571747779846, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.06919431279620854, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.567047655582428 }, { "epoch": 0.6919431279620853, "step": 292, "train/alpha": 0.2, "train/clean_ce_loss": 0.111328125, "train/clean_loss_scale": 8.041887283325195, "train/clean_loss_weight": 1.608377456665039, "train/combined_loss": 1.0743458271026611, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.06919431279620854, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8952882289886475 }, { "epoch": 0.6919431279620853, "step": 292, "train/alpha": 0.2, "train/clean_ce_loss": 0.11083984375, "train/clean_loss_scale": 7.95795202255249, "train/clean_loss_weight": 1.591590404510498, "train/combined_loss": 1.0584697723388672, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.06919431279620854, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8820581436157227 }, { "epoch": 0.6919431279620853, "step": 292, "train/alpha": 0.2, "train/clean_ce_loss": 0.056884765625, "train/clean_loss_scale": 9.296154975891113, "train/clean_loss_weight": 1.8592309951782227, "train/combined_loss": 0.6345715522766113, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.06919431279620854, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5288096070289612 }, { "epoch": 0.6943127962085308, "step": 293, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 9.782625198364258, "train/clean_loss_weight": 1.9565250873565674, "train/combined_loss": 0.7623568773269653, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.06943127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.635297417640686 }, { "epoch": 0.6943127962085308, "step": 293, "train/alpha": 0.2, "train/clean_ce_loss": 0.171875, "train/clean_loss_scale": 7.127521991729736, "train/clean_loss_weight": 1.425504446029663, "train/combined_loss": 1.4700514078140259, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.06943127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2250428199768066 }, { "epoch": 0.6943127962085308, "step": 293, "train/alpha": 0.2, "train/clean_ce_loss": 0.166015625, "train/clean_loss_scale": 6.861875057220459, "train/clean_loss_weight": 1.3723750114440918, "train/combined_loss": 1.3670141696929932, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.06943127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1391785144805908 }, { "epoch": 0.6943127962085308, "step": 293, "train/alpha": 0.2, "train/clean_ce_loss": 0.057861328125, "train/clean_loss_scale": 11.05775260925293, "train/clean_loss_weight": 2.21155047416687, "train/combined_loss": 0.7677794694900513, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.06943127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6398162245750427 }, { "epoch": 0.6966824644549763, "step": 294, "train/alpha": 0.2, "train/clean_ce_loss": 0.09521484375, "train/clean_loss_scale": 7.730891704559326, "train/clean_loss_weight": 1.5461783409118652, "train/combined_loss": 0.8833147883415222, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.06966824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.736095666885376 }, { "epoch": 0.6966824644549763, "step": 294, "train/alpha": 0.2, "train/clean_ce_loss": 0.08056640625, "train/clean_loss_scale": 9.029006004333496, "train/clean_loss_weight": 1.805801272392273, "train/combined_loss": 0.8729214668273926, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.06966824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7274345755577087 }, { "epoch": 0.6966824644549763, "step": 294, "train/alpha": 0.2, "train/clean_ce_loss": 0.126953125, "train/clean_loss_scale": 11.805288314819336, "train/clean_loss_weight": 2.361057758331299, "train/combined_loss": 1.7984619140625, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.06966824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.49871826171875 }, { "epoch": 0.6966824644549763, "step": 294, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 7.345574855804443, "train/clean_loss_weight": 1.4691150188446045, "train/combined_loss": 0.7704245448112488, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.06966824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6420204639434814 }, { "epoch": 0.6990521327014217, "grad_norm": 11.693081855773926, "learning_rate": 2e-06, "loss": 3.7882, "step": 295 }, { "epoch": 0.6990521327014217, "step": 295, "train/alpha": 0.2, "train/clean_ce_loss": 0.03369140625, "train/clean_loss_scale": 12.977949142456055, "train/clean_loss_weight": 2.5955898761749268, "train/combined_loss": 0.5246944427490234, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.06990521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43724536895751953 }, { "epoch": 0.6990521327014217, "step": 295, "train/alpha": 0.2, "train/clean_ce_loss": 0.068359375, "train/clean_loss_scale": 9.748416900634766, "train/clean_loss_weight": 1.949683427810669, "train/combined_loss": 0.7996748089790344, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.06990521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6663956642150879 }, { "epoch": 0.6990521327014217, "step": 295, "train/alpha": 0.2, "train/clean_ce_loss": 0.08251953125, "train/clean_loss_scale": 8.326116561889648, "train/clean_loss_weight": 1.6652233600616455, "train/combined_loss": 0.8244806528091431, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.06990521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6870672106742859 }, { "epoch": 0.6990521327014217, "step": 295, "train/alpha": 0.2, "train/clean_ce_loss": 0.171875, "train/clean_loss_scale": 7.246648788452148, "train/clean_loss_weight": 1.4493297338485718, "train/combined_loss": 1.4946212768554688, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.06990521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2455177307128906 }, { "epoch": 0.7014218009478673, "step": 296, "train/alpha": 0.2, "train/clean_ce_loss": 0.1689453125, "train/clean_loss_scale": 9.902219772338867, "train/clean_loss_weight": 1.9804439544677734, "train/combined_loss": 2.0075201988220215, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.07014218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.672933578491211 }, { "epoch": 0.7014218009478673, "step": 296, "train/alpha": 0.2, "train/clean_ce_loss": 0.109375, "train/clean_loss_scale": 7.31749963760376, "train/clean_loss_weight": 1.463499903678894, "train/combined_loss": 0.9604218006134033, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.07014218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8003515005111694 }, { "epoch": 0.7014218009478673, "step": 296, "train/alpha": 0.2, "train/clean_ce_loss": 0.130859375, "train/clean_loss_scale": 10.369349479675293, "train/clean_loss_weight": 2.0738699436187744, "train/combined_loss": 1.6283118724822998, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.07014218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3569265604019165 }, { "epoch": 0.7014218009478673, "step": 296, "train/alpha": 0.2, "train/clean_ce_loss": 0.11669921875, "train/clean_loss_scale": 11.21755599975586, "train/clean_loss_weight": 2.243511199951172, "train/combined_loss": 1.570896029472351, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.07014218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3090800046920776 }, { "epoch": 0.7037914691943128, "step": 297, "train/alpha": 0.2, "train/clean_ce_loss": 0.028564453125, "train/clean_loss_scale": 15.702104568481445, "train/clean_loss_weight": 3.140420913696289, "train/combined_loss": 0.5382264256477356, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.07037914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.44852203130722046 }, { "epoch": 0.7037914691943128, "step": 297, "train/alpha": 0.2, "train/clean_ce_loss": 0.11328125, "train/clean_loss_scale": 9.92341136932373, "train/clean_loss_weight": 1.984682321548462, "train/combined_loss": 1.348963737487793, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.07037914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1241364479064941 }, { "epoch": 0.7037914691943128, "step": 297, "train/alpha": 0.2, "train/clean_ce_loss": 0.07666015625, "train/clean_loss_scale": 9.092280387878418, "train/clean_loss_weight": 1.8184560537338257, "train/combined_loss": 0.8364187479019165, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.07037914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.697015643119812 }, { "epoch": 0.7037914691943128, "step": 297, "train/alpha": 0.2, "train/clean_ce_loss": 0.1513671875, "train/clean_loss_scale": 8.020398139953613, "train/clean_loss_weight": 1.6040796041488647, "train/combined_loss": 1.4568301439285278, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.07037914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2140251398086548 }, { "epoch": 0.7061611374407583, "step": 298, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 6.958696365356445, "train/clean_loss_weight": 1.3917392492294312, "train/combined_loss": 0.6686872243881226, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.07061611374407582, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5572393536567688 }, { "epoch": 0.7061611374407583, "step": 298, "train/alpha": 0.2, "train/clean_ce_loss": 0.1650390625, "train/clean_loss_scale": 7.76496696472168, "train/clean_loss_weight": 1.5529934167861938, "train/combined_loss": 1.537827491760254, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.07061611374407582, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2815228700637817 }, { "epoch": 0.7061611374407583, "step": 298, "train/alpha": 0.2, "train/clean_ce_loss": 0.12890625, "train/clean_loss_scale": 10.551836967468262, "train/clean_loss_weight": 2.1103675365448, "train/combined_loss": 1.6322373151779175, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.07061611374407582, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3601977825164795 }, { "epoch": 0.7061611374407583, "step": 298, "train/alpha": 0.2, "train/clean_ce_loss": 0.119140625, "train/clean_loss_scale": 7.867295265197754, "train/clean_loss_weight": 1.5734590291976929, "train/combined_loss": 1.1247773170471191, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.07061611374407582, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9373144507408142 }, { "epoch": 0.7085308056872038, "step": 299, "train/alpha": 0.2, "train/clean_ce_loss": 0.0498046875, "train/clean_loss_scale": 9.4194917678833, "train/clean_loss_weight": 1.883898377418518, "train/combined_loss": 0.5629618167877197, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.07085308056872038, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4691348373889923 }, { "epoch": 0.7085308056872038, "step": 299, "train/alpha": 0.2, "train/clean_ce_loss": 0.162109375, "train/clean_loss_scale": 7.960659503936768, "train/clean_loss_weight": 1.5921319723129272, "train/combined_loss": 1.5485970973968506, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.07085308056872038, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2904975414276123 }, { "epoch": 0.7085308056872038, "step": 299, "train/alpha": 0.2, "train/clean_ce_loss": 0.1337890625, "train/clean_loss_scale": 7.1832756996154785, "train/clean_loss_weight": 1.4366551637649536, "train/combined_loss": 1.1532524824142456, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.07085308056872038, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9610437154769897 }, { "epoch": 0.7085308056872038, "step": 299, "train/alpha": 0.2, "train/clean_ce_loss": 0.021240234375, "train/clean_loss_scale": 20.90025520324707, "train/clean_loss_weight": 4.180051326751709, "train/combined_loss": 0.5327116250991821, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.07085308056872038, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4439263343811035 }, { "epoch": 0.7109004739336493, "grad_norm": 12.405698776245117, "learning_rate": 2e-06, "loss": 3.8061, "step": 300 }, { "epoch": 0.7109004739336493, "step": 300, "train/alpha": 0.2, "train/clean_ce_loss": 0.06103515625, "train/clean_loss_scale": 9.342954635620117, "train/clean_loss_weight": 1.8685909509658813, "train/combined_loss": 0.6842984557151794, "train/pos_loss/pos0": 0.21607735753059387, "train/pos_loss/pos1": 0.39298826456069946, "train/pos_loss/pos2": 0.7181379795074463, "train/pos_loss/pos3": 1.03764808177948, "train/pos_loss/pos4": 1.530633807182312, "train/pos_loss/pos5": 1.8440624475479126, "train/pos_loss/pos6": 2.354074001312256, "train/pos_loss/weight0": 2.540029525756836, "train/pos_loss/weight1": 1.8561444282531738, "train/pos_loss/weight2": 1.2680084705352783, "train/pos_loss/weight3": 0.7185155153274536, "train/pos_loss/weight4": 0.3761143982410431, "train/pos_loss/weight5": 0.16607141494750977, "train/pos_loss/weight6": 0.0751166045665741, "train/pos_weight_progress": 0.07109004739336493, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5702487230300903 }, { "epoch": 0.7109004739336493, "step": 300, "train/alpha": 0.2, "train/clean_ce_loss": 0.09765625, "train/clean_loss_scale": 9.244242668151855, "train/clean_loss_weight": 1.848848581314087, "train/combined_loss": 1.0833096504211426, "train/pos_loss/pos0": 0.3383486866950989, "train/pos_loss/pos1": 0.626747727394104, "train/pos_loss/pos2": 0.9906226992607117, "train/pos_loss/pos3": 1.4751635789871216, "train/pos_loss/pos4": 2.128462791442871, "train/pos_loss/pos5": 2.7286884784698486, "train/pos_loss/pos6": 3.2034177780151367, "train/pos_loss/weight0": 2.392669439315796, "train/pos_loss/weight1": 1.778955101966858, "train/pos_loss/weight2": 1.2280402183532715, "train/pos_loss/weight3": 0.7843155860900879, "train/pos_loss/weight4": 0.44714146852493286, "train/pos_loss/weight5": 0.23461589217185974, "train/pos_loss/weight6": 0.1342623233795166, "train/pos_weight_progress": 0.07109004739336493, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9027580618858337 }, { "epoch": 0.7109004739336493, "step": 300, "train/alpha": 0.2, "train/clean_ce_loss": 0.140625, "train/clean_loss_scale": 8.014962196350098, "train/clean_loss_weight": 1.6029924154281616, "train/combined_loss": 1.3525248765945435, "train/pos_loss/pos0": 0.5293543338775635, "train/pos_loss/pos1": 0.9930457472801208, "train/pos_loss/pos2": 1.4351756572723389, "train/pos_loss/pos3": 2.0614511966705322, "train/pos_loss/pos4": 2.8017537593841553, "train/pos_loss/pos5": 3.4352424144744873, "train/pos_loss/pos6": 3.972121000289917, "train/pos_loss/weight0": 2.727327823638916, "train/pos_loss/weight1": 1.9339767694473267, "train/pos_loss/weight2": 1.2135298252105713, "train/pos_loss/weight3": 0.6638830900192261, "train/pos_loss/weight4": 0.30178579688072205, "train/pos_loss/weight5": 0.11232864111661911, "train/pos_loss/weight6": 0.04716794937849045, "train/pos_weight_progress": 0.07109004739336493, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1271040439605713 }, { "epoch": 0.7109004739336493, "step": 300, "train/alpha": 0.2, "train/clean_ce_loss": 0.1708984375, "train/clean_loss_scale": 7.063406944274902, "train/clean_loss_weight": 1.4126814603805542, "train/combined_loss": 1.4485502243041992, "train/pos_loss/pos0": 0.6841578483581543, "train/pos_loss/pos1": 1.212119221687317, "train/pos_loss/pos2": 1.8550816774368286, "train/pos_loss/pos3": 2.3345067501068115, "train/pos_loss/pos4": 2.9493157863616943, "train/pos_loss/pos5": 3.3043181896209717, "train/pos_loss/pos6": 3.7417373657226562, "train/pos_loss/weight0": 3.1722042560577393, "train/pos_loss/weight1": 2.032339334487915, "train/pos_loss/weight2": 1.0950396060943604, "train/pos_loss/weight3": 0.47608092427253723, "train/pos_loss/weight4": 0.1643134355545044, "train/pos_loss/weight5": 0.04716969281435013, "train/pos_loss/weight6": 0.012852665968239307, "train/pos_weight_progress": 0.07109004739336493, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.207125186920166 }, { "epoch": 0.7132701421800948, "step": 301, "train/alpha": 0.2, "train/clean_ce_loss": 0.1083984375, "train/clean_loss_scale": 7.266012668609619, "train/clean_loss_weight": 1.4532026052474976, "train/combined_loss": 0.945149302482605, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07132701421800948, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7876244187355042 }, { "epoch": 0.7132701421800948, "step": 301, "train/alpha": 0.2, "train/clean_ce_loss": 0.11865234375, "train/clean_loss_scale": 9.706578254699707, "train/clean_loss_weight": 1.9413156509399414, "train/combined_loss": 1.3820499181747437, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07132701421800948, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1517082452774048 }, { "epoch": 0.7132701421800948, "step": 301, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 5.931117534637451, "train/clean_loss_weight": 1.1862235069274902, "train/combined_loss": 0.6603001952171326, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07132701421800948, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5502501726150513 }, { "epoch": 0.7132701421800948, "step": 301, "train/alpha": 0.2, "train/clean_ce_loss": 0.1259765625, "train/clean_loss_scale": 6.731776237487793, "train/clean_loss_weight": 1.3463553190231323, "train/combined_loss": 1.0176552534103394, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07132701421800948, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8480460047721863 }, { "epoch": 0.7156398104265402, "step": 302, "train/alpha": 0.2, "train/clean_ce_loss": 0.13671875, "train/clean_loss_scale": 8.170029640197754, "train/clean_loss_weight": 1.6340059041976929, "train/combined_loss": 1.3403955698013306, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07156398104265403, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1169962882995605 }, { "epoch": 0.7156398104265402, "step": 302, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 6.533695220947266, "train/clean_loss_weight": 1.306739091873169, "train/combined_loss": 0.5819072723388672, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07156398104265403, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.48492270708084106 }, { "epoch": 0.7156398104265402, "step": 302, "train/alpha": 0.2, "train/clean_ce_loss": 0.09814453125, "train/clean_loss_scale": 5.790398597717285, "train/clean_loss_weight": 1.158079743385315, "train/combined_loss": 0.6819551587104797, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07156398104265403, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.568295955657959 }, { "epoch": 0.7156398104265402, "step": 302, "train/alpha": 0.2, "train/clean_ce_loss": 0.033935546875, "train/clean_loss_scale": 10.70934772491455, "train/clean_loss_weight": 2.14186954498291, "train/combined_loss": 0.4361130893230438, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07156398104265403, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3634275794029236 }, { "epoch": 0.7180094786729858, "step": 303, "train/alpha": 0.2, "train/clean_ce_loss": 0.11669921875, "train/clean_loss_scale": 8.259106636047363, "train/clean_loss_weight": 1.6518213748931885, "train/combined_loss": 1.15659761428833, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07180094786729858, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9638313055038452 }, { "epoch": 0.7180094786729858, "step": 303, "train/alpha": 0.2, "train/clean_ce_loss": 0.06298828125, "train/clean_loss_scale": 6.564981937408447, "train/clean_loss_weight": 1.3129963874816895, "train/combined_loss": 0.49622032046318054, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07180094786729858, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.41351693868637085 }, { "epoch": 0.7180094786729858, "step": 303, "train/alpha": 0.2, "train/clean_ce_loss": 0.1416015625, "train/clean_loss_scale": 8.956459999084473, "train/clean_loss_weight": 1.7912920713424683, "train/combined_loss": 1.5218985080718994, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07180094786729858, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2682487964630127 }, { "epoch": 0.7180094786729858, "step": 303, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 6.437779426574707, "train/clean_loss_weight": 1.2875559329986572, "train/combined_loss": 0.5054662823677063, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07180094786729858, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.42122188210487366 }, { "epoch": 0.7203791469194313, "step": 304, "train/alpha": 0.2, "train/clean_ce_loss": 0.059814453125, "train/clean_loss_scale": 4.574822902679443, "train/clean_loss_weight": 0.9149646162986755, "train/combined_loss": 0.3283686637878418, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07203791469194312, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.27364054322242737 }, { "epoch": 0.7203791469194313, "step": 304, "train/alpha": 0.2, "train/clean_ce_loss": 0.169921875, "train/clean_loss_scale": 5.5785040855407715, "train/clean_loss_weight": 1.1157008409500122, "train/combined_loss": 1.1374918222427368, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07203791469194312, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9479098916053772 }, { "epoch": 0.7203791469194313, "step": 304, "train/alpha": 0.2, "train/clean_ce_loss": 0.10986328125, "train/clean_loss_scale": 5.093074321746826, "train/clean_loss_weight": 1.0186148881912231, "train/combined_loss": 0.6714502573013306, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07203791469194312, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5595418810844421 }, { "epoch": 0.7203791469194313, "step": 304, "train/alpha": 0.2, "train/clean_ce_loss": 0.12890625, "train/clean_loss_scale": 7.365815162658691, "train/clean_loss_weight": 1.4731630086898804, "train/combined_loss": 1.139399528503418, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07203791469194312, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9494996070861816 }, { "epoch": 0.7227488151658767, "grad_norm": 17.511390686035156, "learning_rate": 2e-06, "loss": 3.6991, "step": 305 }, { "epoch": 0.7227488151658767, "step": 305, "train/alpha": 0.2, "train/clean_ce_loss": 0.10009765625, "train/clean_loss_scale": 7.900421619415283, "train/clean_loss_weight": 1.5800843238830566, "train/combined_loss": 0.9489763975143433, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07227488151658767, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.790813684463501 }, { "epoch": 0.7227488151658767, "step": 305, "train/alpha": 0.2, "train/clean_ce_loss": 0.12890625, "train/clean_loss_scale": 5.888127326965332, "train/clean_loss_weight": 1.1776255369186401, "train/combined_loss": 0.9108196496963501, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07227488151658767, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7590163946151733 }, { "epoch": 0.7227488151658767, "step": 305, "train/alpha": 0.2, "train/clean_ce_loss": 0.1015625, "train/clean_loss_scale": 5.942908763885498, "train/clean_loss_weight": 1.1885818243026733, "train/combined_loss": 0.724291980266571, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07227488151658767, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.60357666015625 }, { "epoch": 0.7227488151658767, "step": 305, "train/alpha": 0.2, "train/clean_ce_loss": 0.10205078125, "train/clean_loss_scale": 6.602678298950195, "train/clean_loss_weight": 1.320535659790039, "train/combined_loss": 0.808570146560669, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07227488151658767, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6738084554672241 }, { "epoch": 0.7251184834123223, "step": 306, "train/alpha": 0.2, "train/clean_ce_loss": 0.05908203125, "train/clean_loss_scale": 7.072569847106934, "train/clean_loss_weight": 1.4145139455795288, "train/combined_loss": 0.5014341473579407, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07251184834123223, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.41786178946495056 }, { "epoch": 0.7251184834123223, "step": 306, "train/alpha": 0.2, "train/clean_ce_loss": 0.083984375, "train/clean_loss_scale": 6.005763053894043, "train/clean_loss_weight": 1.2011526823043823, "train/combined_loss": 0.6052682995796204, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07251184834123223, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5043902397155762 }, { "epoch": 0.7251184834123223, "step": 306, "train/alpha": 0.2, "train/clean_ce_loss": 0.07861328125, "train/clean_loss_scale": 7.1581501960754395, "train/clean_loss_weight": 1.43163001537323, "train/combined_loss": 0.6752707958221436, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07251184834123223, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5627256631851196 }, { "epoch": 0.7251184834123223, "step": 306, "train/alpha": 0.2, "train/clean_ce_loss": 0.0634765625, "train/clean_loss_scale": 7.424139976501465, "train/clean_loss_weight": 1.484827995300293, "train/combined_loss": 0.565510630607605, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07251184834123223, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.47125887870788574 }, { "epoch": 0.7274881516587678, "step": 307, "train/alpha": 0.2, "train/clean_ce_loss": 0.040283203125, "train/clean_loss_scale": 9.121847152709961, "train/clean_loss_weight": 1.8243694305419922, "train/combined_loss": 0.4409486949443817, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07274881516587678, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.36745724081993103 }, { "epoch": 0.7274881516587678, "step": 307, "train/alpha": 0.2, "train/clean_ce_loss": 0.06298828125, "train/clean_loss_scale": 6.784060955047607, "train/clean_loss_weight": 1.3568122386932373, "train/combined_loss": 0.5127795934677124, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07274881516587678, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4273163378238678 }, { "epoch": 0.7274881516587678, "step": 307, "train/alpha": 0.2, "train/clean_ce_loss": 0.0150146484375, "train/clean_loss_scale": 19.809396743774414, "train/clean_loss_weight": 3.9618794918060303, "train/combined_loss": 0.3569173812866211, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07274881516587678, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2974311411380768 }, { "epoch": 0.7274881516587678, "step": 307, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 7.390880107879639, "train/clean_loss_weight": 1.4781759977340698, "train/combined_loss": 0.7102174162864685, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07274881516587678, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.591847836971283 }, { "epoch": 0.7298578199052133, "step": 308, "train/alpha": 0.2, "train/clean_ce_loss": 0.05322265625, "train/clean_loss_scale": 7.6186628341674805, "train/clean_loss_weight": 1.5237325429916382, "train/combined_loss": 0.48658257722854614, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07298578199052133, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.40548548102378845 }, { "epoch": 0.7298578199052133, "step": 308, "train/alpha": 0.2, "train/clean_ce_loss": 0.1337890625, "train/clean_loss_scale": 6.62235164642334, "train/clean_loss_weight": 1.3244704008102417, "train/combined_loss": 1.0631978511810303, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07298578199052133, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8859981894493103 }, { "epoch": 0.7298578199052133, "step": 308, "train/alpha": 0.2, "train/clean_ce_loss": 0.146484375, "train/clean_loss_scale": 6.956305503845215, "train/clean_loss_weight": 1.391261100769043, "train/combined_loss": 1.2227880954742432, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07298578199052133, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0189900398254395 }, { "epoch": 0.7298578199052133, "step": 308, "train/alpha": 0.2, "train/clean_ce_loss": 0.01031494140625, "train/clean_loss_scale": 24.777626037597656, "train/clean_loss_weight": 4.9555253982543945, "train/combined_loss": 0.30669572949409485, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07298578199052133, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.25557976961135864 }, { "epoch": 0.7322274881516587, "step": 309, "train/alpha": 0.2, "train/clean_ce_loss": 0.08544921875, "train/clean_loss_scale": 5.277835845947266, "train/clean_loss_weight": 1.0555671453475952, "train/combined_loss": 0.5411843061447144, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07322274881516588, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45098695158958435 }, { "epoch": 0.7322274881516587, "step": 309, "train/alpha": 0.2, "train/clean_ce_loss": 0.08544921875, "train/clean_loss_scale": 6.473220348358154, "train/clean_loss_weight": 1.2946441173553467, "train/combined_loss": 0.6637579798698425, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07322274881516588, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.553131639957428 }, { "epoch": 0.7322274881516587, "step": 309, "train/alpha": 0.2, "train/clean_ce_loss": 0.11865234375, "train/clean_loss_scale": 6.613566875457764, "train/clean_loss_weight": 1.3227133750915527, "train/combined_loss": 0.9416582584381104, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07322274881516588, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7847152352333069 }, { "epoch": 0.7322274881516587, "step": 309, "train/alpha": 0.2, "train/clean_ce_loss": 0.1884765625, "train/clean_loss_scale": 5.8341779708862305, "train/clean_loss_weight": 1.1668356657028198, "train/combined_loss": 1.3195269107818604, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07322274881516588, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0996057987213135 }, { "epoch": 0.7345971563981043, "grad_norm": 13.494294166564941, "learning_rate": 2e-06, "loss": 3.4534, "step": 310 }, { "epoch": 0.7345971563981043, "step": 310, "train/alpha": 0.2, "train/clean_ce_loss": 0.053466796875, "train/clean_loss_scale": 8.237420082092285, "train/clean_loss_weight": 1.6474840641021729, "train/combined_loss": 0.5285141468048096, "train/pos_loss/pos0": 0.2015574425458908, "train/pos_loss/pos1": 0.3925747573375702, "train/pos_loss/pos2": 0.7629894614219666, "train/pos_loss/pos3": 1.3055154085159302, "train/pos_loss/pos4": 1.7526888847351074, "train/pos_loss/pos5": 2.3319225311279297, "train/pos_loss/pos6": 2.632561445236206, "train/pos_loss/weight0": 3.3734278678894043, "train/pos_loss/weight1": 2.032618999481201, "train/pos_loss/weight2": 1.0348104238510132, "train/pos_loss/weight3": 0.4082425534725189, "train/pos_loss/weight4": 0.11692892014980316, "train/pos_loss/weight5": 0.027627497911453247, "train/pos_loss/weight6": 0.006343487650156021, "train/pos_weight_progress": 0.07345971563981042, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4404284656047821 }, { "epoch": 0.7345971563981043, "step": 310, "train/alpha": 0.2, "train/clean_ce_loss": 0.142578125, "train/clean_loss_scale": 9.412863731384277, "train/clean_loss_weight": 1.8825727701187134, "train/combined_loss": 1.610482096672058, "train/pos_loss/pos0": 0.6561887860298157, "train/pos_loss/pos1": 1.1906472444534302, "train/pos_loss/pos2": 1.743606448173523, "train/pos_loss/pos3": 2.2051992416381836, "train/pos_loss/pos4": 2.675363063812256, "train/pos_loss/pos5": 2.913097858428955, "train/pos_loss/pos6": 3.266584873199463, "train/pos_loss/weight0": 2.5574722290039062, "train/pos_loss/weight1": 1.8416695594787598, "train/pos_loss/weight2": 1.2497633695602417, "train/pos_loss/weight3": 0.7473865151405334, "train/pos_loss/weight4": 0.36218979954719543, "train/pos_loss/weight5": 0.1631743609905243, "train/pos_loss/weight6": 0.07834412902593613, "train/pos_weight_progress": 0.07345971563981042, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3420684337615967 }, { "epoch": 0.7345971563981043, "step": 310, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 6.669741153717041, "train/clean_loss_weight": 1.3339482545852661, "train/combined_loss": 0.6331043243408203, "train/pos_loss/pos0": 0.2733866274356842, "train/pos_loss/pos1": 0.5590404868125916, "train/pos_loss/pos2": 0.8376547694206238, "train/pos_loss/pos3": 1.124129295349121, "train/pos_loss/pos4": 1.5047389268875122, "train/pos_loss/pos5": 1.8987951278686523, "train/pos_loss/pos6": 2.286878824234009, "train/pos_loss/weight0": 3.3124046325683594, "train/pos_loss/weight1": 2.03204345703125, "train/pos_loss/weight2": 1.0462486743927002, "train/pos_loss/weight3": 0.42394328117370605, "train/pos_loss/weight4": 0.13925829529762268, "train/pos_loss/weight5": 0.03597472608089447, "train/pos_loss/weight6": 0.010126748122274876, "train/pos_weight_progress": 0.07345971563981042, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5275869369506836 }, { "epoch": 0.7345971563981043, "step": 310, "train/alpha": 0.2, "train/clean_ce_loss": 0.09326171875, "train/clean_loss_scale": 10.316636085510254, "train/clean_loss_weight": 2.0633273124694824, "train/combined_loss": 1.1545766592025757, "train/pos_loss/pos0": 0.36967265605926514, "train/pos_loss/pos1": 0.7722523212432861, "train/pos_loss/pos2": 1.2004919052124023, "train/pos_loss/pos3": 1.6646604537963867, "train/pos_loss/pos4": 2.0443994998931885, "train/pos_loss/pos5": 2.55505633354187, "train/pos_loss/pos6": 2.9276232719421387, "train/pos_loss/weight0": 2.4804270267486572, "train/pos_loss/weight1": 1.8152012825012207, "train/pos_loss/weight2": 1.2214558124542236, "train/pos_loss/weight3": 0.750694751739502, "train/pos_loss/weight4": 0.410534530878067, "train/pos_loss/weight5": 0.21032018959522247, "train/pos_loss/weight6": 0.11136652529239655, "train/pos_weight_progress": 0.07345971563981042, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9621472358703613 }, { "epoch": 0.7369668246445498, "step": 311, "train/alpha": 0.2, "train/clean_ce_loss": 0.09912109375, "train/clean_loss_scale": 6.444619655609131, "train/clean_loss_weight": 1.288923978805542, "train/combined_loss": 0.7665573358535767, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07369668246445497, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6387977600097656 }, { "epoch": 0.7369668246445498, "step": 311, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 6.588024139404297, "train/clean_loss_weight": 1.317604899406433, "train/combined_loss": 0.5211229920387268, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07369668246445497, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43426916003227234 }, { "epoch": 0.7369668246445498, "step": 311, "train/alpha": 0.2, "train/clean_ce_loss": 0.06396484375, "train/clean_loss_scale": 8.569229125976562, "train/clean_loss_weight": 1.7138458490371704, "train/combined_loss": 0.6577552556991577, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07369668246445497, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5481293797492981 }, { "epoch": 0.7369668246445498, "step": 311, "train/alpha": 0.2, "train/clean_ce_loss": 0.06689453125, "train/clean_loss_scale": 6.814096927642822, "train/clean_loss_weight": 1.3628194332122803, "train/combined_loss": 0.5469909906387329, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07369668246445497, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4558258056640625 }, { "epoch": 0.7393364928909952, "step": 312, "train/alpha": 0.2, "train/clean_ce_loss": 0.1630859375, "train/clean_loss_scale": 6.368649959564209, "train/clean_loss_weight": 1.2737300395965576, "train/combined_loss": 1.246364712715149, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07393364928909953, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0386372804641724 }, { "epoch": 0.7393364928909952, "step": 312, "train/alpha": 0.2, "train/clean_ce_loss": 0.0634765625, "train/clean_loss_scale": 8.20685863494873, "train/clean_loss_weight": 1.641371726989746, "train/combined_loss": 0.6251317858695984, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07393364928909953, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5209431648254395 }, { "epoch": 0.7393364928909952, "step": 312, "train/alpha": 0.2, "train/clean_ce_loss": 0.024169921875, "train/clean_loss_scale": 12.855607986450195, "train/clean_loss_weight": 2.5711216926574707, "train/combined_loss": 0.372862845659256, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07393364928909953, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3107190430164337 }, { "epoch": 0.7393364928909952, "step": 312, "train/alpha": 0.2, "train/clean_ce_loss": 0.0303955078125, "train/clean_loss_scale": 9.197425842285156, "train/clean_loss_weight": 1.8394851684570312, "train/combined_loss": 0.3354724943637848, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07393364928909953, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2795604169368744 }, { "epoch": 0.7417061611374408, "step": 313, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 9.813612937927246, "train/clean_loss_weight": 1.962722659111023, "train/combined_loss": 0.7820222973823547, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07417061611374408, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6516852378845215 }, { "epoch": 0.7417061611374408, "step": 313, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 6.731002330780029, "train/clean_loss_weight": 1.3462004661560059, "train/combined_loss": 0.5994799137115479, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07417061611374408, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.49956658482551575 }, { "epoch": 0.7417061611374408, "step": 313, "train/alpha": 0.2, "train/clean_ce_loss": 0.130859375, "train/clean_loss_scale": 8.474137306213379, "train/clean_loss_weight": 1.6948274374008179, "train/combined_loss": 1.3307044506072998, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07417061611374408, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1089203357696533 }, { "epoch": 0.7417061611374408, "step": 313, "train/alpha": 0.2, "train/clean_ce_loss": 0.13671875, "train/clean_loss_scale": 7.734029769897461, "train/clean_loss_weight": 1.54680597782135, "train/combined_loss": 1.2688642740249634, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07417061611374408, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.057386875152588 }, { "epoch": 0.7440758293838863, "step": 314, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 8.337129592895508, "train/clean_loss_weight": 1.6674259901046753, "train/combined_loss": 0.6448248624801636, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07440758293838863, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.537354052066803 }, { "epoch": 0.7440758293838863, "step": 314, "train/alpha": 0.2, "train/clean_ce_loss": 0.0634765625, "train/clean_loss_scale": 7.93927001953125, "train/clean_loss_weight": 1.587854027748108, "train/combined_loss": 0.6047490835189819, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07440758293838863, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5039575695991516 }, { "epoch": 0.7440758293838863, "step": 314, "train/alpha": 0.2, "train/clean_ce_loss": 0.05615234375, "train/clean_loss_scale": 7.245141506195068, "train/clean_loss_weight": 1.4490283727645874, "train/combined_loss": 0.48819801211357117, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07440758293838863, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.40683168172836304 }, { "epoch": 0.7440758293838863, "step": 314, "train/alpha": 0.2, "train/clean_ce_loss": 0.02099609375, "train/clean_loss_scale": 18.48902702331543, "train/clean_loss_weight": 3.697805404663086, "train/combined_loss": 0.4658367931842804, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07440758293838863, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3881973326206207 }, { "epoch": 0.7464454976303317, "grad_norm": 20.489961624145508, "learning_rate": 2e-06, "loss": 3.8328, "step": 315 }, { "epoch": 0.7464454976303317, "step": 315, "train/alpha": 0.2, "train/clean_ce_loss": 0.12060546875, "train/clean_loss_scale": 7.8793487548828125, "train/clean_loss_weight": 1.5758697986602783, "train/combined_loss": 1.1403510570526123, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07464454976303317, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9502925276756287 }, { "epoch": 0.7464454976303317, "step": 315, "train/alpha": 0.2, "train/clean_ce_loss": 0.011962890625, "train/clean_loss_scale": 33.7834587097168, "train/clean_loss_weight": 6.756691932678223, "train/combined_loss": 0.4849773943424225, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07464454976303317, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.40414783358573914 }, { "epoch": 0.7464454976303317, "step": 315, "train/alpha": 0.2, "train/clean_ce_loss": 0.1005859375, "train/clean_loss_scale": 6.446770191192627, "train/clean_loss_weight": 1.2893540859222412, "train/combined_loss": 0.7781453132629395, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07464454976303317, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6484544277191162 }, { "epoch": 0.7464454976303317, "step": 315, "train/alpha": 0.2, "train/clean_ce_loss": 0.111328125, "train/clean_loss_scale": 8.125982284545898, "train/clean_loss_weight": 1.6251964569091797, "train/combined_loss": 1.0855804681777954, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07464454976303317, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9046503901481628 }, { "epoch": 0.7488151658767772, "step": 316, "train/alpha": 0.2, "train/clean_ce_loss": 0.0478515625, "train/clean_loss_scale": 8.06873893737793, "train/clean_loss_weight": 1.6137478351593018, "train/combined_loss": 0.46332210302352905, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07488151658767772, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.38610175251960754 }, { "epoch": 0.7488151658767772, "step": 316, "train/alpha": 0.2, "train/clean_ce_loss": 0.1455078125, "train/clean_loss_scale": 6.289878845214844, "train/clean_loss_weight": 1.2579758167266846, "train/combined_loss": 1.0982718467712402, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07488151658767772, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9152265191078186 }, { "epoch": 0.7488151658767772, "step": 316, "train/alpha": 0.2, "train/clean_ce_loss": 0.1376953125, "train/clean_loss_scale": 7.560723304748535, "train/clean_loss_weight": 1.512144684791565, "train/combined_loss": 1.2492914199829102, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07488151658767772, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0410761833190918 }, { "epoch": 0.7488151658767772, "step": 316, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 8.705902099609375, "train/clean_loss_weight": 1.741180419921875, "train/combined_loss": 0.6835492849349976, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07488151658767772, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5696244239807129 }, { "epoch": 0.7511848341232228, "step": 317, "train/alpha": 0.2, "train/clean_ce_loss": 0.049072265625, "train/clean_loss_scale": 9.76330280303955, "train/clean_loss_weight": 1.9526605606079102, "train/combined_loss": 0.574928879737854, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07511848341232227, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4791073799133301 }, { "epoch": 0.7511848341232228, "step": 317, "train/alpha": 0.2, "train/clean_ce_loss": 0.0732421875, "train/clean_loss_scale": 7.397915840148926, "train/clean_loss_weight": 1.4795831441879272, "train/combined_loss": 0.6502074599266052, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07511848341232227, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5418395400047302 }, { "epoch": 0.7511848341232228, "step": 317, "train/alpha": 0.2, "train/clean_ce_loss": 0.0240478515625, "train/clean_loss_scale": 17.762964248657227, "train/clean_loss_weight": 3.5525929927825928, "train/combined_loss": 0.5125933885574341, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07511848341232227, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4271611273288727 }, { "epoch": 0.7511848341232228, "step": 317, "train/alpha": 0.2, "train/clean_ce_loss": 0.0771484375, "train/clean_loss_scale": 7.220791816711426, "train/clean_loss_weight": 1.4441584348678589, "train/combined_loss": 0.6684873700141907, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07511848341232227, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5570728182792664 }, { "epoch": 0.7535545023696683, "step": 318, "train/alpha": 0.2, "train/clean_ce_loss": 0.0162353515625, "train/clean_loss_scale": 23.937795639038086, "train/clean_loss_weight": 4.7875590324401855, "train/combined_loss": 0.4663662314414978, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07535545023696683, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.38863852620124817 }, { "epoch": 0.7535545023696683, "step": 318, "train/alpha": 0.2, "train/clean_ce_loss": 0.1357421875, "train/clean_loss_scale": 9.060428619384766, "train/clean_loss_weight": 1.812085747718811, "train/combined_loss": 1.4758588075637817, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07535545023696683, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2298823595046997 }, { "epoch": 0.7535545023696683, "step": 318, "train/alpha": 0.2, "train/clean_ce_loss": 0.1455078125, "train/clean_loss_scale": 7.829023838043213, "train/clean_loss_weight": 1.5658048391342163, "train/combined_loss": 1.3670209646224976, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07535545023696683, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1391841173171997 }, { "epoch": 0.7535545023696683, "step": 318, "train/alpha": 0.2, "train/clean_ce_loss": 0.0458984375, "train/clean_loss_scale": 8.12070369720459, "train/clean_loss_weight": 1.624140739440918, "train/combined_loss": 0.4472731649875641, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07535545023696683, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.372727632522583 }, { "epoch": 0.7559241706161137, "step": 319, "train/alpha": 0.2, "train/clean_ce_loss": 0.11376953125, "train/clean_loss_scale": 7.970168113708496, "train/clean_loss_weight": 1.5940335988998413, "train/combined_loss": 1.0881147384643555, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07559241706161138, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9067623019218445 }, { "epoch": 0.7559241706161137, "step": 319, "train/alpha": 0.2, "train/clean_ce_loss": 0.078125, "train/clean_loss_scale": 7.810330867767334, "train/clean_loss_weight": 1.5620661973953247, "train/combined_loss": 0.7322185039520264, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07559241706161138, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6101821064949036 }, { "epoch": 0.7559241706161137, "step": 319, "train/alpha": 0.2, "train/clean_ce_loss": 0.169921875, "train/clean_loss_scale": 7.378766059875488, "train/clean_loss_weight": 1.4757531881332397, "train/combined_loss": 1.5045764446258545, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07559241706161138, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2538137435913086 }, { "epoch": 0.7559241706161137, "step": 319, "train/alpha": 0.2, "train/clean_ce_loss": 0.027099609375, "train/clean_loss_scale": 12.481138229370117, "train/clean_loss_weight": 2.496227741241455, "train/combined_loss": 0.40588077902793884, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07559241706161138, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.33823397755622864 }, { "epoch": 0.7582938388625592, "grad_norm": 16.332469940185547, "learning_rate": 2e-06, "loss": 3.6878, "step": 320 }, { "epoch": 0.7582938388625592, "step": 320, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 5.765263080596924, "train/clean_loss_weight": 1.1530526876449585, "train/combined_loss": 0.5033344626426697, "train/pos_loss/pos0": 0.1258726865053177, "train/pos_loss/pos1": 0.2968999147415161, "train/pos_loss/pos2": 0.6498072147369385, "train/pos_loss/pos3": 1.083555817604065, "train/pos_loss/pos4": 1.4599195718765259, "train/pos_loss/pos5": 1.8309874534606934, "train/pos_loss/pos6": 2.4368438720703125, "train/pos_loss/weight0": 2.9120371341705322, "train/pos_loss/weight1": 1.9755854606628418, "train/pos_loss/weight2": 1.164415955543518, "train/pos_loss/weight3": 0.5806435346603394, "train/pos_loss/weight4": 0.24817553162574768, "train/pos_loss/weight5": 0.08645480871200562, "train/pos_loss/weight6": 0.0326872281730175, "train/pos_weight_progress": 0.07582938388625593, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4194453954696655 }, { "epoch": 0.7582938388625592, "step": 320, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 13.348699569702148, "train/clean_loss_weight": 2.6697399616241455, "train/combined_loss": 1.1419395208358765, "train/pos_loss/pos0": 0.2950090169906616, "train/pos_loss/pos1": 0.6647759675979614, "train/pos_loss/pos2": 1.0640857219696045, "train/pos_loss/pos3": 1.8103423118591309, "train/pos_loss/pos4": 2.477663040161133, "train/pos_loss/pos5": 2.5741443634033203, "train/pos_loss/pos6": 2.933292865753174, "train/pos_loss/weight0": 2.376797676086426, "train/pos_loss/weight1": 1.7988977432250977, "train/pos_loss/weight2": 1.2505457401275635, "train/pos_loss/weight3": 0.7982134222984314, "train/pos_loss/weight4": 0.4390619993209839, "train/pos_loss/weight5": 0.214073047041893, "train/pos_loss/weight6": 0.12241007387638092, "train/pos_weight_progress": 0.07582938388625593, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9516162872314453 }, { "epoch": 0.7582938388625592, "step": 320, "train/alpha": 0.2, "train/clean_ce_loss": 0.0966796875, "train/clean_loss_scale": 8.942998886108398, "train/clean_loss_weight": 1.7885998487472534, "train/combined_loss": 1.0375276803970337, "train/pos_loss/pos0": 0.3206855058670044, "train/pos_loss/pos1": 0.693931519985199, "train/pos_loss/pos2": 1.2261412143707275, "train/pos_loss/pos3": 1.6931043863296509, "train/pos_loss/pos4": 2.282724142074585, "train/pos_loss/pos5": 2.898789405822754, "train/pos_loss/pos6": 3.3566558361053467, "train/pos_loss/weight0": 2.70829439163208, "train/pos_loss/weight1": 1.9252104759216309, "train/pos_loss/weight2": 1.214550256729126, "train/pos_loss/weight3": 0.6669637560844421, "train/pos_loss/weight4": 0.310522198677063, "train/pos_loss/weight5": 0.11998005956411362, "train/pos_loss/weight6": 0.054479166865348816, "train/pos_weight_progress": 0.07582938388625593, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8646063804626465 }, { "epoch": 0.7582938388625592, "step": 320, "train/alpha": 0.2, "train/clean_ce_loss": 0.142578125, "train/clean_loss_scale": 10.013713836669922, "train/clean_loss_weight": 2.0027427673339844, "train/combined_loss": 1.7132837772369385, "train/pos_loss/pos0": 0.7008581757545471, "train/pos_loss/pos1": 1.569582462310791, "train/pos_loss/pos2": 1.9322844743728638, "train/pos_loss/pos3": 2.6109719276428223, "train/pos_loss/pos4": 3.2673356533050537, "train/pos_loss/pos5": 3.853543996810913, "train/pos_loss/pos6": 4.138981342315674, "train/pos_loss/weight0": 2.960644245147705, "train/pos_loss/weight1": 2.0002410411834717, "train/pos_loss/weight2": 1.1737697124481201, "train/pos_loss/weight3": 0.5587189793586731, "train/pos_loss/weight4": 0.21921013295650482, "train/pos_loss/weight5": 0.06595098227262497, "train/pos_loss/weight6": 0.021464945748448372, "train/pos_weight_progress": 0.07582938388625593, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.427736520767212 }, { "epoch": 0.7606635071090048, "step": 321, "train/alpha": 0.2, "train/clean_ce_loss": 0.09423828125, "train/clean_loss_scale": 5.592366695404053, "train/clean_loss_weight": 1.1184734106063843, "train/combined_loss": 0.6324180364608765, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07606635071090047, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5270150303840637 }, { "epoch": 0.7606635071090048, "step": 321, "train/alpha": 0.2, "train/clean_ce_loss": 0.107421875, "train/clean_loss_scale": 6.103739261627197, "train/clean_loss_weight": 1.2207478284835815, "train/combined_loss": 0.7868101596832275, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07606635071090047, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6556751132011414 }, { "epoch": 0.7606635071090048, "step": 321, "train/alpha": 0.2, "train/clean_ce_loss": 0.1328125, "train/clean_loss_scale": 4.660192489624023, "train/clean_loss_weight": 0.9320384860038757, "train/combined_loss": 0.7427182197570801, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07606635071090047, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6189318299293518 }, { "epoch": 0.7606635071090048, "step": 321, "train/alpha": 0.2, "train/clean_ce_loss": 0.17578125, "train/clean_loss_scale": 5.588799476623535, "train/clean_loss_weight": 1.1177599430084229, "train/combined_loss": 1.1788874864578247, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07606635071090047, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9824061989784241 }, { "epoch": 0.7630331753554502, "step": 322, "train/alpha": 0.2, "train/clean_ce_loss": 0.03955078125, "train/clean_loss_scale": 4.907861232757568, "train/clean_loss_weight": 0.9815722703933716, "train/combined_loss": 0.2329317033290863, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07630331753554502, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1941097527742386 }, { "epoch": 0.7630331753554502, "step": 322, "train/alpha": 0.2, "train/clean_ce_loss": 0.06982421875, "train/clean_loss_scale": 6.286251544952393, "train/clean_loss_weight": 1.2572503089904785, "train/combined_loss": 0.5267190933227539, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07630331753554502, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4389325976371765 }, { "epoch": 0.7630331753554502, "step": 322, "train/alpha": 0.2, "train/clean_ce_loss": 0.1171875, "train/clean_loss_scale": 6.278142929077148, "train/clean_loss_weight": 1.2556285858154297, "train/combined_loss": 0.8828638195991516, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07630331753554502, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7357198596000671 }, { "epoch": 0.7630331753554502, "step": 322, "train/alpha": 0.2, "train/clean_ce_loss": 0.1171875, "train/clean_loss_scale": 7.428396701812744, "train/clean_loss_weight": 1.4856793880462646, "train/combined_loss": 1.0446182489395142, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07630331753554502, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8705152273178101 }, { "epoch": 0.7654028436018957, "step": 323, "train/alpha": 0.2, "train/clean_ce_loss": 0.1357421875, "train/clean_loss_scale": 6.829977035522461, "train/clean_loss_weight": 1.3659954071044922, "train/combined_loss": 1.112539291381836, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07654028436018957, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9271160364151001 }, { "epoch": 0.7654028436018957, "step": 323, "train/alpha": 0.2, "train/clean_ce_loss": 0.11474609375, "train/clean_loss_scale": 6.546718597412109, "train/clean_loss_weight": 1.309343695640564, "train/combined_loss": 0.9014524817466736, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07654028436018957, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7512103915214539 }, { "epoch": 0.7654028436018957, "step": 323, "train/alpha": 0.2, "train/clean_ce_loss": 0.1376953125, "train/clean_loss_scale": 7.523955345153809, "train/clean_loss_weight": 1.5047911405563354, "train/combined_loss": 1.2432160377502441, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07654028436018957, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0360133647918701 }, { "epoch": 0.7654028436018957, "step": 323, "train/alpha": 0.2, "train/clean_ce_loss": 0.08984375, "train/clean_loss_scale": 6.20271110534668, "train/clean_loss_weight": 1.2405422925949097, "train/combined_loss": 0.6687297821044922, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07654028436018957, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5572748184204102 }, { "epoch": 0.7677725118483413, "step": 324, "train/alpha": 0.2, "train/clean_ce_loss": 0.0057373046875, "train/clean_loss_scale": 33.9740104675293, "train/clean_loss_weight": 6.794802188873291, "train/combined_loss": 0.23390311002731323, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07677725118483412, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.19491925835609436 }, { "epoch": 0.7677725118483413, "step": 324, "train/alpha": 0.2, "train/clean_ce_loss": 0.08447265625, "train/clean_loss_scale": 6.449542045593262, "train/clean_loss_weight": 1.2899084091186523, "train/combined_loss": 0.6537719368934631, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07677725118483412, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5448099374771118 }, { "epoch": 0.7677725118483413, "step": 324, "train/alpha": 0.2, "train/clean_ce_loss": 0.15625, "train/clean_loss_scale": 6.298046588897705, "train/clean_loss_weight": 1.259609341621399, "train/combined_loss": 1.180883765220642, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07677725118483412, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9840697646141052 }, { "epoch": 0.7677725118483413, "step": 324, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 5.73323917388916, "train/clean_loss_weight": 1.1466478109359741, "train/combined_loss": 0.5744436979293823, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07677725118483412, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4787030816078186 }, { "epoch": 0.7701421800947867, "grad_norm": 18.999595642089844, "learning_rate": 2e-06, "loss": 3.8976, "step": 325 }, { "epoch": 0.7701421800947867, "step": 325, "train/alpha": 0.2, "train/clean_ce_loss": 0.1298828125, "train/clean_loss_scale": 5.6764726638793945, "train/clean_loss_weight": 1.1352945566177368, "train/combined_loss": 0.8847315311431885, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07701421800947868, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7372762560844421 }, { "epoch": 0.7701421800947867, "step": 325, "train/alpha": 0.2, "train/clean_ce_loss": 0.16015625, "train/clean_loss_scale": 7.037825584411621, "train/clean_loss_weight": 1.4075651168823242, "train/combined_loss": 1.3525820970535278, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07701421800947868, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1271517276763916 }, { "epoch": 0.7701421800947867, "step": 325, "train/alpha": 0.2, "train/clean_ce_loss": 0.1826171875, "train/clean_loss_scale": 5.485163688659668, "train/clean_loss_weight": 1.0970327854156494, "train/combined_loss": 1.2020221948623657, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07701421800947868, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0016851425170898 }, { "epoch": 0.7701421800947867, "step": 325, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 6.806795120239258, "train/clean_loss_weight": 1.3613590002059937, "train/combined_loss": 0.5424164533615112, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07701421800947868, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4520137310028076 }, { "epoch": 0.7725118483412322, "step": 326, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 5.6917595863342285, "train/clean_loss_weight": 1.1383519172668457, "train/combined_loss": 0.5569475889205933, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07725118483412323, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4641229808330536 }, { "epoch": 0.7725118483412322, "step": 326, "train/alpha": 0.2, "train/clean_ce_loss": 0.01531982421875, "train/clean_loss_scale": 19.766984939575195, "train/clean_loss_weight": 3.953397035598755, "train/combined_loss": 0.36339208483695984, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07725118483412323, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.30282673239707947 }, { "epoch": 0.7725118483412322, "step": 326, "train/alpha": 0.2, "train/clean_ce_loss": 0.05859375, "train/clean_loss_scale": 7.488492488861084, "train/clean_loss_weight": 1.4976985454559326, "train/combined_loss": 0.5265346169471741, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07725118483412323, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4387788474559784 }, { "epoch": 0.7725118483412322, "step": 326, "train/alpha": 0.2, "train/clean_ce_loss": 0.014892578125, "train/clean_loss_scale": 15.299015045166016, "train/clean_loss_weight": 3.059803009033203, "train/combined_loss": 0.2734101414680481, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07725118483412323, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.22784177958965302 }, { "epoch": 0.7748815165876777, "step": 327, "train/alpha": 0.2, "train/clean_ce_loss": 0.1162109375, "train/clean_loss_scale": 7.2375688552856445, "train/clean_loss_weight": 1.4475138187408447, "train/combined_loss": 1.0093016624450684, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07748815165876777, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8410846590995789 }, { "epoch": 0.7748815165876777, "step": 327, "train/alpha": 0.2, "train/clean_ce_loss": 0.037353515625, "train/clean_loss_scale": 7.952373504638672, "train/clean_loss_weight": 1.5904747247695923, "train/combined_loss": 0.3564589321613312, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07748815165876777, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2970491051673889 }, { "epoch": 0.7748815165876777, "step": 327, "train/alpha": 0.2, "train/clean_ce_loss": 0.09521484375, "train/clean_loss_scale": 3.9614410400390625, "train/clean_loss_weight": 0.7922882437705994, "train/combined_loss": 0.4526256024837494, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07748815165876777, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3771879971027374 }, { "epoch": 0.7748815165876777, "step": 327, "train/alpha": 0.2, "train/clean_ce_loss": 0.0556640625, "train/clean_loss_scale": 7.067131519317627, "train/clean_loss_weight": 1.4134262800216675, "train/combined_loss": 0.4720623195171356, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07748815165876777, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.39338526129722595 }, { "epoch": 0.7772511848341233, "step": 328, "train/alpha": 0.2, "train/clean_ce_loss": 0.083984375, "train/clean_loss_scale": 5.463759899139404, "train/clean_loss_weight": 1.0927519798278809, "train/combined_loss": 0.5506445169448853, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07772511848341232, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45887044072151184 }, { "epoch": 0.7772511848341233, "step": 328, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 6.300912380218506, "train/clean_loss_weight": 1.260182499885559, "train/combined_loss": 0.5353314280509949, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07772511848341232, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4461095333099365 }, { "epoch": 0.7772511848341233, "step": 328, "train/alpha": 0.2, "train/clean_ce_loss": 0.04833984375, "train/clean_loss_scale": 6.661466121673584, "train/clean_loss_weight": 1.3322932720184326, "train/combined_loss": 0.38641709089279175, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07772511848341232, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3220142424106598 }, { "epoch": 0.7772511848341233, "step": 328, "train/alpha": 0.2, "train/clean_ce_loss": 0.125, "train/clean_loss_scale": 5.523862838745117, "train/clean_loss_weight": 1.1047725677490234, "train/combined_loss": 0.8285794258117676, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07772511848341232, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6904828548431396 }, { "epoch": 0.7796208530805687, "step": 329, "train/alpha": 0.2, "train/clean_ce_loss": 0.052978515625, "train/clean_loss_scale": 6.914021015167236, "train/clean_loss_weight": 1.382804274559021, "train/combined_loss": 0.43955349922180176, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07796208530805687, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3662945628166199 }, { "epoch": 0.7796208530805687, "step": 329, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 5.239588737487793, "train/clean_loss_weight": 1.0479177236557007, "train/combined_loss": 0.5127019882202148, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07796208530805687, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4272516369819641 }, { "epoch": 0.7796208530805687, "step": 329, "train/alpha": 0.2, "train/clean_ce_loss": 0.17578125, "train/clean_loss_scale": 5.719731330871582, "train/clean_loss_weight": 1.1439462900161743, "train/combined_loss": 1.2065057754516602, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07796208530805687, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.00542151927948 }, { "epoch": 0.7796208530805687, "step": 329, "train/alpha": 0.2, "train/clean_ce_loss": 0.0859375, "train/clean_loss_scale": 5.48164176940918, "train/clean_loss_weight": 1.0963283777236938, "train/combined_loss": 0.5652943253517151, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07796208530805687, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4710786044597626 }, { "epoch": 0.7819905213270142, "grad_norm": 18.39928436279297, "learning_rate": 2e-06, "loss": 3.5332, "step": 330 }, { "epoch": 0.7819905213270142, "step": 330, "train/alpha": 0.2, "train/clean_ce_loss": 0.021484375, "train/clean_loss_scale": 11.13971996307373, "train/clean_loss_weight": 2.2279441356658936, "train/combined_loss": 0.28719592094421387, "train/pos_loss/pos0": 0.05209112539887428, "train/pos_loss/pos1": 0.2324971854686737, "train/pos_loss/pos2": 0.5012861490249634, "train/pos_loss/pos3": 1.1074779033660889, "train/pos_loss/pos4": 1.8166550397872925, "train/pos_loss/pos5": 2.580941677093506, "train/pos_loss/pos6": 3.161827564239502, "train/pos_loss/weight0": 3.553351402282715, "train/pos_loss/weight1": 2.115854024887085, "train/pos_loss/weight2": 0.9140415787696838, "train/pos_loss/weight3": 0.3218204975128174, "train/pos_loss/weight4": 0.0789102241396904, "train/pos_loss/weight5": 0.013889314606785774, "train/pos_loss/weight6": 0.0021327605936676264, "train/pos_weight_progress": 0.07819905213270142, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.23932991921901703 }, { "epoch": 0.7819905213270142, "step": 330, "train/alpha": 0.2, "train/clean_ce_loss": 0.138671875, "train/clean_loss_scale": 5.8811140060424805, "train/clean_loss_weight": 1.176222801208496, "train/combined_loss": 0.9786540865898132, "train/pos_loss/pos0": 0.27097147703170776, "train/pos_loss/pos1": 0.6507612466812134, "train/pos_loss/pos2": 1.0328598022460938, "train/pos_loss/pos3": 1.4454208612442017, "train/pos_loss/pos4": 1.9239252805709839, "train/pos_loss/pos5": 2.5881357192993164, "train/pos_loss/pos6": 2.937577486038208, "train/pos_loss/weight0": 2.4748568534851074, "train/pos_loss/weight1": 1.841626763343811, "train/pos_loss/weight2": 1.2625837326049805, "train/pos_loss/weight3": 0.7895315885543823, "train/pos_loss/weight4": 0.3897264301776886, "train/pos_loss/weight5": 0.16709579527378082, "train/pos_loss/weight6": 0.07457876950502396, "train/pos_weight_progress": 0.07819905213270142, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8155450820922852 }, { "epoch": 0.7819905213270142, "step": 330, "train/alpha": 0.2, "train/clean_ce_loss": 0.06103515625, "train/clean_loss_scale": 3.6397876739501953, "train/clean_loss_weight": 0.727957546710968, "train/combined_loss": 0.2665860056877136, "train/pos_loss/pos0": 0.0029625841416418552, "train/pos_loss/pos1": 0.05358830839395523, "train/pos_loss/pos2": 0.20602864027023315, "train/pos_loss/pos3": 0.6281592845916748, "train/pos_loss/pos4": 1.1779224872589111, "train/pos_loss/pos5": 1.9446567296981812, "train/pos_loss/pos6": 2.4819178581237793, "train/pos_loss/weight0": 2.667422294616699, "train/pos_loss/weight1": 1.9117248058319092, "train/pos_loss/weight2": 1.2226413488388062, "train/pos_loss/weight3": 0.6977509260177612, "train/pos_loss/weight4": 0.3207440972328186, "train/pos_loss/weight5": 0.12423819303512573, "train/pos_loss/weight6": 0.05547846108675003, "train/pos_weight_progress": 0.07819905213270142, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.22215500473976135 }, { "epoch": 0.7819905213270142, "step": 330, "train/alpha": 0.2, "train/clean_ce_loss": 0.053466796875, "train/clean_loss_scale": 16.511703491210938, "train/clean_loss_weight": 3.3023407459259033, "train/combined_loss": 1.0593935251235962, "train/pos_loss/pos0": 0.16477754712104797, "train/pos_loss/pos1": 0.4098471403121948, "train/pos_loss/pos2": 0.8578314185142517, "train/pos_loss/pos3": 1.4207462072372437, "train/pos_loss/pos4": 2.0704305171966553, "train/pos_loss/pos5": 2.6203603744506836, "train/pos_loss/pos6": 2.991610288619995, "train/pos_loss/weight0": 2.0819473266601562, "train/pos_loss/weight1": 1.6278506517410278, "train/pos_loss/weight2": 1.2159650325775146, "train/pos_loss/weight3": 0.891290009021759, "train/pos_loss/weight4": 0.5932852029800415, "train/pos_loss/weight5": 0.3488735854625702, "train/pos_loss/weight6": 0.24078842997550964, "train/pos_weight_progress": 0.07819905213270142, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8828279376029968 }, { "epoch": 0.7843601895734598, "step": 331, "train/alpha": 0.2, "train/clean_ce_loss": 0.0947265625, "train/clean_loss_scale": 12.051273345947266, "train/clean_loss_weight": 2.410254716873169, "train/combined_loss": 1.369890809059143, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07843601895734598, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1415756940841675 }, { "epoch": 0.7843601895734598, "step": 331, "train/alpha": 0.2, "train/clean_ce_loss": 0.01324462890625, "train/clean_loss_scale": 27.83672523498535, "train/clean_loss_weight": 5.567345142364502, "train/combined_loss": 0.4424245059490204, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07843601895734598, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.36868709325790405 }, { "epoch": 0.7843601895734598, "step": 331, "train/alpha": 0.2, "train/clean_ce_loss": 0.142578125, "train/clean_loss_scale": 11.379464149475098, "train/clean_loss_weight": 2.275892972946167, "train/combined_loss": 1.9469552040100098, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07843601895734598, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.6224626302719116 }, { "epoch": 0.7843601895734598, "step": 331, "train/alpha": 0.2, "train/clean_ce_loss": 0.004669189453125, "train/clean_loss_scale": 79.46305847167969, "train/clean_loss_weight": 15.892611503601074, "train/combined_loss": 0.44523367285728455, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07843601895734598, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3710280656814575 }, { "epoch": 0.7867298578199052, "step": 332, "train/alpha": 0.2, "train/clean_ce_loss": 0.1494140625, "train/clean_loss_scale": 9.300321578979492, "train/clean_loss_weight": 1.8600643873214722, "train/combined_loss": 1.6675186157226562, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07867298578199051, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3895988464355469 }, { "epoch": 0.7867298578199052, "step": 332, "train/alpha": 0.2, "train/clean_ce_loss": 0.0869140625, "train/clean_loss_scale": 7.518254280090332, "train/clean_loss_weight": 1.5036509037017822, "train/combined_loss": 0.7841304540634155, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07867298578199051, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6534420251846313 }, { "epoch": 0.7867298578199052, "step": 332, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 6.431471824645996, "train/clean_loss_weight": 1.2862943410873413, "train/combined_loss": 0.5614976286888123, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07867298578199051, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4679147005081177 }, { "epoch": 0.7867298578199052, "step": 332, "train/alpha": 0.2, "train/clean_ce_loss": 0.0908203125, "train/clean_loss_scale": 8.54728889465332, "train/clean_loss_weight": 1.7094577550888062, "train/combined_loss": 0.931520938873291, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07867298578199051, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7762674689292908 }, { "epoch": 0.7890995260663507, "step": 333, "train/alpha": 0.2, "train/clean_ce_loss": 0.0771484375, "train/clean_loss_scale": 9.419917106628418, "train/clean_loss_weight": 1.8839834928512573, "train/combined_loss": 0.8720782995223999, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07890995260663507, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7267318964004517 }, { "epoch": 0.7890995260663507, "step": 333, "train/alpha": 0.2, "train/clean_ce_loss": 0.0225830078125, "train/clean_loss_scale": 17.567420959472656, "train/clean_loss_weight": 3.513484239578247, "train/combined_loss": 0.47607025504112244, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07890995260663507, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.39672520756721497 }, { "epoch": 0.7890995260663507, "step": 333, "train/alpha": 0.2, "train/clean_ce_loss": 0.1044921875, "train/clean_loss_scale": 8.853968620300293, "train/clean_loss_weight": 1.7707937955856323, "train/combined_loss": 1.1102046966552734, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07890995260663507, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9251705408096313 }, { "epoch": 0.7890995260663507, "step": 333, "train/alpha": 0.2, "train/clean_ce_loss": 0.107421875, "train/clean_loss_scale": 11.707330703735352, "train/clean_loss_weight": 2.341466188430786, "train/combined_loss": 1.509148120880127, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07890995260663507, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2576234340667725 }, { "epoch": 0.7914691943127962, "step": 334, "train/alpha": 0.2, "train/clean_ce_loss": 0.0771484375, "train/clean_loss_scale": 8.078749656677246, "train/clean_loss_weight": 1.6157499551773071, "train/combined_loss": 0.7479154467582703, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07914691943127962, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.623262882232666 }, { "epoch": 0.7914691943127962, "step": 334, "train/alpha": 0.2, "train/clean_ce_loss": 0.11279296875, "train/clean_loss_scale": 6.859761714935303, "train/clean_loss_weight": 1.3719524145126343, "train/combined_loss": 0.9284794926643372, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07914691943127962, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7737329006195068 }, { "epoch": 0.7914691943127962, "step": 334, "train/alpha": 0.2, "train/clean_ce_loss": 0.1640625, "train/clean_loss_scale": 8.828496932983398, "train/clean_loss_weight": 1.7656993865966797, "train/combined_loss": 1.7381103038787842, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07914691943127962, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.44842529296875 }, { "epoch": 0.7914691943127962, "step": 334, "train/alpha": 0.2, "train/clean_ce_loss": 0.031982421875, "train/clean_loss_scale": 14.119885444641113, "train/clean_loss_weight": 2.82397723197937, "train/combined_loss": 0.5419057607650757, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07914691943127962, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45158812403678894 }, { "epoch": 0.7938388625592417, "grad_norm": 15.721272468566895, "learning_rate": 2e-06, "loss": 3.7774, "step": 335 }, { "epoch": 0.7938388625592417, "step": 335, "train/alpha": 0.2, "train/clean_ce_loss": 0.09130859375, "train/clean_loss_scale": 7.307189464569092, "train/clean_loss_weight": 1.4614379405975342, "train/combined_loss": 0.8006510734558105, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07938388625592417, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6672092080116272 }, { "epoch": 0.7938388625592417, "step": 335, "train/alpha": 0.2, "train/clean_ce_loss": 0.169921875, "train/clean_loss_scale": 6.57762336730957, "train/clean_loss_weight": 1.315524697303772, "train/combined_loss": 1.3412184715270996, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07938388625592417, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1176820993423462 }, { "epoch": 0.7938388625592417, "step": 335, "train/alpha": 0.2, "train/clean_ce_loss": 0.08251953125, "train/clean_loss_scale": 6.9232916831970215, "train/clean_loss_weight": 1.3846583366394043, "train/combined_loss": 0.68556809425354, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07938388625592417, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5713067650794983 }, { "epoch": 0.7938388625592417, "step": 335, "train/alpha": 0.2, "train/clean_ce_loss": 0.0634765625, "train/clean_loss_scale": 7.854705333709717, "train/clean_loss_weight": 1.5709410905838013, "train/combined_loss": 0.5983076095581055, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07938388625592417, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4985896944999695 }, { "epoch": 0.7962085308056872, "step": 336, "train/alpha": 0.2, "train/clean_ce_loss": 0.166015625, "train/clean_loss_scale": 8.837276458740234, "train/clean_loss_weight": 1.7674553394317627, "train/combined_loss": 1.7605512142181396, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07962085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4671260118484497 }, { "epoch": 0.7962085308056872, "step": 336, "train/alpha": 0.2, "train/clean_ce_loss": 0.087890625, "train/clean_loss_scale": 8.588492393493652, "train/clean_loss_weight": 1.7176984548568726, "train/combined_loss": 0.9058176279067993, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07962085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7548480033874512 }, { "epoch": 0.7962085308056872, "step": 336, "train/alpha": 0.2, "train/clean_ce_loss": 0.09765625, "train/clean_loss_scale": 8.203836441040039, "train/clean_loss_weight": 1.6407673358917236, "train/combined_loss": 0.961387038230896, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07962085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8011558651924133 }, { "epoch": 0.7962085308056872, "step": 336, "train/alpha": 0.2, "train/clean_ce_loss": 0.177734375, "train/clean_loss_scale": 7.080219745635986, "train/clean_loss_weight": 1.416043996810913, "train/combined_loss": 1.5100780725479126, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07962085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.258398413658142 }, { "epoch": 0.7985781990521327, "step": 337, "train/alpha": 0.2, "train/clean_ce_loss": 0.0028533935546875, "train/clean_loss_scale": 117.07320404052734, "train/clean_loss_weight": 23.414640426635742, "train/combined_loss": 0.4008671045303345, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07985781990521328, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.33405593037605286 }, { "epoch": 0.7985781990521327, "step": 337, "train/alpha": 0.2, "train/clean_ce_loss": 0.08935546875, "train/clean_loss_scale": 9.684809684753418, "train/clean_loss_weight": 1.9369620084762573, "train/combined_loss": 1.038468837738037, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07985781990521328, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8653907179832458 }, { "epoch": 0.7985781990521327, "step": 337, "train/alpha": 0.2, "train/clean_ce_loss": 0.054443359375, "train/clean_loss_scale": 9.03731918334961, "train/clean_loss_weight": 1.8074638843536377, "train/combined_loss": 0.5904264450073242, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07985781990521328, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4920220375061035 }, { "epoch": 0.7985781990521327, "step": 337, "train/alpha": 0.2, "train/clean_ce_loss": 0.0255126953125, "train/clean_loss_scale": 15.112346649169922, "train/clean_loss_weight": 3.0224692821502686, "train/combined_loss": 0.4626680314540863, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.07985781990521328, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.385556697845459 }, { "epoch": 0.8009478672985783, "step": 338, "train/alpha": 0.2, "train/clean_ce_loss": 0.04052734375, "train/clean_loss_scale": 8.15917682647705, "train/clean_loss_weight": 1.6318353414535522, "train/combined_loss": 0.39680370688438416, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.08009478672985781, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3306697607040405 }, { "epoch": 0.8009478672985783, "step": 338, "train/alpha": 0.2, "train/clean_ce_loss": 0.006500244140625, "train/clean_loss_scale": 46.411231994628906, "train/clean_loss_weight": 9.282246589660645, "train/combined_loss": 0.36202120780944824, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.08009478672985781, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.30168434977531433 }, { "epoch": 0.8009478672985783, "step": 338, "train/alpha": 0.2, "train/clean_ce_loss": 0.0322265625, "train/clean_loss_scale": 21.37076187133789, "train/clean_loss_weight": 4.2741522789001465, "train/combined_loss": 0.8264474868774414, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.08009478672985781, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6887062191963196 }, { "epoch": 0.8009478672985783, "step": 338, "train/alpha": 0.2, "train/clean_ce_loss": 0.1083984375, "train/clean_loss_scale": 9.22957706451416, "train/clean_loss_weight": 1.84591543674469, "train/combined_loss": 1.200566053390503, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.08009478672985781, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0004717111587524 }, { "epoch": 0.8033175355450237, "step": 339, "train/alpha": 0.2, "train/clean_ce_loss": 0.04736328125, "train/clean_loss_scale": 6.499338626861572, "train/clean_loss_weight": 1.2998677492141724, "train/combined_loss": 0.36939600110054016, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.08033175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.30783000588417053 }, { "epoch": 0.8033175355450237, "step": 339, "train/alpha": 0.2, "train/clean_ce_loss": 0.12060546875, "train/clean_loss_scale": 9.38276195526123, "train/clean_loss_weight": 1.8765524625778198, "train/combined_loss": 1.3579349517822266, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.08033175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1316124200820923 }, { "epoch": 0.8033175355450237, "step": 339, "train/alpha": 0.2, "train/clean_ce_loss": 0.04931640625, "train/clean_loss_scale": 9.91741943359375, "train/clean_loss_weight": 1.983483910560608, "train/combined_loss": 0.5869097709655762, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.08033175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4890914857387543 }, { "epoch": 0.8033175355450237, "step": 339, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 7.223421573638916, "train/clean_loss_weight": 1.444684386253357, "train/combined_loss": 0.7068231105804443, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.08033175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.589019238948822 }, { "epoch": 0.8056872037914692, "grad_norm": 13.165879249572754, "learning_rate": 2e-06, "loss": 3.7724, "step": 340 }, { "epoch": 0.8056872037914692, "step": 340, "train/alpha": 0.2, "train/clean_ce_loss": 0.0771484375, "train/clean_loss_scale": 8.525568008422852, "train/clean_loss_weight": 1.7051136493682861, "train/combined_loss": 0.7892811298370361, "train/pos_loss/pos0": 0.2787603735923767, "train/pos_loss/pos1": 0.46864163875579834, "train/pos_loss/pos2": 0.9025201797485352, "train/pos_loss/pos3": 1.3493479490280151, "train/pos_loss/pos4": 1.6870027780532837, "train/pos_loss/pos5": 2.032604455947876, "train/pos_loss/pos6": 2.4964516162872314, "train/pos_loss/weight0": 2.657118797302246, "train/pos_loss/weight1": 1.9104924201965332, "train/pos_loss/weight2": 1.2442808151245117, "train/pos_loss/weight3": 0.7043412327766418, "train/pos_loss/weight4": 0.31627047061920166, "train/pos_loss/weight5": 0.11939558386802673, "train/pos_loss/weight6": 0.04810084402561188, "train/pos_weight_progress": 0.08056872037914692, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6577342748641968 }, { "epoch": 0.8056872037914692, "step": 340, "train/alpha": 0.2, "train/clean_ce_loss": 0.13671875, "train/clean_loss_scale": 8.156475067138672, "train/clean_loss_weight": 1.631295084953308, "train/combined_loss": 1.3381717205047607, "train/pos_loss/pos0": 0.45743727684020996, "train/pos_loss/pos1": 0.837609589099884, "train/pos_loss/pos2": 1.3531248569488525, "train/pos_loss/pos3": 2.264237403869629, "train/pos_loss/pos4": 2.5458335876464844, "train/pos_loss/pos5": 3.196343421936035, "train/pos_loss/pos6": 3.526384115219116, "train/pos_loss/weight0": 2.5933682918548584, "train/pos_loss/weight1": 1.8331600427627563, "train/pos_loss/weight2": 1.2172186374664307, "train/pos_loss/weight3": 0.728173553943634, "train/pos_loss/weight4": 0.3720749616622925, "train/pos_loss/weight5": 0.16963228583335876, "train/pos_loss/weight6": 0.086372010409832, "train/pos_weight_progress": 0.08056872037914692, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.115143060684204 }, { "epoch": 0.8056872037914692, "step": 340, "train/alpha": 0.2, "train/clean_ce_loss": 0.10400390625, "train/clean_loss_scale": 7.09254264831543, "train/clean_loss_weight": 1.418508529663086, "train/combined_loss": 0.8851825594902039, "train/pos_loss/pos0": 0.3461715281009674, "train/pos_loss/pos1": 0.7534334659576416, "train/pos_loss/pos2": 1.1127371788024902, "train/pos_loss/pos3": 1.7207239866256714, "train/pos_loss/pos4": 2.2525556087493896, "train/pos_loss/pos5": 2.758002519607544, "train/pos_loss/pos6": 3.1727123260498047, "train/pos_loss/weight0": 3.2019829750061035, "train/pos_loss/weight1": 2.012632369995117, "train/pos_loss/weight2": 1.1003165245056152, "train/pos_loss/weight3": 0.48199644684791565, "train/pos_loss/weight4": 0.15157045423984528, "train/pos_loss/weight5": 0.04064080864191055, "train/pos_loss/weight6": 0.010860755108296871, "train/pos_weight_progress": 0.08056872037914692, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7376521229743958 }, { "epoch": 0.8056872037914692, "step": 340, "train/alpha": 0.2, "train/clean_ce_loss": 0.037353515625, "train/clean_loss_scale": 9.728797912597656, "train/clean_loss_weight": 1.945759654045105, "train/combined_loss": 0.4360857605934143, "train/pos_loss/pos0": 0.10688037425279617, "train/pos_loss/pos1": 0.2781110107898712, "train/pos_loss/pos2": 0.561856210231781, "train/pos_loss/pos3": 0.8730675578117371, "train/pos_loss/pos4": 1.3350566625595093, "train/pos_loss/pos5": 1.8289841413497925, "train/pos_loss/pos6": 2.1225924491882324, "train/pos_loss/weight0": 2.9112772941589355, "train/pos_loss/weight1": 1.9924132823944092, "train/pos_loss/weight2": 1.1706433296203613, "train/pos_loss/weight3": 0.5877605080604553, "train/pos_loss/weight4": 0.23626591265201569, "train/pos_loss/weight5": 0.07573952525854111, "train/pos_loss/weight6": 0.025899630039930344, "train/pos_weight_progress": 0.08056872037914692, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3634048104286194 }, { "epoch": 0.8080568720379147, "step": 341, "train/alpha": 0.2, "train/clean_ce_loss": 0.09716796875, "train/clean_loss_scale": 8.069129943847656, "train/clean_loss_weight": 1.613826036453247, "train/combined_loss": 0.9408731460571289, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08080568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7840609550476074 }, { "epoch": 0.8080568720379147, "step": 341, "train/alpha": 0.2, "train/clean_ce_loss": 0.095703125, "train/clean_loss_scale": 9.059085845947266, "train/clean_loss_weight": 1.8118171691894531, "train/combined_loss": 1.0403794050216675, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08080568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8669828176498413 }, { "epoch": 0.8080568720379147, "step": 341, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 9.197832107543945, "train/clean_loss_weight": 1.8395664691925049, "train/combined_loss": 0.7868457436561584, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08080568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6557047963142395 }, { "epoch": 0.8080568720379147, "step": 341, "train/alpha": 0.2, "train/clean_ce_loss": 0.078125, "train/clean_loss_scale": 9.808980941772461, "train/clean_loss_weight": 1.9617961645126343, "train/combined_loss": 0.9195919036865234, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08080568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7663266062736511 }, { "epoch": 0.8104265402843602, "step": 342, "train/alpha": 0.2, "train/clean_ce_loss": 0.0966796875, "train/clean_loss_scale": 14.626935005187988, "train/clean_loss_weight": 2.925387144088745, "train/combined_loss": 1.6969530582427979, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08104265402843602, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4141274690628052 }, { "epoch": 0.8104265402843602, "step": 342, "train/alpha": 0.2, "train/clean_ce_loss": 0.072265625, "train/clean_loss_scale": 10.630414009094238, "train/clean_loss_weight": 2.1260828971862793, "train/combined_loss": 0.9218562245368958, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08104265402843602, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7682135105133057 }, { "epoch": 0.8104265402843602, "step": 342, "train/alpha": 0.2, "train/clean_ce_loss": 0.126953125, "train/clean_loss_scale": 9.773648262023926, "train/clean_loss_weight": 1.954729676246643, "train/combined_loss": 1.4889541864395142, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08104265402843602, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2407951354980469 }, { "epoch": 0.8104265402843602, "step": 342, "train/alpha": 0.2, "train/clean_ce_loss": 0.115234375, "train/clean_loss_scale": 12.032944679260254, "train/clean_loss_weight": 2.4065890312194824, "train/combined_loss": 1.6639306545257568, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08104265402843602, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3866088390350342 }, { "epoch": 0.8127962085308057, "step": 343, "train/alpha": 0.2, "train/clean_ce_loss": 0.0615234375, "train/clean_loss_scale": 11.378132820129395, "train/clean_loss_weight": 2.2756266593933105, "train/combined_loss": 0.8400262594223022, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08127962085308058, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7000218629837036 }, { "epoch": 0.8127962085308057, "step": 343, "train/alpha": 0.2, "train/clean_ce_loss": 0.19140625, "train/clean_loss_scale": 7.8148512840271, "train/clean_loss_weight": 1.5629702806472778, "train/combined_loss": 1.794973611831665, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08127962085308058, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4958113431930542 }, { "epoch": 0.8127962085308057, "step": 343, "train/alpha": 0.2, "train/clean_ce_loss": 0.1357421875, "train/clean_loss_scale": 10.358719825744629, "train/clean_loss_weight": 2.071743965148926, "train/combined_loss": 1.6873383522033691, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08127962085308058, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4061152935028076 }, { "epoch": 0.8127962085308057, "step": 343, "train/alpha": 0.2, "train/clean_ce_loss": 0.11474609375, "train/clean_loss_scale": 10.649310111999512, "train/clean_loss_weight": 2.129862070083618, "train/combined_loss": 1.466360092163086, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08127962085308058, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2219667434692383 }, { "epoch": 0.8151658767772512, "step": 344, "train/alpha": 0.2, "train/clean_ce_loss": 0.057373046875, "train/clean_loss_scale": 14.031458854675293, "train/clean_loss_weight": 2.8062918186187744, "train/combined_loss": 0.9660330414772034, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08151658767772511, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8050275444984436 }, { "epoch": 0.8151658767772512, "step": 344, "train/alpha": 0.2, "train/clean_ce_loss": 0.1748046875, "train/clean_loss_scale": 9.133411407470703, "train/clean_loss_weight": 1.8266823291778564, "train/combined_loss": 1.9158756732940674, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08151658767772511, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5965631008148193 }, { "epoch": 0.8151658767772512, "step": 344, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 10.503463745117188, "train/clean_loss_weight": 2.1006927490234375, "train/combined_loss": 0.9170016050338745, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08151658767772511, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7641680240631104 }, { "epoch": 0.8151658767772512, "step": 344, "train/alpha": 0.2, "train/clean_ce_loss": 0.06982421875, "train/clean_loss_scale": 9.851866722106934, "train/clean_loss_weight": 1.9703733921051025, "train/combined_loss": 0.8254786729812622, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08151658767772511, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6878988742828369 }, { "epoch": 0.8175355450236966, "grad_norm": 25.309282302856445, "learning_rate": 2e-06, "loss": 3.6919, "step": 345 }, { "epoch": 0.8175355450236966, "step": 345, "train/alpha": 0.2, "train/clean_ce_loss": 0.1123046875, "train/clean_loss_scale": 9.900096893310547, "train/clean_loss_weight": 1.980019450187683, "train/combined_loss": 1.3341927528381348, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08175355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1118272542953491 }, { "epoch": 0.8175355450236966, "step": 345, "train/alpha": 0.2, "train/clean_ce_loss": 0.078125, "train/clean_loss_scale": 10.532340049743652, "train/clean_loss_weight": 2.1064679622650146, "train/combined_loss": 0.9874069094657898, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08175355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.822839081287384 }, { "epoch": 0.8175355450236966, "step": 345, "train/alpha": 0.2, "train/clean_ce_loss": 0.061279296875, "train/clean_loss_scale": 10.692790985107422, "train/clean_loss_weight": 2.1385581493377686, "train/combined_loss": 0.7862960696220398, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08175355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6552467346191406 }, { "epoch": 0.8175355450236966, "step": 345, "train/alpha": 0.2, "train/clean_ce_loss": 0.12353515625, "train/clean_loss_scale": 11.81685733795166, "train/clean_loss_weight": 2.3633716106414795, "train/combined_loss": 1.7517567873001099, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08175355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4597972631454468 }, { "epoch": 0.8199052132701422, "step": 346, "train/alpha": 0.2, "train/clean_ce_loss": 0.09423828125, "train/clean_loss_scale": 10.345868110656738, "train/clean_loss_weight": 2.069173574447632, "train/combined_loss": 1.1699721813201904, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08199052132701422, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9749768376350403 }, { "epoch": 0.8199052132701422, "step": 346, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 11.363945007324219, "train/clean_loss_weight": 2.2727890014648438, "train/combined_loss": 0.9055643081665039, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08199052132701422, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7546369433403015 }, { "epoch": 0.8199052132701422, "step": 346, "train/alpha": 0.2, "train/clean_ce_loss": 0.171875, "train/clean_loss_scale": 10.242264747619629, "train/clean_loss_weight": 2.0484530925750732, "train/combined_loss": 2.1124672889709473, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08199052132701422, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.7603893280029297 }, { "epoch": 0.8199052132701422, "step": 346, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 9.3140869140625, "train/clean_loss_weight": 1.862817406654358, "train/combined_loss": 0.8841105699539185, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08199052132701422, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.736758828163147 }, { "epoch": 0.8222748815165877, "step": 347, "train/alpha": 0.2, "train/clean_ce_loss": 0.16015625, "train/clean_loss_scale": 9.781758308410645, "train/clean_loss_weight": 1.956351637840271, "train/combined_loss": 1.8799316883087158, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08222748815165877, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5666097402572632 }, { "epoch": 0.8222748815165877, "step": 347, "train/alpha": 0.2, "train/clean_ce_loss": 0.11279296875, "train/clean_loss_scale": 10.50886344909668, "train/clean_loss_weight": 2.1017727851867676, "train/combined_loss": 1.4223910570144653, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08222748815165877, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1853258609771729 }, { "epoch": 0.8222748815165877, "step": 347, "train/alpha": 0.2, "train/clean_ce_loss": 0.1201171875, "train/clean_loss_scale": 8.08836841583252, "train/clean_loss_weight": 1.6176737546920776, "train/combined_loss": 1.1658625602722168, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08222748815165877, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9715520739555359 }, { "epoch": 0.8222748815165877, "step": 347, "train/alpha": 0.2, "train/clean_ce_loss": 0.177734375, "train/clean_loss_scale": 8.495943069458008, "train/clean_loss_weight": 1.6991885900497437, "train/combined_loss": 1.8120254278182983, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08222748815165877, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5100212097167969 }, { "epoch": 0.8246445497630331, "step": 348, "train/alpha": 0.2, "train/clean_ce_loss": 0.039794921875, "train/clean_loss_scale": 15.474777221679688, "train/clean_loss_weight": 3.0949554443359375, "train/combined_loss": 0.7389810681343079, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08246445497630332, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6158175468444824 }, { "epoch": 0.8246445497630331, "step": 348, "train/alpha": 0.2, "train/clean_ce_loss": 0.1474609375, "train/clean_loss_scale": 11.822057723999023, "train/clean_loss_weight": 2.3644115924835205, "train/combined_loss": 2.0919501781463623, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08246445497630332, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.7432917356491089 }, { "epoch": 0.8246445497630331, "step": 348, "train/alpha": 0.2, "train/clean_ce_loss": 0.00689697265625, "train/clean_loss_scale": 85.54460906982422, "train/clean_loss_weight": 17.108922958374023, "train/combined_loss": 0.7079986333847046, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08246445497630332, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5899988412857056 }, { "epoch": 0.8246445497630331, "step": 348, "train/alpha": 0.2, "train/clean_ce_loss": 0.024169921875, "train/clean_loss_scale": 22.517770767211914, "train/clean_loss_weight": 4.503554344177246, "train/combined_loss": 0.6531032919883728, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08246445497630332, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5442527532577515 }, { "epoch": 0.8270142180094787, "step": 349, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 9.815658569335938, "train/clean_loss_weight": 1.9631317853927612, "train/combined_loss": 1.029493808746338, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08270142180094786, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8579115271568298 }, { "epoch": 0.8270142180094787, "step": 349, "train/alpha": 0.2, "train/clean_ce_loss": 0.0498046875, "train/clean_loss_scale": 9.72938346862793, "train/clean_loss_weight": 1.9458767175674438, "train/combined_loss": 0.5814826488494873, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08270142180094786, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.48456889390945435 }, { "epoch": 0.8270142180094787, "step": 349, "train/alpha": 0.2, "train/clean_ce_loss": 0.0223388671875, "train/clean_loss_scale": 23.27654266357422, "train/clean_loss_weight": 4.655308723449707, "train/combined_loss": 0.6239659190177917, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08270142180094786, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5199716091156006 }, { "epoch": 0.8270142180094787, "step": 349, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 8.932186126708984, "train/clean_loss_weight": 1.7864372730255127, "train/combined_loss": 0.8530935049057007, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08270142180094786, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7109112739562988 }, { "epoch": 0.8293838862559242, "grad_norm": 15.231898307800293, "learning_rate": 2e-06, "loss": 3.702, "step": 350 }, { "epoch": 0.8293838862559242, "step": 350, "train/alpha": 0.2, "train/clean_ce_loss": 0.1103515625, "train/clean_loss_scale": 8.322184562683105, "train/clean_loss_weight": 1.664436936378479, "train/combined_loss": 1.1020393371582031, "train/pos_loss/pos0": 0.3185139298439026, "train/pos_loss/pos1": 0.7091562747955322, "train/pos_loss/pos2": 0.9796954393386841, "train/pos_loss/pos3": 1.3999907970428467, "train/pos_loss/pos4": 1.8587783575057983, "train/pos_loss/pos5": 2.2916436195373535, "train/pos_loss/pos6": 2.7335052490234375, "train/pos_loss/weight0": 2.227724313735962, "train/pos_loss/weight1": 1.7014060020446777, "train/pos_loss/weight2": 1.2453627586364746, "train/pos_loss/weight3": 0.8361098766326904, "train/pos_loss/weight4": 0.5152076482772827, "train/pos_loss/weight5": 0.29734304547309875, "train/pos_loss/weight6": 0.1768464744091034, "train/pos_weight_progress": 0.08293838862559241, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9183660745620728 }, { "epoch": 0.8293838862559242, "step": 350, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 7.214352130889893, "train/clean_loss_weight": 1.4428704977035522, "train/combined_loss": 0.7566615343093872, "train/pos_loss/pos0": 0.21404001116752625, "train/pos_loss/pos1": 0.5050605535507202, "train/pos_loss/pos2": 0.8718189597129822, "train/pos_loss/pos3": 1.3857232332229614, "train/pos_loss/pos4": 1.8628768920898438, "train/pos_loss/pos5": 2.41675066947937, "train/pos_loss/pos6": 2.9468066692352295, "train/pos_loss/weight0": 2.8026788234710693, "train/pos_loss/weight1": 1.9339439868927002, "train/pos_loss/weight2": 1.194673776626587, "train/pos_loss/weight3": 0.6349119544029236, "train/pos_loss/weight4": 0.279238760471344, "train/pos_loss/weight5": 0.10858414322137833, "train/pos_loss/weight6": 0.04596838727593422, "train/pos_weight_progress": 0.08293838862559241, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.630551278591156 }, { "epoch": 0.8293838862559242, "step": 350, "train/alpha": 0.2, "train/clean_ce_loss": 0.06689453125, "train/clean_loss_scale": 8.466883659362793, "train/clean_loss_weight": 1.6933767795562744, "train/combined_loss": 0.6796658039093018, "train/pos_loss/pos0": 0.21341484785079956, "train/pos_loss/pos1": 0.4880157709121704, "train/pos_loss/pos2": 0.7076008915901184, "train/pos_loss/pos3": 1.1113255023956299, "train/pos_loss/pos4": 1.444210171699524, "train/pos_loss/pos5": 1.7877086400985718, "train/pos_loss/pos6": 2.239912748336792, "train/pos_loss/weight0": 2.6686606407165527, "train/pos_loss/weight1": 1.8989589214324951, "train/pos_loss/weight2": 1.2271645069122314, "train/pos_loss/weight3": 0.6916490197181702, "train/pos_loss/weight4": 0.3237761855125427, "train/pos_loss/weight5": 0.13351371884346008, "train/pos_loss/weight6": 0.0562770701944828, "train/pos_weight_progress": 0.08293838862559241, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5663881897926331 }, { "epoch": 0.8293838862559242, "step": 350, "train/alpha": 0.2, "train/clean_ce_loss": 0.12158203125, "train/clean_loss_scale": 7.342807292938232, "train/clean_loss_weight": 1.4685615301132202, "train/combined_loss": 1.0713040828704834, "train/pos_loss/pos0": 0.3427468240261078, "train/pos_loss/pos1": 0.6502020359039307, "train/pos_loss/pos2": 1.0281562805175781, "train/pos_loss/pos3": 1.5259690284729004, "train/pos_loss/pos4": 2.0149636268615723, "train/pos_loss/pos5": 2.4459986686706543, "train/pos_loss/pos6": 2.8053970336914062, "train/pos_loss/weight0": 2.4019532203674316, "train/pos_loss/weight1": 1.7949415445327759, "train/pos_loss/weight2": 1.2278317213058472, "train/pos_loss/weight3": 0.7757798433303833, "train/pos_loss/weight4": 0.4375098645687103, "train/pos_loss/weight5": 0.23033174872398376, "train/pos_loss/weight6": 0.13165165483951569, "train/pos_weight_progress": 0.08293838862559241, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8927534222602844 }, { "epoch": 0.8317535545023697, "step": 351, "train/alpha": 0.2, "train/clean_ce_loss": 0.0203857421875, "train/clean_loss_scale": 15.081388473510742, "train/clean_loss_weight": 3.01627779006958, "train/combined_loss": 0.3689343333244324, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08317535545023697, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3074452877044678 }, { "epoch": 0.8317535545023697, "step": 351, "train/alpha": 0.2, "train/clean_ce_loss": 0.09228515625, "train/clean_loss_scale": 10.220094680786133, "train/clean_loss_weight": 2.0440189838409424, "train/combined_loss": 1.1317956447601318, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08317535545023697, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9431630373001099 }, { "epoch": 0.8317535545023697, "step": 351, "train/alpha": 0.2, "train/clean_ce_loss": 0.0771484375, "train/clean_loss_scale": 7.9303154945373535, "train/clean_loss_weight": 1.5860631465911865, "train/combined_loss": 0.7341737747192383, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08317535545023697, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6118114590644836 }, { "epoch": 0.8317535545023697, "step": 351, "train/alpha": 0.2, "train/clean_ce_loss": 0.083984375, "train/clean_loss_scale": 8.410368919372559, "train/clean_loss_weight": 1.6820738315582275, "train/combined_loss": 0.8476074934005737, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08317535545023697, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7063395380973816 }, { "epoch": 0.8341232227488151, "step": 352, "train/alpha": 0.2, "train/clean_ce_loss": 0.1416015625, "train/clean_loss_scale": 8.697391510009766, "train/clean_loss_weight": 1.739478349685669, "train/combined_loss": 1.477877140045166, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08341232227488152, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2315642833709717 }, { "epoch": 0.8341232227488151, "step": 352, "train/alpha": 0.2, "train/clean_ce_loss": 0.07568359375, "train/clean_loss_scale": 8.633882522583008, "train/clean_loss_weight": 1.7267764806747437, "train/combined_loss": 0.7841319441795349, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08341232227488152, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6534432768821716 }, { "epoch": 0.8341232227488151, "step": 352, "train/alpha": 0.2, "train/clean_ce_loss": 0.01153564453125, "train/clean_loss_scale": 33.480628967285156, "train/clean_loss_weight": 6.6961259841918945, "train/combined_loss": 0.46346476674079895, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08341232227488152, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.38622063398361206 }, { "epoch": 0.8341232227488151, "step": 352, "train/alpha": 0.2, "train/clean_ce_loss": 0.1513671875, "train/clean_loss_scale": 6.313776969909668, "train/clean_loss_weight": 1.2627553939819336, "train/combined_loss": 1.1468384265899658, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08341232227488152, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9556986689567566 }, { "epoch": 0.8364928909952607, "step": 353, "train/alpha": 0.2, "train/clean_ce_loss": 0.1357421875, "train/clean_loss_scale": 8.065844535827637, "train/clean_loss_weight": 1.6131689548492432, "train/combined_loss": 1.3138504028320312, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08364928909952607, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0948753356933594 }, { "epoch": 0.8364928909952607, "step": 353, "train/alpha": 0.2, "train/clean_ce_loss": 0.1279296875, "train/clean_loss_scale": 11.525688171386719, "train/clean_loss_weight": 2.3051376342773438, "train/combined_loss": 1.7693731784820557, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08364928909952607, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4744776487350464 }, { "epoch": 0.8364928909952607, "step": 353, "train/alpha": 0.2, "train/clean_ce_loss": 0.1298828125, "train/clean_loss_scale": 9.016003608703613, "train/clean_loss_weight": 1.8032007217407227, "train/combined_loss": 1.4052287340164185, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08364928909952607, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.171023964881897 }, { "epoch": 0.8364928909952607, "step": 353, "train/alpha": 0.2, "train/clean_ce_loss": 0.0194091796875, "train/clean_loss_scale": 17.698530197143555, "train/clean_loss_weight": 3.539705991744995, "train/combined_loss": 0.4122167229652405, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08364928909952607, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.34351393580436707 }, { "epoch": 0.8388625592417062, "step": 354, "train/alpha": 0.2, "train/clean_ce_loss": 0.18359375, "train/clean_loss_scale": 7.0603532791137695, "train/clean_loss_weight": 1.412070631980896, "train/combined_loss": 1.5554840564727783, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08388625592417062, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2962367534637451 }, { "epoch": 0.8388625592417062, "step": 354, "train/alpha": 0.2, "train/clean_ce_loss": 0.1416015625, "train/clean_loss_scale": 6.731838703155518, "train/clean_loss_weight": 1.3463677167892456, "train/combined_loss": 1.143886685371399, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08388625592417062, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9532389044761658 }, { "epoch": 0.8388625592417062, "step": 354, "train/alpha": 0.2, "train/clean_ce_loss": 0.126953125, "train/clean_loss_scale": 9.050899505615234, "train/clean_loss_weight": 1.8101799488067627, "train/combined_loss": 1.3788479566574097, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08388625592417062, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1490399837493896 }, { "epoch": 0.8388625592417062, "step": 354, "train/alpha": 0.2, "train/clean_ce_loss": 0.0576171875, "train/clean_loss_scale": 7.277076244354248, "train/clean_loss_weight": 1.4554152488708496, "train/combined_loss": 0.5031415820121765, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08388625592417062, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4192846715450287 }, { "epoch": 0.8412322274881516, "grad_norm": 14.042078971862793, "learning_rate": 2e-06, "loss": 3.5598, "step": 355 }, { "epoch": 0.8412322274881516, "step": 355, "train/alpha": 0.2, "train/clean_ce_loss": 0.1240234375, "train/clean_loss_scale": 10.962647438049316, "train/clean_loss_weight": 2.1925294399261475, "train/combined_loss": 1.6315503120422363, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08412322274881516, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.359625220298767 }, { "epoch": 0.8412322274881516, "step": 355, "train/alpha": 0.2, "train/clean_ce_loss": 0.003662109375, "train/clean_loss_scale": 62.58448791503906, "train/clean_loss_weight": 12.516898155212402, "train/combined_loss": 0.2750294804573059, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08412322274881516, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.22919124364852905 }, { "epoch": 0.8412322274881516, "step": 355, "train/alpha": 0.2, "train/clean_ce_loss": 0.15234375, "train/clean_loss_scale": 7.841347694396973, "train/clean_loss_weight": 1.5682696104049683, "train/combined_loss": 1.433496356010437, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08412322274881516, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.194580316543579 }, { "epoch": 0.8412322274881516, "step": 355, "train/alpha": 0.2, "train/clean_ce_loss": 0.1552734375, "train/clean_loss_scale": 9.47116470336914, "train/clean_loss_weight": 1.894232988357544, "train/combined_loss": 1.7647442817687988, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08412322274881516, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4706202745437622 }, { "epoch": 0.8436018957345972, "step": 356, "train/alpha": 0.2, "train/clean_ce_loss": 0.09619140625, "train/clean_loss_scale": 6.878085613250732, "train/clean_loss_weight": 1.3756171464920044, "train/combined_loss": 0.7939352989196777, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08436018957345971, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6616127490997314 }, { "epoch": 0.8436018957345972, "step": 356, "train/alpha": 0.2, "train/clean_ce_loss": 0.1142578125, "train/clean_loss_scale": 7.861452102661133, "train/clean_loss_weight": 1.5722904205322266, "train/combined_loss": 1.0778788328170776, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08436018957345971, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8982323408126831 }, { "epoch": 0.8436018957345972, "step": 356, "train/alpha": 0.2, "train/clean_ce_loss": 0.0947265625, "train/clean_loss_scale": 11.761460304260254, "train/clean_loss_weight": 2.352292060852051, "train/combined_loss": 1.3369473218917847, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08436018957345971, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.114122748374939 }, { "epoch": 0.8436018957345972, "step": 356, "train/alpha": 0.2, "train/clean_ce_loss": 0.08251953125, "train/clean_loss_scale": 7.667582988739014, "train/clean_loss_weight": 1.5335166454315186, "train/combined_loss": 0.759270429611206, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08436018957345971, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6327253580093384 }, { "epoch": 0.8459715639810427, "step": 357, "train/alpha": 0.2, "train/clean_ce_loss": 0.12451171875, "train/clean_loss_scale": 7.6436767578125, "train/clean_loss_weight": 1.5287353992462158, "train/combined_loss": 1.1420727968215942, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08459715639810426, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9517273306846619 }, { "epoch": 0.8459715639810427, "step": 357, "train/alpha": 0.2, "train/clean_ce_loss": 0.115234375, "train/clean_loss_scale": 9.20955753326416, "train/clean_loss_weight": 1.8419115543365479, "train/combined_loss": 1.27350914478302, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08459715639810426, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0612576007843018 }, { "epoch": 0.8459715639810427, "step": 357, "train/alpha": 0.2, "train/clean_ce_loss": 0.06005859375, "train/clean_loss_scale": 9.3234281539917, "train/clean_loss_weight": 1.8646856546401978, "train/combined_loss": 0.6719423532485962, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08459715639810426, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5599519610404968 }, { "epoch": 0.8459715639810427, "step": 357, "train/alpha": 0.2, "train/clean_ce_loss": 0.07861328125, "train/clean_loss_scale": 7.388426780700684, "train/clean_loss_weight": 1.4776853322982788, "train/combined_loss": 0.6969941854476929, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08459715639810426, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5808284878730774 }, { "epoch": 0.8483412322274881, "step": 358, "train/alpha": 0.2, "train/clean_ce_loss": 0.06787109375, "train/clean_loss_scale": 6.856991767883301, "train/clean_loss_weight": 1.3713983297348022, "train/combined_loss": 0.558469831943512, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08483412322274882, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4653915464878082 }, { "epoch": 0.8483412322274881, "step": 358, "train/alpha": 0.2, "train/clean_ce_loss": 0.0179443359375, "train/clean_loss_scale": 24.229251861572266, "train/clean_loss_weight": 4.845850467681885, "train/combined_loss": 0.5217334032058716, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08483412322274882, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4347778260707855 }, { "epoch": 0.8483412322274881, "step": 358, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 13.053213119506836, "train/clean_loss_weight": 2.610642671585083, "train/combined_loss": 1.0095844268798828, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08483412322274882, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8413203954696655 }, { "epoch": 0.8483412322274881, "step": 358, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 7.957507610321045, "train/clean_loss_weight": 1.5915015935897827, "train/combined_loss": 0.6154634952545166, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08483412322274882, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5128862261772156 }, { "epoch": 0.8507109004739336, "step": 359, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 7.60206937789917, "train/clean_loss_weight": 1.520413875579834, "train/combined_loss": 0.8463241457939148, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08507109004739337, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7052701115608215 }, { "epoch": 0.8507109004739336, "step": 359, "train/alpha": 0.2, "train/clean_ce_loss": 0.11279296875, "train/clean_loss_scale": 8.462504386901855, "train/clean_loss_weight": 1.6925009489059448, "train/combined_loss": 1.1454132795333862, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08507109004739337, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9545110464096069 }, { "epoch": 0.8507109004739336, "step": 359, "train/alpha": 0.2, "train/clean_ce_loss": 0.119140625, "train/clean_loss_scale": 8.363608360290527, "train/clean_loss_weight": 1.6727217435836792, "train/combined_loss": 1.1957346200942993, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08507109004739337, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9964454770088196 }, { "epoch": 0.8507109004739336, "step": 359, "train/alpha": 0.2, "train/clean_ce_loss": 0.05810546875, "train/clean_loss_scale": 7.607511043548584, "train/clean_loss_weight": 1.5215022563934326, "train/combined_loss": 0.5304455757141113, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08507109004739337, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.44203799962997437 }, { "epoch": 0.8530805687203792, "grad_norm": 11.239232063293457, "learning_rate": 2e-06, "loss": 3.8663, "step": 360 }, { "epoch": 0.8530805687203792, "step": 360, "train/alpha": 0.2, "train/clean_ce_loss": 0.1396484375, "train/clean_loss_scale": 6.378082752227783, "train/clean_loss_weight": 1.2756165266036987, "train/combined_loss": 1.0688271522521973, "train/pos_loss/pos0": 0.40736377239227295, "train/pos_loss/pos1": 0.7613580226898193, "train/pos_loss/pos2": 1.2151931524276733, "train/pos_loss/pos3": 1.7208142280578613, "train/pos_loss/pos4": 2.26798677444458, "train/pos_loss/pos5": 2.6253774166107178, "train/pos_loss/pos6": 3.1703341007232666, "train/pos_loss/weight0": 2.802645683288574, "train/pos_loss/weight1": 1.9286577701568604, "train/pos_loss/weight2": 1.1869620084762573, "train/pos_loss/weight3": 0.6361316442489624, "train/pos_loss/weight4": 0.2889775335788727, "train/pos_loss/weight5": 0.1114356741309166, "train/pos_loss/weight6": 0.045189134776592255, "train/pos_weight_progress": 0.08530805687203792, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8906893134117126 }, { "epoch": 0.8530805687203792, "step": 360, "train/alpha": 0.2, "train/clean_ce_loss": 0.049072265625, "train/clean_loss_scale": 7.316080093383789, "train/clean_loss_weight": 1.4632160663604736, "train/combined_loss": 0.43081995844841003, "train/pos_loss/pos0": 0.10106336325407028, "train/pos_loss/pos1": 0.2659316956996918, "train/pos_loss/pos2": 0.5937700271606445, "train/pos_loss/pos3": 0.9383164048194885, "train/pos_loss/pos4": 1.4286903142929077, "train/pos_loss/pos5": 1.7451362609863281, "train/pos_loss/pos6": 2.19120192527771, "train/pos_loss/weight0": 3.009376287460327, "train/pos_loss/weight1": 1.978206753730774, "train/pos_loss/weight2": 1.1471014022827148, "train/pos_loss/weight3": 0.5577157139778137, "train/pos_loss/weight4": 0.2148808389902115, "train/pos_loss/weight5": 0.06861061602830887, "train/pos_loss/weight6": 0.024107761681079865, "train/pos_weight_progress": 0.08530805687203792, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.35901662707328796 }, { "epoch": 0.8530805687203792, "step": 360, "train/alpha": 0.2, "train/clean_ce_loss": 0.146484375, "train/clean_loss_scale": 12.148783683776855, "train/clean_loss_weight": 2.4297568798065186, "train/combined_loss": 2.135528326034546, "train/pos_loss/pos0": 0.8334970474243164, "train/pos_loss/pos1": 1.5844858884811401, "train/pos_loss/pos2": 2.2076051235198975, "train/pos_loss/pos3": 2.5647711753845215, "train/pos_loss/pos4": 2.888042449951172, "train/pos_loss/pos5": 3.6212055683135986, "train/pos_loss/pos6": 3.868600368499756, "train/pos_loss/weight0": 2.279747486114502, "train/pos_loss/weight1": 1.7519614696502686, "train/pos_loss/weight2": 1.2486138343811035, "train/pos_loss/weight3": 0.8134567141532898, "train/pos_loss/weight4": 0.4891834259033203, "train/pos_loss/weight5": 0.26135674118995667, "train/pos_loss/weight6": 0.1556801050901413, "train/pos_weight_progress": 0.08530805687203792, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.7796069383621216 }, { "epoch": 0.8530805687203792, "step": 360, "train/alpha": 0.2, "train/clean_ce_loss": 0.14453125, "train/clean_loss_scale": 6.158048629760742, "train/clean_loss_weight": 1.2316097021102905, "train/combined_loss": 1.0680365562438965, "train/pos_loss/pos0": 0.47550544142723083, "train/pos_loss/pos1": 0.9366321563720703, "train/pos_loss/pos2": 1.4557244777679443, "train/pos_loss/pos3": 1.9085054397583008, "train/pos_loss/pos4": 2.7683029174804688, "train/pos_loss/pos5": 3.1854114532470703, "train/pos_loss/pos6": 3.525038719177246, "train/pos_loss/weight0": 3.3759233951568604, "train/pos_loss/weight1": 2.0400092601776123, "train/pos_loss/weight2": 1.0054758787155151, "train/pos_loss/weight3": 0.40906453132629395, "train/pos_loss/weight4": 0.13079394400119781, "train/pos_loss/weight5": 0.031065210700035095, "train/pos_loss/weight6": 0.007667733822017908, "train/pos_weight_progress": 0.08530805687203792, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8900304436683655 }, { "epoch": 0.8554502369668247, "step": 361, "train/alpha": 0.2, "train/clean_ce_loss": 0.11181640625, "train/clean_loss_scale": 6.1506524085998535, "train/clean_loss_weight": 1.2301305532455444, "train/combined_loss": 0.8252925872802734, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08554502369668246, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6877438426017761 }, { "epoch": 0.8554502369668247, "step": 361, "train/alpha": 0.2, "train/clean_ce_loss": 0.1435546875, "train/clean_loss_scale": 6.546275615692139, "train/clean_loss_weight": 1.3092551231384277, "train/combined_loss": 1.1276981830596924, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08554502369668246, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9397485256195068 }, { "epoch": 0.8554502369668247, "step": 361, "train/alpha": 0.2, "train/clean_ce_loss": 0.0859375, "train/clean_loss_scale": 8.137450218200684, "train/clean_loss_weight": 1.6274900436401367, "train/combined_loss": 0.8391745090484619, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08554502369668246, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6993120908737183 }, { "epoch": 0.8554502369668247, "step": 361, "train/alpha": 0.2, "train/clean_ce_loss": 0.1435546875, "train/clean_loss_scale": 5.930200099945068, "train/clean_loss_weight": 1.1860400438308716, "train/combined_loss": 1.0215696096420288, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08554502369668246, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8513080477714539 }, { "epoch": 0.8578199052132701, "step": 362, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 6.810234069824219, "train/clean_loss_weight": 1.3620468378067017, "train/combined_loss": 0.6663920283317566, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08578199052132701, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5553267002105713 }, { "epoch": 0.8578199052132701, "step": 362, "train/alpha": 0.2, "train/clean_ce_loss": 0.09130859375, "train/clean_loss_scale": 11.647364616394043, "train/clean_loss_weight": 2.3294730186462402, "train/combined_loss": 1.27620530128479, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08578199052132701, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0635044574737549 }, { "epoch": 0.8578199052132701, "step": 362, "train/alpha": 0.2, "train/clean_ce_loss": 0.1435546875, "train/clean_loss_scale": 6.692255973815918, "train/clean_loss_weight": 1.3384512662887573, "train/combined_loss": 1.1528456211090088, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08578199052132701, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9607046842575073 }, { "epoch": 0.8578199052132701, "step": 362, "train/alpha": 0.2, "train/clean_ce_loss": 0.09912109375, "train/clean_loss_scale": 6.972920894622803, "train/clean_loss_weight": 1.3945841789245605, "train/combined_loss": 0.8293962478637695, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08578199052132701, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6911635398864746 }, { "epoch": 0.8601895734597157, "step": 363, "train/alpha": 0.2, "train/clean_ce_loss": 0.1513671875, "train/clean_loss_scale": 7.159002304077148, "train/clean_loss_weight": 1.4318004846572876, "train/combined_loss": 1.300365686416626, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08601895734597156, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.083638072013855 }, { "epoch": 0.8601895734597157, "step": 363, "train/alpha": 0.2, "train/clean_ce_loss": 0.091796875, "train/clean_loss_scale": 7.426540851593018, "train/clean_loss_weight": 1.4853081703186035, "train/combined_loss": 0.8180798888206482, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08601895734597156, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.681733250617981 }, { "epoch": 0.8601895734597157, "step": 363, "train/alpha": 0.2, "train/clean_ce_loss": 0.009033203125, "train/clean_loss_scale": 30.819923400878906, "train/clean_loss_weight": 6.163984775543213, "train/combined_loss": 0.3340831398963928, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08601895734597156, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2784026265144348 }, { "epoch": 0.8601895734597157, "step": 363, "train/alpha": 0.2, "train/clean_ce_loss": 0.048095703125, "train/clean_loss_scale": 7.5891852378845215, "train/clean_loss_weight": 1.5178370475769043, "train/combined_loss": 0.4380086362361908, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08601895734597156, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3650071918964386 }, { "epoch": 0.8625592417061612, "step": 364, "train/alpha": 0.2, "train/clean_ce_loss": 0.035400390625, "train/clean_loss_scale": 5.598339080810547, "train/clean_loss_weight": 1.119667887687683, "train/combined_loss": 0.23782005906105042, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08625592417061612, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.19818338751792908 }, { "epoch": 0.8625592417061612, "step": 364, "train/alpha": 0.2, "train/clean_ce_loss": 0.0986328125, "train/clean_loss_scale": 7.822299480438232, "train/clean_loss_weight": 1.5644599199295044, "train/combined_loss": 0.9258424639701843, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08625592417061612, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7715353965759277 }, { "epoch": 0.8625592417061612, "step": 364, "train/alpha": 0.2, "train/clean_ce_loss": 0.05859375, "train/clean_loss_scale": 6.477060794830322, "train/clean_loss_weight": 1.2954121828079224, "train/combined_loss": 0.45541834831237793, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08625592417061612, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.37951529026031494 }, { "epoch": 0.8625592417061612, "step": 364, "train/alpha": 0.2, "train/clean_ce_loss": 0.02392578125, "train/clean_loss_scale": 15.441170692443848, "train/clean_loss_weight": 3.0882341861724854, "train/combined_loss": 0.4433304965496063, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08625592417061612, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.36944207549095154 }, { "epoch": 0.8649289099526066, "grad_norm": 22.739524841308594, "learning_rate": 2e-06, "loss": 3.8072, "step": 365 }, { "epoch": 0.8649289099526066, "step": 365, "train/alpha": 0.2, "train/clean_ce_loss": 0.10791015625, "train/clean_loss_scale": 4.555052757263184, "train/clean_loss_weight": 0.9110105633735657, "train/combined_loss": 0.58984375, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08649289099526067, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4915364384651184 }, { "epoch": 0.8649289099526066, "step": 365, "train/alpha": 0.2, "train/clean_ce_loss": 0.045654296875, "train/clean_loss_scale": 6.797926425933838, "train/clean_loss_weight": 1.3595852851867676, "train/combined_loss": 0.37242546677589417, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08649289099526067, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3103545606136322 }, { "epoch": 0.8649289099526066, "step": 365, "train/alpha": 0.2, "train/clean_ce_loss": 0.1396484375, "train/clean_loss_scale": 7.310164451599121, "train/clean_loss_weight": 1.4620329141616821, "train/combined_loss": 1.225023627281189, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08649289099526067, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.020853042602539 }, { "epoch": 0.8649289099526066, "step": 365, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 7.170375823974609, "train/clean_loss_weight": 1.4340752363204956, "train/combined_loss": 0.6386115550994873, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08649289099526067, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5321763157844543 }, { "epoch": 0.8672985781990521, "step": 366, "train/alpha": 0.2, "train/clean_ce_loss": 0.12890625, "train/clean_loss_scale": 7.093866348266602, "train/clean_loss_weight": 1.4187732934951782, "train/combined_loss": 1.09733247756958, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.0867298578199052, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9144437313079834 }, { "epoch": 0.8672985781990521, "step": 366, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 7.655171871185303, "train/clean_loss_weight": 1.5310343503952026, "train/combined_loss": 0.5920796990394592, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.0867298578199052, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4933997392654419 }, { "epoch": 0.8672985781990521, "step": 366, "train/alpha": 0.2, "train/clean_ce_loss": 0.0703125, "train/clean_loss_scale": 6.229081153869629, "train/clean_loss_weight": 1.2458162307739258, "train/combined_loss": 0.5255787372589111, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.0867298578199052, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4379822611808777 }, { "epoch": 0.8672985781990521, "step": 366, "train/alpha": 0.2, "train/clean_ce_loss": 0.166015625, "train/clean_loss_scale": 6.16788911819458, "train/clean_loss_weight": 1.233577847480774, "train/combined_loss": 1.2287591695785522, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.0867298578199052, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0239659547805786 }, { "epoch": 0.8696682464454977, "step": 367, "train/alpha": 0.2, "train/clean_ce_loss": 0.16796875, "train/clean_loss_scale": 5.803079605102539, "train/clean_loss_weight": 1.1606159210205078, "train/combined_loss": 1.1696832180023193, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08696682464454976, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9747360348701477 }, { "epoch": 0.8696682464454977, "step": 367, "train/alpha": 0.2, "train/clean_ce_loss": 0.0849609375, "train/clean_loss_scale": 6.747742176055908, "train/clean_loss_weight": 1.3495484590530396, "train/combined_loss": 0.6879534125328064, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08696682464454976, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5732945203781128 }, { "epoch": 0.8696682464454977, "step": 367, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 7.13319206237793, "train/clean_loss_weight": 1.4266384840011597, "train/combined_loss": 0.7147124409675598, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08696682464454976, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5955936908721924 }, { "epoch": 0.8696682464454977, "step": 367, "train/alpha": 0.2, "train/clean_ce_loss": 0.0191650390625, "train/clean_loss_scale": 11.685219764709473, "train/clean_loss_weight": 2.3370440006256104, "train/combined_loss": 0.26873722672462463, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08696682464454976, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2239476889371872 }, { "epoch": 0.8720379146919431, "step": 368, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 5.791496276855469, "train/clean_loss_weight": 1.1582993268966675, "train/combined_loss": 0.4615098536014557, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08720379146919431, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.38459154963493347 }, { "epoch": 0.8720379146919431, "step": 368, "train/alpha": 0.2, "train/clean_ce_loss": 0.10986328125, "train/clean_loss_scale": 8.023172378540039, "train/clean_loss_weight": 1.6046345233917236, "train/combined_loss": 1.0577424764633179, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08720379146919431, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8814520239830017 }, { "epoch": 0.8720379146919431, "step": 368, "train/alpha": 0.2, "train/clean_ce_loss": 0.10693359375, "train/clean_loss_scale": 9.611616134643555, "train/clean_loss_weight": 1.922323226928711, "train/combined_loss": 1.233365535736084, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08720379146919431, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0278046131134033 }, { "epoch": 0.8720379146919431, "step": 368, "train/alpha": 0.2, "train/clean_ce_loss": 0.1484375, "train/clean_loss_scale": 5.760981559753418, "train/clean_loss_weight": 1.1521962881088257, "train/combined_loss": 1.026174783706665, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08720379146919431, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8551456928253174 }, { "epoch": 0.8744075829383886, "step": 369, "train/alpha": 0.2, "train/clean_ce_loss": 0.05908203125, "train/clean_loss_scale": 9.357394218444824, "train/clean_loss_weight": 1.8714789152145386, "train/combined_loss": 0.6634246706962585, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08744075829383886, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5528538823127747 }, { "epoch": 0.8744075829383886, "step": 369, "train/alpha": 0.2, "train/clean_ce_loss": 0.0133056640625, "train/clean_loss_scale": 15.500899314880371, "train/clean_loss_weight": 3.10017991065979, "train/combined_loss": 0.24749970436096191, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08744075829383886, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.20624975860118866 }, { "epoch": 0.8744075829383886, "step": 369, "train/alpha": 0.2, "train/clean_ce_loss": 0.05029296875, "train/clean_loss_scale": 7.514733791351318, "train/clean_loss_weight": 1.5029467344284058, "train/combined_loss": 0.45352593064308167, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08744075829383886, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.37793827056884766 }, { "epoch": 0.8744075829383886, "step": 369, "train/alpha": 0.2, "train/clean_ce_loss": 0.08447265625, "train/clean_loss_scale": 7.454582214355469, "train/clean_loss_weight": 1.4909164905548096, "train/combined_loss": 0.7556500434875488, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08744075829383886, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6297083497047424 }, { "epoch": 0.8767772511848341, "grad_norm": 16.19022560119629, "learning_rate": 2e-06, "loss": 3.7032, "step": 370 }, { "epoch": 0.8767772511848341, "step": 370, "train/alpha": 0.2, "train/clean_ce_loss": 0.146484375, "train/clean_loss_scale": 8.08691120147705, "train/clean_loss_weight": 1.617382287979126, "train/combined_loss": 1.4215272665023804, "train/pos_loss/pos0": 0.557805597782135, "train/pos_loss/pos1": 1.302854061126709, "train/pos_loss/pos2": 1.8437421321868896, "train/pos_loss/pos3": 2.623053789138794, "train/pos_loss/pos4": 3.1047019958496094, "train/pos_loss/pos5": 3.9112391471862793, "train/pos_loss/pos6": 4.002279281616211, "train/pos_loss/weight0": 3.231051445007324, "train/pos_loss/weight1": 2.037402391433716, "train/pos_loss/weight2": 1.0733904838562012, "train/pos_loss/weight3": 0.4559612274169922, "train/pos_loss/weight4": 0.15766388177871704, "train/pos_loss/weight5": 0.0357181616127491, "train/pos_loss/weight6": 0.008812075480818748, "train/pos_weight_progress": 0.08767772511848342, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1846060752868652 }, { "epoch": 0.8767772511848341, "step": 370, "train/alpha": 0.2, "train/clean_ce_loss": 0.14453125, "train/clean_loss_scale": 8.618721961975098, "train/clean_loss_weight": 1.7237443923950195, "train/combined_loss": 1.4948095083236694, "train/pos_loss/pos0": 0.6698837876319885, "train/pos_loss/pos1": 1.3405883312225342, "train/pos_loss/pos2": 2.0977048873901367, "train/pos_loss/pos3": 3.068769693374634, "train/pos_loss/pos4": 3.7100253105163574, "train/pos_loss/pos5": 4.104005813598633, "train/pos_loss/pos6": 4.571703910827637, "train/pos_loss/weight0": 3.395143747329712, "train/pos_loss/weight1": 2.1407110691070557, "train/pos_loss/weight2": 0.9993244409561157, "train/pos_loss/weight3": 0.35239821672439575, "train/pos_loss/weight4": 0.0923427864909172, "train/pos_loss/weight5": 0.017004255205392838, "train/pos_loss/weight6": 0.003075078595429659, "train/pos_weight_progress": 0.08767772511848342, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2456746101379395 }, { "epoch": 0.8767772511848341, "step": 370, "train/alpha": 0.2, "train/clean_ce_loss": 0.10546875, "train/clean_loss_scale": 7.716093063354492, "train/clean_loss_weight": 1.5432186126708984, "train/combined_loss": 0.9765680432319641, "train/pos_loss/pos0": 0.48506224155426025, "train/pos_loss/pos1": 0.8801489472389221, "train/pos_loss/pos2": 1.384840965270996, "train/pos_loss/pos3": 1.6175298690795898, "train/pos_loss/pos4": 1.8800288438796997, "train/pos_loss/pos5": 2.801621437072754, "train/pos_loss/pos6": 3.171825408935547, "train/pos_loss/weight0": 3.418358087539673, "train/pos_loss/weight1": 2.0960822105407715, "train/pos_loss/weight2": 1.0012882947921753, "train/pos_loss/weight3": 0.37039676308631897, "train/pos_loss/weight4": 0.09545903652906418, "train/pos_loss/weight5": 0.015889327973127365, "train/pos_loss/weight6": 0.002526006195694208, "train/pos_weight_progress": 0.08767772511848342, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8138067126274109 }, { "epoch": 0.8767772511848341, "step": 370, "train/alpha": 0.2, "train/clean_ce_loss": 0.0018157958984375, "train/clean_loss_scale": 165.21435546875, "train/clean_loss_weight": 33.04287338256836, "train/combined_loss": 0.35999464988708496, "train/pos_loss/pos0": 0.041218943893909454, "train/pos_loss/pos1": 0.12655431032180786, "train/pos_loss/pos2": 0.4058767557144165, "train/pos_loss/pos3": 1.1545275449752808, "train/pos_loss/pos4": 1.7355104684829712, "train/pos_loss/pos5": 2.270871162414551, "train/pos_loss/pos6": 2.6251797676086426, "train/pos_loss/weight0": 3.002636432647705, "train/pos_loss/weight1": 1.9977695941925049, "train/pos_loss/weight2": 1.1274293661117554, "train/pos_loss/weight3": 0.5440512895584106, "train/pos_loss/weight4": 0.22216594219207764, "train/pos_loss/weight5": 0.07666642963886261, "train/pos_loss/weight6": 0.029281198978424072, "train/pos_weight_progress": 0.08767772511848342, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2999955415725708 }, { "epoch": 0.8791469194312796, "step": 371, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 8.684842109680176, "train/clean_loss_weight": 1.7369683980941772, "train/combined_loss": 0.9108906388282776, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08791469194312797, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7590755224227905 }, { "epoch": 0.8791469194312796, "step": 371, "train/alpha": 0.2, "train/clean_ce_loss": 0.07666015625, "train/clean_loss_scale": 9.230081558227539, "train/clean_loss_weight": 1.84601628780365, "train/combined_loss": 0.8490954041481018, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08791469194312797, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.707579493522644 }, { "epoch": 0.8791469194312796, "step": 371, "train/alpha": 0.2, "train/clean_ce_loss": 0.125, "train/clean_loss_scale": 11.113369941711426, "train/clean_loss_weight": 2.2226741313934326, "train/combined_loss": 1.6670055389404297, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08791469194312797, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3891712427139282 }, { "epoch": 0.8791469194312796, "step": 371, "train/alpha": 0.2, "train/clean_ce_loss": 0.12060546875, "train/clean_loss_scale": 10.455341339111328, "train/clean_loss_weight": 2.0910682678222656, "train/combined_loss": 1.5131655931472778, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08791469194312797, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2609713077545166 }, { "epoch": 0.8815165876777251, "step": 372, "train/alpha": 0.2, "train/clean_ce_loss": 0.09130859375, "train/clean_loss_scale": 13.109294891357422, "train/clean_loss_weight": 2.621859073638916, "train/combined_loss": 1.4363895654678345, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.0881516587677725, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1969913244247437 }, { "epoch": 0.8815165876777251, "step": 372, "train/alpha": 0.2, "train/clean_ce_loss": 0.09765625, "train/clean_loss_scale": 10.09546947479248, "train/clean_loss_weight": 2.0190939903259277, "train/combined_loss": 1.1830627918243408, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.0881516587677725, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9858856797218323 }, { "epoch": 0.8815165876777251, "step": 372, "train/alpha": 0.2, "train/clean_ce_loss": 0.05322265625, "train/clean_loss_scale": 13.84890079498291, "train/clean_loss_weight": 2.769780158996582, "train/combined_loss": 0.8844903111457825, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.0881516587677725, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7370752692222595 }, { "epoch": 0.8815165876777251, "step": 372, "train/alpha": 0.2, "train/clean_ce_loss": 0.15234375, "train/clean_loss_scale": 9.80888843536377, "train/clean_loss_weight": 1.961777687072754, "train/combined_loss": 1.7931873798370361, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.0881516587677725, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4943227767944336 }, { "epoch": 0.8838862559241706, "step": 373, "train/alpha": 0.2, "train/clean_ce_loss": 0.15234375, "train/clean_loss_scale": 9.18445110321045, "train/clean_loss_weight": 1.8368902206420898, "train/combined_loss": 1.679032564163208, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08838862559241706, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3991937637329102 }, { "epoch": 0.8838862559241706, "step": 373, "train/alpha": 0.2, "train/clean_ce_loss": 0.119140625, "train/clean_loss_scale": 11.188164710998535, "train/clean_loss_weight": 2.237632989883423, "train/combined_loss": 1.599557876586914, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08838862559241706, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3329648971557617 }, { "epoch": 0.8838862559241706, "step": 373, "train/alpha": 0.2, "train/clean_ce_loss": 0.1396484375, "train/clean_loss_scale": 10.483814239501953, "train/clean_loss_weight": 2.0967628955841064, "train/combined_loss": 1.7568578720092773, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08838862559241706, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4640482664108276 }, { "epoch": 0.8838862559241706, "step": 373, "train/alpha": 0.2, "train/clean_ce_loss": 0.1142578125, "train/clean_loss_scale": 11.289995193481445, "train/clean_loss_weight": 2.2579991817474365, "train/combined_loss": 1.5479642152786255, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08838862559241706, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2899701595306396 }, { "epoch": 0.8862559241706162, "step": 374, "train/alpha": 0.2, "train/clean_ce_loss": 0.0615234375, "train/clean_loss_scale": 10.811568260192871, "train/clean_loss_weight": 2.16231369972229, "train/combined_loss": 0.7981978058815002, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08862559241706161, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6651648283004761 }, { "epoch": 0.8862559241706162, "step": 374, "train/alpha": 0.2, "train/clean_ce_loss": 0.0869140625, "train/clean_loss_scale": 7.593337059020996, "train/clean_loss_weight": 1.518667459487915, "train/combined_loss": 0.7919613122940063, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08862559241706161, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6599677801132202 }, { "epoch": 0.8862559241706162, "step": 374, "train/alpha": 0.2, "train/clean_ce_loss": 0.06005859375, "train/clean_loss_scale": 12.703182220458984, "train/clean_loss_weight": 2.5406365394592285, "train/combined_loss": 0.9155223369598389, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08862559241706161, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7629352807998657 }, { "epoch": 0.8862559241706162, "step": 374, "train/alpha": 0.2, "train/clean_ce_loss": 0.12255859375, "train/clean_loss_scale": 11.457035064697266, "train/clean_loss_weight": 2.2914071083068848, "train/combined_loss": 1.6849896907806396, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08862559241706161, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.404158115386963 }, { "epoch": 0.8886255924170616, "grad_norm": 14.818435668945312, "learning_rate": 2e-06, "loss": 3.8495, "step": 375 }, { "epoch": 0.8886255924170616, "step": 375, "train/alpha": 0.2, "train/clean_ce_loss": 0.04248046875, "train/clean_loss_scale": 11.972566604614258, "train/clean_loss_weight": 2.3945133686065674, "train/combined_loss": 0.6103202700614929, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08886255924170616, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5086002349853516 }, { "epoch": 0.8886255924170616, "step": 375, "train/alpha": 0.2, "train/clean_ce_loss": 0.0830078125, "train/clean_loss_scale": 10.512737274169922, "train/clean_loss_weight": 2.1025474071502686, "train/combined_loss": 1.047167181968689, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08886255924170616, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8726392984390259 }, { "epoch": 0.8886255924170616, "step": 375, "train/alpha": 0.2, "train/clean_ce_loss": 0.0703125, "train/clean_loss_scale": 9.272350311279297, "train/clean_loss_weight": 1.854470133781433, "train/combined_loss": 0.7823545336723328, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08886255924170616, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6519621014595032 }, { "epoch": 0.8886255924170616, "step": 375, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 10.104134559631348, "train/clean_loss_weight": 2.020827054977417, "train/combined_loss": 0.9650237560272217, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08886255924170616, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8041864633560181 }, { "epoch": 0.8909952606635071, "step": 376, "train/alpha": 0.2, "train/clean_ce_loss": 0.1015625, "train/clean_loss_scale": 10.227177619934082, "train/clean_loss_weight": 2.045435667037964, "train/combined_loss": 1.2464373111724854, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08909952606635071, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0386977195739746 }, { "epoch": 0.8909952606635071, "step": 376, "train/alpha": 0.2, "train/clean_ce_loss": 0.1640625, "train/clean_loss_scale": 8.603218078613281, "train/clean_loss_weight": 1.7206436395645142, "train/combined_loss": 1.693758487701416, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08909952606635071, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4114654064178467 }, { "epoch": 0.8909952606635071, "step": 376, "train/alpha": 0.2, "train/clean_ce_loss": 0.1064453125, "train/clean_loss_scale": 9.603365898132324, "train/clean_loss_weight": 1.9206732511520386, "train/combined_loss": 1.2266799211502075, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08909952606635071, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.022233247756958 }, { "epoch": 0.8909952606635071, "step": 376, "train/alpha": 0.2, "train/clean_ce_loss": 0.12109375, "train/clean_loss_scale": 11.10019302368164, "train/clean_loss_weight": 2.220038652420044, "train/combined_loss": 1.6129968166351318, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08909952606635071, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3441640138626099 }, { "epoch": 0.8933649289099526, "step": 377, "train/alpha": 0.2, "train/clean_ce_loss": 0.1298828125, "train/clean_loss_scale": 10.232701301574707, "train/clean_loss_weight": 2.0465402603149414, "train/combined_loss": 1.5948623418807983, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08933649289099527, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3290519714355469 }, { "epoch": 0.8933649289099526, "step": 377, "train/alpha": 0.2, "train/clean_ce_loss": 0.07861328125, "train/clean_loss_scale": 10.050883293151855, "train/clean_loss_weight": 2.010176658630371, "train/combined_loss": 0.9481595158576965, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08933649289099527, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7901329398155212 }, { "epoch": 0.8933649289099526, "step": 377, "train/alpha": 0.2, "train/clean_ce_loss": 0.0260009765625, "train/clean_loss_scale": 20.417448043823242, "train/clean_loss_weight": 4.083489894866943, "train/combined_loss": 0.6370483040809631, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08933649289099527, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5308735966682434 }, { "epoch": 0.8933649289099526, "step": 377, "train/alpha": 0.2, "train/clean_ce_loss": 0.10498046875, "train/clean_loss_scale": 11.153373718261719, "train/clean_loss_weight": 2.2306747436523438, "train/combined_loss": 1.4050636291503906, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08933649289099527, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1708863973617554 }, { "epoch": 0.8957345971563981, "step": 378, "train/alpha": 0.2, "train/clean_ce_loss": 0.09814453125, "train/clean_loss_scale": 9.049991607666016, "train/clean_loss_weight": 1.8099983930587769, "train/combined_loss": 1.0658485889434814, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.0895734597156398, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8882071375846863 }, { "epoch": 0.8957345971563981, "step": 378, "train/alpha": 0.2, "train/clean_ce_loss": 0.061279296875, "train/clean_loss_scale": 10.902314186096191, "train/clean_loss_weight": 2.1804628372192383, "train/combined_loss": 0.8017033934593201, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.0895734597156398, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6680861711502075 }, { "epoch": 0.8957345971563981, "step": 378, "train/alpha": 0.2, "train/clean_ce_loss": 0.0380859375, "train/clean_loss_scale": 12.128634452819824, "train/clean_loss_weight": 2.425726890563965, "train/combined_loss": 0.554316520690918, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.0895734597156398, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46193042397499084 }, { "epoch": 0.8957345971563981, "step": 378, "train/alpha": 0.2, "train/clean_ce_loss": 0.12890625, "train/clean_loss_scale": 8.205278396606445, "train/clean_loss_weight": 1.641055703163147, "train/combined_loss": 1.2692540884017944, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.0895734597156398, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0577117204666138 }, { "epoch": 0.8981042654028436, "step": 379, "train/alpha": 0.2, "train/clean_ce_loss": 0.013671875, "train/clean_loss_scale": 42.377166748046875, "train/clean_loss_weight": 8.475433349609375, "train/combined_loss": 0.695250391960144, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08981042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5793753266334534 }, { "epoch": 0.8981042654028436, "step": 379, "train/alpha": 0.2, "train/clean_ce_loss": 0.04443359375, "train/clean_loss_scale": 10.169918060302734, "train/clean_loss_weight": 2.0339837074279785, "train/combined_loss": 0.5422632098197937, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08981042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45188599824905396 }, { "epoch": 0.8981042654028436, "step": 379, "train/alpha": 0.2, "train/clean_ce_loss": 0.062255859375, "train/clean_loss_scale": 8.838212013244629, "train/clean_loss_weight": 1.7676423788070679, "train/combined_loss": 0.6602765917778015, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08981042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5502305030822754 }, { "epoch": 0.8981042654028436, "step": 379, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 9.872961044311523, "train/clean_loss_weight": 1.9745922088623047, "train/combined_loss": 0.9660847187042236, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.08981042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8050705790519714 }, { "epoch": 0.9004739336492891, "grad_norm": 21.52027702331543, "learning_rate": 2e-06, "loss": 4.094, "step": 380 }, { "epoch": 0.9004739336492891, "step": 380, "train/alpha": 0.2, "train/clean_ce_loss": 0.01324462890625, "train/clean_loss_scale": 34.948204040527344, "train/clean_loss_weight": 6.989640712738037, "train/combined_loss": 0.5554512143135071, "train/pos_loss/pos0": 0.04346903786063194, "train/pos_loss/pos1": 0.18460607528686523, "train/pos_loss/pos2": 0.47699907422065735, "train/pos_loss/pos3": 0.7779886722564697, "train/pos_loss/pos4": 1.443471908569336, "train/pos_loss/pos5": 2.18403697013855, "train/pos_loss/pos6": 2.6141042709350586, "train/pos_loss/weight0": 2.3018317222595215, "train/pos_loss/weight1": 1.7401236295700073, "train/pos_loss/weight2": 1.2576868534088135, "train/pos_loss/weight3": 0.8444778919219971, "train/pos_loss/weight4": 0.4849546253681183, "train/pos_loss/weight5": 0.24551531672477722, "train/pos_loss/weight6": 0.12540991604328156, "train/pos_weight_progress": 0.09004739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46287599205970764 }, { "epoch": 0.9004739336492891, "step": 380, "train/alpha": 0.2, "train/clean_ce_loss": 0.000698089599609375, "train/clean_loss_scale": 313.37017822265625, "train/clean_loss_weight": 62.67403793334961, "train/combined_loss": 0.26251256465911865, "train/pos_loss/pos0": 0.0016981156077235937, "train/pos_loss/pos1": 0.025355208665132523, "train/pos_loss/pos2": 0.1355283409357071, "train/pos_loss/pos3": 0.39967668056488037, "train/pos_loss/pos4": 0.7819820642471313, "train/pos_loss/pos5": 1.103375792503357, "train/pos_loss/pos6": 1.7833958864212036, "train/pos_loss/weight0": 2.243907928466797, "train/pos_loss/weight1": 1.7392619848251343, "train/pos_loss/weight2": 1.2460365295410156, "train/pos_loss/weight3": 0.8334802985191345, "train/pos_loss/weight4": 0.500676691532135, "train/pos_loss/weight5": 0.2762772738933563, "train/pos_loss/weight6": 0.16035890579223633, "train/pos_weight_progress": 0.09004739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.21876046061515808 }, { "epoch": 0.9004739336492891, "step": 380, "train/alpha": 0.2, "train/clean_ce_loss": 0.027587890625, "train/clean_loss_scale": 22.04963493347168, "train/clean_loss_weight": 4.409926891326904, "train/combined_loss": 0.729963481426239, "train/pos_loss/pos0": 0.04232489690184593, "train/pos_loss/pos1": 0.14921589195728302, "train/pos_loss/pos2": 0.4771871268749237, "train/pos_loss/pos3": 0.897469162940979, "train/pos_loss/pos4": 1.479881763458252, "train/pos_loss/pos5": 1.7316311597824097, "train/pos_loss/pos6": 2.3286705017089844, "train/pos_loss/weight0": 1.9080537557601929, "train/pos_loss/weight1": 1.5321521759033203, "train/pos_loss/weight2": 1.2065311670303345, "train/pos_loss/weight3": 0.9031600952148438, "train/pos_loss/weight4": 0.6455156207084656, "train/pos_loss/weight5": 0.44701066613197327, "train/pos_loss/weight6": 0.3575766682624817, "train/pos_weight_progress": 0.09004739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.608302891254425 }, { "epoch": 0.9004739336492891, "step": 380, "train/alpha": 0.2, "train/clean_ce_loss": 0.0869140625, "train/clean_loss_scale": 10.28016471862793, "train/clean_loss_weight": 2.05603289604187, "train/combined_loss": 1.0721889734268188, "train/pos_loss/pos0": 0.19363044202327728, "train/pos_loss/pos1": 0.5383918881416321, "train/pos_loss/pos2": 0.9635924696922302, "train/pos_loss/pos3": 1.5612592697143555, "train/pos_loss/pos4": 2.0549631118774414, "train/pos_loss/pos5": 2.584972620010376, "train/pos_loss/pos6": 2.829498291015625, "train/pos_loss/weight0": 2.197204113006592, "train/pos_loss/weight1": 1.6872103214263916, "train/pos_loss/weight2": 1.2381032705307007, "train/pos_loss/weight3": 0.8732219338417053, "train/pos_loss/weight4": 0.5228365063667297, "train/pos_loss/weight5": 0.2925718426704407, "train/pos_loss/weight6": 0.18885183334350586, "train/pos_weight_progress": 0.09004739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8934908509254456 }, { "epoch": 0.9028436018957346, "step": 381, "train/alpha": 0.2, "train/clean_ce_loss": 0.1416015625, "train/clean_loss_scale": 8.518317222595215, "train/clean_loss_weight": 1.7036634683609009, "train/combined_loss": 1.447448492050171, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09028436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.206207036972046 }, { "epoch": 0.9028436018957346, "step": 381, "train/alpha": 0.2, "train/clean_ce_loss": 0.1298828125, "train/clean_loss_scale": 10.522183418273926, "train/clean_loss_weight": 2.1044366359710693, "train/combined_loss": 1.6399810314178467, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09028436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3666508197784424 }, { "epoch": 0.9028436018957346, "step": 381, "train/alpha": 0.2, "train/clean_ce_loss": 0.00958251953125, "train/clean_loss_scale": 28.639020919799805, "train/clean_loss_weight": 5.727804183959961, "train/combined_loss": 0.3293207585811615, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09028436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.274433970451355 }, { "epoch": 0.9028436018957346, "step": 381, "train/alpha": 0.2, "train/clean_ce_loss": 0.12890625, "train/clean_loss_scale": 9.926589012145996, "train/clean_loss_weight": 1.9853178262710571, "train/combined_loss": 1.5355191230773926, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09028436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2795993089675903 }, { "epoch": 0.9052132701421801, "step": 382, "train/alpha": 0.2, "train/clean_ce_loss": 0.052978515625, "train/clean_loss_scale": 9.19256591796875, "train/clean_loss_weight": 1.8385132551193237, "train/combined_loss": 0.5844101905822754, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09052132701421801, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4870085120201111 }, { "epoch": 0.9052132701421801, "step": 382, "train/alpha": 0.2, "train/clean_ce_loss": 0.12451171875, "train/clean_loss_scale": 6.5315842628479, "train/clean_loss_weight": 1.30631685256958, "train/combined_loss": 0.9759105443954468, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09052132701421801, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8132587671279907 }, { "epoch": 0.9052132701421801, "step": 382, "train/alpha": 0.2, "train/clean_ce_loss": 0.1181640625, "train/clean_loss_scale": 8.074834823608398, "train/clean_loss_weight": 1.6149669885635376, "train/combined_loss": 1.1449863910675049, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09052132701421801, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9541552662849426 }, { "epoch": 0.9052132701421801, "step": 382, "train/alpha": 0.2, "train/clean_ce_loss": 0.158203125, "train/clean_loss_scale": 6.797636032104492, "train/clean_loss_weight": 1.3595272302627563, "train/combined_loss": 1.2904887199401855, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09052132701421801, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0754072666168213 }, { "epoch": 0.9075829383886256, "step": 383, "train/alpha": 0.2, "train/clean_ce_loss": 0.1083984375, "train/clean_loss_scale": 7.036355495452881, "train/clean_loss_weight": 1.407271146774292, "train/combined_loss": 0.9152759313583374, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09075829383886255, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7627299427986145 }, { "epoch": 0.9075829383886256, "step": 383, "train/alpha": 0.2, "train/clean_ce_loss": 0.0223388671875, "train/clean_loss_scale": 22.024202346801758, "train/clean_loss_weight": 4.404840469360352, "train/combined_loss": 0.590394914150238, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09075829383886255, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.49199575185775757 }, { "epoch": 0.9075829383886256, "step": 383, "train/alpha": 0.2, "train/clean_ce_loss": 0.060302734375, "train/clean_loss_scale": 9.088749885559082, "train/clean_loss_weight": 1.8177499771118164, "train/combined_loss": 0.6576917171478271, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09075829383886255, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5480764508247375 }, { "epoch": 0.9075829383886256, "step": 383, "train/alpha": 0.2, "train/clean_ce_loss": 0.07666015625, "train/clean_loss_scale": 7.05774450302124, "train/clean_loss_weight": 1.4115489721298218, "train/combined_loss": 0.6492573618888855, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09075829383886255, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5410478115081787 }, { "epoch": 0.909952606635071, "step": 384, "train/alpha": 0.2, "train/clean_ce_loss": 0.10888671875, "train/clean_loss_scale": 7.271521091461182, "train/clean_loss_weight": 1.4543042182922363, "train/combined_loss": 0.9501264691352844, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.0909952606635071, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7917720675468445 }, { "epoch": 0.909952606635071, "step": 384, "train/alpha": 0.2, "train/clean_ce_loss": 0.1591796875, "train/clean_loss_scale": 6.563594341278076, "train/clean_loss_weight": 1.3127188682556152, "train/combined_loss": 1.2537490129470825, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.0909952606635071, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0447908639907837 }, { "epoch": 0.909952606635071, "step": 384, "train/alpha": 0.2, "train/clean_ce_loss": 0.12890625, "train/clean_loss_scale": 9.63057804107666, "train/clean_loss_weight": 1.92611563205719, "train/combined_loss": 1.4897301197052002, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.0909952606635071, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2414417266845703 }, { "epoch": 0.909952606635071, "step": 384, "train/alpha": 0.2, "train/clean_ce_loss": 0.169921875, "train/clean_loss_scale": 7.503978729248047, "train/clean_loss_weight": 1.5007957220077515, "train/combined_loss": 1.530108094215393, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.0909952606635071, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2750900983810425 }, { "epoch": 0.9123222748815166, "grad_norm": 16.505428314208984, "learning_rate": 2e-06, "loss": 3.5285, "step": 385 }, { "epoch": 0.9123222748815166, "step": 385, "train/alpha": 0.2, "train/clean_ce_loss": 0.00494384765625, "train/clean_loss_scale": 27.074899673461914, "train/clean_loss_weight": 5.414979934692383, "train/combined_loss": 0.16062501072883606, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09123222748815166, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.13385418057441711 }, { "epoch": 0.9123222748815166, "step": 385, "train/alpha": 0.2, "train/clean_ce_loss": 0.017333984375, "train/clean_loss_scale": 19.178955078125, "train/clean_loss_weight": 3.8357911109924316, "train/combined_loss": 0.39893725514411926, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09123222748815166, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.33244770765304565 }, { "epoch": 0.9123222748815166, "step": 385, "train/alpha": 0.2, "train/clean_ce_loss": 0.0732421875, "train/clean_loss_scale": 8.856081008911133, "train/clean_loss_weight": 1.7712162733078003, "train/combined_loss": 0.7783664464950562, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09123222748815166, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6486387252807617 }, { "epoch": 0.9123222748815166, "step": 385, "train/alpha": 0.2, "train/clean_ce_loss": 0.07177734375, "train/clean_loss_scale": 9.021868705749512, "train/clean_loss_weight": 1.8043737411499023, "train/combined_loss": 0.7770789265632629, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09123222748815166, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6475657820701599 }, { "epoch": 0.9146919431279621, "step": 386, "train/alpha": 0.2, "train/clean_ce_loss": 0.09326171875, "train/clean_loss_scale": 7.709771633148193, "train/clean_loss_weight": 1.5419543981552124, "train/combined_loss": 0.8628318905830383, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09146919431279621, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7190265655517578 }, { "epoch": 0.9146919431279621, "step": 386, "train/alpha": 0.2, "train/clean_ce_loss": 0.0966796875, "train/clean_loss_scale": 5.9395623207092285, "train/clean_loss_weight": 1.1879124641418457, "train/combined_loss": 0.6890820264816284, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09146919431279621, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5742350220680237 }, { "epoch": 0.9146919431279621, "step": 386, "train/alpha": 0.2, "train/clean_ce_loss": 0.0238037109375, "train/clean_loss_scale": 18.360933303833008, "train/clean_loss_weight": 3.6721866130828857, "train/combined_loss": 0.5244699716567993, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09146919431279621, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43705832958221436 }, { "epoch": 0.9146919431279621, "step": 386, "train/alpha": 0.2, "train/clean_ce_loss": 0.099609375, "train/clean_loss_scale": 12.242321968078613, "train/clean_loss_weight": 2.4484643936157227, "train/combined_loss": 1.4633400440216064, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09146919431279621, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2194499969482422 }, { "epoch": 0.9170616113744076, "step": 387, "train/alpha": 0.2, "train/clean_ce_loss": 0.12353515625, "train/clean_loss_scale": 5.054435729980469, "train/clean_loss_weight": 1.0108871459960938, "train/combined_loss": 0.749280571937561, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09170616113744076, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6244004964828491 }, { "epoch": 0.9170616113744076, "step": 387, "train/alpha": 0.2, "train/clean_ce_loss": 0.0859375, "train/clean_loss_scale": 9.46627426147461, "train/clean_loss_weight": 1.8932548761367798, "train/combined_loss": 0.9762095808982849, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09170616113744076, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8135079741477966 }, { "epoch": 0.9170616113744076, "step": 387, "train/alpha": 0.2, "train/clean_ce_loss": 0.07861328125, "train/clean_loss_scale": 8.708423614501953, "train/clean_loss_weight": 1.7416847944259644, "train/combined_loss": 0.8215173482894897, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09170616113744076, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6845977902412415 }, { "epoch": 0.9170616113744076, "step": 387, "train/alpha": 0.2, "train/clean_ce_loss": 0.033203125, "train/clean_loss_scale": 11.565524101257324, "train/clean_loss_weight": 2.3131048679351807, "train/combined_loss": 0.46081385016441345, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09170616113744076, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.38401153683662415 }, { "epoch": 0.919431279620853, "step": 388, "train/alpha": 0.2, "train/clean_ce_loss": 0.0693359375, "train/clean_loss_scale": 8.085525512695312, "train/clean_loss_weight": 1.6171051263809204, "train/combined_loss": 0.6727409958839417, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09194312796208531, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5606175065040588 }, { "epoch": 0.919431279620853, "step": 388, "train/alpha": 0.2, "train/clean_ce_loss": 0.1611328125, "train/clean_loss_scale": 6.936374664306641, "train/clean_loss_weight": 1.387274980545044, "train/combined_loss": 1.3412131071090698, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09194312796208531, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1176775693893433 }, { "epoch": 0.919431279620853, "step": 388, "train/alpha": 0.2, "train/clean_ce_loss": 0.1640625, "train/clean_loss_scale": 6.320250034332275, "train/clean_loss_weight": 1.264050006866455, "train/combined_loss": 1.2442991733551025, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09194312796208531, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0369160175323486 }, { "epoch": 0.919431279620853, "step": 388, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 8.004193305969238, "train/clean_loss_weight": 1.6008386611938477, "train/combined_loss": 0.7691529989242554, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09194312796208531, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6409608125686646 }, { "epoch": 0.9218009478672986, "step": 389, "train/alpha": 0.2, "train/clean_ce_loss": 0.08251953125, "train/clean_loss_scale": 10.187047004699707, "train/clean_loss_weight": 2.037409543991089, "train/combined_loss": 1.008756399154663, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09218009478672985, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8406303524971008 }, { "epoch": 0.9218009478672986, "step": 389, "train/alpha": 0.2, "train/clean_ce_loss": 0.01318359375, "train/clean_loss_scale": 28.781307220458984, "train/clean_loss_weight": 5.756261348724365, "train/combined_loss": 0.4553292691707611, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09218009478672985, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3794410526752472 }, { "epoch": 0.9218009478672986, "step": 389, "train/alpha": 0.2, "train/clean_ce_loss": 0.016357421875, "train/clean_loss_scale": 21.666894912719727, "train/clean_loss_weight": 4.33337926864624, "train/combined_loss": 0.42529746890068054, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09218009478672985, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.35441455245018005 }, { "epoch": 0.9218009478672986, "step": 389, "train/alpha": 0.2, "train/clean_ce_loss": 0.1650390625, "train/clean_loss_scale": 10.105202674865723, "train/clean_loss_weight": 2.021040678024292, "train/combined_loss": 2.0013039112091064, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.09218009478672985, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.6677532196044922 }, { "epoch": 0.9241706161137441, "grad_norm": 16.40512466430664, "learning_rate": 2e-06, "loss": 3.6263, "step": 390 }, { "epoch": 0.9241706161137441, "step": 390, "train/alpha": 0.2, "train/clean_ce_loss": 0.0252685546875, "train/clean_loss_scale": 9.165070533752441, "train/clean_loss_weight": 1.8330141305923462, "train/combined_loss": 0.2779057025909424, "train/pos_loss/pos0": 0.003548968583345413, "train/pos_loss/pos1": 0.0691966712474823, "train/pos_loss/pos2": 0.18069134652614594, "train/pos_loss/pos3": 0.5528567433357239, "train/pos_loss/pos4": 1.5249992609024048, "train/pos_loss/pos5": 2.076845407485962, "train/pos_loss/pos6": 2.495225429534912, "train/pos_loss/weight0": 2.676863431930542, "train/pos_loss/weight1": 1.9221962690353394, "train/pos_loss/weight2": 1.2212661504745483, "train/pos_loss/weight3": 0.680386483669281, "train/pos_loss/weight4": 0.31453001499176025, "train/pos_loss/weight5": 0.12872782349586487, "train/pos_loss/weight6": 0.05602999031543732, "train/pos_weight_progress": 0.0924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.23158808052539825 }, { "epoch": 0.9241706161137441, "step": 390, "train/alpha": 0.2, "train/clean_ce_loss": 0.0830078125, "train/clean_loss_scale": 10.557596206665039, "train/clean_loss_weight": 2.1115193367004395, "train/combined_loss": 1.0516356229782104, "train/pos_loss/pos0": 0.16785015165805817, "train/pos_loss/pos1": 0.5157317519187927, "train/pos_loss/pos2": 0.9509205222129822, "train/pos_loss/pos3": 1.5490617752075195, "train/pos_loss/pos4": 2.0301101207733154, "train/pos_loss/pos5": 2.571826457977295, "train/pos_loss/pos6": 2.8187825679779053, "train/pos_loss/weight0": 2.1885998249053955, "train/pos_loss/weight1": 1.69672691822052, "train/pos_loss/weight2": 1.225462555885315, "train/pos_loss/weight3": 0.8545821905136108, "train/pos_loss/weight4": 0.5561536550521851, "train/pos_loss/weight5": 0.30034133791923523, "train/pos_loss/weight6": 0.1781335175037384, "train/pos_weight_progress": 0.0924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8763629794120789 }, { "epoch": 0.9241706161137441, "step": 390, "train/alpha": 0.2, "train/clean_ce_loss": 0.07568359375, "train/clean_loss_scale": 7.842981338500977, "train/clean_loss_weight": 1.5685962438583374, "train/combined_loss": 0.712302029132843, "train/pos_loss/pos0": 0.26392269134521484, "train/pos_loss/pos1": 0.5471810102462769, "train/pos_loss/pos2": 0.7727083563804626, "train/pos_loss/pos3": 1.0235592126846313, "train/pos_loss/pos4": 1.3030951023101807, "train/pos_loss/pos5": 1.596915364265442, "train/pos_loss/pos6": 2.07275128364563, "train/pos_loss/weight0": 2.6472158432006836, "train/pos_loss/weight1": 1.9222811460494995, "train/pos_loss/weight2": 1.2307015657424927, "train/pos_loss/weight3": 0.6885655522346497, "train/pos_loss/weight4": 0.32012268900871277, "train/pos_loss/weight5": 0.13293655216693878, "train/pos_loss/weight6": 0.058177050203084946, "train/pos_weight_progress": 0.0924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5935850143432617 }, { "epoch": 0.9241706161137441, "step": 390, "train/alpha": 0.2, "train/clean_ce_loss": 0.0986328125, "train/clean_loss_scale": 9.643738746643066, "train/clean_loss_weight": 1.9287477731704712, "train/combined_loss": 1.141426920890808, "train/pos_loss/pos0": 0.3795260787010193, "train/pos_loss/pos1": 0.7266243696212769, "train/pos_loss/pos2": 1.1140034198760986, "train/pos_loss/pos3": 1.537255048751831, "train/pos_loss/pos4": 2.0016613006591797, "train/pos_loss/pos5": 2.3876898288726807, "train/pos_loss/pos6": 2.704961061477661, "train/pos_loss/weight0": 2.3722875118255615, "train/pos_loss/weight1": 1.774654507637024, "train/pos_loss/weight2": 1.2224154472351074, "train/pos_loss/weight3": 0.7725609540939331, "train/pos_loss/weight4": 0.454043984413147, "train/pos_loss/weight5": 0.2528885006904602, "train/pos_loss/weight6": 0.1511492282152176, "train/pos_weight_progress": 0.0924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9511891007423401 }, { "epoch": 0.9265402843601895, "step": 391, "train/alpha": 0.2, "train/clean_ce_loss": 0.10498046875, "train/clean_loss_scale": 6.662181377410889, "train/clean_loss_weight": 1.3324363231658936, "train/combined_loss": 0.8392786979675293, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09265402843601896, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.699398934841156 }, { "epoch": 0.9265402843601895, "step": 391, "train/alpha": 0.2, "train/clean_ce_loss": 0.054443359375, "train/clean_loss_scale": 9.478377342224121, "train/clean_loss_weight": 1.895675539970398, "train/combined_loss": 0.6192416548728943, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09265402843601896, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.516034722328186 }, { "epoch": 0.9265402843601895, "step": 391, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 7.088705539703369, "train/clean_loss_weight": 1.4177411794662476, "train/combined_loss": 0.5565741658210754, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09265402843601896, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4638117849826813 }, { "epoch": 0.9265402843601895, "step": 391, "train/alpha": 0.2, "train/clean_ce_loss": 0.138671875, "train/clean_loss_scale": 7.098207950592041, "train/clean_loss_weight": 1.4196416139602661, "train/combined_loss": 1.1811861991882324, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09265402843601896, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9843217730522156 }, { "epoch": 0.9289099526066351, "step": 392, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 7.634796142578125, "train/clean_loss_weight": 1.5269593000411987, "train/combined_loss": 0.7291827201843262, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09289099526066351, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6076522469520569 }, { "epoch": 0.9289099526066351, "step": 392, "train/alpha": 0.2, "train/clean_ce_loss": 0.04736328125, "train/clean_loss_scale": 9.989056587219238, "train/clean_loss_weight": 1.9978113174438477, "train/combined_loss": 0.5677374005317688, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09289099526066351, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4731144905090332 }, { "epoch": 0.9289099526066351, "step": 392, "train/alpha": 0.2, "train/clean_ce_loss": 0.01397705078125, "train/clean_loss_scale": 26.26496124267578, "train/clean_loss_weight": 5.252992153167725, "train/combined_loss": 0.4405280351638794, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09289099526066351, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.36710670590400696 }, { "epoch": 0.9289099526066351, "step": 392, "train/alpha": 0.2, "train/clean_ce_loss": 0.111328125, "train/clean_loss_scale": 7.190284252166748, "train/clean_loss_weight": 1.4380568265914917, "train/combined_loss": 0.9605770111083984, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09289099526066351, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.800480842590332 }, { "epoch": 0.9312796208530806, "step": 393, "train/alpha": 0.2, "train/clean_ce_loss": 0.134765625, "train/clean_loss_scale": 7.810028553009033, "train/clean_loss_weight": 1.5620057582855225, "train/combined_loss": 1.2630280256271362, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09312796208530806, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0525233745574951 }, { "epoch": 0.9312796208530806, "step": 393, "train/alpha": 0.2, "train/clean_ce_loss": 0.1767578125, "train/clean_loss_scale": 7.252824306488037, "train/clean_loss_weight": 1.4505648612976074, "train/combined_loss": 1.5383920669555664, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09312796208530806, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2819933891296387 }, { "epoch": 0.9312796208530806, "step": 393, "train/alpha": 0.2, "train/clean_ce_loss": 0.115234375, "train/clean_loss_scale": 6.731531143188477, "train/clean_loss_weight": 1.3463062047958374, "train/combined_loss": 0.9308445453643799, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09312796208530806, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7757037878036499 }, { "epoch": 0.9312796208530806, "step": 393, "train/alpha": 0.2, "train/clean_ce_loss": 0.1845703125, "train/clean_loss_scale": 5.979974746704102, "train/clean_loss_weight": 1.1959949731826782, "train/combined_loss": 1.3244709968566895, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09312796208530806, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.103725790977478 }, { "epoch": 0.933649289099526, "step": 394, "train/alpha": 0.2, "train/clean_ce_loss": 0.0301513671875, "train/clean_loss_scale": 13.912078857421875, "train/clean_loss_weight": 2.7824158668518066, "train/combined_loss": 0.5033618211746216, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09336492890995261, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4194681942462921 }, { "epoch": 0.933649289099526, "step": 394, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 7.974842071533203, "train/clean_loss_weight": 1.5949684381484985, "train/combined_loss": 0.7569869756698608, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09336492890995261, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.630822479724884 }, { "epoch": 0.933649289099526, "step": 394, "train/alpha": 0.2, "train/clean_ce_loss": 0.1142578125, "train/clean_loss_scale": 10.226890563964844, "train/clean_loss_weight": 2.0453782081604004, "train/combined_loss": 1.4022024869918823, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09336492890995261, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1685020923614502 }, { "epoch": 0.933649289099526, "step": 394, "train/alpha": 0.2, "train/clean_ce_loss": 0.03076171875, "train/clean_loss_scale": 16.319185256958008, "train/clean_loss_weight": 3.2638370990753174, "train/combined_loss": 0.6024073958396912, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09336492890995261, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5020061731338501 }, { "epoch": 0.9360189573459715, "grad_norm": 13.682769775390625, "learning_rate": 2e-06, "loss": 3.7194, "step": 395 }, { "epoch": 0.9360189573459715, "step": 395, "train/alpha": 0.2, "train/clean_ce_loss": 0.07470703125, "train/clean_loss_scale": 10.009214401245117, "train/clean_loss_weight": 2.001842975616455, "train/combined_loss": 0.8973104357719421, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09360189573459715, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.747758686542511 }, { "epoch": 0.9360189573459715, "step": 395, "train/alpha": 0.2, "train/clean_ce_loss": 0.087890625, "train/clean_loss_scale": 13.711291313171387, "train/clean_loss_weight": 2.742258310317993, "train/combined_loss": 1.4461127519607544, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09360189573459715, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2050939798355103 }, { "epoch": 0.9360189573459715, "step": 395, "train/alpha": 0.2, "train/clean_ce_loss": 0.1806640625, "train/clean_loss_scale": 6.657471179962158, "train/clean_loss_weight": 1.3314942121505737, "train/combined_loss": 1.4433189630508423, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09360189573459715, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2027658224105835 }, { "epoch": 0.9360189573459715, "step": 395, "train/alpha": 0.2, "train/clean_ce_loss": 0.1630859375, "train/clean_loss_scale": 8.649080276489258, "train/clean_loss_weight": 1.7298160791397095, "train/combined_loss": 1.6926519870758057, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09360189573459715, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4105433225631714 }, { "epoch": 0.9383886255924171, "step": 396, "train/alpha": 0.2, "train/clean_ce_loss": 0.099609375, "train/clean_loss_scale": 7.415823936462402, "train/clean_loss_weight": 1.4831647872924805, "train/combined_loss": 0.8864227533340454, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.0938388625592417, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7386856079101562 }, { "epoch": 0.9383886255924171, "step": 396, "train/alpha": 0.2, "train/clean_ce_loss": 0.07861328125, "train/clean_loss_scale": 8.895854949951172, "train/clean_loss_weight": 1.7791709899902344, "train/combined_loss": 0.8391988277435303, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.0938388625592417, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6993323564529419 }, { "epoch": 0.9383886255924171, "step": 396, "train/alpha": 0.2, "train/clean_ce_loss": 0.041015625, "train/clean_loss_scale": 9.606696128845215, "train/clean_loss_weight": 1.9213392734527588, "train/combined_loss": 0.47282958030700684, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.0938388625592417, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.39402464032173157 }, { "epoch": 0.9383886255924171, "step": 396, "train/alpha": 0.2, "train/clean_ce_loss": 0.08642578125, "train/clean_loss_scale": 8.632678985595703, "train/clean_loss_weight": 1.7265357971191406, "train/combined_loss": 0.8953031897544861, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.0938388625592417, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7460860013961792 }, { "epoch": 0.9407582938388626, "step": 397, "train/alpha": 0.2, "train/clean_ce_loss": 0.169921875, "train/clean_loss_scale": 8.901520729064941, "train/clean_loss_weight": 1.780304193496704, "train/combined_loss": 1.8150757551193237, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09407582938388626, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5125631093978882 }, { "epoch": 0.9407582938388626, "step": 397, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 10.404391288757324, "train/clean_loss_weight": 2.080878257751465, "train/combined_loss": 0.9937006235122681, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09407582938388626, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.828083872795105 }, { "epoch": 0.9407582938388626, "step": 397, "train/alpha": 0.2, "train/clean_ce_loss": 0.126953125, "train/clean_loss_scale": 9.170573234558105, "train/clean_loss_weight": 1.834114670753479, "train/combined_loss": 1.397079586982727, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09407582938388626, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1642329692840576 }, { "epoch": 0.9407582938388626, "step": 397, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 9.077999114990234, "train/clean_loss_weight": 1.815599799156189, "train/combined_loss": 0.8085092902183533, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09407582938388626, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6737577319145203 }, { "epoch": 0.943127962085308, "step": 398, "train/alpha": 0.2, "train/clean_ce_loss": 0.1494140625, "train/clean_loss_scale": 8.489357948303223, "train/clean_loss_weight": 1.6978715658187866, "train/combined_loss": 1.5221152305603027, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09431279620853081, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2684293985366821 }, { "epoch": 0.943127962085308, "step": 398, "train/alpha": 0.2, "train/clean_ce_loss": 0.1298828125, "train/clean_loss_scale": 11.551741600036621, "train/clean_loss_weight": 2.3103482723236084, "train/combined_loss": 1.8004472255706787, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09431279620853081, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5003726482391357 }, { "epoch": 0.943127962085308, "step": 398, "train/alpha": 0.2, "train/clean_ce_loss": 0.099609375, "train/clean_loss_scale": 6.713871479034424, "train/clean_loss_weight": 1.3427742719650269, "train/combined_loss": 0.8025174140930176, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09431279620853081, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6687645316123962 }, { "epoch": 0.943127962085308, "step": 398, "train/alpha": 0.2, "train/clean_ce_loss": 0.05419921875, "train/clean_loss_scale": 9.053932189941406, "train/clean_loss_weight": 1.810786485671997, "train/combined_loss": 0.5888592600822449, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09431279620853081, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4907160699367523 }, { "epoch": 0.9454976303317536, "step": 399, "train/alpha": 0.2, "train/clean_ce_loss": 0.1142578125, "train/clean_loss_scale": 7.461524486541748, "train/clean_loss_weight": 1.4923049211502075, "train/combined_loss": 1.0230449438095093, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09454976303317536, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8525374531745911 }, { "epoch": 0.9454976303317536, "step": 399, "train/alpha": 0.2, "train/clean_ce_loss": 0.09326171875, "train/clean_loss_scale": 8.83701229095459, "train/clean_loss_weight": 1.7674025297164917, "train/combined_loss": 0.9889859557151794, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09454976303317536, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8241549730300903 }, { "epoch": 0.9454976303317536, "step": 399, "train/alpha": 0.2, "train/clean_ce_loss": 0.16015625, "train/clean_loss_scale": 6.52653169631958, "train/clean_loss_weight": 1.305306315422058, "train/combined_loss": 1.2543177604675293, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09454976303317536, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0452648401260376 }, { "epoch": 0.9454976303317536, "step": 399, "train/alpha": 0.2, "train/clean_ce_loss": 0.0216064453125, "train/clean_loss_scale": 15.16562271118164, "train/clean_loss_weight": 3.0331246852874756, "train/combined_loss": 0.393210232257843, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.09454976303317536, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3276751935482025 }, { "epoch": 0.9478672985781991, "grad_norm": 20.694034576416016, "learning_rate": 2e-06, "loss": 3.6205, "step": 400 }, { "epoch": 0.9478672985781991, "step": 400, "train/alpha": 0.2, "train/clean_ce_loss": 0.06689453125, "train/clean_loss_scale": 7.963391304016113, "train/clean_loss_weight": 1.5926783084869385, "train/combined_loss": 0.639248788356781, "train/pos_loss/pos0": 0.20401476323604584, "train/pos_loss/pos1": 0.46126389503479004, "train/pos_loss/pos2": 0.7175135016441345, "train/pos_loss/pos3": 1.0375293493270874, "train/pos_loss/pos4": 1.4852895736694336, "train/pos_loss/pos5": 1.9670048952102661, "train/pos_loss/pos6": 2.227534294128418, "train/pos_loss/weight0": 2.798125982284546, "train/pos_loss/weight1": 1.927965760231018, "train/pos_loss/weight2": 1.1900206804275513, "train/pos_loss/weight3": 0.632128894329071, "train/pos_loss/weight4": 0.2911336421966553, "train/pos_loss/weight5": 0.11385978013277054, "train/pos_loss/weight6": 0.04676475375890732, "train/pos_weight_progress": 0.0947867298578199, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5327073335647583 }, { "epoch": 0.9478672985781991, "step": 400, "train/alpha": 0.2, "train/clean_ce_loss": 0.173828125, "train/clean_loss_scale": 7.253743648529053, "train/clean_loss_weight": 1.4507488012313843, "train/combined_loss": 1.5130856037139893, "train/pos_loss/pos0": 0.5247541666030884, "train/pos_loss/pos1": 1.1182091236114502, "train/pos_loss/pos2": 1.4987866878509521, "train/pos_loss/pos3": 1.9616683721542358, "train/pos_loss/pos4": 2.54457950592041, "train/pos_loss/pos5": 2.7812204360961914, "train/pos_loss/pos6": 3.5495712757110596, "train/pos_loss/weight0": 2.4043967723846436, "train/pos_loss/weight1": 1.8022778034210205, "train/pos_loss/weight2": 1.229846715927124, "train/pos_loss/weight3": 0.7637118101119995, "train/pos_loss/weight4": 0.44035691022872925, "train/pos_loss/weight5": 0.23354527354240417, "train/pos_loss/weight6": 0.125865176320076, "train/pos_weight_progress": 0.0947867298578199, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2609046697616577 }, { "epoch": 0.9478672985781991, "step": 400, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 7.181691646575928, "train/clean_loss_weight": 1.4363383054733276, "train/combined_loss": 0.5722910761833191, "train/pos_loss/pos0": 0.2567758858203888, "train/pos_loss/pos1": 0.4145158529281616, "train/pos_loss/pos2": 0.7194736003875732, "train/pos_loss/pos3": 1.0711783170700073, "train/pos_loss/pos4": 1.4881004095077515, "train/pos_loss/pos5": 1.8681226968765259, "train/pos_loss/pos6": 2.2119076251983643, "train/pos_loss/weight0": 3.1469852924346924, "train/pos_loss/weight1": 2.0563063621520996, "train/pos_loss/weight2": 1.0702272653579712, "train/pos_loss/weight3": 0.4848019480705261, "train/pos_loss/weight4": 0.17590270936489105, "train/pos_loss/weight5": 0.05056704208254814, "train/pos_loss/weight6": 0.015209503471851349, "train/pos_weight_progress": 0.0947867298578199, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.47690922021865845 }, { "epoch": 0.9478672985781991, "step": 400, "train/alpha": 0.2, "train/clean_ce_loss": 0.06884765625, "train/clean_loss_scale": 9.300277709960938, "train/clean_loss_weight": 1.8600555658340454, "train/combined_loss": 0.7683627605438232, "train/pos_loss/pos0": 0.22903892397880554, "train/pos_loss/pos1": 0.4520241320133209, "train/pos_loss/pos2": 0.7860225439071655, "train/pos_loss/pos3": 1.0905570983886719, "train/pos_loss/pos4": 1.492998719215393, "train/pos_loss/pos5": 2.0088651180267334, "train/pos_loss/pos6": 2.2325656414031982, "train/pos_loss/weight0": 2.4692811965942383, "train/pos_loss/weight1": 1.7951353788375854, "train/pos_loss/weight2": 1.2264175415039062, "train/pos_loss/weight3": 0.7577676177024841, "train/pos_loss/weight4": 0.42452847957611084, "train/pos_loss/weight5": 0.2130790650844574, "train/pos_loss/weight6": 0.11379053443670273, "train/pos_weight_progress": 0.0947867298578199, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.640302300453186 }, { "epoch": 0.9502369668246445, "step": 401, "train/alpha": 0.2, "train/clean_ce_loss": 0.1318359375, "train/clean_loss_scale": 8.425941467285156, "train/clean_loss_weight": 1.6851882934570312, "train/combined_loss": 1.333010196685791, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.09502369668246445, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1108418703079224 }, { "epoch": 0.9502369668246445, "step": 401, "train/alpha": 0.2, "train/clean_ce_loss": 0.125, "train/clean_loss_scale": 9.278287887573242, "train/clean_loss_weight": 1.8556575775146484, "train/combined_loss": 1.3917431831359863, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.09502369668246445, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1597859859466553 }, { "epoch": 0.9502369668246445, "step": 401, "train/alpha": 0.2, "train/clean_ce_loss": 0.1044921875, "train/clean_loss_scale": 8.900857925415039, "train/clean_loss_weight": 1.7801716327667236, "train/combined_loss": 1.1160842180252075, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.09502369668246445, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.930070161819458 }, { "epoch": 0.9502369668246445, "step": 401, "train/alpha": 0.2, "train/clean_ce_loss": 0.1923828125, "train/clean_loss_scale": 7.829015254974365, "train/clean_loss_weight": 1.565803050994873, "train/combined_loss": 1.8074016571044922, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.09502369668246445, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.506168007850647 }, { "epoch": 0.95260663507109, "step": 402, "train/alpha": 0.2, "train/clean_ce_loss": 0.1201171875, "train/clean_loss_scale": 10.711641311645508, "train/clean_loss_weight": 2.1423282623291016, "train/combined_loss": 1.5439826250076294, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.095260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2866522073745728 }, { "epoch": 0.95260663507109, "step": 402, "train/alpha": 0.2, "train/clean_ce_loss": 0.10400390625, "train/clean_loss_scale": 8.852154731750488, "train/clean_loss_weight": 1.7704309225082397, "train/combined_loss": 1.1047903299331665, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.095260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9206586480140686 }, { "epoch": 0.95260663507109, "step": 402, "train/alpha": 0.2, "train/clean_ce_loss": 0.07568359375, "train/clean_loss_scale": 7.996636390686035, "train/clean_loss_weight": 1.5993273258209229, "train/combined_loss": 0.7262570261955261, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.095260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6052141785621643 }, { "epoch": 0.95260663507109, "step": 402, "train/alpha": 0.2, "train/clean_ce_loss": 0.107421875, "train/clean_loss_scale": 10.520407676696777, "train/clean_loss_weight": 2.104081630706787, "train/combined_loss": 1.3561463356018066, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.095260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1301219463348389 }, { "epoch": 0.9549763033175356, "step": 403, "train/alpha": 0.2, "train/clean_ce_loss": 0.138671875, "train/clean_loss_scale": 5.683866024017334, "train/clean_loss_weight": 1.1367732286453247, "train/combined_loss": 0.9458308219909668, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.09549763033175356, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7881923317909241 }, { "epoch": 0.9549763033175356, "step": 403, "train/alpha": 0.2, "train/clean_ce_loss": 0.057861328125, "train/clean_loss_scale": 10.741645812988281, "train/clean_loss_weight": 2.148329257965088, "train/combined_loss": 0.7458310723304749, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.09549763033175356, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6215258836746216 }, { "epoch": 0.9549763033175356, "step": 403, "train/alpha": 0.2, "train/clean_ce_loss": 0.06298828125, "train/clean_loss_scale": 10.230950355529785, "train/clean_loss_weight": 2.046190023422241, "train/combined_loss": 0.773315966129303, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.09549763033175356, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6444299817085266 }, { "epoch": 0.9549763033175356, "step": 403, "train/alpha": 0.2, "train/clean_ce_loss": 0.03857421875, "train/clean_loss_scale": 13.75910758972168, "train/clean_loss_weight": 2.751821517944336, "train/combined_loss": 0.6368961930274963, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.09549763033175356, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5307468175888062 }, { "epoch": 0.957345971563981, "step": 404, "train/alpha": 0.2, "train/clean_ce_loss": 0.10107421875, "train/clean_loss_scale": 8.939682960510254, "train/clean_loss_weight": 1.7879365682601929, "train/combined_loss": 1.0842857360839844, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.09573459715639811, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9035714864730835 }, { "epoch": 0.957345971563981, "step": 404, "train/alpha": 0.2, "train/clean_ce_loss": 0.15625, "train/clean_loss_scale": 8.49985122680664, "train/clean_loss_weight": 1.6999702453613281, "train/combined_loss": 1.5937221050262451, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.09573459715639811, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3281017541885376 }, { "epoch": 0.957345971563981, "step": 404, "train/alpha": 0.2, "train/clean_ce_loss": 0.10693359375, "train/clean_loss_scale": 9.451580047607422, "train/clean_loss_weight": 1.8903160095214844, "train/combined_loss": 1.2128297090530396, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.09573459715639811, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0106914043426514 }, { "epoch": 0.957345971563981, "step": 404, "train/alpha": 0.2, "train/clean_ce_loss": 0.126953125, "train/clean_loss_scale": 10.507364273071289, "train/clean_loss_weight": 2.101472854614258, "train/combined_loss": 1.600731372833252, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.09573459715639811, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3339427709579468 }, { "epoch": 0.9597156398104265, "grad_norm": 12.981256484985352, "learning_rate": 2e-06, "loss": 3.5948, "step": 405 }, { "epoch": 0.9597156398104265, "step": 405, "train/alpha": 0.2, "train/clean_ce_loss": 0.00604248046875, "train/clean_loss_scale": 79.23739624023438, "train/clean_loss_weight": 15.847479820251465, "train/combined_loss": 0.5745485424995422, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.09597156398104266, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.47879043221473694 }, { "epoch": 0.9597156398104265, "step": 405, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 7.628871440887451, "train/clean_loss_weight": 1.525774359703064, "train/combined_loss": 0.8001375198364258, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.09597156398104266, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6667812466621399 }, { "epoch": 0.9597156398104265, "step": 405, "train/alpha": 0.2, "train/clean_ce_loss": 0.032470703125, "train/clean_loss_scale": 20.037281036376953, "train/clean_loss_weight": 4.007456302642822, "train/combined_loss": 0.7807495594024658, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.09597156398104266, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6506246328353882 }, { "epoch": 0.9597156398104265, "step": 405, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 7.933310508728027, "train/clean_loss_weight": 1.5866621732711792, "train/combined_loss": 0.7576931715011597, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.09597156398104266, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6314109563827515 }, { "epoch": 0.9620853080568721, "step": 406, "train/alpha": 0.2, "train/clean_ce_loss": 0.1650390625, "train/clean_loss_scale": 9.3538179397583, "train/clean_loss_weight": 1.8707636594772339, "train/combined_loss": 1.8524943590164185, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.0962085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5437452793121338 }, { "epoch": 0.9620853080568721, "step": 406, "train/alpha": 0.2, "train/clean_ce_loss": 0.035400390625, "train/clean_loss_scale": 16.330286026000977, "train/clean_loss_weight": 3.266057252883911, "train/combined_loss": 0.6937181949615479, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.0962085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.578098475933075 }, { "epoch": 0.9620853080568721, "step": 406, "train/alpha": 0.2, "train/clean_ce_loss": 0.1259765625, "train/clean_loss_scale": 9.773529052734375, "train/clean_loss_weight": 1.954705834388733, "train/combined_loss": 1.477482795715332, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.0962085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2312356233596802 }, { "epoch": 0.9620853080568721, "step": 406, "train/alpha": 0.2, "train/clean_ce_loss": 0.0228271484375, "train/clean_loss_scale": 18.035625457763672, "train/clean_loss_weight": 3.6071250438690186, "train/combined_loss": 0.4940422773361206, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.0962085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.41170188784599304 }, { "epoch": 0.9644549763033176, "step": 407, "train/alpha": 0.2, "train/clean_ce_loss": 0.0703125, "train/clean_loss_scale": 9.632391929626465, "train/clean_loss_weight": 1.926478385925293, "train/combined_loss": 0.8127330541610718, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.09644549763033175, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6772775650024414 }, { "epoch": 0.9644549763033176, "step": 407, "train/alpha": 0.2, "train/clean_ce_loss": 0.09326171875, "train/clean_loss_scale": 8.926398277282715, "train/clean_loss_weight": 1.785279631614685, "train/combined_loss": 0.998989462852478, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.09644549763033175, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8324912190437317 }, { "epoch": 0.9644549763033176, "step": 407, "train/alpha": 0.2, "train/clean_ce_loss": 0.1162109375, "train/clean_loss_scale": 9.611342430114746, "train/clean_loss_weight": 1.9222685098648071, "train/combined_loss": 1.340331792831421, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.09644549763033175, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.116943120956421 }, { "epoch": 0.9644549763033176, "step": 407, "train/alpha": 0.2, "train/clean_ce_loss": 0.12158203125, "train/clean_loss_scale": 7.455819606781006, "train/clean_loss_weight": 1.491163969039917, "train/combined_loss": 1.0877923965454102, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.09644549763033175, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9064936637878418 }, { "epoch": 0.966824644549763, "step": 408, "train/alpha": 0.2, "train/clean_ce_loss": 0.060302734375, "train/clean_loss_scale": 9.211294174194336, "train/clean_loss_weight": 1.8422588109970093, "train/combined_loss": 0.6665594577789307, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.0966824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5554662346839905 }, { "epoch": 0.966824644549763, "step": 408, "train/alpha": 0.2, "train/clean_ce_loss": 0.039306640625, "train/clean_loss_scale": 15.084671020507812, "train/clean_loss_weight": 3.0169341564178467, "train/combined_loss": 0.7115132808685303, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.0966824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5929277539253235 }, { "epoch": 0.966824644549763, "step": 408, "train/alpha": 0.2, "train/clean_ce_loss": 0.0198974609375, "train/clean_loss_scale": 15.866447448730469, "train/clean_loss_weight": 3.1732895374298096, "train/combined_loss": 0.37884241342544556, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.0966824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.31570202112197876 }, { "epoch": 0.966824644549763, "step": 408, "train/alpha": 0.2, "train/clean_ce_loss": 0.1640625, "train/clean_loss_scale": 8.708921432495117, "train/clean_loss_weight": 1.7417843341827393, "train/combined_loss": 1.714568853378296, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.0966824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4288073778152466 }, { "epoch": 0.9691943127962085, "step": 409, "train/alpha": 0.2, "train/clean_ce_loss": 0.07861328125, "train/clean_loss_scale": 8.64244270324707, "train/clean_loss_weight": 1.728488564491272, "train/combined_loss": 0.8152928948402405, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.09691943127962085, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6794107556343079 }, { "epoch": 0.9691943127962085, "step": 409, "train/alpha": 0.2, "train/clean_ce_loss": 0.181640625, "train/clean_loss_scale": 9.154406547546387, "train/clean_loss_weight": 1.8308813571929932, "train/combined_loss": 1.9953745603561401, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.09691943127962085, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.6628121137619019 }, { "epoch": 0.9691943127962085, "step": 409, "train/alpha": 0.2, "train/clean_ce_loss": 0.125, "train/clean_loss_scale": 11.197447776794434, "train/clean_loss_weight": 2.2394895553588867, "train/combined_loss": 1.679617166519165, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.09691943127962085, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3996809720993042 }, { "epoch": 0.9691943127962085, "step": 409, "train/alpha": 0.2, "train/clean_ce_loss": 0.099609375, "train/clean_loss_scale": 7.559225082397461, "train/clean_loss_weight": 1.5118449926376343, "train/combined_loss": 0.9035636186599731, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.09691943127962085, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7529696822166443 }, { "epoch": 0.9715639810426541, "grad_norm": 14.604696273803711, "learning_rate": 2e-06, "loss": 3.286, "step": 410 }, { "epoch": 0.9715639810426541, "step": 410, "train/alpha": 0.2, "train/clean_ce_loss": 0.00189208984375, "train/clean_loss_scale": 183.1365509033203, "train/clean_loss_weight": 36.62731170654297, "train/combined_loss": 0.41581296920776367, "train/pos_loss/pos0": 0.0014469678280875087, "train/pos_loss/pos1": 0.03554851561784744, "train/pos_loss/pos2": 0.3136785924434662, "train/pos_loss/pos3": 0.9215026497840881, "train/pos_loss/pos4": 1.3800384998321533, "train/pos_loss/pos5": 1.9980525970458984, "train/pos_loss/pos6": 2.4381661415100098, "train/pos_loss/weight0": 2.4520223140716553, "train/pos_loss/weight1": 1.8193824291229248, "train/pos_loss/weight2": 1.226851224899292, "train/pos_loss/weight3": 0.7581425905227661, "train/pos_loss/weight4": 0.41848117113113403, "train/pos_loss/weight5": 0.21051788330078125, "train/pos_loss/weight6": 0.11460253596305847, "train/pos_weight_progress": 0.0971563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.34651079773902893 }, { "epoch": 0.9715639810426541, "step": 410, "train/alpha": 0.2, "train/clean_ce_loss": 0.1435546875, "train/clean_loss_scale": 8.277764320373535, "train/clean_loss_weight": 1.655552864074707, "train/combined_loss": 1.4259743690490723, "train/pos_loss/pos0": 0.48709723353385925, "train/pos_loss/pos1": 0.7823089957237244, "train/pos_loss/pos2": 1.3098961114883423, "train/pos_loss/pos3": 1.8745468854904175, "train/pos_loss/pos4": 2.3321080207824707, "train/pos_loss/pos5": 2.8806092739105225, "train/pos_loss/pos6": 3.303359270095825, "train/pos_loss/weight0": 2.202463388442993, "train/pos_loss/weight1": 1.6650432348251343, "train/pos_loss/weight2": 1.2572871446609497, "train/pos_loss/weight3": 0.8789477348327637, "train/pos_loss/weight4": 0.5341921448707581, "train/pos_loss/weight5": 0.2854098677635193, "train/pos_loss/weight6": 0.17665627598762512, "train/pos_weight_progress": 0.0971563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1883119344711304 }, { "epoch": 0.9715639810426541, "step": 410, "train/alpha": 0.2, "train/clean_ce_loss": 0.150390625, "train/clean_loss_scale": 6.9737548828125, "train/clean_loss_weight": 1.394750952720642, "train/combined_loss": 1.2585448026657104, "train/pos_loss/pos0": 0.5193415880203247, "train/pos_loss/pos1": 1.0471819639205933, "train/pos_loss/pos2": 1.5937473773956299, "train/pos_loss/pos3": 2.013606309890747, "train/pos_loss/pos4": 2.5701370239257812, "train/pos_loss/pos5": 3.1306331157684326, "train/pos_loss/pos6": 3.5309627056121826, "train/pos_loss/weight0": 3.067610025405884, "train/pos_loss/weight1": 1.9854793548583984, "train/pos_loss/weight2": 1.118874192237854, "train/pos_loss/weight3": 0.5345796942710876, "train/pos_loss/weight4": 0.20781384408473969, "train/pos_loss/weight5": 0.06437934190034866, "train/pos_loss/weight6": 0.02126314863562584, "train/pos_weight_progress": 0.0971563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0487873554229736 }, { "epoch": 0.9715639810426541, "step": 410, "train/alpha": 0.2, "train/clean_ce_loss": 0.05908203125, "train/clean_loss_scale": 7.0967793464660645, "train/clean_loss_weight": 1.419355869293213, "train/combined_loss": 0.5031505823135376, "train/pos_loss/pos0": 0.1609470695257187, "train/pos_loss/pos1": 0.3832107186317444, "train/pos_loss/pos2": 0.6913536787033081, "train/pos_loss/pos3": 0.9578429460525513, "train/pos_loss/pos4": 1.3591477870941162, "train/pos_loss/pos5": 1.7232227325439453, "train/pos_loss/pos6": 1.9973492622375488, "train/pos_loss/weight0": 3.0995795726776123, "train/pos_loss/weight1": 1.9949593544006348, "train/pos_loss/weight2": 1.1159064769744873, "train/pos_loss/weight3": 0.5143685340881348, "train/pos_loss/weight4": 0.1956045925617218, "train/pos_loss/weight5": 0.0603860467672348, "train/pos_loss/weight6": 0.01919585093855858, "train/pos_weight_progress": 0.0971563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.419292151927948 }, { "epoch": 0.9739336492890995, "step": 411, "train/alpha": 0.2, "train/clean_ce_loss": 0.0189208984375, "train/clean_loss_scale": 28.422319412231445, "train/clean_loss_weight": 5.684463977813721, "train/combined_loss": 0.6453309655189514, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.09739336492890996, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5377758145332336 }, { "epoch": 0.9739336492890995, "step": 411, "train/alpha": 0.2, "train/clean_ce_loss": 0.0751953125, "train/clean_loss_scale": 8.645822525024414, "train/clean_loss_weight": 1.729164481163025, "train/combined_loss": 0.7801504135131836, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.09739336492890996, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6501253247261047 }, { "epoch": 0.9739336492890995, "step": 411, "train/alpha": 0.2, "train/clean_ce_loss": 0.140625, "train/clean_loss_scale": 7.121827125549316, "train/clean_loss_weight": 1.4243654012680054, "train/combined_loss": 1.2018083333969116, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.09739336492890996, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0015069246292114 }, { "epoch": 0.9739336492890995, "step": 411, "train/alpha": 0.2, "train/clean_ce_loss": 0.1923828125, "train/clean_loss_scale": 8.664360046386719, "train/clean_loss_weight": 1.7328720092773438, "train/combined_loss": 2.000248670578003, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.09739336492890996, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.6668739318847656 }, { "epoch": 0.976303317535545, "step": 412, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 9.82054328918457, "train/clean_loss_weight": 1.9641087055206299, "train/combined_loss": 0.8343625664710999, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.0976303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6953021287918091 }, { "epoch": 0.976303317535545, "step": 412, "train/alpha": 0.2, "train/clean_ce_loss": 0.08544921875, "train/clean_loss_scale": 7.831779956817627, "train/clean_loss_weight": 1.5663560628890991, "train/combined_loss": 0.8030633926391602, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.0976303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6692194938659668 }, { "epoch": 0.976303317535545, "step": 412, "train/alpha": 0.2, "train/clean_ce_loss": 0.1611328125, "train/clean_loss_scale": 8.296058654785156, "train/clean_loss_weight": 1.6592117547988892, "train/combined_loss": 1.6041206121444702, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.0976303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3367671966552734 }, { "epoch": 0.976303317535545, "step": 412, "train/alpha": 0.2, "train/clean_ce_loss": 0.0498046875, "train/clean_loss_scale": 12.14549446105957, "train/clean_loss_weight": 2.4290988445281982, "train/combined_loss": 0.7258830666542053, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.0976303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6049025654792786 }, { "epoch": 0.9786729857819905, "step": 413, "train/alpha": 0.2, "train/clean_ce_loss": 0.09375, "train/clean_loss_scale": 10.190974235534668, "train/clean_loss_weight": 2.0381948947906494, "train/combined_loss": 1.146484613418579, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.09786729857819905, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9554038643836975 }, { "epoch": 0.9786729857819905, "step": 413, "train/alpha": 0.2, "train/clean_ce_loss": 0.09619140625, "train/clean_loss_scale": 10.967591285705566, "train/clean_loss_weight": 2.1935184001922607, "train/combined_loss": 1.2659856081008911, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.09786729857819905, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0549880266189575 }, { "epoch": 0.9786729857819905, "step": 413, "train/alpha": 0.2, "train/clean_ce_loss": 0.09716796875, "train/clean_loss_scale": 9.27877140045166, "train/clean_loss_weight": 1.8557542562484741, "train/combined_loss": 1.0819191932678223, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.09786729857819905, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9015993475914001 }, { "epoch": 0.9786729857819905, "step": 413, "train/alpha": 0.2, "train/clean_ce_loss": 0.12060546875, "train/clean_loss_scale": 8.807174682617188, "train/clean_loss_weight": 1.7614349126815796, "train/combined_loss": 1.2746320962905884, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.09786729857819905, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0621933937072754 }, { "epoch": 0.981042654028436, "step": 414, "train/alpha": 0.2, "train/clean_ce_loss": 0.162109375, "train/clean_loss_scale": 8.49604320526123, "train/clean_loss_weight": 1.6992086172103882, "train/combined_loss": 1.6527458429336548, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.0981042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3772882223129272 }, { "epoch": 0.981042654028436, "step": 414, "train/alpha": 0.2, "train/clean_ce_loss": 0.06982421875, "train/clean_loss_scale": 10.434162139892578, "train/clean_loss_weight": 2.0868325233459473, "train/combined_loss": 0.874268651008606, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.0981042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7285572290420532 }, { "epoch": 0.981042654028436, "step": 414, "train/alpha": 0.2, "train/clean_ce_loss": 0.0078125, "train/clean_loss_scale": 46.68949890136719, "train/clean_loss_weight": 9.337900161743164, "train/combined_loss": 0.4377140402793884, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.0981042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.36476171016693115 }, { "epoch": 0.981042654028436, "step": 414, "train/alpha": 0.2, "train/clean_ce_loss": 0.0245361328125, "train/clean_loss_scale": 21.603994369506836, "train/clean_loss_weight": 4.320798873901367, "train/combined_loss": 0.6360941529273987, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.0981042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5300784707069397 }, { "epoch": 0.9834123222748815, "grad_norm": 18.466800689697266, "learning_rate": 2e-06, "loss": 3.7787, "step": 415 }, { "epoch": 0.9834123222748815, "step": 415, "train/alpha": 0.2, "train/clean_ce_loss": 0.01446533203125, "train/clean_loss_scale": 28.775299072265625, "train/clean_loss_weight": 5.755059719085693, "train/combined_loss": 0.4994931221008301, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.09834123222748815, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4162442684173584 }, { "epoch": 0.9834123222748815, "step": 415, "train/alpha": 0.2, "train/clean_ce_loss": 0.10498046875, "train/clean_loss_scale": 12.775235176086426, "train/clean_loss_weight": 2.555047035217285, "train/combined_loss": 1.6093802452087402, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.09834123222748815, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.341150164604187 }, { "epoch": 0.9834123222748815, "step": 415, "train/alpha": 0.2, "train/clean_ce_loss": 0.1513671875, "train/clean_loss_scale": 9.084169387817383, "train/clean_loss_weight": 1.8168338537216187, "train/combined_loss": 1.6500542163848877, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.09834123222748815, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3750451803207397 }, { "epoch": 0.9834123222748815, "step": 415, "train/alpha": 0.2, "train/clean_ce_loss": 0.09765625, "train/clean_loss_scale": 8.118736267089844, "train/clean_loss_weight": 1.6237472295761108, "train/combined_loss": 0.9514143466949463, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.09834123222748815, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7928453087806702 }, { "epoch": 0.985781990521327, "step": 416, "train/alpha": 0.2, "train/clean_ce_loss": 0.00131988525390625, "train/clean_loss_scale": 311.9067687988281, "train/clean_loss_weight": 62.38135528564453, "train/combined_loss": 0.4940173625946045, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.0985781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4116811454296112 }, { "epoch": 0.985781990521327, "step": 416, "train/alpha": 0.2, "train/clean_ce_loss": 0.158203125, "train/clean_loss_scale": 10.620424270629883, "train/clean_loss_weight": 2.124084949493408, "train/combined_loss": 2.016221284866333, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.0985781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.6801843643188477 }, { "epoch": 0.985781990521327, "step": 416, "train/alpha": 0.2, "train/clean_ce_loss": 0.14453125, "train/clean_loss_scale": 10.910435676574707, "train/clean_loss_weight": 2.1820871829986572, "train/combined_loss": 1.8922786712646484, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.0985781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5768989324569702 }, { "epoch": 0.985781990521327, "step": 416, "train/alpha": 0.2, "train/clean_ce_loss": 0.07177734375, "train/clean_loss_scale": 11.979796409606934, "train/clean_loss_weight": 2.3959593772888184, "train/combined_loss": 1.0318535566329956, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.0985781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8598779439926147 }, { "epoch": 0.9881516587677726, "step": 417, "train/alpha": 0.2, "train/clean_ce_loss": 0.1591796875, "train/clean_loss_scale": 10.00711441040039, "train/clean_loss_weight": 2.001422882080078, "train/combined_loss": 1.911515235900879, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.09881516587677724, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5929293632507324 }, { "epoch": 0.9881516587677726, "step": 417, "train/alpha": 0.2, "train/clean_ce_loss": 0.11279296875, "train/clean_loss_scale": 9.725010871887207, "train/clean_loss_weight": 1.9450021982192993, "train/combined_loss": 1.3162953853607178, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.09881516587677724, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0969128608703613 }, { "epoch": 0.9881516587677726, "step": 417, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 12.097012519836426, "train/clean_loss_weight": 2.419402599334717, "train/combined_loss": 1.2120639085769653, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.09881516587677724, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.010053277015686 }, { "epoch": 0.9881516587677726, "step": 417, "train/alpha": 0.2, "train/clean_ce_loss": 0.1181640625, "train/clean_loss_scale": 10.881589889526367, "train/clean_loss_weight": 2.1763179302215576, "train/combined_loss": 1.5429754257202148, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.09881516587677724, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2858128547668457 }, { "epoch": 0.990521327014218, "step": 418, "train/alpha": 0.2, "train/clean_ce_loss": 0.005096435546875, "train/clean_loss_scale": 94.12071990966797, "train/clean_loss_weight": 18.82414436340332, "train/combined_loss": 0.5756162405014038, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.0990521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4796801805496216 }, { "epoch": 0.990521327014218, "step": 418, "train/alpha": 0.2, "train/clean_ce_loss": 0.01434326171875, "train/clean_loss_scale": 39.77065658569336, "train/clean_loss_weight": 7.954131603240967, "train/combined_loss": 0.6845291256904602, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.0990521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.570440948009491 }, { "epoch": 0.990521327014218, "step": 418, "train/alpha": 0.2, "train/clean_ce_loss": 0.057861328125, "train/clean_loss_scale": 9.817476272583008, "train/clean_loss_weight": 1.9634952545166016, "train/combined_loss": 0.6816626787185669, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.0990521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5680522322654724 }, { "epoch": 0.990521327014218, "step": 418, "train/alpha": 0.2, "train/clean_ce_loss": 0.033447265625, "train/clean_loss_scale": 10.534178733825684, "train/clean_loss_weight": 2.1068358421325684, "train/combined_loss": 0.42280736565589905, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.0990521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3523394763469696 }, { "epoch": 0.9928909952606635, "step": 419, "train/alpha": 0.2, "train/clean_ce_loss": 0.027587890625, "train/clean_loss_scale": 18.685813903808594, "train/clean_loss_weight": 3.7371628284454346, "train/combined_loss": 0.6186026334762573, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.09928909952606635, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5155022144317627 }, { "epoch": 0.9928909952606635, "step": 419, "train/alpha": 0.2, "train/clean_ce_loss": 0.0517578125, "train/clean_loss_scale": 12.099727630615234, "train/clean_loss_weight": 2.419945478439331, "train/combined_loss": 0.7515065670013428, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.09928909952606635, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.626255452632904 }, { "epoch": 0.9928909952606635, "step": 419, "train/alpha": 0.2, "train/clean_ce_loss": 0.0830078125, "train/clean_loss_scale": 9.646744728088379, "train/clean_loss_weight": 1.9293489456176758, "train/combined_loss": 0.9609062671661377, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.09928909952606635, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8007552027702332 }, { "epoch": 0.9928909952606635, "step": 419, "train/alpha": 0.2, "train/clean_ce_loss": 0.1748046875, "train/clean_loss_scale": 7.483652591705322, "train/clean_loss_weight": 1.4967305660247803, "train/combined_loss": 1.5698131322860718, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.09928909952606635, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3081775903701782 }, { "epoch": 0.995260663507109, "grad_norm": 13.632644653320312, "learning_rate": 2e-06, "loss": 3.8854, "step": 420 }, { "epoch": 0.995260663507109, "step": 420, "train/alpha": 0.2, "train/clean_ce_loss": 0.07177734375, "train/clean_loss_scale": 8.9391508102417, "train/clean_loss_weight": 1.7878302335739136, "train/combined_loss": 0.7699542045593262, "train/pos_loss/pos0": 0.18665574491024017, "train/pos_loss/pos1": 0.37843888998031616, "train/pos_loss/pos2": 0.6955010890960693, "train/pos_loss/pos3": 1.1805927753448486, "train/pos_loss/pos4": 1.557021141052246, "train/pos_loss/pos5": 2.087883234024048, "train/pos_loss/pos6": 2.5008065700531006, "train/pos_loss/weight0": 2.3446476459503174, "train/pos_loss/weight1": 1.781735897064209, "train/pos_loss/weight2": 1.2329216003417969, "train/pos_loss/weight3": 0.7919583916664124, "train/pos_loss/weight4": 0.45709463953971863, "train/pos_loss/weight5": 0.2511916160583496, "train/pos_loss/weight6": 0.14045032858848572, "train/pos_weight_progress": 0.0995260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6416285037994385 }, { "epoch": 0.995260663507109, "step": 420, "train/alpha": 0.2, "train/clean_ce_loss": 0.056396484375, "train/clean_loss_scale": 11.07593822479248, "train/clean_loss_weight": 2.2151877880096436, "train/combined_loss": 0.74957275390625, "train/pos_loss/pos0": 0.14941853284835815, "train/pos_loss/pos1": 0.4251525104045868, "train/pos_loss/pos2": 0.735166609287262, "train/pos_loss/pos3": 1.2785693407058716, "train/pos_loss/pos4": 1.7939443588256836, "train/pos_loss/pos5": 2.2937405109405518, "train/pos_loss/pos6": 2.7568018436431885, "train/pos_loss/weight0": 2.5327768325805664, "train/pos_loss/weight1": 1.8235716819763184, "train/pos_loss/weight2": 1.2423479557037354, "train/pos_loss/weight3": 0.7454328536987305, "train/pos_loss/weight4": 0.38612276315689087, "train/pos_loss/weight5": 0.1806345134973526, "train/pos_loss/weight6": 0.08911355584859848, "train/pos_weight_progress": 0.0995260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6246439814567566 }, { "epoch": 0.995260663507109, "step": 420, "train/alpha": 0.2, "train/clean_ce_loss": 0.1181640625, "train/clean_loss_scale": 8.846857070922852, "train/clean_loss_weight": 1.7693713903427124, "train/combined_loss": 1.2544567584991455, "train/pos_loss/pos0": 0.4076938033103943, "train/pos_loss/pos1": 0.7233487963676453, "train/pos_loss/pos2": 1.3878368139266968, "train/pos_loss/pos3": 2.0026583671569824, "train/pos_loss/pos4": 2.6713945865631104, "train/pos_loss/pos5": 2.8809258937835693, "train/pos_loss/pos6": 3.3425281047821045, "train/pos_loss/weight0": 2.5296921730041504, "train/pos_loss/weight1": 1.8763976097106934, "train/pos_loss/weight2": 1.2359740734100342, "train/pos_loss/weight3": 0.7459254264831543, "train/pos_loss/weight4": 0.37151455879211426, "train/pos_loss/weight5": 0.1632072478532791, "train/pos_loss/weight6": 0.07728807628154755, "train/pos_weight_progress": 0.0995260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0453805923461914 }, { "epoch": 0.995260663507109, "step": 420, "train/alpha": 0.2, "train/clean_ce_loss": 0.0257568359375, "train/clean_loss_scale": 15.03927230834961, "train/clean_loss_weight": 3.007854461669922, "train/combined_loss": 0.46483686566352844, "train/pos_loss/pos0": 0.06695621460676193, "train/pos_loss/pos1": 0.16931484639644623, "train/pos_loss/pos2": 0.5978855490684509, "train/pos_loss/pos3": 1.4841389656066895, "train/pos_loss/pos4": 2.0043580532073975, "train/pos_loss/pos5": 2.822064161300659, "train/pos_loss/pos6": 3.4731216430664062, "train/pos_loss/weight0": 3.0213890075683594, "train/pos_loss/weight1": 1.9751489162445068, "train/pos_loss/weight2": 1.1616415977478027, "train/pos_loss/weight3": 0.5516023635864258, "train/pos_loss/weight4": 0.20985357463359833, "train/pos_loss/weight5": 0.06057499721646309, "train/pos_loss/weight6": 0.019789505749940872, "train/pos_weight_progress": 0.0995260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.38736405968666077 }, { "epoch": 0.9976303317535545, "step": 421, "train/alpha": 0.2, "train/clean_ce_loss": 0.06884765625, "train/clean_loss_scale": 10.260125160217285, "train/clean_loss_weight": 2.052025079727173, "train/combined_loss": 0.8476626873016357, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.09976303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7063855528831482 }, { "epoch": 0.9976303317535545, "step": 421, "train/alpha": 0.2, "train/clean_ce_loss": 0.0859375, "train/clean_loss_scale": 6.456823825836182, "train/clean_loss_weight": 1.2913647890090942, "train/combined_loss": 0.6658599376678467, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.09976303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5548833012580872 }, { "epoch": 0.9976303317535545, "step": 421, "train/alpha": 0.2, "train/clean_ce_loss": 0.1083984375, "train/clean_loss_scale": 7.664973735809326, "train/clean_loss_weight": 1.5329947471618652, "train/combined_loss": 0.9970453977584839, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.09976303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8308711647987366 }, { "epoch": 0.9976303317535545, "step": 421, "train/alpha": 0.2, "train/clean_ce_loss": 0.1669921875, "train/clean_loss_scale": 8.758868217468262, "train/clean_loss_weight": 1.751773715019226, "train/combined_loss": 1.755195140838623, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.09976303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4626625776290894 }, { "epoch": 1.0, "step": 422, "train/alpha": 0.2, "train/clean_ce_loss": 0.062255859375, "train/clean_loss_scale": 9.898885726928711, "train/clean_loss_weight": 1.979777216911316, "train/combined_loss": 0.7395163774490356, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.1, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6162636280059814 }, { "epoch": 1.0, "step": 422, "train/alpha": 0.2, "train/clean_ce_loss": 0.08984375, "train/clean_loss_scale": 9.358368873596191, "train/clean_loss_weight": 1.871673822402954, "train/combined_loss": 1.0089491605758667, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.1, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8407909870147705 }, { "epoch": 1.0, "step": 422, "train/alpha": 0.2, "train/clean_ce_loss": 0.1484375, "train/clean_loss_scale": 8.198071479797363, "train/clean_loss_weight": 1.6396143436431885, "train/combined_loss": 1.4602816104888916, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.1, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2169013023376465 }, { "epoch": 1.0, "step": 422, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 10.526114463806152, "train/clean_loss_weight": 2.1052229404449463, "train/combined_loss": 1.0053261518478394, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.1, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8377718329429626 }, { "epoch": 1.0023696682464456, "step": 423, "train/alpha": 0.2, "train/clean_ce_loss": 0.08837890625, "train/clean_loss_scale": 7.760880947113037, "train/clean_loss_weight": 1.5521762371063232, "train/combined_loss": 0.8230777978897095, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.10023696682464454, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6858981847763062 }, { "epoch": 1.0023696682464456, "step": 423, "train/alpha": 0.2, "train/clean_ce_loss": 0.0125732421875, "train/clean_loss_scale": 25.0982608795166, "train/clean_loss_weight": 5.019652366638184, "train/combined_loss": 0.3786798119544983, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.10023696682464454, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3155665099620819 }, { "epoch": 1.0023696682464456, "step": 423, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 9.825016975402832, "train/clean_loss_weight": 1.9650033712387085, "train/combined_loss": 0.7829310297966003, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.10023696682464454, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6524425148963928 }, { "epoch": 1.0023696682464456, "step": 423, "train/alpha": 0.2, "train/clean_ce_loss": 0.045166015625, "train/clean_loss_scale": 10.919351577758789, "train/clean_loss_weight": 2.183870315551758, "train/combined_loss": 0.5918203592300415, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.10023696682464454, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.49318361282348633 }, { "epoch": 1.004739336492891, "step": 424, "train/alpha": 0.2, "train/clean_ce_loss": 0.1640625, "train/clean_loss_scale": 6.008321285247803, "train/clean_loss_weight": 1.2016643285751343, "train/combined_loss": 1.1828882694244385, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.1004739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9857402443885803 }, { "epoch": 1.004739336492891, "step": 424, "train/alpha": 0.2, "train/clean_ce_loss": 0.01556396484375, "train/clean_loss_scale": 26.319318771362305, "train/clean_loss_weight": 5.263864040374756, "train/combined_loss": 0.49155953526496887, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.1004739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.40963295102119446 }, { "epoch": 1.004739336492891, "step": 424, "train/alpha": 0.2, "train/clean_ce_loss": 0.01531982421875, "train/clean_loss_scale": 25.26156234741211, "train/clean_loss_weight": 5.05231237411499, "train/combined_loss": 0.4644032418727875, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.1004739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.38700270652770996 }, { "epoch": 1.004739336492891, "step": 424, "train/alpha": 0.2, "train/clean_ce_loss": 0.07177734375, "train/clean_loss_scale": 8.521759033203125, "train/clean_loss_weight": 1.704351782798767, "train/combined_loss": 0.7340030670166016, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.1004739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6116692423820496 }, { "epoch": 1.0071090047393365, "grad_norm": 12.286144256591797, "learning_rate": 2e-06, "loss": 3.5139, "step": 425 }, { "epoch": 1.0071090047393365, "step": 425, "train/alpha": 0.2, "train/clean_ce_loss": 0.0439453125, "train/clean_loss_scale": 11.056090354919434, "train/clean_loss_weight": 2.2112181186676025, "train/combined_loss": 0.5830360054969788, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.10071090047393365, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4858633577823639 }, { "epoch": 1.0071090047393365, "step": 425, "train/alpha": 0.2, "train/clean_ce_loss": 0.10546875, "train/clean_loss_scale": 5.966135025024414, "train/clean_loss_weight": 1.1932270526885986, "train/combined_loss": 0.7550889849662781, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.10071090047393365, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6292408108711243 }, { "epoch": 1.0071090047393365, "step": 425, "train/alpha": 0.2, "train/clean_ce_loss": 0.06103515625, "train/clean_loss_scale": 6.920407295227051, "train/clean_loss_weight": 1.384081482887268, "train/combined_loss": 0.5068657398223877, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.10071090047393365, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.42238813638687134 }, { "epoch": 1.0071090047393365, "step": 425, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 9.454741477966309, "train/clean_loss_weight": 1.8909482955932617, "train/combined_loss": 0.7423449158668518, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.10071090047393365, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.618620753288269 }, { "epoch": 1.009478672985782, "step": 426, "train/alpha": 0.2, "train/clean_ce_loss": 0.07177734375, "train/clean_loss_scale": 8.232007026672363, "train/clean_loss_weight": 1.6464014053344727, "train/combined_loss": 0.709045946598053, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.1009478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5908716320991516 }, { "epoch": 1.009478672985782, "step": 426, "train/alpha": 0.2, "train/clean_ce_loss": 0.0634765625, "train/clean_loss_scale": 18.09632682800293, "train/clean_loss_weight": 3.61926531791687, "train/combined_loss": 1.3784310817718506, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.1009478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1486926078796387 }, { "epoch": 1.009478672985782, "step": 426, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 10.161094665527344, "train/clean_loss_weight": 2.0322189331054688, "train/combined_loss": 0.8097121715545654, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.1009478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6747601628303528 }, { "epoch": 1.009478672985782, "step": 426, "train/alpha": 0.2, "train/clean_ce_loss": 0.115234375, "train/clean_loss_scale": 7.8848161697387695, "train/clean_loss_weight": 1.5769633054733276, "train/combined_loss": 1.0903222560882568, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.1009478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9086018800735474 }, { "epoch": 1.0118483412322274, "step": 427, "train/alpha": 0.2, "train/clean_ce_loss": 0.01055908203125, "train/clean_loss_scale": 34.85128402709961, "train/clean_loss_weight": 6.970256805419922, "train/combined_loss": 0.4415970742702484, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.10118483412322275, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.36799755692481995 }, { "epoch": 1.0118483412322274, "step": 427, "train/alpha": 0.2, "train/clean_ce_loss": 0.08251953125, "train/clean_loss_scale": 12.399194717407227, "train/clean_loss_weight": 2.4798390865325928, "train/combined_loss": 1.2278108596801758, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.10118483412322275, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0231757164001465 }, { "epoch": 1.0118483412322274, "step": 427, "train/alpha": 0.2, "train/clean_ce_loss": 0.0255126953125, "train/clean_loss_scale": 15.01028823852539, "train/clean_loss_weight": 3.0020577907562256, "train/combined_loss": 0.45954346656799316, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.10118483412322275, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.38295289874076843 }, { "epoch": 1.0118483412322274, "step": 427, "train/alpha": 0.2, "train/clean_ce_loss": 0.1513671875, "train/clean_loss_scale": 8.316217422485352, "train/clean_loss_weight": 1.6632435321807861, "train/combined_loss": 1.5105628967285156, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.10118483412322275, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2588024139404297 }, { "epoch": 1.014218009478673, "step": 428, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 12.313858032226562, "train/clean_loss_weight": 2.4627716541290283, "train/combined_loss": 1.1832847595214844, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.1014218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9860706329345703 }, { "epoch": 1.014218009478673, "step": 428, "train/alpha": 0.2, "train/clean_ce_loss": 0.03857421875, "train/clean_loss_scale": 8.771523475646973, "train/clean_loss_weight": 1.7543047666549683, "train/combined_loss": 0.40602558851242065, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.1014218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3383546471595764 }, { "epoch": 1.014218009478673, "step": 428, "train/alpha": 0.2, "train/clean_ce_loss": 0.0888671875, "train/clean_loss_scale": 13.156590461730957, "train/clean_loss_weight": 2.6313180923461914, "train/combined_loss": 1.403027057647705, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.1014218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.169189214706421 }, { "epoch": 1.014218009478673, "step": 428, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 15.860377311706543, "train/clean_loss_weight": 3.1720755100250244, "train/combined_loss": 1.4125648736953735, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.1014218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1771373748779297 }, { "epoch": 1.0165876777251184, "step": 429, "train/alpha": 0.2, "train/clean_ce_loss": 0.047607421875, "train/clean_loss_scale": 9.1714506149292, "train/clean_loss_weight": 1.8342901468276978, "train/combined_loss": 0.5239549279212952, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.10165876777251184, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43662911653518677 }, { "epoch": 1.0165876777251184, "step": 429, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 7.973874092102051, "train/clean_loss_weight": 1.594774842262268, "train/combined_loss": 0.6260737180709839, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.10165876777251184, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5217280983924866 }, { "epoch": 1.0165876777251184, "step": 429, "train/alpha": 0.2, "train/clean_ce_loss": 0.103515625, "train/clean_loss_scale": 12.757692337036133, "train/clean_loss_weight": 2.5515384674072266, "train/combined_loss": 1.5847446918487549, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.10165876777251184, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3206205368041992 }, { "epoch": 1.0165876777251184, "step": 429, "train/alpha": 0.2, "train/clean_ce_loss": 0.056640625, "train/clean_loss_scale": 9.587787628173828, "train/clean_loss_weight": 1.9175575971603394, "train/combined_loss": 0.6516699194908142, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.10165876777251184, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5430582761764526 }, { "epoch": 1.018957345971564, "grad_norm": 29.417476654052734, "learning_rate": 2e-06, "loss": 3.4467, "step": 430 }, { "epoch": 1.018957345971564, "step": 430, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 6.473298072814941, "train/clean_loss_weight": 1.2946596145629883, "train/combined_loss": 0.6220434904098511, "train/pos_loss/pos0": 0.08239364624023438, "train/pos_loss/pos1": 0.291731059551239, "train/pos_loss/pos2": 0.6977788209915161, "train/pos_loss/pos3": 1.1768525838851929, "train/pos_loss/pos4": 1.7917424440383911, "train/pos_loss/pos5": 2.0595266819000244, "train/pos_loss/pos6": 2.4951212406158447, "train/pos_loss/weight0": 2.5409133434295654, "train/pos_loss/weight1": 1.8498804569244385, "train/pos_loss/weight2": 1.2728004455566406, "train/pos_loss/weight3": 0.7786510586738586, "train/pos_loss/weight4": 0.36698973178863525, "train/pos_loss/weight5": 0.1354348063468933, "train/pos_loss/weight6": 0.055329568684101105, "train/pos_weight_progress": 0.1018957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5183695554733276 }, { "epoch": 1.018957345971564, "step": 430, "train/alpha": 0.2, "train/clean_ce_loss": 0.1513671875, "train/clean_loss_scale": 9.794759750366211, "train/clean_loss_weight": 1.9589519500732422, "train/combined_loss": 1.7791262865066528, "train/pos_loss/pos0": 0.538058876991272, "train/pos_loss/pos1": 1.1120306253433228, "train/pos_loss/pos2": 1.7735662460327148, "train/pos_loss/pos3": 2.6116063594818115, "train/pos_loss/pos4": 3.128512382507324, "train/pos_loss/pos5": 3.668175458908081, "train/pos_loss/pos6": 3.9752817153930664, "train/pos_loss/weight0": 2.341610908508301, "train/pos_loss/weight1": 1.7550220489501953, "train/pos_loss/weight2": 1.2705994844436646, "train/pos_loss/weight3": 0.8092097640037537, "train/pos_loss/weight4": 0.4594549834728241, "train/pos_loss/weight5": 0.23351621627807617, "train/pos_loss/weight6": 0.13058599829673767, "train/pos_weight_progress": 0.1018957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.482605218887329 }, { "epoch": 1.018957345971564, "step": 430, "train/alpha": 0.2, "train/clean_ce_loss": 0.07177734375, "train/clean_loss_scale": 6.868298530578613, "train/clean_loss_weight": 1.3736597299575806, "train/combined_loss": 0.5915858745574951, "train/pos_loss/pos0": 0.26347053050994873, "train/pos_loss/pos1": 0.48491665720939636, "train/pos_loss/pos2": 0.7868090867996216, "train/pos_loss/pos3": 1.2492015361785889, "train/pos_loss/pos4": 1.6309852600097656, "train/pos_loss/pos5": 1.9657421112060547, "train/pos_loss/pos6": 2.24674129486084, "train/pos_loss/weight0": 3.358618974685669, "train/pos_loss/weight1": 2.0613698959350586, "train/pos_loss/weight2": 1.021240234375, "train/pos_loss/weight3": 0.4129811227321625, "train/pos_loss/weight4": 0.11592499166727066, "train/pos_loss/weight5": 0.024775952100753784, "train/pos_loss/weight6": 0.0050893607549369335, "train/pos_weight_progress": 0.1018957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4929882287979126 }, { "epoch": 1.018957345971564, "step": 430, "train/alpha": 0.2, "train/clean_ce_loss": 0.01495361328125, "train/clean_loss_scale": 24.534122467041016, "train/clean_loss_weight": 4.906824588775635, "train/combined_loss": 0.4402485191822052, "train/pos_loss/pos0": 0.020879337564110756, "train/pos_loss/pos1": 0.08692293614149094, "train/pos_loss/pos2": 0.2907700538635254, "train/pos_loss/pos3": 0.8097570538520813, "train/pos_loss/pos4": 1.622434139251709, "train/pos_loss/pos5": 2.148116111755371, "train/pos_loss/pos6": 2.3715219497680664, "train/pos_loss/weight0": 2.4719908237457275, "train/pos_loss/weight1": 1.816904067993164, "train/pos_loss/weight2": 1.2196364402770996, "train/pos_loss/weight3": 0.7546485066413879, "train/pos_loss/weight4": 0.41243046522140503, "train/pos_loss/weight5": 0.21143443882465363, "train/pos_loss/weight6": 0.11295541375875473, "train/pos_weight_progress": 0.1018957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.36687377095222473 }, { "epoch": 1.0213270142180095, "step": 431, "train/alpha": 0.2, "train/clean_ce_loss": 0.023193359375, "train/clean_loss_scale": 27.231021881103516, "train/clean_loss_weight": 5.446204662322998, "train/combined_loss": 0.7578946352005005, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10213270142180095, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6315788626670837 }, { "epoch": 1.0213270142180095, "step": 431, "train/alpha": 0.2, "train/clean_ce_loss": 2.300739288330078e-05, "train/clean_loss_scale": 9724.9404296875, "train/clean_loss_weight": 1944.9881591796875, "train/combined_loss": 0.2684946358203888, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10213270142180095, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.22374552488327026 }, { "epoch": 1.0213270142180095, "step": 431, "train/alpha": 0.2, "train/clean_ce_loss": 0.0576171875, "train/clean_loss_scale": 13.08649730682373, "train/clean_loss_weight": 2.6172995567321777, "train/combined_loss": 0.9048085808753967, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10213270142180095, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7540071606636047 }, { "epoch": 1.0213270142180095, "step": 431, "train/alpha": 0.2, "train/clean_ce_loss": 0.00897216796875, "train/clean_loss_scale": 44.54008865356445, "train/clean_loss_weight": 8.908018112182617, "train/combined_loss": 0.47954538464546204, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10213270142180095, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3996211588382721 }, { "epoch": 1.0236966824644549, "step": 432, "train/alpha": 0.2, "train/clean_ce_loss": 0.01263427734375, "train/clean_loss_scale": 35.63115310668945, "train/clean_loss_weight": 7.126230716705322, "train/combined_loss": 0.540208637714386, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.1023696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45017385482788086 }, { "epoch": 1.0236966824644549, "step": 432, "train/alpha": 0.2, "train/clean_ce_loss": 0.10400390625, "train/clean_loss_scale": 10.77980899810791, "train/clean_loss_weight": 2.155961751937866, "train/combined_loss": 1.3453707695007324, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.1023696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1211422681808472 }, { "epoch": 1.0236966824644549, "step": 432, "train/alpha": 0.2, "train/clean_ce_loss": 0.00119781494140625, "train/clean_loss_scale": 297.1139221191406, "train/clean_loss_weight": 59.422786712646484, "train/combined_loss": 0.42706501483917236, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.1023696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3558875024318695 }, { "epoch": 1.0236966824644549, "step": 432, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 11.627366065979004, "train/clean_loss_weight": 2.3254733085632324, "train/combined_loss": 1.1105042695999146, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.1023696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9254202246665955 }, { "epoch": 1.0260663507109005, "step": 433, "train/alpha": 0.2, "train/clean_ce_loss": 0.1787109375, "train/clean_loss_scale": 3.382784128189087, "train/clean_loss_weight": 0.6765568256378174, "train/combined_loss": 0.7254486083984375, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10260663507109005, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6045405268669128 }, { "epoch": 1.0260663507109005, "step": 433, "train/alpha": 0.2, "train/clean_ce_loss": 0.1357421875, "train/clean_loss_scale": 9.416961669921875, "train/clean_loss_weight": 1.883392333984375, "train/combined_loss": 1.5339347124099731, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10260663507109005, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2782789468765259 }, { "epoch": 1.0260663507109005, "step": 433, "train/alpha": 0.2, "train/clean_ce_loss": 0.05029296875, "train/clean_loss_scale": 11.11921501159668, "train/clean_loss_weight": 2.2238430976867676, "train/combined_loss": 0.6710619926452637, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10260663507109005, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5592183470726013 }, { "epoch": 1.0260663507109005, "step": 433, "train/alpha": 0.2, "train/clean_ce_loss": 0.09912109375, "train/clean_loss_scale": 10.438404083251953, "train/clean_loss_weight": 2.0876808166503906, "train/combined_loss": 1.2415993213653564, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10260663507109005, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0346660614013672 }, { "epoch": 1.028436018957346, "step": 434, "train/alpha": 0.2, "train/clean_ce_loss": 0.0546875, "train/clean_loss_scale": 10.424051284790039, "train/clean_loss_weight": 2.084810256958008, "train/combined_loss": 0.6840783953666687, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10284360189573459, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5700653195381165 }, { "epoch": 1.028436018957346, "step": 434, "train/alpha": 0.2, "train/clean_ce_loss": 0.154296875, "train/clean_loss_scale": 9.432321548461914, "train/clean_loss_weight": 1.8864643573760986, "train/combined_loss": 1.7464532852172852, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10284360189573459, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4553776979446411 }, { "epoch": 1.028436018957346, "step": 434, "train/alpha": 0.2, "train/clean_ce_loss": 0.12060546875, "train/clean_loss_scale": 9.200778007507324, "train/clean_loss_weight": 1.8401556015014648, "train/combined_loss": 1.331597089767456, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10284360189573459, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1096642017364502 }, { "epoch": 1.028436018957346, "step": 434, "train/alpha": 0.2, "train/clean_ce_loss": 0.02587890625, "train/clean_loss_scale": 10.307518005371094, "train/clean_loss_weight": 2.0615036487579346, "train/combined_loss": 0.3200967609882355, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10284360189573459, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.26674729585647583 }, { "epoch": 1.0308056872037914, "grad_norm": 12.957075119018555, "learning_rate": 2e-06, "loss": 3.3119, "step": 435 }, { "epoch": 1.0308056872037914, "step": 435, "train/alpha": 0.2, "train/clean_ce_loss": 2.7060508728027344e-05, "train/clean_loss_scale": 11229.505859375, "train/clean_loss_weight": 2245.901123046875, "train/combined_loss": 0.3646513521671295, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10308056872037914, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.303876131772995 }, { "epoch": 1.0308056872037914, "step": 435, "train/alpha": 0.2, "train/clean_ce_loss": 0.00421142578125, "train/clean_loss_scale": 155.69346618652344, "train/clean_loss_weight": 31.13869285583496, "train/combined_loss": 0.7868298292160034, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10308056872037914, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6556915044784546 }, { "epoch": 1.0308056872037914, "step": 435, "train/alpha": 0.2, "train/clean_ce_loss": 0.0849609375, "train/clean_loss_scale": 10.592358589172363, "train/clean_loss_weight": 2.11847186088562, "train/combined_loss": 1.0799239873886108, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10308056872037914, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8999366760253906 }, { "epoch": 1.0308056872037914, "step": 435, "train/alpha": 0.2, "train/clean_ce_loss": 0.06982421875, "train/clean_loss_scale": 10.953458786010742, "train/clean_loss_weight": 2.1906917095184326, "train/combined_loss": 0.9177800416946411, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10308056872037914, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7648167014122009 }, { "epoch": 1.033175355450237, "step": 436, "train/alpha": 0.2, "train/clean_ce_loss": 0.09228515625, "train/clean_loss_scale": 8.795415878295898, "train/clean_loss_weight": 1.7590831518173218, "train/combined_loss": 0.9740235805511475, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.1033175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8116863369941711 }, { "epoch": 1.033175355450237, "step": 436, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 10.199302673339844, "train/clean_loss_weight": 2.039860486984253, "train/combined_loss": 0.788852334022522, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.1033175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6573769450187683 }, { "epoch": 1.033175355450237, "step": 436, "train/alpha": 0.2, "train/clean_ce_loss": 0.10302734375, "train/clean_loss_scale": 8.971555709838867, "train/clean_loss_weight": 1.7943111658096313, "train/combined_loss": 1.1091786623001099, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.1033175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9243155717849731 }, { "epoch": 1.033175355450237, "step": 436, "train/alpha": 0.2, "train/clean_ce_loss": 0.01361083984375, "train/clean_loss_scale": 40.180728912353516, "train/clean_loss_weight": 8.03614616394043, "train/combined_loss": 0.6562721729278564, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.1033175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5468934774398804 }, { "epoch": 1.0355450236966826, "step": 437, "train/alpha": 0.2, "train/clean_ce_loss": 0.0859375, "train/clean_loss_scale": 9.542168617248535, "train/clean_loss_weight": 1.9084337949752808, "train/combined_loss": 0.9840362071990967, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10355450236966825, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.820030152797699 }, { "epoch": 1.0355450236966826, "step": 437, "train/alpha": 0.2, "train/clean_ce_loss": 0.1484375, "train/clean_loss_scale": 8.008749008178711, "train/clean_loss_weight": 1.6017497777938843, "train/combined_loss": 1.4265583753585815, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10355450236966825, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1887986660003662 }, { "epoch": 1.0355450236966826, "step": 437, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 14.68506145477295, "train/clean_loss_weight": 2.9370124340057373, "train/combined_loss": 1.3939335346221924, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10355450236966825, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1616113185882568 }, { "epoch": 1.0355450236966826, "step": 437, "train/alpha": 0.2, "train/clean_ce_loss": 0.03564453125, "train/clean_loss_scale": 15.433553695678711, "train/clean_loss_weight": 3.0867106914520264, "train/combined_loss": 0.6601461172103882, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10355450236966825, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5501217842102051 }, { "epoch": 1.037914691943128, "step": 438, "train/alpha": 0.2, "train/clean_ce_loss": 0.023681640625, "train/clean_loss_scale": 24.302200317382812, "train/clean_loss_weight": 4.860440254211426, "train/combined_loss": 0.690619170665741, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.1037914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5755159854888916 }, { "epoch": 1.037914691943128, "step": 438, "train/alpha": 0.2, "train/clean_ce_loss": 0.09375, "train/clean_loss_scale": 11.177454948425293, "train/clean_loss_weight": 2.2354910373687744, "train/combined_loss": 1.2574636936187744, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.1037914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0478863716125488 }, { "epoch": 1.037914691943128, "step": 438, "train/alpha": 0.2, "train/clean_ce_loss": 0.08056640625, "train/clean_loss_scale": 10.80997371673584, "train/clean_loss_weight": 2.161994695663452, "train/combined_loss": 1.0451048612594604, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.1037914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.870920717716217 }, { "epoch": 1.037914691943128, "step": 438, "train/alpha": 0.2, "train/clean_ce_loss": 0.1083984375, "train/clean_loss_scale": 13.165367126464844, "train/clean_loss_weight": 2.633073568344116, "train/combined_loss": 1.7125263214111328, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.1037914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.427105188369751 }, { "epoch": 1.0402843601895735, "step": 439, "train/alpha": 0.2, "train/clean_ce_loss": 0.1123046875, "train/clean_loss_scale": 10.654685020446777, "train/clean_loss_weight": 2.130937099456787, "train/combined_loss": 1.4358853101730347, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10402843601895735, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1965711116790771 }, { "epoch": 1.0402843601895735, "step": 439, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 10.631738662719727, "train/clean_loss_weight": 2.126347780227661, "train/combined_loss": 0.8409870862960815, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10402843601895735, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7008225917816162 }, { "epoch": 1.0402843601895735, "step": 439, "train/alpha": 0.2, "train/clean_ce_loss": 0.09033203125, "train/clean_loss_scale": 8.714451789855957, "train/clean_loss_weight": 1.7428903579711914, "train/combined_loss": 0.9446329474449158, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10402843601895735, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7871941328048706 }, { "epoch": 1.0402843601895735, "step": 439, "train/alpha": 0.2, "train/clean_ce_loss": 0.1484375, "train/clean_loss_scale": 9.247154235839844, "train/clean_loss_weight": 1.8494309186935425, "train/combined_loss": 1.6471494436264038, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10402843601895735, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3726245164871216 }, { "epoch": 1.042654028436019, "grad_norm": 12.29018783569336, "learning_rate": 2e-06, "loss": 3.287, "step": 440 }, { "epoch": 1.042654028436019, "step": 440, "train/alpha": 0.2, "train/clean_ce_loss": 0.10107421875, "train/clean_loss_scale": 13.397621154785156, "train/clean_loss_weight": 2.6795241832733154, "train/combined_loss": 1.6249849796295166, "train/pos_loss/pos0": 0.4633261263370514, "train/pos_loss/pos1": 0.8764916658401489, "train/pos_loss/pos2": 1.4016488790512085, "train/pos_loss/pos3": 2.2246017456054688, "train/pos_loss/pos4": 2.7749969959259033, "train/pos_loss/pos5": 3.369932174682617, "train/pos_loss/pos6": 3.890653610229492, "train/pos_loss/weight0": 2.171215534210205, "train/pos_loss/weight1": 1.6586024761199951, "train/pos_loss/weight2": 1.2289432287216187, "train/pos_loss/weight3": 0.8805280327796936, "train/pos_loss/weight4": 0.551849365234375, "train/pos_loss/weight5": 0.31952008605003357, "train/pos_loss/weight6": 0.18934117257595062, "train/pos_weight_progress": 0.10426540284360189, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.354154109954834 }, { "epoch": 1.042654028436019, "step": 440, "train/alpha": 0.2, "train/clean_ce_loss": 0.052734375, "train/clean_loss_scale": 7.94103479385376, "train/clean_loss_weight": 1.5882070064544678, "train/combined_loss": 0.5025185942649841, "train/pos_loss/pos0": 0.15437045693397522, "train/pos_loss/pos1": 0.3337060511112213, "train/pos_loss/pos2": 0.7206480503082275, "train/pos_loss/pos3": 1.20595121383667, "train/pos_loss/pos4": 1.6217341423034668, "train/pos_loss/pos5": 2.2241594791412354, "train/pos_loss/pos6": 2.6010918617248535, "train/pos_loss/weight0": 3.152517080307007, "train/pos_loss/weight1": 2.025261163711548, "train/pos_loss/weight2": 1.133329153060913, "train/pos_loss/weight3": 0.4799642562866211, "train/pos_loss/weight4": 0.15839126706123352, "train/pos_loss/weight5": 0.039549414068460464, "train/pos_loss/weight6": 0.010987397283315659, "train/pos_weight_progress": 0.10426540284360189, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.41876551508903503 }, { "epoch": 1.042654028436019, "step": 440, "train/alpha": 0.2, "train/clean_ce_loss": 0.042724609375, "train/clean_loss_scale": 9.874855995178223, "train/clean_loss_weight": 1.9749711751937866, "train/combined_loss": 0.5062792301177979, "train/pos_loss/pos0": 0.11256452649831772, "train/pos_loss/pos1": 0.23844873905181885, "train/pos_loss/pos2": 0.48340290784835815, "train/pos_loss/pos3": 0.786785364151001, "train/pos_loss/pos4": 1.2385586500167847, "train/pos_loss/pos5": 1.441268801689148, "train/pos_loss/pos6": 1.8124029636383057, "train/pos_loss/weight0": 2.4170823097229004, "train/pos_loss/weight1": 1.7925631999969482, "train/pos_loss/weight2": 1.257746696472168, "train/pos_loss/weight3": 0.7921902537345886, "train/pos_loss/weight4": 0.41927123069763184, "train/pos_loss/weight5": 0.21211393177509308, "train/pos_loss/weight6": 0.1090322807431221, "train/pos_weight_progress": 0.10426540284360189, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.42189934849739075 }, { "epoch": 1.042654028436019, "step": 440, "train/alpha": 0.2, "train/clean_ce_loss": 0.029052734375, "train/clean_loss_scale": 23.86320686340332, "train/clean_loss_weight": 4.772641658782959, "train/combined_loss": 0.8319497108459473, "train/pos_loss/pos0": 0.10498879849910736, "train/pos_loss/pos1": 0.28774118423461914, "train/pos_loss/pos2": 0.5932618379592896, "train/pos_loss/pos3": 1.189462423324585, "train/pos_loss/pos4": 1.7639901638031006, "train/pos_loss/pos5": 2.3201816082000732, "train/pos_loss/pos6": 2.6962807178497314, "train/pos_loss/weight0": 2.122758388519287, "train/pos_loss/weight1": 1.6479953527450562, "train/pos_loss/weight2": 1.238084077835083, "train/pos_loss/weight3": 0.8568239212036133, "train/pos_loss/weight4": 0.5628310441970825, "train/pos_loss/weight5": 0.3427923023700714, "train/pos_loss/weight6": 0.22871482372283936, "train/pos_weight_progress": 0.10426540284360189, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.693291425704956 }, { "epoch": 1.0450236966824644, "step": 441, "train/alpha": 0.2, "train/clean_ce_loss": 0.08984375, "train/clean_loss_scale": 12.158215522766113, "train/clean_loss_weight": 2.43164324760437, "train/combined_loss": 1.3108075857162476, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.10450236966824644, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0923396348953247 }, { "epoch": 1.0450236966824644, "step": 441, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 9.525981903076172, "train/clean_loss_weight": 1.9051964282989502, "train/combined_loss": 0.9321322441101074, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.10450236966824644, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7767768502235413 }, { "epoch": 1.0450236966824644, "step": 441, "train/alpha": 0.2, "train/clean_ce_loss": 0.0869140625, "train/clean_loss_scale": 9.392044067382812, "train/clean_loss_weight": 1.8784087896347046, "train/combined_loss": 0.9795608520507812, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.10450236966824644, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8163006901741028 }, { "epoch": 1.0450236966824644, "step": 441, "train/alpha": 0.2, "train/clean_ce_loss": 0.07763671875, "train/clean_loss_scale": 9.960710525512695, "train/clean_loss_weight": 1.9921420812606812, "train/combined_loss": 0.9279802441596985, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.10450236966824644, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7733168601989746 }, { "epoch": 1.04739336492891, "step": 442, "train/alpha": 0.2, "train/clean_ce_loss": 0.123046875, "train/clean_loss_scale": 9.352110862731934, "train/clean_loss_weight": 1.8704222440719604, "train/combined_loss": 1.3808976411819458, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.104739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1507480144500732 }, { "epoch": 1.04739336492891, "step": 442, "train/alpha": 0.2, "train/clean_ce_loss": 0.142578125, "train/clean_loss_scale": 8.639941215515137, "train/clean_loss_weight": 1.7279882431030273, "train/combined_loss": 1.478239893913269, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.104739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2318665981292725 }, { "epoch": 1.04739336492891, "step": 442, "train/alpha": 0.2, "train/clean_ce_loss": 0.1015625, "train/clean_loss_scale": 8.122793197631836, "train/clean_loss_weight": 1.624558687210083, "train/combined_loss": 0.9899653792381287, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.104739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8249711394309998 }, { "epoch": 1.04739336492891, "step": 442, "train/alpha": 0.2, "train/clean_ce_loss": 0.08203125, "train/clean_loss_scale": 8.538276672363281, "train/clean_loss_weight": 1.7076553106307983, "train/combined_loss": 0.8404865860939026, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.104739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.700405478477478 }, { "epoch": 1.0497630331753554, "step": 443, "train/alpha": 0.2, "train/clean_ce_loss": 0.06689453125, "train/clean_loss_scale": 8.259726524353027, "train/clean_loss_weight": 1.6519453525543213, "train/combined_loss": 0.6630366444587708, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.10497630331753555, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5525305271148682 }, { "epoch": 1.0497630331753554, "step": 443, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 14.196313858032227, "train/clean_loss_weight": 2.8392627239227295, "train/combined_loss": 1.2144503593444824, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.10497630331753555, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0120419263839722 }, { "epoch": 1.0497630331753554, "step": 443, "train/alpha": 0.2, "train/clean_ce_loss": 0.0751953125, "train/clean_loss_scale": 16.074186325073242, "train/clean_loss_weight": 3.2148373126983643, "train/combined_loss": 1.4504441022872925, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.10497630331753555, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2087033987045288 }, { "epoch": 1.0497630331753554, "step": 443, "train/alpha": 0.2, "train/clean_ce_loss": 0.06201171875, "train/clean_loss_scale": 10.448357582092285, "train/clean_loss_weight": 2.0896716117858887, "train/combined_loss": 0.7775047421455383, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.10497630331753555, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6479206085205078 }, { "epoch": 1.052132701421801, "step": 444, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 9.991451263427734, "train/clean_loss_weight": 1.9982903003692627, "train/combined_loss": 1.1123294830322266, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.1052132701421801, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9269412755966187 }, { "epoch": 1.052132701421801, "step": 444, "train/alpha": 0.2, "train/clean_ce_loss": 0.06396484375, "train/clean_loss_scale": 10.279032707214355, "train/clean_loss_weight": 2.0558066368103027, "train/combined_loss": 0.7889961004257202, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.1052132701421801, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6574967503547668 }, { "epoch": 1.052132701421801, "step": 444, "train/alpha": 0.2, "train/clean_ce_loss": 0.0064697265625, "train/clean_loss_scale": 72.07068634033203, "train/clean_loss_weight": 14.414137840270996, "train/combined_loss": 0.5595331788063049, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.1052132701421801, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4662776291370392 }, { "epoch": 1.052132701421801, "step": 444, "train/alpha": 0.2, "train/clean_ce_loss": 0.05859375, "train/clean_loss_scale": 17.128145217895508, "train/clean_loss_weight": 3.425629138946533, "train/combined_loss": 1.2043226957321167, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.1052132701421801, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0036022663116455 }, { "epoch": 1.0545023696682465, "grad_norm": 22.368511199951172, "learning_rate": 2e-06, "loss": 3.5961, "step": 445 }, { "epoch": 1.0545023696682465, "step": 445, "train/alpha": 0.2, "train/clean_ce_loss": 0.00872802734375, "train/clean_loss_scale": 55.47901916503906, "train/clean_loss_weight": 11.095804214477539, "train/combined_loss": 0.581066906452179, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.10545023696682465, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.484222412109375 }, { "epoch": 1.0545023696682465, "step": 445, "train/alpha": 0.2, "train/clean_ce_loss": 0.0751953125, "train/clean_loss_scale": 9.983438491821289, "train/clean_loss_weight": 1.9966877698898315, "train/combined_loss": 0.9008493423461914, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.10545023696682465, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7507078051567078 }, { "epoch": 1.0545023696682465, "step": 445, "train/alpha": 0.2, "train/clean_ce_loss": 0.1318359375, "train/clean_loss_scale": 7.162620544433594, "train/clean_loss_weight": 1.4325240850448608, "train/combined_loss": 1.1331489086151123, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.10545023696682465, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.944290816783905 }, { "epoch": 1.0545023696682465, "step": 445, "train/alpha": 0.2, "train/clean_ce_loss": 0.052001953125, "train/clean_loss_scale": 11.778422355651855, "train/clean_loss_weight": 2.355684518814087, "train/combined_loss": 0.7350011467933655, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.10545023696682465, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6125009655952454 }, { "epoch": 1.0568720379146919, "step": 446, "train/alpha": 0.2, "train/clean_ce_loss": 0.0205078125, "train/clean_loss_scale": 19.53019142150879, "train/clean_loss_weight": 3.906038284301758, "train/combined_loss": 0.48062577843666077, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.10568720379146919, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.40052148699760437 }, { "epoch": 1.0568720379146919, "step": 446, "train/alpha": 0.2, "train/clean_ce_loss": 0.0179443359375, "train/clean_loss_scale": 26.510156631469727, "train/clean_loss_weight": 5.302031517028809, "train/combined_loss": 0.5708485841751099, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.10568720379146919, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.47570714354515076 }, { "epoch": 1.0568720379146919, "step": 446, "train/alpha": 0.2, "train/clean_ce_loss": 0.043212890625, "train/clean_loss_scale": 14.418290138244629, "train/clean_loss_weight": 2.8836581707000732, "train/combined_loss": 0.7476671934127808, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.10568720379146919, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6230559945106506 }, { "epoch": 1.0568720379146919, "step": 446, "train/alpha": 0.2, "train/clean_ce_loss": 0.0849609375, "train/clean_loss_scale": 8.317110061645508, "train/clean_loss_weight": 1.6634219884872437, "train/combined_loss": 0.8479553461074829, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.10568720379146919, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7066294550895691 }, { "epoch": 1.0592417061611374, "step": 447, "train/alpha": 0.2, "train/clean_ce_loss": 0.1572265625, "train/clean_loss_scale": 8.006196022033691, "train/clean_loss_weight": 1.6012392044067383, "train/combined_loss": 1.5105440616607666, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.10592417061611374, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.258786678314209 }, { "epoch": 1.0592417061611374, "step": 447, "train/alpha": 0.2, "train/clean_ce_loss": 0.06201171875, "train/clean_loss_scale": 8.23660945892334, "train/clean_loss_weight": 1.6473219394683838, "train/combined_loss": 0.6129195690155029, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.10592417061611374, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.510766327381134 }, { "epoch": 1.0592417061611374, "step": 447, "train/alpha": 0.2, "train/clean_ce_loss": 0.07373046875, "train/clean_loss_scale": 10.893976211547852, "train/clean_loss_weight": 2.178795337677002, "train/combined_loss": 0.9638615846633911, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.10592417061611374, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8032179474830627 }, { "epoch": 1.0592417061611374, "step": 447, "train/alpha": 0.2, "train/clean_ce_loss": 0.05810546875, "train/clean_loss_scale": 9.668285369873047, "train/clean_loss_weight": 1.9336570501327515, "train/combined_loss": 0.6741363406181335, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.10592417061611374, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5617802739143372 }, { "epoch": 1.061611374407583, "step": 448, "train/alpha": 0.2, "train/clean_ce_loss": 0.06298828125, "train/clean_loss_scale": 8.637919425964355, "train/clean_loss_weight": 1.727583885192871, "train/combined_loss": 0.6529052257537842, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.1061611374407583, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5440877079963684 }, { "epoch": 1.061611374407583, "step": 448, "train/alpha": 0.2, "train/clean_ce_loss": 0.0693359375, "train/clean_loss_scale": 10.357215881347656, "train/clean_loss_weight": 2.0714433193206787, "train/combined_loss": 0.8617527484893799, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.1061611374407583, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7181272506713867 }, { "epoch": 1.061611374407583, "step": 448, "train/alpha": 0.2, "train/clean_ce_loss": 0.055908203125, "train/clean_loss_scale": 17.483110427856445, "train/clean_loss_weight": 3.496622085571289, "train/combined_loss": 1.1729391813278198, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.1061611374407583, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9774492979049683 }, { "epoch": 1.061611374407583, "step": 448, "train/alpha": 0.2, "train/clean_ce_loss": 0.16015625, "train/clean_loss_scale": 7.373528480529785, "train/clean_loss_weight": 1.474705696105957, "train/combined_loss": 1.417099952697754, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.1061611374407583, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.180916666984558 }, { "epoch": 1.0639810426540284, "step": 449, "train/alpha": 0.2, "train/clean_ce_loss": 0.0133056640625, "train/clean_loss_scale": 38.081722259521484, "train/clean_loss_weight": 7.616344451904297, "train/combined_loss": 0.6080431342124939, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.10639810426540285, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5067026019096375 }, { "epoch": 1.0639810426540284, "step": 449, "train/alpha": 0.2, "train/clean_ce_loss": 0.06298828125, "train/clean_loss_scale": 13.718438148498535, "train/clean_loss_weight": 2.743687629699707, "train/combined_loss": 1.0369210243225098, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.10639810426540285, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8641008138656616 }, { "epoch": 1.0639810426540284, "step": 449, "train/alpha": 0.2, "train/clean_ce_loss": 0.1044921875, "train/clean_loss_scale": 7.8251166343688965, "train/clean_loss_weight": 1.5650233030319214, "train/combined_loss": 0.9811962842941284, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.10639810426540285, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8176635503768921 }, { "epoch": 1.0639810426540284, "step": 449, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 8.971809387207031, "train/clean_loss_weight": 1.79436194896698, "train/combined_loss": 0.7096841335296631, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.10639810426540285, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5914034247398376 }, { "epoch": 1.066350710900474, "grad_norm": 15.946478843688965, "learning_rate": 2e-06, "loss": 3.4434, "step": 450 }, { "epoch": 1.066350710900474, "step": 450, "train/alpha": 0.2, "train/clean_ce_loss": 0.1044921875, "train/clean_loss_scale": 10.584992408752441, "train/clean_loss_weight": 2.1169984340667725, "train/combined_loss": 1.327258825302124, "train/pos_loss/pos0": 0.4282459616661072, "train/pos_loss/pos1": 0.8266298770904541, "train/pos_loss/pos2": 1.2758846282958984, "train/pos_loss/pos3": 1.9224672317504883, "train/pos_loss/pos4": 2.564478874206543, "train/pos_loss/pos5": 3.0692315101623535, "train/pos_loss/pos6": 3.4489545822143555, "train/pos_loss/weight0": 2.401019811630249, "train/pos_loss/weight1": 1.819629192352295, "train/pos_loss/weight2": 1.2519549131393433, "train/pos_loss/weight3": 0.7777373194694519, "train/pos_loss/weight4": 0.43836596608161926, "train/pos_loss/weight5": 0.2025911808013916, "train/pos_loss/weight6": 0.10870186984539032, "train/pos_weight_progress": 0.1066350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1060490608215332 }, { "epoch": 1.066350710900474, "step": 450, "train/alpha": 0.2, "train/clean_ce_loss": 0.09619140625, "train/clean_loss_scale": 7.225630760192871, "train/clean_loss_weight": 1.4451261758804321, "train/combined_loss": 0.8340522646903992, "train/pos_loss/pos0": 0.32282131910324097, "train/pos_loss/pos1": 0.6843230724334717, "train/pos_loss/pos2": 1.0401742458343506, "train/pos_loss/pos3": 1.3153990507125854, "train/pos_loss/pos4": 1.9129482507705688, "train/pos_loss/pos5": 2.321242094039917, "train/pos_loss/pos6": 2.6380774974823, "train/pos_loss/weight0": 2.996016025543213, "train/pos_loss/weight1": 2.010918378829956, "train/pos_loss/weight2": 1.1350585222244263, "train/pos_loss/weight3": 0.5616515278816223, "train/pos_loss/weight4": 0.2161942571401596, "train/pos_loss/weight5": 0.061831023544073105, "train/pos_loss/weight6": 0.01832963153719902, "train/pos_weight_progress": 0.1066350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6950435638427734 }, { "epoch": 1.066350710900474, "step": 450, "train/alpha": 0.2, "train/clean_ce_loss": 0.021484375, "train/clean_loss_scale": 20.194129943847656, "train/clean_loss_weight": 4.038825988769531, "train/combined_loss": 0.5206299424171448, "train/pos_loss/pos0": 0.06337503343820572, "train/pos_loss/pos1": 0.24127644300460815, "train/pos_loss/pos2": 0.603674590587616, "train/pos_loss/pos3": 1.0626888275146484, "train/pos_loss/pos4": 1.6826295852661133, "train/pos_loss/pos5": 2.005796194076538, "train/pos_loss/pos6": 2.5640532970428467, "train/pos_loss/weight0": 2.707225799560547, "train/pos_loss/weight1": 1.9044363498687744, "train/pos_loss/weight2": 1.2112101316452026, "train/pos_loss/weight3": 0.6540467739105225, "train/pos_loss/weight4": 0.3214888274669647, "train/pos_loss/weight5": 0.14050687849521637, "train/pos_loss/weight6": 0.06108541041612625, "train/pos_weight_progress": 0.1066350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4338582754135132 }, { "epoch": 1.066350710900474, "step": 450, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 8.322672843933105, "train/clean_loss_weight": 1.664534568786621, "train/combined_loss": 0.7412380576133728, "train/pos_loss/pos0": 0.2132120579481125, "train/pos_loss/pos1": 0.4060080647468567, "train/pos_loss/pos2": 0.7356610298156738, "train/pos_loss/pos3": 1.065446138381958, "train/pos_loss/pos4": 1.4217270612716675, "train/pos_loss/pos5": 1.9206342697143555, "train/pos_loss/pos6": 2.167912721633911, "train/pos_loss/weight0": 2.363222599029541, "train/pos_loss/weight1": 1.7884504795074463, "train/pos_loss/weight2": 1.252516508102417, "train/pos_loss/weight3": 0.8122223019599915, "train/pos_loss/weight4": 0.4511992931365967, "train/pos_loss/weight5": 0.21392934024333954, "train/pos_loss/weight6": 0.11845916509628296, "train/pos_weight_progress": 0.1066350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6176983714103699 }, { "epoch": 1.0687203791469195, "step": 451, "train/alpha": 0.2, "train/clean_ce_loss": 7.724761962890625e-05, "train/clean_loss_scale": 3140.381103515625, "train/clean_loss_weight": 628.0762329101562, "train/combined_loss": 0.29110437631607056, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10687203791469194, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.24258697032928467 }, { "epoch": 1.0687203791469195, "step": 451, "train/alpha": 0.2, "train/clean_ce_loss": 0.1552734375, "train/clean_loss_scale": 8.24605655670166, "train/clean_loss_weight": 1.6492112874984741, "train/combined_loss": 1.536472201347351, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10687203791469194, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2803934812545776 }, { "epoch": 1.0687203791469195, "step": 451, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 7.776668548583984, "train/clean_loss_weight": 1.5553337335586548, "train/combined_loss": 0.7791857123374939, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10687203791469194, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.649321436882019 }, { "epoch": 1.0687203791469195, "step": 451, "train/alpha": 0.2, "train/clean_ce_loss": 0.01300048828125, "train/clean_loss_scale": 42.188934326171875, "train/clean_loss_weight": 8.437787055969238, "train/combined_loss": 0.6581721305847168, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10687203791469194, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5484767556190491 }, { "epoch": 1.0710900473933649, "step": 452, "train/alpha": 0.2, "train/clean_ce_loss": 0.053955078125, "train/clean_loss_scale": 10.988158226013184, "train/clean_loss_weight": 2.197631597518921, "train/combined_loss": 0.7114403247833252, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10710900473933649, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5928669571876526 }, { "epoch": 1.0710900473933649, "step": 452, "train/alpha": 0.2, "train/clean_ce_loss": 0.1044921875, "train/clean_loss_scale": 11.733041763305664, "train/clean_loss_weight": 2.3466084003448486, "train/combined_loss": 1.4712133407592773, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10710900473933649, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2260111570358276 }, { "epoch": 1.0710900473933649, "step": 452, "train/alpha": 0.2, "train/clean_ce_loss": 0.057373046875, "train/clean_loss_scale": 11.823347091674805, "train/clean_loss_weight": 2.3646695613861084, "train/combined_loss": 0.8140097260475159, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10710900473933649, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6783414483070374 }, { "epoch": 1.0710900473933649, "step": 452, "train/alpha": 0.2, "train/clean_ce_loss": 0.003173828125, "train/clean_loss_scale": 136.36831665039062, "train/clean_loss_weight": 27.273664474487305, "train/combined_loss": 0.5193715691566467, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10710900473933649, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43280962109565735 }, { "epoch": 1.0734597156398105, "step": 453, "train/alpha": 0.2, "train/clean_ce_loss": 0.059326171875, "train/clean_loss_scale": 10.83281135559082, "train/clean_loss_weight": 2.16656231880188, "train/combined_loss": 0.7712030410766602, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10734597156398104, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6426692008972168 }, { "epoch": 1.0734597156398105, "step": 453, "train/alpha": 0.2, "train/clean_ce_loss": 0.0966796875, "train/clean_loss_scale": 7.7410759925842285, "train/clean_loss_weight": 1.5482152700424194, "train/combined_loss": 0.8980857729911804, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10734597156398104, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7484048008918762 }, { "epoch": 1.0734597156398105, "step": 453, "train/alpha": 0.2, "train/clean_ce_loss": 0.07568359375, "train/clean_loss_scale": 8.932647705078125, "train/clean_loss_weight": 1.786529541015625, "train/combined_loss": 0.8112658858299255, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10734597156398104, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6760548949241638 }, { "epoch": 1.0734597156398105, "step": 453, "train/alpha": 0.2, "train/clean_ce_loss": 0.061767578125, "train/clean_loss_scale": 15.98293399810791, "train/clean_loss_weight": 3.196586847305298, "train/combined_loss": 1.1846725940704346, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10734597156398104, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9872271418571472 }, { "epoch": 1.0758293838862558, "step": 454, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 14.1029052734375, "train/clean_loss_weight": 2.8205811977386475, "train/combined_loss": 1.3717279434204102, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10758293838862559, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1431065797805786 }, { "epoch": 1.0758293838862558, "step": 454, "train/alpha": 0.2, "train/clean_ce_loss": 0.03662109375, "train/clean_loss_scale": 10.348353385925293, "train/clean_loss_weight": 2.0696706771850586, "train/combined_loss": 0.4547616243362427, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10758293838862559, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3789680302143097 }, { "epoch": 1.0758293838862558, "step": 454, "train/alpha": 0.2, "train/clean_ce_loss": 0.15625, "train/clean_loss_scale": 7.797328948974609, "train/clean_loss_weight": 1.559465765953064, "train/combined_loss": 1.4619991779327393, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10758293838862559, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2183326482772827 }, { "epoch": 1.0758293838862558, "step": 454, "train/alpha": 0.2, "train/clean_ce_loss": 0.08837890625, "train/clean_loss_scale": 12.156607627868652, "train/clean_loss_weight": 2.431321620941162, "train/combined_loss": 1.2892651557922363, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10758293838862559, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0743876695632935 }, { "epoch": 1.0781990521327014, "grad_norm": 14.851945877075195, "learning_rate": 2e-06, "loss": 3.5284, "step": 455 }, { "epoch": 1.0781990521327014, "step": 455, "train/alpha": 0.2, "train/clean_ce_loss": 0.07763671875, "train/clean_loss_scale": 8.45985221862793, "train/clean_loss_weight": 1.6919704675674438, "train/combined_loss": 0.7881542444229126, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10781990521327015, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6567952036857605 }, { "epoch": 1.0781990521327014, "step": 455, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 11.001433372497559, "train/clean_loss_weight": 2.200286626815796, "train/combined_loss": 1.1022920608520508, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10781990521327015, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.918576717376709 }, { "epoch": 1.0781990521327014, "step": 455, "train/alpha": 0.2, "train/clean_ce_loss": 0.025146484375, "train/clean_loss_scale": 13.853214263916016, "train/clean_loss_weight": 2.7706429958343506, "train/combined_loss": 0.41803157329559326, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10781990521327015, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3483596444129944 }, { "epoch": 1.0781990521327014, "step": 455, "train/alpha": 0.2, "train/clean_ce_loss": 0.1171875, "train/clean_loss_scale": 9.38422966003418, "train/clean_loss_weight": 1.8768459558486938, "train/combined_loss": 1.319657325744629, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10781990521327015, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0997143983840942 }, { "epoch": 1.080568720379147, "step": 456, "train/alpha": 0.2, "train/clean_ce_loss": 0.130859375, "train/clean_loss_scale": 8.840938568115234, "train/clean_loss_weight": 1.7681877613067627, "train/combined_loss": 1.3883036375045776, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.1080568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1569197177886963 }, { "epoch": 1.080568720379147, "step": 456, "train/alpha": 0.2, "train/clean_ce_loss": 0.10107421875, "train/clean_loss_scale": 8.315546035766602, "train/clean_loss_weight": 1.6631091833114624, "train/combined_loss": 1.0085848569869995, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.1080568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8404873609542847 }, { "epoch": 1.080568720379147, "step": 456, "train/alpha": 0.2, "train/clean_ce_loss": 0.09619140625, "train/clean_loss_scale": 12.305218696594238, "train/clean_loss_weight": 2.4610438346862793, "train/combined_loss": 1.420387625694275, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.1080568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1836563348770142 }, { "epoch": 1.080568720379147, "step": 456, "train/alpha": 0.2, "train/clean_ce_loss": 0.1083984375, "train/clean_loss_scale": 9.839423179626465, "train/clean_loss_weight": 1.9678846597671509, "train/combined_loss": 1.2798937559127808, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.1080568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0665781497955322 }, { "epoch": 1.0829383886255923, "step": 457, "train/alpha": 0.2, "train/clean_ce_loss": 0.046630859375, "train/clean_loss_scale": 10.075295448303223, "train/clean_loss_weight": 2.015059232711792, "train/combined_loss": 0.563783586025238, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10829383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46981966495513916 }, { "epoch": 1.0829383886255923, "step": 457, "train/alpha": 0.2, "train/clean_ce_loss": 0.056884765625, "train/clean_loss_scale": 9.916560173034668, "train/clean_loss_weight": 1.9833120107650757, "train/combined_loss": 0.6769214868545532, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10829383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5641012191772461 }, { "epoch": 1.0829383886255923, "step": 457, "train/alpha": 0.2, "train/clean_ce_loss": 0.01544189453125, "train/clean_loss_scale": 24.490699768066406, "train/clean_loss_weight": 4.898139953613281, "train/combined_loss": 0.4538193643093109, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10829383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3781827986240387 }, { "epoch": 1.0829383886255923, "step": 457, "train/alpha": 0.2, "train/clean_ce_loss": 0.06103515625, "train/clean_loss_scale": 14.917165756225586, "train/clean_loss_weight": 2.983433246612549, "train/combined_loss": 1.092565894126892, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10829383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9104715585708618 }, { "epoch": 1.085308056872038, "step": 458, "train/alpha": 0.2, "train/clean_ce_loss": 0.1279296875, "train/clean_loss_scale": 11.154596328735352, "train/clean_loss_weight": 2.230919361114502, "train/combined_loss": 1.712404727935791, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10853080568720379, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4270039796829224 }, { "epoch": 1.085308056872038, "step": 458, "train/alpha": 0.2, "train/clean_ce_loss": 0.037353515625, "train/clean_loss_scale": 13.852154731750488, "train/clean_loss_weight": 2.7704310417175293, "train/combined_loss": 0.6209120154380798, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10853080568720379, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5174266695976257 }, { "epoch": 1.085308056872038, "step": 458, "train/alpha": 0.2, "train/clean_ce_loss": 0.103515625, "train/clean_loss_scale": 8.320516586303711, "train/clean_loss_weight": 1.664103388786316, "train/combined_loss": 1.0335640907287598, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10853080568720379, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.861303448677063 }, { "epoch": 1.085308056872038, "step": 458, "train/alpha": 0.2, "train/clean_ce_loss": 0.10546875, "train/clean_loss_scale": 10.628401756286621, "train/clean_loss_weight": 2.125680446624756, "train/combined_loss": 1.3451571464538574, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10853080568720379, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1209642887115479 }, { "epoch": 1.0876777251184835, "step": 459, "train/alpha": 0.2, "train/clean_ce_loss": 0.107421875, "train/clean_loss_scale": 9.3529634475708, "train/clean_loss_weight": 1.870592713356018, "train/combined_loss": 1.2056553363800049, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10876777251184834, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0047128200531006 }, { "epoch": 1.0876777251184835, "step": 459, "train/alpha": 0.2, "train/clean_ce_loss": 0.014892578125, "train/clean_loss_scale": 24.636104583740234, "train/clean_loss_weight": 4.927220821380615, "train/combined_loss": 0.44027411937713623, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10876777251184834, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3668951094150543 }, { "epoch": 1.0876777251184835, "step": 459, "train/alpha": 0.2, "train/clean_ce_loss": 0.06787109375, "train/clean_loss_scale": 9.228551864624023, "train/clean_loss_weight": 1.8457103967666626, "train/combined_loss": 0.7516222596168518, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10876777251184834, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6263518929481506 }, { "epoch": 1.0876777251184835, "step": 459, "train/alpha": 0.2, "train/clean_ce_loss": 0.11572265625, "train/clean_loss_scale": 8.247812271118164, "train/clean_loss_weight": 1.6495624780654907, "train/combined_loss": 1.1453505754470825, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10876777251184834, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9544587731361389 }, { "epoch": 1.0900473933649288, "grad_norm": 23.257240295410156, "learning_rate": 2e-06, "loss": 3.5537, "step": 460 }, { "epoch": 1.0900473933649288, "step": 460, "train/alpha": 0.2, "train/clean_ce_loss": 0.010986328125, "train/clean_loss_scale": 33.18556213378906, "train/clean_loss_weight": 6.637112617492676, "train/combined_loss": 0.43750497698783875, "train/pos_loss/pos0": 0.018994202837347984, "train/pos_loss/pos1": 0.09834370017051697, "train/pos_loss/pos2": 0.3883710205554962, "train/pos_loss/pos3": 0.8156739473342896, "train/pos_loss/pos4": 1.385622262954712, "train/pos_loss/pos5": 1.9997700452804565, "train/pos_loss/pos6": 2.4473459720611572, "train/pos_loss/weight0": 2.4712958335876465, "train/pos_loss/weight1": 1.8136210441589355, "train/pos_loss/weight2": 1.2463006973266602, "train/pos_loss/weight3": 0.7531614303588867, "train/pos_loss/weight4": 0.40773335099220276, "train/pos_loss/weight5": 0.20177669823169708, "train/pos_loss/weight6": 0.10611104220151901, "train/pos_weight_progress": 0.10900473933649289, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3645874857902527 }, { "epoch": 1.0900473933649288, "step": 460, "train/alpha": 0.2, "train/clean_ce_loss": 0.185546875, "train/clean_loss_scale": 7.927001476287842, "train/clean_loss_weight": 1.5854003429412842, "train/combined_loss": 1.7649964094161987, "train/pos_loss/pos0": 0.5752732157707214, "train/pos_loss/pos1": 1.064940333366394, "train/pos_loss/pos2": 1.7474746704101562, "train/pos_loss/pos3": 2.3244783878326416, "train/pos_loss/pos4": 2.6841862201690674, "train/pos_loss/pos5": 3.155426263809204, "train/pos_loss/pos6": 3.5939929485321045, "train/pos_loss/weight0": 2.185953378677368, "train/pos_loss/weight1": 1.6692922115325928, "train/pos_loss/weight2": 1.2575334310531616, "train/pos_loss/weight3": 0.8645390272140503, "train/pos_loss/weight4": 0.5386756658554077, "train/pos_loss/weight5": 0.29676997661590576, "train/pos_loss/weight6": 0.1872357428073883, "train/pos_weight_progress": 0.10900473933649289, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4708303213119507 }, { "epoch": 1.0900473933649288, "step": 460, "train/alpha": 0.2, "train/clean_ce_loss": 0.14453125, "train/clean_loss_scale": 7.8302531242370605, "train/clean_loss_weight": 1.56605064868927, "train/combined_loss": 1.3580595254898071, "train/pos_loss/pos0": 0.5483095645904541, "train/pos_loss/pos1": 1.1250298023223877, "train/pos_loss/pos2": 1.889626145362854, "train/pos_loss/pos3": 2.670672655105591, "train/pos_loss/pos4": 3.332420587539673, "train/pos_loss/pos5": 3.563455820083618, "train/pos_loss/pos6": 3.948328733444214, "train/pos_loss/weight0": 3.258174419403076, "train/pos_loss/weight1": 2.0470893383026123, "train/pos_loss/weight2": 1.060530185699463, "train/pos_loss/weight3": 0.4479670524597168, "train/pos_loss/weight4": 0.14066943526268005, "train/pos_loss/weight5": 0.03617769479751587, "train/pos_loss/weight6": 0.00939198862761259, "train/pos_weight_progress": 0.10900473933649289, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.131716251373291 }, { "epoch": 1.0900473933649288, "step": 460, "train/alpha": 0.2, "train/clean_ce_loss": 0.058837890625, "train/clean_loss_scale": 9.215110778808594, "train/clean_loss_weight": 1.8430222272872925, "train/combined_loss": 0.6506372690200806, "train/pos_loss/pos0": 0.18507088720798492, "train/pos_loss/pos1": 0.5013359189033508, "train/pos_loss/pos2": 0.9954662919044495, "train/pos_loss/pos3": 2.0073018074035645, "train/pos_loss/pos4": 2.405122756958008, "train/pos_loss/pos5": 2.9911675453186035, "train/pos_loss/pos6": 3.635892152786255, "train/pos_loss/weight0": 3.3577942848205566, "train/pos_loss/weight1": 2.0699920654296875, "train/pos_loss/weight2": 1.052382230758667, "train/pos_loss/weight3": 0.39483317732810974, "train/pos_loss/weight4": 0.10200390219688416, "train/pos_loss/weight5": 0.01946619525551796, "train/pos_loss/weight6": 0.003528211498633027, "train/pos_weight_progress": 0.10900473933649289, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5421977043151855 }, { "epoch": 1.0924170616113744, "step": 461, "train/alpha": 0.2, "train/clean_ce_loss": 0.050537109375, "train/clean_loss_scale": 9.592406272888184, "train/clean_loss_weight": 1.9184812307357788, "train/combined_loss": 0.5817269682884216, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.10924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4847724735736847 }, { "epoch": 1.0924170616113744, "step": 461, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 8.371004104614258, "train/clean_loss_weight": 1.6742008924484253, "train/combined_loss": 0.7112083435058594, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.10924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5926735997200012 }, { "epoch": 1.0924170616113744, "step": 461, "train/alpha": 0.2, "train/clean_ce_loss": 0.0211181640625, "train/clean_loss_scale": 11.498618125915527, "train/clean_loss_weight": 2.2997236251831055, "train/combined_loss": 0.2913956344127655, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.10924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.24282969534397125 }, { "epoch": 1.0924170616113744, "step": 461, "train/alpha": 0.2, "train/clean_ce_loss": 0.0859375, "train/clean_loss_scale": 8.844688415527344, "train/clean_loss_weight": 1.7689377069473267, "train/combined_loss": 0.9121084809303284, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.10924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7600904107093811 }, { "epoch": 1.09478672985782, "step": 462, "train/alpha": 0.2, "train/clean_ce_loss": 0.0015411376953125, "train/clean_loss_scale": 191.2582244873047, "train/clean_loss_weight": 38.251644134521484, "train/combined_loss": 0.3537063002586365, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.109478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2947552502155304 }, { "epoch": 1.09478672985782, "step": 462, "train/alpha": 0.2, "train/clean_ce_loss": 0.06787109375, "train/clean_loss_scale": 12.59361457824707, "train/clean_loss_weight": 2.5187230110168457, "train/combined_loss": 1.0256909132003784, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.109478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8547424077987671 }, { "epoch": 1.09478672985782, "step": 462, "train/alpha": 0.2, "train/clean_ce_loss": 0.07763671875, "train/clean_loss_scale": 10.202048301696777, "train/clean_loss_weight": 2.040409803390503, "train/combined_loss": 0.9504643082618713, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.109478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7920535802841187 }, { "epoch": 1.09478672985782, "step": 462, "train/alpha": 0.2, "train/clean_ce_loss": 0.12353515625, "train/clean_loss_scale": 7.796938419342041, "train/clean_loss_weight": 1.5593876838684082, "train/combined_loss": 1.1558351516723633, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.109478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9631959795951843 }, { "epoch": 1.0971563981042654, "step": 463, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 12.178642272949219, "train/clean_loss_weight": 2.4357285499572754, "train/combined_loss": 1.034708857536316, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.10971563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8622573614120483 }, { "epoch": 1.0971563981042654, "step": 463, "train/alpha": 0.2, "train/clean_ce_loss": 0.0284423828125, "train/clean_loss_scale": 12.587756156921387, "train/clean_loss_weight": 2.5175511837005615, "train/combined_loss": 0.42963093519210815, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.10971563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.35802578926086426 }, { "epoch": 1.0971563981042654, "step": 463, "train/alpha": 0.2, "train/clean_ce_loss": 0.07763671875, "train/clean_loss_scale": 10.08292007446289, "train/clean_loss_weight": 2.0165841579437256, "train/combined_loss": 0.939365804195404, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.10971563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7828048467636108 }, { "epoch": 1.0971563981042654, "step": 463, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 7.097811698913574, "train/clean_loss_weight": 1.4195623397827148, "train/combined_loss": 0.6071956157684326, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.10971563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5059963464736938 }, { "epoch": 1.099526066350711, "step": 464, "train/alpha": 0.2, "train/clean_ce_loss": 0.130859375, "train/clean_loss_scale": 6.20021915435791, "train/clean_loss_weight": 1.2400438785552979, "train/combined_loss": 0.9736281633377075, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.10995260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.811356782913208 }, { "epoch": 1.099526066350711, "step": 464, "train/alpha": 0.2, "train/clean_ce_loss": 0.061279296875, "train/clean_loss_scale": 8.998719215393066, "train/clean_loss_weight": 1.799743890762329, "train/combined_loss": 0.6617221832275391, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.10995260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5514351725578308 }, { "epoch": 1.099526066350711, "step": 464, "train/alpha": 0.2, "train/clean_ce_loss": 0.146484375, "train/clean_loss_scale": 7.795589923858643, "train/clean_loss_weight": 1.5591180324554443, "train/combined_loss": 1.3703185319900513, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.10995260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1419321298599243 }, { "epoch": 1.099526066350711, "step": 464, "train/alpha": 0.2, "train/clean_ce_loss": 0.06787109375, "train/clean_loss_scale": 11.766131401062012, "train/clean_loss_weight": 2.35322642326355, "train/combined_loss": 0.9582962989807129, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.10995260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7985802292823792 }, { "epoch": 1.1018957345971563, "grad_norm": 16.403989791870117, "learning_rate": 2e-06, "loss": 3.7775, "step": 465 }, { "epoch": 1.1018957345971563, "step": 465, "train/alpha": 0.2, "train/clean_ce_loss": 0.11083984375, "train/clean_loss_scale": 7.809811592102051, "train/clean_loss_weight": 1.561962366104126, "train/combined_loss": 1.0387660264968872, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.11018957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8656383156776428 }, { "epoch": 1.1018957345971563, "step": 465, "train/alpha": 0.2, "train/clean_ce_loss": 0.130859375, "train/clean_loss_scale": 7.046401500701904, "train/clean_loss_weight": 1.4092803001403809, "train/combined_loss": 1.1065051555633545, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.11018957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9220876693725586 }, { "epoch": 1.1018957345971563, "step": 465, "train/alpha": 0.2, "train/clean_ce_loss": 0.034912109375, "train/clean_loss_scale": 9.485259056091309, "train/clean_loss_weight": 1.8970518112182617, "train/combined_loss": 0.3973805010318756, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.11018957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3311504125595093 }, { "epoch": 1.1018957345971563, "step": 465, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 8.336108207702637, "train/clean_loss_weight": 1.6672216653823853, "train/combined_loss": 0.810816764831543, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.11018957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6756806373596191 }, { "epoch": 1.1042654028436019, "step": 466, "train/alpha": 0.2, "train/clean_ce_loss": 0.049072265625, "train/clean_loss_scale": 10.253777503967285, "train/clean_loss_weight": 2.050755500793457, "train/combined_loss": 0.6038113236427307, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.11042654028436019, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5031760931015015 }, { "epoch": 1.1042654028436019, "step": 466, "train/alpha": 0.2, "train/clean_ce_loss": 0.057861328125, "train/clean_loss_scale": 7.74299430847168, "train/clean_loss_weight": 1.5485988855361938, "train/combined_loss": 0.5376238822937012, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.11042654028436019, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.44801992177963257 }, { "epoch": 1.1042654028436019, "step": 466, "train/alpha": 0.2, "train/clean_ce_loss": 0.103515625, "train/clean_loss_scale": 7.324090480804443, "train/clean_loss_weight": 1.4648181200027466, "train/combined_loss": 0.9097893238067627, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.11042654028436019, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7581577897071838 }, { "epoch": 1.1042654028436019, "step": 466, "train/alpha": 0.2, "train/clean_ce_loss": 0.08203125, "train/clean_loss_scale": 7.636754512786865, "train/clean_loss_weight": 1.527350902557373, "train/combined_loss": 0.7517430186271667, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.11042654028436019, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6264525055885315 }, { "epoch": 1.1066350710900474, "step": 467, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 11.076340675354004, "train/clean_loss_weight": 2.215268135070801, "train/combined_loss": 0.8761558532714844, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.11066350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7301298975944519 }, { "epoch": 1.1066350710900474, "step": 467, "train/alpha": 0.2, "train/clean_ce_loss": 0.078125, "train/clean_loss_scale": 11.683847427368164, "train/clean_loss_weight": 2.3367695808410645, "train/combined_loss": 1.0953606367111206, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.11066350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9128005504608154 }, { "epoch": 1.1066350710900474, "step": 467, "train/alpha": 0.2, "train/clean_ce_loss": 0.0986328125, "train/clean_loss_scale": 9.028454780578613, "train/clean_loss_weight": 1.8056910037994385, "train/combined_loss": 1.0686023235321045, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.11066350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8905019164085388 }, { "epoch": 1.1066350710900474, "step": 467, "train/alpha": 0.2, "train/clean_ce_loss": 0.1376953125, "train/clean_loss_scale": 6.979990482330322, "train/clean_loss_weight": 1.3959981203079224, "train/combined_loss": 1.153334379196167, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.11066350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9611119627952576 }, { "epoch": 1.1090047393364928, "step": 468, "train/alpha": 0.2, "train/clean_ce_loss": 0.05322265625, "train/clean_loss_scale": 19.87714195251465, "train/clean_loss_weight": 3.975428342819214, "train/combined_loss": 1.2694971561431885, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.11090047393364928, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0579142570495605 }, { "epoch": 1.1090047393364928, "step": 468, "train/alpha": 0.2, "train/clean_ce_loss": 0.09619140625, "train/clean_loss_scale": 9.935690879821777, "train/clean_loss_weight": 1.9871381521224976, "train/combined_loss": 1.1468737125396729, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.11090047393364928, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9557280540466309 }, { "epoch": 1.1090047393364928, "step": 468, "train/alpha": 0.2, "train/clean_ce_loss": 0.09033203125, "train/clean_loss_scale": 7.598790645599365, "train/clean_loss_weight": 1.5197581052780151, "train/combined_loss": 0.823697030544281, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.11090047393364928, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6864141821861267 }, { "epoch": 1.1090047393364928, "step": 468, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 8.400134086608887, "train/clean_loss_weight": 1.680026888847351, "train/combined_loss": 0.6546198129653931, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.11090047393364928, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5455164909362793 }, { "epoch": 1.1113744075829384, "step": 469, "train/alpha": 0.2, "train/clean_ce_loss": 0.04345703125, "train/clean_loss_scale": 11.02537727355957, "train/clean_loss_weight": 2.20507550239563, "train/combined_loss": 0.5749562382698059, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.11113744075829383, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.47913017868995667 }, { "epoch": 1.1113744075829384, "step": 469, "train/alpha": 0.2, "train/clean_ce_loss": 0.146484375, "train/clean_loss_scale": 6.408013820648193, "train/clean_loss_weight": 1.2816027402877808, "train/combined_loss": 1.1264086961746216, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.11113744075829383, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9386739134788513 }, { "epoch": 1.1113744075829384, "step": 469, "train/alpha": 0.2, "train/clean_ce_loss": 0.072265625, "train/clean_loss_scale": 6.803713321685791, "train/clean_loss_weight": 1.3607426881790161, "train/combined_loss": 0.5900095105171204, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.11113744075829383, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.49167460203170776 }, { "epoch": 1.1113744075829384, "step": 469, "train/alpha": 0.2, "train/clean_ce_loss": 0.1337890625, "train/clean_loss_scale": 7.525247097015381, "train/clean_loss_weight": 1.505049467086792, "train/combined_loss": 1.2081549167633057, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.11113744075829383, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0067957639694214 }, { "epoch": 1.113744075829384, "grad_norm": 20.240812301635742, "learning_rate": 2e-06, "loss": 3.5917, "step": 470 }, { "epoch": 1.113744075829384, "step": 470, "train/alpha": 0.2, "train/clean_ce_loss": 0.0007476806640625, "train/clean_loss_scale": 272.917236328125, "train/clean_loss_weight": 54.58344650268555, "train/combined_loss": 0.24486592411994934, "train/pos_loss/pos0": 0.01173006184399128, "train/pos_loss/pos1": 0.06975618749856949, "train/pos_loss/pos2": 0.26183196902275085, "train/pos_loss/pos3": 0.6860134601593018, "train/pos_loss/pos4": 1.2038613557815552, "train/pos_loss/pos5": 1.9127105474472046, "train/pos_loss/pos6": 2.426983118057251, "train/pos_loss/weight0": 2.834319591522217, "train/pos_loss/weight1": 1.9606211185455322, "train/pos_loss/weight2": 1.2002322673797607, "train/pos_loss/weight3": 0.6746123433113098, "train/pos_loss/weight4": 0.23114365339279175, "train/pos_loss/weight5": 0.07553845643997192, "train/pos_loss/weight6": 0.023532560095191002, "train/pos_weight_progress": 0.11137440758293839, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.20405493676662445 }, { "epoch": 1.113744075829384, "step": 470, "train/alpha": 0.2, "train/clean_ce_loss": 0.09814453125, "train/clean_loss_scale": 10.375014305114746, "train/clean_loss_weight": 2.075002908706665, "train/combined_loss": 1.2219011783599854, "train/pos_loss/pos0": 0.33517253398895264, "train/pos_loss/pos1": 0.6237970590591431, "train/pos_loss/pos2": 1.073553442955017, "train/pos_loss/pos3": 1.5199934244155884, "train/pos_loss/pos4": 2.072924852371216, "train/pos_loss/pos5": 2.670900821685791, "train/pos_loss/pos6": 2.995065450668335, "train/pos_loss/weight0": 2.1207258701324463, "train/pos_loss/weight1": 1.6466566324234009, "train/pos_loss/weight2": 1.2235862016677856, "train/pos_loss/weight3": 0.8920375108718872, "train/pos_loss/weight4": 0.5544310808181763, "train/pos_loss/weight5": 0.34942683577537537, "train/pos_loss/weight6": 0.21313628554344177, "train/pos_weight_progress": 0.11137440758293839, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0182509422302246 }, { "epoch": 1.113744075829384, "step": 470, "train/alpha": 0.2, "train/clean_ce_loss": 0.1650390625, "train/clean_loss_scale": 6.593648910522461, "train/clean_loss_weight": 1.3187297582626343, "train/combined_loss": 1.3058515787124634, "train/pos_loss/pos0": 0.5347605347633362, "train/pos_loss/pos1": 0.9789335131645203, "train/pos_loss/pos2": 1.4415802955627441, "train/pos_loss/pos3": 1.99173903465271, "train/pos_loss/pos4": 2.681018352508545, "train/pos_loss/pos5": 3.0075809955596924, "train/pos_loss/pos6": 3.4406044483184814, "train/pos_loss/weight0": 2.791712999343872, "train/pos_loss/weight1": 1.9395251274108887, "train/pos_loss/weight2": 1.195281744003296, "train/pos_loss/weight3": 0.6387494802474976, "train/pos_loss/weight4": 0.28361326456069946, "train/pos_loss/weight5": 0.10908298194408417, "train/pos_loss/weight6": 0.042034536600112915, "train/pos_weight_progress": 0.11137440758293839, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.088209629058838 }, { "epoch": 1.113744075829384, "step": 470, "train/alpha": 0.2, "train/clean_ce_loss": 0.1328125, "train/clean_loss_scale": 7.941218852996826, "train/clean_loss_weight": 1.588243842124939, "train/combined_loss": 1.2656316757202148, "train/pos_loss/pos0": 0.6423002481460571, "train/pos_loss/pos1": 1.0063323974609375, "train/pos_loss/pos2": 1.5468600988388062, "train/pos_loss/pos3": 2.268521308898926, "train/pos_loss/pos4": 2.6407649517059326, "train/pos_loss/pos5": 2.989325761795044, "train/pos_loss/pos6": 3.8009469509124756, "train/pos_loss/weight0": 3.2257282733917236, "train/pos_loss/weight1": 2.0404069423675537, "train/pos_loss/weight2": 1.0651969909667969, "train/pos_loss/weight3": 0.4609152376651764, "train/pos_loss/weight4": 0.15934312343597412, "train/pos_loss/weight5": 0.038524698466062546, "train/pos_loss/weight6": 0.009884531609714031, "train/pos_weight_progress": 0.11137440758293839, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0546931028366089 }, { "epoch": 1.1161137440758293, "step": 471, "train/alpha": 0.2, "train/clean_ce_loss": 0.1064453125, "train/clean_loss_scale": 7.545910835266113, "train/clean_loss_weight": 1.5091822147369385, "train/combined_loss": 0.9638721942901611, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11161137440758294, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8032268285751343 }, { "epoch": 1.1161137440758293, "step": 471, "train/alpha": 0.2, "train/clean_ce_loss": 0.0830078125, "train/clean_loss_scale": 9.2897367477417, "train/clean_loss_weight": 1.8579473495483398, "train/combined_loss": 0.9253448843955994, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11161137440758294, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7711207270622253 }, { "epoch": 1.1161137440758293, "step": 471, "train/alpha": 0.2, "train/clean_ce_loss": 0.1044921875, "train/clean_loss_scale": 4.9486470222473145, "train/clean_loss_weight": 0.9897294044494629, "train/combined_loss": 0.6205139756202698, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11161137440758294, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5170949697494507 }, { "epoch": 1.1161137440758293, "step": 471, "train/alpha": 0.2, "train/clean_ce_loss": 0.08203125, "train/clean_loss_scale": 9.058632850646973, "train/clean_loss_weight": 1.8117265701293945, "train/combined_loss": 0.8917092084884644, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11161137440758294, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7430909872055054 }, { "epoch": 1.1184834123222749, "step": 472, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 8.17923641204834, "train/clean_loss_weight": 1.6358473300933838, "train/combined_loss": 0.7811809778213501, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11184834123222749, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6509841680526733 }, { "epoch": 1.1184834123222749, "step": 472, "train/alpha": 0.2, "train/clean_ce_loss": 0.1318359375, "train/clean_loss_scale": 7.360081672668457, "train/clean_loss_weight": 1.4720163345336914, "train/combined_loss": 1.1643879413604736, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11184834123222749, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9703232645988464 }, { "epoch": 1.1184834123222749, "step": 472, "train/alpha": 0.2, "train/clean_ce_loss": 0.02587890625, "train/clean_loss_scale": 6.689792633056641, "train/clean_loss_weight": 1.337958574295044, "train/combined_loss": 0.20774942636489868, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11184834123222749, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1731245219707489 }, { "epoch": 1.1184834123222749, "step": 472, "train/alpha": 0.2, "train/clean_ce_loss": 0.072265625, "train/clean_loss_scale": 8.656378746032715, "train/clean_loss_weight": 1.7312757968902588, "train/combined_loss": 0.750670313835144, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11184834123222749, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.625558614730835 }, { "epoch": 1.1208530805687205, "step": 473, "train/alpha": 0.2, "train/clean_ce_loss": 0.08642578125, "train/clean_loss_scale": 8.32347297668457, "train/clean_loss_weight": 1.6646946668624878, "train/combined_loss": 0.8632351756095886, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11208530805687204, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7193626165390015 }, { "epoch": 1.1208530805687205, "step": 473, "train/alpha": 0.2, "train/clean_ce_loss": 0.10205078125, "train/clean_loss_scale": 7.162360668182373, "train/clean_loss_weight": 1.4324721097946167, "train/combined_loss": 0.8771094083786011, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11208530805687204, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7309244871139526 }, { "epoch": 1.1208530805687205, "step": 473, "train/alpha": 0.2, "train/clean_ce_loss": 0.02197265625, "train/clean_loss_scale": 12.726110458374023, "train/clean_loss_weight": 2.545222043991089, "train/combined_loss": 0.33555173873901367, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11208530805687204, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2796264588832855 }, { "epoch": 1.1208530805687205, "step": 473, "train/alpha": 0.2, "train/clean_ce_loss": 0.021728515625, "train/clean_loss_scale": 5.402248859405518, "train/clean_loss_weight": 1.0804498195648193, "train/combined_loss": 0.1408594250679016, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11208530805687204, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.11738284677267075 }, { "epoch": 1.1232227488151658, "step": 474, "train/alpha": 0.2, "train/clean_ce_loss": 0.07568359375, "train/clean_loss_scale": 6.530500411987305, "train/clean_loss_weight": 1.3061001300811768, "train/combined_loss": 0.5931020975112915, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11232227488151658, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.49425172805786133 }, { "epoch": 1.1232227488151658, "step": 474, "train/alpha": 0.2, "train/clean_ce_loss": 0.1484375, "train/clean_loss_scale": 6.525442600250244, "train/clean_loss_weight": 1.3050885200500488, "train/combined_loss": 1.1623444557189941, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11232227488151658, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9686204195022583 }, { "epoch": 1.1232227488151658, "step": 474, "train/alpha": 0.2, "train/clean_ce_loss": 0.07177734375, "train/clean_loss_scale": 6.2412943840026855, "train/clean_loss_weight": 1.2482589483261108, "train/combined_loss": 0.5375802516937256, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11232227488151658, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4479835331439972 }, { "epoch": 1.1232227488151658, "step": 474, "train/alpha": 0.2, "train/clean_ce_loss": 0.0218505859375, "train/clean_loss_scale": 6.03949499130249, "train/clean_loss_weight": 1.2078989744186401, "train/combined_loss": 0.15835979580879211, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11232227488151658, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1319665014743805 }, { "epoch": 1.1255924170616114, "grad_norm": 25.91683578491211, "learning_rate": 2e-06, "loss": 3.3522, "step": 475 }, { "epoch": 1.1255924170616114, "step": 475, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 5.8953704833984375, "train/clean_loss_weight": 1.1790741682052612, "train/combined_loss": 0.5043305158615112, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11255924170616113, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4202754497528076 }, { "epoch": 1.1255924170616114, "step": 475, "train/alpha": 0.2, "train/clean_ce_loss": 0.06787109375, "train/clean_loss_scale": 6.874258995056152, "train/clean_loss_weight": 1.3748518228530884, "train/combined_loss": 0.5598761439323425, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11255924170616113, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46656346321105957 }, { "epoch": 1.1255924170616114, "step": 475, "train/alpha": 0.2, "train/clean_ce_loss": 0.0341796875, "train/clean_loss_scale": 5.761291027069092, "train/clean_loss_weight": 1.152258276939392, "train/combined_loss": 0.2363029569387436, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11255924170616113, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.19691912829875946 }, { "epoch": 1.1255924170616114, "step": 475, "train/alpha": 0.2, "train/clean_ce_loss": 0.061279296875, "train/clean_loss_scale": 11.544414520263672, "train/clean_loss_weight": 2.30888295173645, "train/combined_loss": 0.8489202857017517, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11255924170616113, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7074335813522339 }, { "epoch": 1.1279620853080567, "step": 476, "train/alpha": 0.2, "train/clean_ce_loss": 0.055419921875, "train/clean_loss_scale": 7.226963996887207, "train/clean_loss_weight": 1.4453928470611572, "train/combined_loss": 0.480621337890625, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11279620853080569, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4005177915096283 }, { "epoch": 1.1279620853080567, "step": 476, "train/alpha": 0.2, "train/clean_ce_loss": 0.005096435546875, "train/clean_loss_scale": 36.24871826171875, "train/clean_loss_weight": 7.249743938446045, "train/combined_loss": 0.22168710827827454, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11279620853080569, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.18473926186561584 }, { "epoch": 1.1279620853080567, "step": 476, "train/alpha": 0.2, "train/clean_ce_loss": 0.12158203125, "train/clean_loss_scale": 5.168830871582031, "train/clean_loss_weight": 1.0337661504745483, "train/combined_loss": 0.7541244029998779, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11279620853080569, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6284369826316833 }, { "epoch": 1.1279620853080567, "step": 476, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 6.447566509246826, "train/clean_loss_weight": 1.289513349533081, "train/combined_loss": 0.5477913022041321, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11279620853080569, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45649275183677673 }, { "epoch": 1.1303317535545023, "step": 477, "train/alpha": 0.2, "train/clean_ce_loss": 0.09423828125, "train/clean_loss_scale": 6.899375915527344, "train/clean_loss_weight": 1.3798751831054688, "train/combined_loss": 0.7802224159240723, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11303317535545024, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6501853466033936 }, { "epoch": 1.1303317535545023, "step": 477, "train/alpha": 0.2, "train/clean_ce_loss": 0.0693359375, "train/clean_loss_scale": 5.951901912689209, "train/clean_loss_weight": 1.1903804540634155, "train/combined_loss": 0.49521684646606445, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11303317535545024, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4126806855201721 }, { "epoch": 1.1303317535545023, "step": 477, "train/alpha": 0.2, "train/clean_ce_loss": 0.060302734375, "train/clean_loss_scale": 5.778961181640625, "train/clean_loss_weight": 1.155792236328125, "train/combined_loss": 0.4181846082210541, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11303317535545024, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.34848716855049133 }, { "epoch": 1.1303317535545023, "step": 477, "train/alpha": 0.2, "train/clean_ce_loss": 0.07373046875, "train/clean_loss_scale": 5.604326248168945, "train/clean_loss_weight": 1.1208652257919312, "train/combined_loss": 0.4958515465259552, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11303317535545024, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.41320961713790894 }, { "epoch": 1.132701421800948, "step": 478, "train/alpha": 0.2, "train/clean_ce_loss": 0.044677734375, "train/clean_loss_scale": 8.451218605041504, "train/clean_loss_weight": 1.6902437210083008, "train/combined_loss": 0.45309755206108093, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11327014218009479, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.37758129835128784 }, { "epoch": 1.132701421800948, "step": 478, "train/alpha": 0.2, "train/clean_ce_loss": 0.078125, "train/clean_loss_scale": 7.215850830078125, "train/clean_loss_weight": 1.443170189857483, "train/combined_loss": 0.6764860153198242, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11327014218009479, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5637383460998535 }, { "epoch": 1.132701421800948, "step": 478, "train/alpha": 0.2, "train/clean_ce_loss": 0.054931640625, "train/clean_loss_scale": 5.286200046539307, "train/clean_loss_weight": 1.0572400093078613, "train/combined_loss": 0.3484555780887604, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11327014218009479, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2903796434402466 }, { "epoch": 1.132701421800948, "step": 478, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 4.976851463317871, "train/clean_loss_weight": 0.9953703284263611, "train/combined_loss": 0.521986186504364, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11327014218009479, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43498849868774414 }, { "epoch": 1.1350710900473935, "step": 479, "train/alpha": 0.2, "train/clean_ce_loss": 0.0400390625, "train/clean_loss_scale": 7.1561689376831055, "train/clean_loss_weight": 1.4312337636947632, "train/combined_loss": 0.34383153915405273, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11350710900473934, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.28652629256248474 }, { "epoch": 1.1350710900473935, "step": 479, "train/alpha": 0.2, "train/clean_ce_loss": 0.044921875, "train/clean_loss_scale": 5.458013534545898, "train/clean_loss_weight": 1.0916026830673218, "train/combined_loss": 0.29422104358673096, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11350710900473934, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.24518419802188873 }, { "epoch": 1.1350710900473935, "step": 479, "train/alpha": 0.2, "train/clean_ce_loss": 0.052490234375, "train/clean_loss_scale": 6.941333770751953, "train/clean_loss_weight": 1.3882668018341064, "train/combined_loss": 0.4372226595878601, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11350710900473934, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3643522262573242 }, { "epoch": 1.1350710900473935, "step": 479, "train/alpha": 0.2, "train/clean_ce_loss": 0.087890625, "train/clean_loss_scale": 8.031668663024902, "train/clean_loss_weight": 1.6063337326049805, "train/combined_loss": 0.8470900058746338, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11350710900473934, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7059083580970764 }, { "epoch": 1.1374407582938388, "grad_norm": 20.371116638183594, "learning_rate": 2e-06, "loss": 3.3061, "step": 480 }, { "epoch": 1.1374407582938388, "step": 480, "train/alpha": 0.2, "train/clean_ce_loss": 0.09619140625, "train/clean_loss_scale": 8.300662994384766, "train/clean_loss_weight": 1.660132646560669, "train/combined_loss": 0.9581429362297058, "train/pos_loss/pos0": 0.45392686128616333, "train/pos_loss/pos1": 0.8050763010978699, "train/pos_loss/pos2": 1.1620678901672363, "train/pos_loss/pos3": 1.7201828956604004, "train/pos_loss/pos4": 2.0302774906158447, "train/pos_loss/pos5": 2.5065536499023438, "train/pos_loss/pos6": 2.9205424785614014, "train/pos_loss/weight0": 3.2491137981414795, "train/pos_loss/weight1": 2.01617169380188, "train/pos_loss/weight2": 1.0506932735443115, "train/pos_loss/weight3": 0.47342121601104736, "train/pos_loss/weight4": 0.15676501393318176, "train/pos_loss/weight5": 0.04122152924537659, "train/pos_loss/weight6": 0.01261388324201107, "train/pos_weight_progress": 0.11374407582938388, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7984524369239807 }, { "epoch": 1.1374407582938388, "step": 480, "train/alpha": 0.2, "train/clean_ce_loss": 0.09716796875, "train/clean_loss_scale": 10.891043663024902, "train/clean_loss_weight": 2.178208827972412, "train/combined_loss": 1.2699127197265625, "train/pos_loss/pos0": 0.4339745342731476, "train/pos_loss/pos1": 0.9245785474777222, "train/pos_loss/pos2": 1.6285967826843262, "train/pos_loss/pos3": 2.067354917526245, "train/pos_loss/pos4": 2.671135187149048, "train/pos_loss/pos5": 3.2585020065307617, "train/pos_loss/pos6": 3.4709110260009766, "train/pos_loss/weight0": 2.855093002319336, "train/pos_loss/weight1": 1.9413306713104248, "train/pos_loss/weight2": 1.1740056276321411, "train/pos_loss/weight3": 0.6221534609794617, "train/pos_loss/weight4": 0.2680943012237549, "train/pos_loss/weight5": 0.09994953870773315, "train/pos_loss/weight6": 0.039373643696308136, "train/pos_weight_progress": 0.11374407582938388, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0582605600357056 }, { "epoch": 1.1374407582938388, "step": 480, "train/alpha": 0.2, "train/clean_ce_loss": 0.09228515625, "train/clean_loss_scale": 4.582406997680664, "train/clean_loss_weight": 0.9164814352989197, "train/combined_loss": 0.5074657797813416, "train/pos_loss/pos0": 0.16333983838558197, "train/pos_loss/pos1": 0.3286018967628479, "train/pos_loss/pos2": 0.7054134011268616, "train/pos_loss/pos3": 1.161280632019043, "train/pos_loss/pos4": 1.7102177143096924, "train/pos_loss/pos5": 2.1599385738372803, "train/pos_loss/pos6": 2.5022008419036865, "train/pos_loss/weight0": 3.0855534076690674, "train/pos_loss/weight1": 2.055009365081787, "train/pos_loss/weight2": 1.1610766649246216, "train/pos_loss/weight3": 0.47304898500442505, "train/pos_loss/weight4": 0.17021633684635162, "train/pos_loss/weight5": 0.04276135191321373, "train/pos_loss/weight6": 0.012333877384662628, "train/pos_weight_progress": 0.11374407582938388, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4228881299495697 }, { "epoch": 1.1374407582938388, "step": 480, "train/alpha": 0.2, "train/clean_ce_loss": 0.1044921875, "train/clean_loss_scale": 11.992886543273926, "train/clean_loss_weight": 2.3985774517059326, "train/combined_loss": 1.5037956237792969, "train/pos_loss/pos0": 0.4690767228603363, "train/pos_loss/pos1": 0.9928938150405884, "train/pos_loss/pos2": 1.631833791732788, "train/pos_loss/pos3": 2.299109935760498, "train/pos_loss/pos4": 3.0955183506011963, "train/pos_loss/pos5": 3.122554302215576, "train/pos_loss/pos6": 3.4930167198181152, "train/pos_loss/weight0": 2.526901960372925, "train/pos_loss/weight1": 1.8374675512313843, "train/pos_loss/weight2": 1.2477937936782837, "train/pos_loss/weight3": 0.7515280842781067, "train/pos_loss/weight4": 0.3818608224391937, "train/pos_loss/weight5": 0.16978846490383148, "train/pos_loss/weight6": 0.0846591368317604, "train/pos_weight_progress": 0.11374407582938388, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2531629800796509 }, { "epoch": 1.1398104265402844, "step": 481, "train/alpha": 0.2, "train/clean_ce_loss": 0.0634765625, "train/clean_loss_scale": 6.169649600982666, "train/clean_loss_weight": 1.233929991722107, "train/combined_loss": 0.4699537754058838, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11398104265402843, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3916281461715698 }, { "epoch": 1.1398104265402844, "step": 481, "train/alpha": 0.2, "train/clean_ce_loss": 0.087890625, "train/clean_loss_scale": 7.485250473022461, "train/clean_loss_weight": 1.497050166130066, "train/combined_loss": 0.7894600033760071, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11398104265402843, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.65788334608078 }, { "epoch": 1.1398104265402844, "step": 481, "train/alpha": 0.2, "train/clean_ce_loss": 0.1220703125, "train/clean_loss_scale": 8.539783477783203, "train/clean_loss_weight": 1.7079566717147827, "train/combined_loss": 1.2509448528289795, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11398104265402843, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0424540042877197 }, { "epoch": 1.1398104265402844, "step": 481, "train/alpha": 0.2, "train/clean_ce_loss": 0.0966796875, "train/clean_loss_scale": 7.997714519500732, "train/clean_loss_weight": 1.5995429754257202, "train/combined_loss": 0.9278598427772522, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11398104265402843, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7732165455818176 }, { "epoch": 1.1421800947867298, "step": 482, "train/alpha": 0.2, "train/clean_ce_loss": 0.123046875, "train/clean_loss_scale": 10.664081573486328, "train/clean_loss_weight": 2.1328163146972656, "train/combined_loss": 1.5746183395385742, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11421800947867299, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3121819496154785 }, { "epoch": 1.1421800947867298, "step": 482, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 12.686230659484863, "train/clean_loss_weight": 2.5372462272644043, "train/combined_loss": 1.0109339952468872, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11421800947867299, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8424450159072876 }, { "epoch": 1.1421800947867298, "step": 482, "train/alpha": 0.2, "train/clean_ce_loss": 0.068359375, "train/clean_loss_scale": 6.530041694641113, "train/clean_loss_weight": 1.3060083389282227, "train/combined_loss": 0.5356674790382385, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11421800947867299, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4463895559310913 }, { "epoch": 1.1421800947867298, "step": 482, "train/alpha": 0.2, "train/clean_ce_loss": 0.10107421875, "train/clean_loss_scale": 6.641232490539551, "train/clean_loss_weight": 1.3282464742660522, "train/combined_loss": 0.8055088520050049, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11421800947867299, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6712573766708374 }, { "epoch": 1.1445497630331753, "step": 483, "train/alpha": 0.2, "train/clean_ce_loss": 0.1005859375, "train/clean_loss_scale": 7.986262798309326, "train/clean_loss_weight": 1.597252607345581, "train/combined_loss": 0.9639669060707092, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11445497630331754, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8033057451248169 }, { "epoch": 1.1445497630331753, "step": 483, "train/alpha": 0.2, "train/clean_ce_loss": 0.1611328125, "train/clean_loss_scale": 6.141130447387695, "train/clean_loss_weight": 1.2282260656356812, "train/combined_loss": 1.1874451637268066, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11445497630331754, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9895375967025757 }, { "epoch": 1.1445497630331753, "step": 483, "train/alpha": 0.2, "train/clean_ce_loss": 0.0157470703125, "train/clean_loss_scale": 16.097930908203125, "train/clean_loss_weight": 3.219586133956909, "train/combined_loss": 0.30419430136680603, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11445497630331754, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2534952461719513 }, { "epoch": 1.1445497630331753, "step": 483, "train/alpha": 0.2, "train/clean_ce_loss": 0.087890625, "train/clean_loss_scale": 6.722198009490967, "train/clean_loss_weight": 1.3444396257400513, "train/combined_loss": 0.7089818120002747, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11445497630331754, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5908181667327881 }, { "epoch": 1.146919431279621, "step": 484, "train/alpha": 0.2, "train/clean_ce_loss": 0.06787109375, "train/clean_loss_scale": 6.350290775299072, "train/clean_loss_weight": 1.2700581550598145, "train/combined_loss": 0.5172014236450195, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11469194312796209, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4310011863708496 }, { "epoch": 1.146919431279621, "step": 484, "train/alpha": 0.2, "train/clean_ce_loss": 0.0096435546875, "train/clean_loss_scale": 20.401330947875977, "train/clean_loss_weight": 4.08026647567749, "train/combined_loss": 0.23608961701393127, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11469194312796209, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.19674134254455566 }, { "epoch": 1.146919431279621, "step": 484, "train/alpha": 0.2, "train/clean_ce_loss": 0.130859375, "train/clean_loss_scale": 6.548804759979248, "train/clean_loss_weight": 1.3097609281539917, "train/combined_loss": 1.028367042541504, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11469194312796209, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8569725155830383 }, { "epoch": 1.146919431279621, "step": 484, "train/alpha": 0.2, "train/clean_ce_loss": 0.0223388671875, "train/clean_loss_scale": 11.777179718017578, "train/clean_loss_weight": 2.355436086654663, "train/combined_loss": 0.31570661067962646, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11469194312796209, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2630888521671295 }, { "epoch": 1.1492890995260663, "grad_norm": 13.44551944732666, "learning_rate": 2e-06, "loss": 3.2461, "step": 485 }, { "epoch": 1.1492890995260663, "step": 485, "train/alpha": 0.2, "train/clean_ce_loss": 0.0673828125, "train/clean_loss_scale": 10.243276596069336, "train/clean_loss_weight": 2.0486552715301514, "train/combined_loss": 0.8282649517059326, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11492890995260663, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6902207732200623 }, { "epoch": 1.1492890995260663, "step": 485, "train/alpha": 0.2, "train/clean_ce_loss": 0.0228271484375, "train/clean_loss_scale": 6.255417346954346, "train/clean_loss_weight": 1.251083493232727, "train/combined_loss": 0.17135201394557953, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11492890995260663, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.14279334247112274 }, { "epoch": 1.1492890995260663, "step": 485, "train/alpha": 0.2, "train/clean_ce_loss": 0.11328125, "train/clean_loss_scale": 7.865237712860107, "train/clean_loss_weight": 1.5730475187301636, "train/combined_loss": 1.0691807270050049, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11492890995260663, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8909839391708374 }, { "epoch": 1.1492890995260663, "step": 485, "train/alpha": 0.2, "train/clean_ce_loss": 0.076171875, "train/clean_loss_scale": 6.151696681976318, "train/clean_loss_weight": 1.2303394079208374, "train/combined_loss": 0.5623035430908203, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11492890995260663, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.468586266040802 }, { "epoch": 1.1516587677725119, "step": 486, "train/alpha": 0.2, "train/clean_ce_loss": 0.061279296875, "train/clean_loss_scale": 6.870931625366211, "train/clean_loss_weight": 1.374186396598816, "train/combined_loss": 0.5052550435066223, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11516587677725118, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.42104586958885193 }, { "epoch": 1.1516587677725119, "step": 486, "train/alpha": 0.2, "train/clean_ce_loss": 0.0673828125, "train/clean_loss_scale": 6.737513065338135, "train/clean_loss_weight": 1.347502589225769, "train/combined_loss": 0.5447911024093628, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11516587677725118, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4539925754070282 }, { "epoch": 1.1516587677725119, "step": 486, "train/alpha": 0.2, "train/clean_ce_loss": 0.0859375, "train/clean_loss_scale": 8.257315635681152, "train/clean_loss_weight": 1.6514631509780884, "train/combined_loss": 0.8515356779098511, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11516587677725118, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7096130847930908 }, { "epoch": 1.1516587677725119, "step": 486, "train/alpha": 0.2, "train/clean_ce_loss": 0.03515625, "train/clean_loss_scale": 6.696783065795898, "train/clean_loss_weight": 1.3393566608428955, "train/combined_loss": 0.2825205326080322, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11516587677725118, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.23543378710746765 }, { "epoch": 1.1540284360189574, "step": 487, "train/alpha": 0.2, "train/clean_ce_loss": 0.01116943359375, "train/clean_loss_scale": 14.759942054748535, "train/clean_loss_weight": 2.951988458633423, "train/combined_loss": 0.19783222675323486, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11540284360189573, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.16486018896102905 }, { "epoch": 1.1540284360189574, "step": 487, "train/alpha": 0.2, "train/clean_ce_loss": 0.042724609375, "train/clean_loss_scale": 8.288762092590332, "train/clean_loss_weight": 1.6577523946762085, "train/combined_loss": 0.4249609708786011, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11540284360189573, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.35413414239883423 }, { "epoch": 1.1540284360189574, "step": 487, "train/alpha": 0.2, "train/clean_ce_loss": 0.08056640625, "train/clean_loss_scale": 7.369451522827148, "train/clean_loss_weight": 1.4738903045654297, "train/combined_loss": 0.7124762535095215, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11540284360189573, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5937302112579346 }, { "epoch": 1.1540284360189574, "step": 487, "train/alpha": 0.2, "train/clean_ce_loss": 0.043212890625, "train/clean_loss_scale": 9.346944808959961, "train/clean_loss_weight": 1.8693889379501343, "train/combined_loss": 0.48469021916389465, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11540284360189573, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.40390852093696594 }, { "epoch": 1.1563981042654028, "step": 488, "train/alpha": 0.2, "train/clean_ce_loss": 0.1513671875, "train/clean_loss_scale": 9.672161102294922, "train/clean_loss_weight": 1.9344322681427002, "train/combined_loss": 1.7568573951721191, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11563981042654028, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4640477895736694 }, { "epoch": 1.1563981042654028, "step": 488, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 6.585608959197998, "train/clean_loss_weight": 1.3171218633651733, "train/combined_loss": 0.5749545097351074, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11563981042654028, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.47912877798080444 }, { "epoch": 1.1563981042654028, "step": 488, "train/alpha": 0.2, "train/clean_ce_loss": 0.054931640625, "train/clean_loss_scale": 6.62684440612793, "train/clean_loss_weight": 1.325368881225586, "train/combined_loss": 0.4368281364440918, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11563981042654028, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.36402344703674316 }, { "epoch": 1.1563981042654028, "step": 488, "train/alpha": 0.2, "train/clean_ce_loss": 0.06103515625, "train/clean_loss_scale": 5.833949089050293, "train/clean_loss_weight": 1.1667898893356323, "train/combined_loss": 0.42729121446609497, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11563981042654028, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3560760021209717 }, { "epoch": 1.1587677725118484, "step": 489, "train/alpha": 0.2, "train/clean_ce_loss": 0.0211181640625, "train/clean_loss_scale": 9.237531661987305, "train/clean_loss_weight": 1.8475064039230347, "train/combined_loss": 0.23409566283226013, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11587677725118484, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1950797140598297 }, { "epoch": 1.1587677725118484, "step": 489, "train/alpha": 0.2, "train/clean_ce_loss": 0.0751953125, "train/clean_loss_scale": 4.7226386070251465, "train/clean_loss_weight": 0.9445277452468872, "train/combined_loss": 0.4261443614959717, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11587677725118484, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.35512030124664307 }, { "epoch": 1.1587677725118484, "step": 489, "train/alpha": 0.2, "train/clean_ce_loss": 0.0262451171875, "train/clean_loss_scale": 14.740897178649902, "train/clean_loss_weight": 2.9481794834136963, "train/combined_loss": 0.46425190567970276, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11587677725118484, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.38687658309936523 }, { "epoch": 1.1587677725118484, "step": 489, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 7.04796838760376, "train/clean_loss_weight": 1.4095937013626099, "train/combined_loss": 0.5616349577903748, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11587677725118484, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4680291414260864 }, { "epoch": 1.161137440758294, "grad_norm": 22.644407272338867, "learning_rate": 2e-06, "loss": 2.9011, "step": 490 }, { "epoch": 1.161137440758294, "step": 490, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 6.577779293060303, "train/clean_loss_weight": 1.3155559301376343, "train/combined_loss": 0.6590626835823059, "train/pos_loss/pos0": 0.25011974573135376, "train/pos_loss/pos1": 0.5547492504119873, "train/pos_loss/pos2": 0.8439621925354004, "train/pos_loss/pos3": 1.248326301574707, "train/pos_loss/pos4": 1.5413508415222168, "train/pos_loss/pos5": 1.749000072479248, "train/pos_loss/pos6": 2.0213871002197266, "train/pos_loss/weight0": 3.1311984062194824, "train/pos_loss/weight1": 2.00321102142334, "train/pos_loss/weight2": 1.1396727561950684, "train/pos_loss/weight3": 0.4957091510295868, "train/pos_loss/weight4": 0.17056383192539215, "train/pos_loss/weight5": 0.046192679554224014, "train/pos_loss/weight6": 0.013452705927193165, "train/pos_weight_progress": 0.11611374407582939, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5492188930511475 }, { "epoch": 1.161137440758294, "step": 490, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 8.375589370727539, "train/clean_loss_weight": 1.67511785030365, "train/combined_loss": 0.7950266599655151, "train/pos_loss/pos0": 0.23663902282714844, "train/pos_loss/pos1": 0.48906585574150085, "train/pos_loss/pos2": 0.765227198600769, "train/pos_loss/pos3": 1.0969939231872559, "train/pos_loss/pos4": 1.536700963973999, "train/pos_loss/pos5": 1.812374472618103, "train/pos_loss/pos6": 2.1124489307403564, "train/pos_loss/weight0": 2.4035117626190186, "train/pos_loss/weight1": 1.7786895036697388, "train/pos_loss/weight2": 1.2134698629379272, "train/pos_loss/weight3": 0.7757503986358643, "train/pos_loss/weight4": 0.4457143545150757, "train/pos_loss/weight5": 0.2464270144701004, "train/pos_loss/weight6": 0.13643711805343628, "train/pos_weight_progress": 0.11611374407582939, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6625221967697144 }, { "epoch": 1.161137440758294, "step": 490, "train/alpha": 0.2, "train/clean_ce_loss": 0.1044921875, "train/clean_loss_scale": 9.47848892211914, "train/clean_loss_weight": 1.895697832107544, "train/combined_loss": 1.1885136365890503, "train/pos_loss/pos0": 0.3105657994747162, "train/pos_loss/pos1": 0.6769627928733826, "train/pos_loss/pos2": 1.0843884944915771, "train/pos_loss/pos3": 1.9508322477340698, "train/pos_loss/pos4": 2.296603202819824, "train/pos_loss/pos5": 2.9575867652893066, "train/pos_loss/pos6": 3.3559606075286865, "train/pos_loss/weight0": 2.405301570892334, "train/pos_loss/weight1": 1.7884557247161865, "train/pos_loss/weight2": 1.2298333644866943, "train/pos_loss/weight3": 0.7673576474189758, "train/pos_loss/weight4": 0.4443631172180176, "train/pos_loss/weight5": 0.23421353101730347, "train/pos_loss/weight6": 0.13047508895397186, "train/pos_weight_progress": 0.11611374407582939, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9904280304908752 }, { "epoch": 1.161137440758294, "step": 490, "train/alpha": 0.2, "train/clean_ce_loss": 0.11279296875, "train/clean_loss_scale": 9.08244514465332, "train/clean_loss_weight": 1.8164891004562378, "train/combined_loss": 1.229323148727417, "train/pos_loss/pos0": 0.4129093289375305, "train/pos_loss/pos1": 0.7960888147354126, "train/pos_loss/pos2": 1.586346983909607, "train/pos_loss/pos3": 2.408555507659912, "train/pos_loss/pos4": 2.9020042419433594, "train/pos_loss/pos5": 3.281860589981079, "train/pos_loss/pos6": 3.5080690383911133, "train/pos_loss/weight0": 2.890653610229492, "train/pos_loss/weight1": 1.9785032272338867, "train/pos_loss/weight2": 1.1672050952911377, "train/pos_loss/weight3": 0.6001352667808533, "train/pos_loss/weight4": 0.24559526145458221, "train/pos_loss/weight5": 0.08591602742671967, "train/pos_loss/weight6": 0.03199204057455063, "train/pos_weight_progress": 0.11611374407582939, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0244359970092773 }, { "epoch": 1.1635071090047393, "step": 491, "train/alpha": 0.2, "train/clean_ce_loss": 0.0233154296875, "train/clean_loss_scale": 9.981651306152344, "train/clean_loss_weight": 1.9963302612304688, "train/combined_loss": 0.27927178144454956, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11635071090047393, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.23272649943828583 }, { "epoch": 1.1635071090047393, "step": 491, "train/alpha": 0.2, "train/clean_ce_loss": 0.07177734375, "train/clean_loss_scale": 7.35765266418457, "train/clean_loss_weight": 1.471530556678772, "train/combined_loss": 0.6337352991104126, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11635071090047393, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5281127691268921 }, { "epoch": 1.1635071090047393, "step": 491, "train/alpha": 0.2, "train/clean_ce_loss": 0.05419921875, "train/clean_loss_scale": 6.369420051574707, "train/clean_loss_weight": 1.2738840579986572, "train/combined_loss": 0.4142611026763916, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11635071090047393, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.34521758556365967 }, { "epoch": 1.1635071090047393, "step": 491, "train/alpha": 0.2, "train/clean_ce_loss": 0.0732421875, "train/clean_loss_scale": 7.095862865447998, "train/clean_loss_weight": 1.4191726446151733, "train/combined_loss": 0.6236597895622253, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11635071090047393, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5197165012359619 }, { "epoch": 1.1658767772511849, "step": 492, "train/alpha": 0.2, "train/clean_ce_loss": 0.0390625, "train/clean_loss_scale": 8.337850570678711, "train/clean_loss_weight": 1.6675701141357422, "train/combined_loss": 0.3908367156982422, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11658767772511848, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.32569727301597595 }, { "epoch": 1.1658767772511849, "step": 492, "train/alpha": 0.2, "train/clean_ce_loss": 0.021484375, "train/clean_loss_scale": 8.949275970458984, "train/clean_loss_weight": 1.7898552417755127, "train/combined_loss": 0.2307235151529312, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11658767772511848, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.19226959347724915 }, { "epoch": 1.1658767772511849, "step": 492, "train/alpha": 0.2, "train/clean_ce_loss": 0.1484375, "train/clean_loss_scale": 5.39096212387085, "train/clean_loss_weight": 1.0781924724578857, "train/combined_loss": 0.9602651000022888, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11658767772511848, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8002209067344666 }, { "epoch": 1.1658767772511849, "step": 492, "train/alpha": 0.2, "train/clean_ce_loss": 0.083984375, "train/clean_loss_scale": 8.13388729095459, "train/clean_loss_weight": 1.6267775297164917, "train/combined_loss": 0.819743275642395, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11658767772511848, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6831194162368774 }, { "epoch": 1.1682464454976302, "step": 493, "train/alpha": 0.2, "train/clean_ce_loss": 0.09423828125, "train/clean_loss_scale": 5.795362949371338, "train/clean_loss_weight": 1.1590726375579834, "train/combined_loss": 0.6553740501403809, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11682464454976303, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5461450219154358 }, { "epoch": 1.1682464454976302, "step": 493, "train/alpha": 0.2, "train/clean_ce_loss": 0.09619140625, "train/clean_loss_scale": 6.386846542358398, "train/clean_loss_weight": 1.2773693799972534, "train/combined_loss": 0.7372316718101501, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11682464454976303, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6143597364425659 }, { "epoch": 1.1682464454976302, "step": 493, "train/alpha": 0.2, "train/clean_ce_loss": 0.107421875, "train/clean_loss_scale": 8.112393379211426, "train/clean_loss_weight": 1.622478723526001, "train/combined_loss": 1.0457382202148438, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11682464454976303, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8714484572410583 }, { "epoch": 1.1682464454976302, "step": 493, "train/alpha": 0.2, "train/clean_ce_loss": 0.03857421875, "train/clean_loss_scale": 9.140623092651367, "train/clean_loss_weight": 1.8281246423721313, "train/combined_loss": 0.42311084270477295, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11682464454976303, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3525923788547516 }, { "epoch": 1.1706161137440758, "step": 494, "train/alpha": 0.2, "train/clean_ce_loss": 0.0380859375, "train/clean_loss_scale": 11.882309913635254, "train/clean_loss_weight": 2.376461982727051, "train/combined_loss": 0.5430586934089661, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11706161137440758, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45254892110824585 }, { "epoch": 1.1706161137440758, "step": 494, "train/alpha": 0.2, "train/clean_ce_loss": 0.08056640625, "train/clean_loss_scale": 11.315134048461914, "train/clean_loss_weight": 2.2630269527435303, "train/combined_loss": 1.0939435958862305, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11706161137440758, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9116196632385254 }, { "epoch": 1.1706161137440758, "step": 494, "train/alpha": 0.2, "train/clean_ce_loss": 0.146484375, "train/clean_loss_scale": 7.493314743041992, "train/clean_loss_weight": 1.4986629486083984, "train/combined_loss": 1.3171842098236084, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11706161137440758, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0976535081863403 }, { "epoch": 1.1706161137440758, "step": 494, "train/alpha": 0.2, "train/clean_ce_loss": 0.08642578125, "train/clean_loss_scale": 7.482100009918213, "train/clean_loss_weight": 1.4964200258255005, "train/combined_loss": 0.7759755849838257, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11706161137440758, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6466463208198547 }, { "epoch": 1.1729857819905214, "grad_norm": 14.015607833862305, "learning_rate": 2e-06, "loss": 3.564, "step": 495 }, { "epoch": 1.1729857819905214, "step": 495, "train/alpha": 0.2, "train/clean_ce_loss": 0.1298828125, "train/clean_loss_scale": 7.983131408691406, "train/clean_loss_weight": 1.5966262817382812, "train/combined_loss": 1.2442458868026733, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11729857819905214, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0368715524673462 }, { "epoch": 1.1729857819905214, "step": 495, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 7.478590488433838, "train/clean_loss_weight": 1.4957181215286255, "train/combined_loss": 0.6660619378089905, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11729857819905214, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5550516247749329 }, { "epoch": 1.1729857819905214, "step": 495, "train/alpha": 0.2, "train/clean_ce_loss": 0.06201171875, "train/clean_loss_scale": 9.796856880187988, "train/clean_loss_weight": 1.9593714475631714, "train/combined_loss": 0.7290239334106445, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11729857819905214, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6075199246406555 }, { "epoch": 1.1729857819905214, "step": 495, "train/alpha": 0.2, "train/clean_ce_loss": 0.0301513671875, "train/clean_loss_scale": 5.5583295822143555, "train/clean_loss_weight": 1.111665964126587, "train/combined_loss": 0.20110946893692017, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11729857819905214, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.16759122908115387 }, { "epoch": 1.1753554502369667, "step": 496, "train/alpha": 0.2, "train/clean_ce_loss": 0.09375, "train/clean_loss_scale": 5.027737617492676, "train/clean_loss_weight": 1.0055475234985352, "train/combined_loss": 0.565620481967926, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11753554502369669, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.47135040163993835 }, { "epoch": 1.1753554502369667, "step": 496, "train/alpha": 0.2, "train/clean_ce_loss": 0.0042724609375, "train/clean_loss_scale": 52.66595458984375, "train/clean_loss_weight": 10.533190727233887, "train/combined_loss": 0.2700158953666687, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11753554502369669, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.22501324117183685 }, { "epoch": 1.1753554502369667, "step": 496, "train/alpha": 0.2, "train/clean_ce_loss": 0.1416015625, "train/clean_loss_scale": 7.125709533691406, "train/clean_loss_weight": 1.4251419305801392, "train/combined_loss": 1.2108139991760254, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11753554502369669, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0090116262435913 }, { "epoch": 1.1753554502369667, "step": 496, "train/alpha": 0.2, "train/clean_ce_loss": 0.0703125, "train/clean_loss_scale": 7.806830406188965, "train/clean_loss_weight": 1.561366081237793, "train/combined_loss": 0.6587013006210327, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11753554502369669, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5489177703857422 }, { "epoch": 1.1777251184834123, "step": 497, "train/alpha": 0.2, "train/clean_ce_loss": 0.051025390625, "train/clean_loss_scale": 6.781112194061279, "train/clean_loss_weight": 1.3562225103378296, "train/combined_loss": 0.41521066427230835, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11777251184834123, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.34600889682769775 }, { "epoch": 1.1777251184834123, "step": 497, "train/alpha": 0.2, "train/clean_ce_loss": 0.0888671875, "train/clean_loss_scale": 6.617799758911133, "train/clean_loss_weight": 1.3235599994659424, "train/combined_loss": 0.7057263255119324, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11777251184834123, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5881052613258362 }, { "epoch": 1.1777251184834123, "step": 497, "train/alpha": 0.2, "train/clean_ce_loss": 0.072265625, "train/clean_loss_scale": 5.648377895355225, "train/clean_loss_weight": 1.1296756267547607, "train/combined_loss": 0.4898202419281006, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11777251184834123, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4081835448741913 }, { "epoch": 1.1777251184834123, "step": 497, "train/alpha": 0.2, "train/clean_ce_loss": 0.05419921875, "train/clean_loss_scale": 6.067505836486816, "train/clean_loss_weight": 1.213501214981079, "train/combined_loss": 0.39462488889694214, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11777251184834123, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3288540840148926 }, { "epoch": 1.180094786729858, "step": 498, "train/alpha": 0.2, "train/clean_ce_loss": 0.0732421875, "train/clean_loss_scale": 7.8523478507995605, "train/clean_loss_weight": 1.570469617843628, "train/combined_loss": 0.6901477575302124, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11800947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.575123131275177 }, { "epoch": 1.180094786729858, "step": 498, "train/alpha": 0.2, "train/clean_ce_loss": 0.08447265625, "train/clean_loss_scale": 8.377223014831543, "train/clean_loss_weight": 1.6754446029663086, "train/combined_loss": 0.8491755723953247, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11800947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7076463103294373 }, { "epoch": 1.180094786729858, "step": 498, "train/alpha": 0.2, "train/clean_ce_loss": 0.1337890625, "train/clean_loss_scale": 7.335200786590576, "train/clean_loss_weight": 1.4670401811599731, "train/combined_loss": 1.1776435375213623, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11800947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9813696146011353 }, { "epoch": 1.180094786729858, "step": 498, "train/alpha": 0.2, "train/clean_ce_loss": 0.12451171875, "train/clean_loss_scale": 6.548670768737793, "train/clean_loss_weight": 1.3097342252731323, "train/combined_loss": 0.9784635305404663, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11800947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8153862357139587 }, { "epoch": 1.1824644549763033, "step": 499, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 7.851173400878906, "train/clean_loss_weight": 1.5702346563339233, "train/combined_loss": 0.6164398193359375, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11824644549763033, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5136998295783997 }, { "epoch": 1.1824644549763033, "step": 499, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 7.188284873962402, "train/clean_loss_weight": 1.4376569986343384, "train/combined_loss": 0.6865373849868774, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11824644549763033, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5721144676208496 }, { "epoch": 1.1824644549763033, "step": 499, "train/alpha": 0.2, "train/clean_ce_loss": 0.11181640625, "train/clean_loss_scale": 6.942636966705322, "train/clean_loss_weight": 1.3885273933410645, "train/combined_loss": 0.9315608739852905, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11824644549763033, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7763007283210754 }, { "epoch": 1.1824644549763033, "step": 499, "train/alpha": 0.2, "train/clean_ce_loss": 0.0693359375, "train/clean_loss_scale": 6.238799571990967, "train/clean_loss_weight": 1.2477599382400513, "train/combined_loss": 0.5190876126289368, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11824644549763033, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43257302045822144 }, { "epoch": 1.1848341232227488, "grad_norm": 18.142566680908203, "learning_rate": 2e-06, "loss": 3.4264, "step": 500 }, { "epoch": 1.1848341232227488, "step": 500, "train/alpha": 0.2, "train/clean_ce_loss": 0.021484375, "train/clean_loss_scale": 6.153864860534668, "train/clean_loss_weight": 1.2307729721069336, "train/combined_loss": 0.1586543321609497, "train/pos_loss/pos0": 0.00434637488797307, "train/pos_loss/pos1": 0.07684454321861267, "train/pos_loss/pos2": 0.19183890521526337, "train/pos_loss/pos3": 0.49674996733665466, "train/pos_loss/pos4": 1.3296771049499512, "train/pos_loss/pos5": 2.2335405349731445, "train/pos_loss/pos6": 2.727508306503296, "train/pos_loss/weight0": 3.195258855819702, "train/pos_loss/weight1": 2.0419633388519287, "train/pos_loss/weight2": 1.0858807563781738, "train/pos_loss/weight3": 0.47671762108802795, "train/pos_loss/weight4": 0.15588437020778656, "train/pos_loss/weight5": 0.03496978431940079, "train/pos_loss/weight6": 0.009325697086751461, "train/pos_weight_progress": 0.11848341232227488, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.13221193850040436 }, { "epoch": 1.1848341232227488, "step": 500, "train/alpha": 0.2, "train/clean_ce_loss": 0.052001953125, "train/clean_loss_scale": 12.718070030212402, "train/clean_loss_weight": 2.543614149093628, "train/combined_loss": 0.7936373949050903, "train/pos_loss/pos0": 0.16227692365646362, "train/pos_loss/pos1": 0.3782075047492981, "train/pos_loss/pos2": 0.6739963293075562, "train/pos_loss/pos3": 0.9924277663230896, "train/pos_loss/pos4": 1.4664441347122192, "train/pos_loss/pos5": 1.8555985689163208, "train/pos_loss/pos6": 2.2221853733062744, "train/pos_loss/weight0": 2.1114797592163086, "train/pos_loss/weight1": 1.6634446382522583, "train/pos_loss/weight2": 1.215790033340454, "train/pos_loss/weight3": 0.865647554397583, "train/pos_loss/weight4": 0.569640576839447, "train/pos_loss/weight5": 0.35588401556015015, "train/pos_loss/weight6": 0.21811378002166748, "train/pos_weight_progress": 0.11848341232227488, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6613644957542419 }, { "epoch": 1.1848341232227488, "step": 500, "train/alpha": 0.2, "train/clean_ce_loss": 0.0849609375, "train/clean_loss_scale": 14.688270568847656, "train/clean_loss_weight": 2.9376542568206787, "train/combined_loss": 1.497515082359314, "train/pos_loss/pos0": 0.5566015243530273, "train/pos_loss/pos1": 0.9302026033401489, "train/pos_loss/pos2": 1.4420900344848633, "train/pos_loss/pos3": 2.011352300643921, "train/pos_loss/pos4": 2.5799622535705566, "train/pos_loss/pos5": 2.9242186546325684, "train/pos_loss/pos6": 3.4085426330566406, "train/pos_loss/weight0": 2.339542865753174, "train/pos_loss/weight1": 1.7666137218475342, "train/pos_loss/weight2": 1.2374787330627441, "train/pos_loss/weight3": 0.7976309061050415, "train/pos_loss/weight4": 0.46365368366241455, "train/pos_loss/weight5": 0.25101926922798157, "train/pos_loss/weight6": 0.1440604329109192, "train/pos_weight_progress": 0.11848341232227488, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2479292154312134 }, { "epoch": 1.1848341232227488, "step": 500, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 6.100171089172363, "train/clean_loss_weight": 1.2200342416763306, "train/combined_loss": 0.5969112515449524, "train/pos_loss/pos0": 0.22076158225536346, "train/pos_loss/pos1": 0.5111326575279236, "train/pos_loss/pos2": 0.7658615708351135, "train/pos_loss/pos3": 1.165523648262024, "train/pos_loss/pos4": 1.480772614479065, "train/pos_loss/pos5": 1.9012302160263062, "train/pos_loss/pos6": 2.12353777885437, "train/pos_loss/weight0": 3.2050328254699707, "train/pos_loss/weight1": 1.9821093082427979, "train/pos_loss/weight2": 1.109496831893921, "train/pos_loss/weight3": 0.4894464313983917, "train/pos_loss/weight4": 0.16018499433994293, "train/pos_loss/weight5": 0.041799090802669525, "train/pos_loss/weight6": 0.011930100619792938, "train/pos_weight_progress": 0.11848341232227488, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4974260628223419 }, { "epoch": 1.1872037914691944, "step": 501, "train/alpha": 0.2, "train/clean_ce_loss": 0.109375, "train/clean_loss_scale": 9.292681694030762, "train/clean_loss_weight": 1.8585363626480103, "train/combined_loss": 1.2196645736694336, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.11872037914691944, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0163871049880981 }, { "epoch": 1.1872037914691944, "step": 501, "train/alpha": 0.2, "train/clean_ce_loss": 0.055908203125, "train/clean_loss_scale": 9.26757526397705, "train/clean_loss_weight": 1.8535150289535522, "train/combined_loss": 0.6217601299285889, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.11872037914691944, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5181334614753723 }, { "epoch": 1.1872037914691944, "step": 501, "train/alpha": 0.2, "train/clean_ce_loss": 0.138671875, "train/clean_loss_scale": 8.539318084716797, "train/clean_loss_weight": 1.707863688468933, "train/combined_loss": 1.420995831489563, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.11872037914691944, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.184163212776184 }, { "epoch": 1.1872037914691944, "step": 501, "train/alpha": 0.2, "train/clean_ce_loss": 0.09521484375, "train/clean_loss_scale": 9.946931838989258, "train/clean_loss_weight": 1.9893864393234253, "train/combined_loss": 1.136514663696289, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.11872037914691944, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9470955729484558 }, { "epoch": 1.1895734597156398, "step": 502, "train/alpha": 0.2, "train/clean_ce_loss": 0.103515625, "train/clean_loss_scale": 10.843228340148926, "train/clean_loss_weight": 2.1686456203460693, "train/combined_loss": 1.346932291984558, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.11895734597156397, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1224435567855835 }, { "epoch": 1.1895734597156398, "step": 502, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 15.262011528015137, "train/clean_loss_weight": 3.0524022579193115, "train/combined_loss": 1.4844690561294556, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.11895734597156397, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2370575666427612 }, { "epoch": 1.1895734597156398, "step": 502, "train/alpha": 0.2, "train/clean_ce_loss": 0.1044921875, "train/clean_loss_scale": 6.909339427947998, "train/clean_loss_weight": 1.3818678855895996, "train/combined_loss": 0.8663663864135742, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.11895734597156397, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7219719886779785 }, { "epoch": 1.1895734597156398, "step": 502, "train/alpha": 0.2, "train/clean_ce_loss": 0.0205078125, "train/clean_loss_scale": 21.17280387878418, "train/clean_loss_weight": 4.234560966491699, "train/combined_loss": 0.5210494995117188, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.11895734597156397, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43420788645744324 }, { "epoch": 1.1919431279620853, "step": 503, "train/alpha": 0.2, "train/clean_ce_loss": 0.06103515625, "train/clean_loss_scale": 10.150083541870117, "train/clean_loss_weight": 2.0300166606903076, "train/combined_loss": 0.7434143424034119, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.11919431279620853, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6195119619369507 }, { "epoch": 1.1919431279620853, "step": 503, "train/alpha": 0.2, "train/clean_ce_loss": 0.1298828125, "train/clean_loss_scale": 7.735574722290039, "train/clean_loss_weight": 1.5471149682998657, "train/combined_loss": 1.2056617736816406, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.11919431279620853, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0047181844711304 }, { "epoch": 1.1919431279620853, "step": 503, "train/alpha": 0.2, "train/clean_ce_loss": 0.07568359375, "train/clean_loss_scale": 10.627903938293457, "train/clean_loss_weight": 2.1255807876586914, "train/combined_loss": 0.9652295112609863, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.11919431279620853, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8043579459190369 }, { "epoch": 1.1919431279620853, "step": 503, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 9.034407615661621, "train/clean_loss_weight": 1.8068815469741821, "train/combined_loss": 0.8046269416809082, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.11919431279620853, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6705224514007568 }, { "epoch": 1.1943127962085307, "step": 504, "train/alpha": 0.2, "train/clean_ce_loss": 0.1162109375, "train/clean_loss_scale": 13.025521278381348, "train/clean_loss_weight": 2.6051042079925537, "train/combined_loss": 1.8164496421813965, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.11943127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5137079954147339 }, { "epoch": 1.1943127962085307, "step": 504, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 14.604154586791992, "train/clean_loss_weight": 2.9208309650421143, "train/combined_loss": 1.1380971670150757, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.11943127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9484143257141113 }, { "epoch": 1.1943127962085307, "step": 504, "train/alpha": 0.2, "train/clean_ce_loss": 0.06787109375, "train/clean_loss_scale": 8.557709693908691, "train/clean_loss_weight": 1.711542010307312, "train/combined_loss": 0.6969853043556213, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.11943127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5808210968971252 }, { "epoch": 1.1943127962085307, "step": 504, "train/alpha": 0.2, "train/clean_ce_loss": 0.00010585784912109375, "train/clean_loss_scale": 1898.3380126953125, "train/clean_loss_weight": 379.6676025390625, "train/combined_loss": 0.24114477634429932, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.11943127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.20095397531986237 }, { "epoch": 1.1966824644549763, "grad_norm": 23.615814208984375, "learning_rate": 2e-06, "loss": 3.2349, "step": 505 }, { "epoch": 1.1966824644549763, "step": 505, "train/alpha": 0.2, "train/clean_ce_loss": 0.04833984375, "train/clean_loss_scale": 22.413442611694336, "train/clean_loss_weight": 4.4826884269714355, "train/combined_loss": 1.3001548051834106, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.11966824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0834623575210571 }, { "epoch": 1.1966824644549763, "step": 505, "train/alpha": 0.2, "train/clean_ce_loss": 0.0166015625, "train/clean_loss_scale": 25.574527740478516, "train/clean_loss_weight": 5.114905834197998, "train/combined_loss": 0.5094925165176392, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.11966824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.42457711696624756 }, { "epoch": 1.1966824644549763, "step": 505, "train/alpha": 0.2, "train/clean_ce_loss": 0.10009765625, "train/clean_loss_scale": 7.827346324920654, "train/clean_loss_weight": 1.5654692649841309, "train/combined_loss": 0.9401987791061401, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.11966824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.783499002456665 }, { "epoch": 1.1966824644549763, "step": 505, "train/alpha": 0.2, "train/clean_ce_loss": 0.05322265625, "train/clean_loss_scale": 9.787849426269531, "train/clean_loss_weight": 1.95756995677948, "train/combined_loss": 0.6251224279403687, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.11966824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5209353566169739 }, { "epoch": 1.1990521327014219, "step": 506, "train/alpha": 0.2, "train/clean_ce_loss": 0.1552734375, "train/clean_loss_scale": 8.339656829833984, "train/clean_loss_weight": 1.6679314374923706, "train/combined_loss": 1.55391263961792, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.11990521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2949272394180298 }, { "epoch": 1.1990521327014219, "step": 506, "train/alpha": 0.2, "train/clean_ce_loss": 0.11962890625, "train/clean_loss_scale": 10.430556297302246, "train/clean_loss_weight": 2.086111307144165, "train/combined_loss": 1.4973552227020264, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.11990521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2477960586547852 }, { "epoch": 1.1990521327014219, "step": 506, "train/alpha": 0.2, "train/clean_ce_loss": 0.060302734375, "train/clean_loss_scale": 10.593201637268066, "train/clean_loss_weight": 2.118640422821045, "train/combined_loss": 0.766558825969696, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.11990521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6387990117073059 }, { "epoch": 1.1990521327014219, "step": 506, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 11.721711158752441, "train/clean_loss_weight": 2.3443422317504883, "train/combined_loss": 1.2294059991836548, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.11990521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0245050191879272 }, { "epoch": 1.2014218009478672, "step": 507, "train/alpha": 0.2, "train/clean_ce_loss": 0.08056640625, "train/clean_loss_scale": 10.03208065032959, "train/clean_loss_weight": 2.006416082382202, "train/combined_loss": 0.9698984622955322, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.12014218009478674, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8082486987113953 }, { "epoch": 1.2014218009478672, "step": 507, "train/alpha": 0.2, "train/clean_ce_loss": 0.0849609375, "train/clean_loss_scale": 8.329224586486816, "train/clean_loss_weight": 1.6658449172973633, "train/combined_loss": 0.8491905331611633, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.12014218009478674, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7076587677001953 }, { "epoch": 1.2014218009478672, "step": 507, "train/alpha": 0.2, "train/clean_ce_loss": 0.058837890625, "train/clean_loss_scale": 8.546428680419922, "train/clean_loss_weight": 1.7092857360839844, "train/combined_loss": 0.6034245491027832, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.12014218009478674, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5028538107872009 }, { "epoch": 1.2014218009478672, "step": 507, "train/alpha": 0.2, "train/clean_ce_loss": 0.06005859375, "train/clean_loss_scale": 9.677194595336914, "train/clean_loss_weight": 1.9354389905929565, "train/combined_loss": 0.6974384188652039, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.12014218009478674, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5811986923217773 }, { "epoch": 1.2037914691943128, "step": 508, "train/alpha": 0.2, "train/clean_ce_loss": 0.1455078125, "train/clean_loss_scale": 11.036139488220215, "train/clean_loss_weight": 2.207227945327759, "train/combined_loss": 1.9270133972167969, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.12037914691943127, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.605844497680664 }, { "epoch": 1.2037914691943128, "step": 508, "train/alpha": 0.2, "train/clean_ce_loss": 0.056640625, "train/clean_loss_scale": 9.896341323852539, "train/clean_loss_weight": 1.9792683124542236, "train/combined_loss": 0.672641932964325, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.12037914691943127, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5605349540710449 }, { "epoch": 1.2037914691943128, "step": 508, "train/alpha": 0.2, "train/clean_ce_loss": 0.05029296875, "train/clean_loss_scale": 7.9234619140625, "train/clean_loss_weight": 1.584692358970642, "train/combined_loss": 0.4781933128833771, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.12037914691943127, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3984944224357605 }, { "epoch": 1.2037914691943128, "step": 508, "train/alpha": 0.2, "train/clean_ce_loss": 0.058349609375, "train/clean_loss_scale": 4.673233985900879, "train/clean_loss_weight": 0.9346467852592468, "train/combined_loss": 0.3272176682949066, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.12037914691943127, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2726813852787018 }, { "epoch": 1.2061611374407584, "step": 509, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 7.007568359375, "train/clean_loss_weight": 1.401513695716858, "train/combined_loss": 0.6733835339546204, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.12061611374407583, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5611529350280762 }, { "epoch": 1.2061611374407584, "step": 509, "train/alpha": 0.2, "train/clean_ce_loss": 0.02001953125, "train/clean_loss_scale": 17.704330444335938, "train/clean_loss_weight": 3.5408661365509033, "train/combined_loss": 0.4253188967704773, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.12061611374407583, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3544324040412903 }, { "epoch": 1.2061611374407584, "step": 509, "train/alpha": 0.2, "train/clean_ce_loss": 0.12109375, "train/clean_loss_scale": 7.515942096710205, "train/clean_loss_weight": 1.5031884908676147, "train/combined_loss": 1.0921603441238403, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.12061611374407583, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9101336002349854 }, { "epoch": 1.2061611374407584, "step": 509, "train/alpha": 0.2, "train/clean_ce_loss": 0.01031494140625, "train/clean_loss_scale": 27.16554832458496, "train/clean_loss_weight": 5.433109760284424, "train/combined_loss": 0.33625322580337524, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.12061611374407583, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.28021103143692017 }, { "epoch": 1.2085308056872037, "grad_norm": 14.063632011413574, "learning_rate": 2e-06, "loss": 3.0868, "step": 510 }, { "epoch": 1.2085308056872037, "step": 510, "train/alpha": 0.2, "train/clean_ce_loss": 0.1337890625, "train/clean_loss_scale": 10.139748573303223, "train/clean_loss_weight": 2.027949810028076, "train/combined_loss": 1.6279048919677734, "train/pos_loss/pos0": 0.6114624738693237, "train/pos_loss/pos1": 1.139844536781311, "train/pos_loss/pos2": 1.6902774572372437, "train/pos_loss/pos3": 2.209774971008301, "train/pos_loss/pos4": 2.567800760269165, "train/pos_loss/pos5": 3.270930528640747, "train/pos_loss/pos6": 3.4526500701904297, "train/pos_loss/weight0": 2.4461171627044678, "train/pos_loss/weight1": 1.8226864337921143, "train/pos_loss/weight2": 1.2167384624481201, "train/pos_loss/weight3": 0.7665144801139832, "train/pos_loss/weight4": 0.4183230698108673, "train/pos_loss/weight5": 0.21405978500843048, "train/pos_loss/weight6": 0.11556022614240646, "train/pos_weight_progress": 0.12085308056872038, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3565874099731445 }, { "epoch": 1.2085308056872037, "step": 510, "train/alpha": 0.2, "train/clean_ce_loss": 0.00897216796875, "train/clean_loss_scale": 21.029422760009766, "train/clean_loss_weight": 4.2058844566345215, "train/combined_loss": 0.22641542553901672, "train/pos_loss/pos0": 0.027218306437134743, "train/pos_loss/pos1": 0.05140754580497742, "train/pos_loss/pos2": 0.28532251715660095, "train/pos_loss/pos3": 0.8964340686798096, "train/pos_loss/pos4": 1.5880364179611206, "train/pos_loss/pos5": 2.0804412364959717, "train/pos_loss/pos6": 2.584610939025879, "train/pos_loss/weight0": 3.202321767807007, "train/pos_loss/weight1": 2.020726203918457, "train/pos_loss/weight2": 1.0837435722351074, "train/pos_loss/weight3": 0.46292370557785034, "train/pos_loss/weight4": 0.16601310670375824, "train/pos_loss/weight5": 0.05007156357169151, "train/pos_loss/weight6": 0.014200061559677124, "train/pos_weight_progress": 0.12085308056872038, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1886795163154602 }, { "epoch": 1.2085308056872037, "step": 510, "train/alpha": 0.2, "train/clean_ce_loss": 0.039794921875, "train/clean_loss_scale": 15.065804481506348, "train/clean_loss_weight": 3.0131609439849854, "train/combined_loss": 0.7194510102272034, "train/pos_loss/pos0": 0.10499107092618942, "train/pos_loss/pos1": 0.2578413188457489, "train/pos_loss/pos2": 0.5808618664741516, "train/pos_loss/pos3": 1.0823392868041992, "train/pos_loss/pos4": 1.7507655620574951, "train/pos_loss/pos5": 2.097668409347534, "train/pos_loss/pos6": 2.5759849548339844, "train/pos_loss/weight0": 2.2611751556396484, "train/pos_loss/weight1": 1.691403865814209, "train/pos_loss/weight2": 1.239292860031128, "train/pos_loss/weight3": 0.8761735558509827, "train/pos_loss/weight4": 0.5068232417106628, "train/pos_loss/weight5": 0.26623839139938354, "train/pos_loss/weight6": 0.15889324247837067, "train/pos_weight_progress": 0.12085308056872038, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.599542498588562 }, { "epoch": 1.2085308056872037, "step": 510, "train/alpha": 0.2, "train/clean_ce_loss": 0.0201416015625, "train/clean_loss_scale": 17.5562801361084, "train/clean_loss_weight": 3.511255979537964, "train/combined_loss": 0.42433393001556396, "train/pos_loss/pos0": 0.02523612417280674, "train/pos_loss/pos1": 0.08598775416612625, "train/pos_loss/pos2": 0.3836786150932312, "train/pos_loss/pos3": 0.7056649327278137, "train/pos_loss/pos4": 1.0555347204208374, "train/pos_loss/pos5": 1.7460497617721558, "train/pos_loss/pos6": 2.4178600311279297, "train/pos_loss/weight0": 2.3496851921081543, "train/pos_loss/weight1": 1.7579985857009888, "train/pos_loss/weight2": 1.2612617015838623, "train/pos_loss/weight3": 0.801584005355835, "train/pos_loss/weight4": 0.4659319519996643, "train/pos_loss/weight5": 0.22989830374717712, "train/pos_loss/weight6": 0.1336403638124466, "train/pos_weight_progress": 0.12085308056872038, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.35361161828041077 }, { "epoch": 1.2109004739336493, "step": 511, "train/alpha": 0.2, "train/clean_ce_loss": 0.1376953125, "train/clean_loss_scale": 8.996108055114746, "train/clean_loss_weight": 1.7992216348648071, "train/combined_loss": 1.486466407775879, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12109004739336493, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2387219667434692 }, { "epoch": 1.2109004739336493, "step": 511, "train/alpha": 0.2, "train/clean_ce_loss": 0.056884765625, "train/clean_loss_scale": 9.753302574157715, "train/clean_loss_weight": 1.9506605863571167, "train/combined_loss": 0.6657772064208984, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12109004739336493, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.554814338684082 }, { "epoch": 1.2109004739336493, "step": 511, "train/alpha": 0.2, "train/clean_ce_loss": 0.193359375, "train/clean_loss_scale": 8.104676246643066, "train/clean_loss_weight": 1.620935320854187, "train/combined_loss": 1.8805381059646606, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12109004739336493, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.567115068435669 }, { "epoch": 1.2109004739336493, "step": 511, "train/alpha": 0.2, "train/clean_ce_loss": 0.1416015625, "train/clean_loss_scale": 9.335245132446289, "train/clean_loss_weight": 1.8670490980148315, "train/combined_loss": 1.5862624645233154, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12109004739336493, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.321885347366333 }, { "epoch": 1.2132701421800949, "step": 512, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 9.764229774475098, "train/clean_loss_weight": 1.9528459310531616, "train/combined_loss": 0.8696267604827881, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12132701421800948, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7246889472007751 }, { "epoch": 1.2132701421800949, "step": 512, "train/alpha": 0.2, "train/clean_ce_loss": 0.0247802734375, "train/clean_loss_scale": 21.95374870300293, "train/clean_loss_weight": 4.390749931335449, "train/combined_loss": 0.6528238654136658, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12132701421800948, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.544019877910614 }, { "epoch": 1.2132701421800949, "step": 512, "train/alpha": 0.2, "train/clean_ce_loss": 0.076171875, "train/clean_loss_scale": 13.227910041809082, "train/clean_loss_weight": 2.6455819606781006, "train/combined_loss": 1.209113597869873, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12132701421800948, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0075947046279907 }, { "epoch": 1.2132701421800949, "step": 512, "train/alpha": 0.2, "train/clean_ce_loss": 0.016357421875, "train/clean_loss_scale": 25.01218605041504, "train/clean_loss_weight": 5.002437114715576, "train/combined_loss": 0.49096184968948364, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12132701421800948, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4091348648071289 }, { "epoch": 1.2156398104265402, "step": 513, "train/alpha": 0.2, "train/clean_ce_loss": 0.0869140625, "train/clean_loss_scale": 10.228857040405273, "train/clean_loss_weight": 2.045771360397339, "train/combined_loss": 1.0668377876281738, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12156398104265403, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8890315294265747 }, { "epoch": 1.2156398104265402, "step": 513, "train/alpha": 0.2, "train/clean_ce_loss": 0.057861328125, "train/clean_loss_scale": 17.805133819580078, "train/clean_loss_weight": 3.5610268115997314, "train/combined_loss": 1.2362744808197021, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12156398104265403, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0302287340164185 }, { "epoch": 1.2156398104265402, "step": 513, "train/alpha": 0.2, "train/clean_ce_loss": 0.044921875, "train/clean_loss_scale": 22.400188446044922, "train/clean_loss_weight": 4.480037689208984, "train/combined_loss": 1.207510232925415, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12156398104265403, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.006258487701416 }, { "epoch": 1.2156398104265402, "step": 513, "train/alpha": 0.2, "train/clean_ce_loss": 0.059326171875, "train/clean_loss_scale": 11.709505081176758, "train/clean_loss_weight": 2.3419010639190674, "train/combined_loss": 0.8336161375045776, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12156398104265403, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6946800947189331 }, { "epoch": 1.2180094786729858, "step": 514, "train/alpha": 0.2, "train/clean_ce_loss": 0.126953125, "train/clean_loss_scale": 10.759532928466797, "train/clean_loss_weight": 2.151906728744507, "train/combined_loss": 1.6391475200653076, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12180094786729857, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3659563064575195 }, { "epoch": 1.2180094786729858, "step": 514, "train/alpha": 0.2, "train/clean_ce_loss": 0.052490234375, "train/clean_loss_scale": 10.806931495666504, "train/clean_loss_weight": 2.161386251449585, "train/combined_loss": 0.6807100176811218, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12180094786729857, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.567258358001709 }, { "epoch": 1.2180094786729858, "step": 514, "train/alpha": 0.2, "train/clean_ce_loss": 0.0361328125, "train/clean_loss_scale": 14.371308326721191, "train/clean_loss_weight": 2.8742616176605225, "train/combined_loss": 0.6231309175491333, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12180094786729857, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5192757844924927 }, { "epoch": 1.2180094786729858, "step": 514, "train/alpha": 0.2, "train/clean_ce_loss": 0.06396484375, "train/clean_loss_scale": 10.492056846618652, "train/clean_loss_weight": 2.0984113216400146, "train/combined_loss": 0.8053473234176636, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12180094786729857, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6711227893829346 }, { "epoch": 1.2203791469194312, "grad_norm": 35.811859130859375, "learning_rate": 2e-06, "loss": 3.6916, "step": 515 }, { "epoch": 1.2203791469194312, "step": 515, "train/alpha": 0.2, "train/clean_ce_loss": 7.009506225585938e-05, "train/clean_loss_scale": 6099.94873046875, "train/clean_loss_weight": 1219.98974609375, "train/combined_loss": 0.5130915641784668, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12203791469194313, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4275762736797333 }, { "epoch": 1.2203791469194312, "step": 515, "train/alpha": 0.2, "train/clean_ce_loss": 0.041748046875, "train/clean_loss_scale": 18.122011184692383, "train/clean_loss_weight": 3.6244022846221924, "train/combined_loss": 0.9078702330589294, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12203791469194313, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7565585374832153 }, { "epoch": 1.2203791469194312, "step": 515, "train/alpha": 0.2, "train/clean_ce_loss": 0.031005859375, "train/clean_loss_scale": 12.785895347595215, "train/clean_loss_weight": 2.5571792125701904, "train/combined_loss": 0.4757252037525177, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12203791469194313, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3964376747608185 }, { "epoch": 1.2203791469194312, "step": 515, "train/alpha": 0.2, "train/clean_ce_loss": 0.026611328125, "train/clean_loss_scale": 22.381486892700195, "train/clean_loss_weight": 4.476297378540039, "train/combined_loss": 0.7147213220596313, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12203791469194313, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5956010818481445 }, { "epoch": 1.2227488151658767, "step": 516, "train/alpha": 0.2, "train/clean_ce_loss": 0.007720947265625, "train/clean_loss_scale": 59.49170684814453, "train/clean_loss_weight": 11.898341178894043, "train/combined_loss": 0.5511987805366516, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12227488151658768, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45933231711387634 }, { "epoch": 1.2227488151658767, "step": 516, "train/alpha": 0.2, "train/clean_ce_loss": 0.03466796875, "train/clean_loss_scale": 22.167016983032227, "train/clean_loss_weight": 4.433403491973877, "train/combined_loss": 0.9221825003623962, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12227488151658768, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.768485426902771 }, { "epoch": 1.2227488151658767, "step": 516, "train/alpha": 0.2, "train/clean_ce_loss": 0.1337890625, "train/clean_loss_scale": 11.211977005004883, "train/clean_loss_weight": 2.2423954010009766, "train/combined_loss": 1.8000478744506836, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12227488151658768, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5000399351119995 }, { "epoch": 1.2227488151658767, "step": 516, "train/alpha": 0.2, "train/clean_ce_loss": 0.045166015625, "train/clean_loss_scale": 14.991131782531738, "train/clean_loss_weight": 2.9982264041900635, "train/combined_loss": 0.8125076293945312, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12227488151658768, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6770896911621094 }, { "epoch": 1.2251184834123223, "step": 517, "train/alpha": 0.2, "train/clean_ce_loss": 0.0869140625, "train/clean_loss_scale": 8.307744979858398, "train/clean_loss_weight": 1.6615489721298218, "train/combined_loss": 0.8664718866348267, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12251184834123223, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7220599055290222 }, { "epoch": 1.2251184834123223, "step": 517, "train/alpha": 0.2, "train/clean_ce_loss": 0.0546875, "train/clean_loss_scale": 11.317667961120605, "train/clean_loss_weight": 2.263533592224121, "train/combined_loss": 0.7427219748497009, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12251184834123223, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6189349889755249 }, { "epoch": 1.2251184834123223, "step": 517, "train/alpha": 0.2, "train/clean_ce_loss": 0.020263671875, "train/clean_loss_scale": 26.419157028198242, "train/clean_loss_weight": 5.283831596374512, "train/combined_loss": 0.6424189805984497, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12251184834123223, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5353491306304932 }, { "epoch": 1.2251184834123223, "step": 517, "train/alpha": 0.2, "train/clean_ce_loss": 0.1337890625, "train/clean_loss_scale": 8.98942756652832, "train/clean_loss_weight": 1.797885537147522, "train/combined_loss": 1.4432244300842285, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12251184834123223, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2026870250701904 }, { "epoch": 1.2274881516587677, "step": 518, "train/alpha": 0.2, "train/clean_ce_loss": 0.16796875, "train/clean_loss_scale": 7.795777797698975, "train/clean_loss_weight": 1.5591555833816528, "train/combined_loss": 1.5713365077972412, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12274881516587678, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3094470500946045 }, { "epoch": 1.2274881516587677, "step": 518, "train/alpha": 0.2, "train/clean_ce_loss": 0.051025390625, "train/clean_loss_scale": 9.45134162902832, "train/clean_loss_weight": 1.890268325805664, "train/combined_loss": 0.5787100791931152, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12274881516587678, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4822583794593811 }, { "epoch": 1.2274881516587677, "step": 518, "train/alpha": 0.2, "train/clean_ce_loss": 0.099609375, "train/clean_loss_scale": 14.551735877990723, "train/clean_loss_weight": 2.9103472232818604, "train/combined_loss": 1.7393872737884521, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12274881516587678, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4494893550872803 }, { "epoch": 1.2274881516587677, "step": 518, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 12.521236419677734, "train/clean_loss_weight": 2.5042474269866943, "train/combined_loss": 0.9757760763168335, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12274881516587678, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.813146710395813 }, { "epoch": 1.2298578199052133, "step": 519, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 9.797528266906738, "train/clean_loss_weight": 1.9595056772232056, "train/combined_loss": 0.8725923299789429, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12298578199052132, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7271602749824524 }, { "epoch": 1.2298578199052133, "step": 519, "train/alpha": 0.2, "train/clean_ce_loss": 0.03173828125, "train/clean_loss_scale": 23.995832443237305, "train/clean_loss_weight": 4.799166679382324, "train/combined_loss": 0.9139037728309631, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12298578199052132, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7615864872932434 }, { "epoch": 1.2298578199052133, "step": 519, "train/alpha": 0.2, "train/clean_ce_loss": 0.11181640625, "train/clean_loss_scale": 8.453597068786621, "train/clean_loss_weight": 1.690719485282898, "train/combined_loss": 1.134300947189331, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12298578199052132, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9452508091926575 }, { "epoch": 1.2298578199052133, "step": 519, "train/alpha": 0.2, "train/clean_ce_loss": 0.099609375, "train/clean_loss_scale": 11.50261116027832, "train/clean_loss_weight": 2.3005223274230957, "train/combined_loss": 1.3749215602874756, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12298578199052132, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1457679271697998 }, { "epoch": 1.2322274881516588, "grad_norm": 17.028148651123047, "learning_rate": 2e-06, "loss": 4.1833, "step": 520 }, { "epoch": 1.2322274881516588, "step": 520, "train/alpha": 0.2, "train/clean_ce_loss": 0.072265625, "train/clean_loss_scale": 10.009368896484375, "train/clean_loss_weight": 2.001873731613159, "train/combined_loss": 0.8679999709129333, "train/pos_loss/pos0": 0.12266893684864044, "train/pos_loss/pos1": 0.3267419636249542, "train/pos_loss/pos2": 0.6629026532173157, "train/pos_loss/pos3": 1.1959353685379028, "train/pos_loss/pos4": 1.7884118556976318, "train/pos_loss/pos5": 2.324979066848755, "train/pos_loss/pos6": 2.8038177490234375, "train/pos_loss/weight0": 2.1226301193237305, "train/pos_loss/weight1": 1.633786678314209, "train/pos_loss/weight2": 1.2586157321929932, "train/pos_loss/weight3": 0.8745549321174622, "train/pos_loss/weight4": 0.5510200262069702, "train/pos_loss/weight5": 0.3413526117801666, "train/pos_loss/weight6": 0.21804016828536987, "train/pos_weight_progress": 0.12322274881516587, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7233332991600037 }, { "epoch": 1.2322274881516588, "step": 520, "train/alpha": 0.2, "train/clean_ce_loss": 0.11328125, "train/clean_loss_scale": 8.330146789550781, "train/clean_loss_weight": 1.6660293340682983, "train/combined_loss": 1.1323792934417725, "train/pos_loss/pos0": 0.2720836102962494, "train/pos_loss/pos1": 0.6940508484840393, "train/pos_loss/pos2": 1.200711727142334, "train/pos_loss/pos3": 1.872078537940979, "train/pos_loss/pos4": 2.303492546081543, "train/pos_loss/pos5": 2.919931650161743, "train/pos_loss/pos6": 3.1963796615600586, "train/pos_loss/weight0": 2.481445789337158, "train/pos_loss/weight1": 1.8437139987945557, "train/pos_loss/weight2": 1.2366162538528442, "train/pos_loss/weight3": 0.7710500359535217, "train/pos_loss/weight4": 0.4017442762851715, "train/pos_loss/weight5": 0.17772094905376434, "train/pos_loss/weight6": 0.08770866692066193, "train/pos_weight_progress": 0.12322274881516587, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.943649411201477 }, { "epoch": 1.2322274881516588, "step": 520, "train/alpha": 0.2, "train/clean_ce_loss": 0.06103515625, "train/clean_loss_scale": 9.31093978881836, "train/clean_loss_weight": 1.8621879816055298, "train/combined_loss": 0.6819535493850708, "train/pos_loss/pos0": 0.1269121915102005, "train/pos_loss/pos1": 0.573635995388031, "train/pos_loss/pos2": 0.9161789417266846, "train/pos_loss/pos3": 1.477749228477478, "train/pos_loss/pos4": 1.9008790254592896, "train/pos_loss/pos5": 2.6181490421295166, "train/pos_loss/pos6": 3.3402528762817383, "train/pos_loss/weight0": 3.0027530193328857, "train/pos_loss/weight1": 2.006126880645752, "train/pos_loss/weight2": 1.1749372482299805, "train/pos_loss/weight3": 0.5471835136413574, "train/pos_loss/weight4": 0.20003212988376617, "train/pos_loss/weight5": 0.05345599725842476, "train/pos_loss/weight6": 0.01551176980137825, "train/pos_weight_progress": 0.12322274881516587, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5682946443557739 }, { "epoch": 1.2322274881516588, "step": 520, "train/alpha": 0.2, "train/clean_ce_loss": 0.09130859375, "train/clean_loss_scale": 7.596041202545166, "train/clean_loss_weight": 1.519208312034607, "train/combined_loss": 0.83230060338974, "train/pos_loss/pos0": 0.14943955838680267, "train/pos_loss/pos1": 0.4439508318901062, "train/pos_loss/pos2": 0.9712926745414734, "train/pos_loss/pos3": 1.682645559310913, "train/pos_loss/pos4": 2.407615900039673, "train/pos_loss/pos5": 2.6942410469055176, "train/pos_loss/pos6": 3.048253059387207, "train/pos_loss/weight0": 2.6277148723602295, "train/pos_loss/weight1": 1.9569381475448608, "train/pos_loss/weight2": 1.2243452072143555, "train/pos_loss/weight3": 0.7106835246086121, "train/pos_loss/weight4": 0.31138527393341064, "train/pos_loss/weight5": 0.11738882213830948, "train/pos_loss/weight6": 0.05154392495751381, "train/pos_weight_progress": 0.12322274881516587, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6935838460922241 }, { "epoch": 1.2345971563981042, "step": 521, "train/alpha": 0.2, "train/clean_ce_loss": 0.059814453125, "train/clean_loss_scale": 9.10618782043457, "train/clean_loss_weight": 1.821237564086914, "train/combined_loss": 0.6536179780960083, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12345971563981042, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5446816682815552 }, { "epoch": 1.2345971563981042, "step": 521, "train/alpha": 0.2, "train/clean_ce_loss": 0.1650390625, "train/clean_loss_scale": 7.773405075073242, "train/clean_loss_weight": 1.5546810626983643, "train/combined_loss": 1.5394985675811768, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12345971563981042, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.282915472984314 }, { "epoch": 1.2345971563981042, "step": 521, "train/alpha": 0.2, "train/clean_ce_loss": 0.039306640625, "train/clean_loss_scale": 11.099004745483398, "train/clean_loss_weight": 2.2198009490966797, "train/combined_loss": 0.5235174894332886, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12345971563981042, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43626460433006287 }, { "epoch": 1.2345971563981042, "step": 521, "train/alpha": 0.2, "train/clean_ce_loss": 0.06298828125, "train/clean_loss_scale": 6.478256702423096, "train/clean_loss_weight": 1.2956513166427612, "train/combined_loss": 0.48966512084007263, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12345971563981042, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.40805426239967346 }, { "epoch": 1.2369668246445498, "step": 522, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 6.945645809173584, "train/clean_loss_weight": 1.3891291618347168, "train/combined_loss": 0.6674331426620483, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12369668246445498, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5561943054199219 }, { "epoch": 1.2369668246445498, "step": 522, "train/alpha": 0.2, "train/clean_ce_loss": 0.0634765625, "train/clean_loss_scale": 8.34183406829834, "train/clean_loss_weight": 1.66836678981781, "train/combined_loss": 0.6354131698608398, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12369668246445498, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5295109748840332 }, { "epoch": 1.2369668246445498, "step": 522, "train/alpha": 0.2, "train/clean_ce_loss": 0.03857421875, "train/clean_loss_scale": 10.659929275512695, "train/clean_loss_weight": 2.131985902786255, "train/combined_loss": 0.49343812465667725, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12369668246445498, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4111984372138977 }, { "epoch": 1.2369668246445498, "step": 522, "train/alpha": 0.2, "train/clean_ce_loss": 0.140625, "train/clean_loss_scale": 6.746198654174805, "train/clean_loss_weight": 1.349239706993103, "train/combined_loss": 1.1384209394454956, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12369668246445498, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9486841559410095 }, { "epoch": 1.2393364928909953, "step": 523, "train/alpha": 0.2, "train/clean_ce_loss": 0.06201171875, "train/clean_loss_scale": 8.725546836853027, "train/clean_loss_weight": 1.7451094388961792, "train/combined_loss": 0.6493033766746521, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12393364928909953, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5410861372947693 }, { "epoch": 1.2393364928909953, "step": 523, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 8.906929016113281, "train/clean_loss_weight": 1.7813857793807983, "train/combined_loss": 0.7097708582878113, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12393364928909953, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5914757251739502 }, { "epoch": 1.2393364928909953, "step": 523, "train/alpha": 0.2, "train/clean_ce_loss": 0.03662109375, "train/clean_loss_scale": 11.692733764648438, "train/clean_loss_weight": 2.3385467529296875, "train/combined_loss": 0.5138408541679382, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12393364928909953, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.42820069193840027 }, { "epoch": 1.2393364928909953, "step": 523, "train/alpha": 0.2, "train/clean_ce_loss": 0.1279296875, "train/clean_loss_scale": 8.308191299438477, "train/clean_loss_weight": 1.6616382598876953, "train/combined_loss": 1.2754371166229248, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12393364928909953, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0628643035888672 }, { "epoch": 1.2417061611374407, "step": 524, "train/alpha": 0.2, "train/clean_ce_loss": 0.1943359375, "train/clean_loss_scale": 6.222206115722656, "train/clean_loss_weight": 1.244441270828247, "train/combined_loss": 1.451037883758545, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12417061611374408, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.209198236465454 }, { "epoch": 1.2417061611374407, "step": 524, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 13.046313285827637, "train/clean_loss_weight": 2.609262704849243, "train/combined_loss": 1.0243394374847412, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12417061611374408, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8536161780357361 }, { "epoch": 1.2417061611374407, "step": 524, "train/alpha": 0.2, "train/clean_ce_loss": 0.0517578125, "train/clean_loss_scale": 14.351712226867676, "train/clean_loss_weight": 2.870342493057251, "train/combined_loss": 0.8913758993148804, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12417061611374408, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.742813229560852 }, { "epoch": 1.2417061611374407, "step": 524, "train/alpha": 0.2, "train/clean_ce_loss": 0.06396484375, "train/clean_loss_scale": 7.139630317687988, "train/clean_loss_weight": 1.4279260635375977, "train/combined_loss": 0.5480223894119263, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12417061611374408, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.456685334444046 }, { "epoch": 1.2440758293838863, "grad_norm": 13.656371116638184, "learning_rate": 2e-06, "loss": 3.6571, "step": 525 }, { "epoch": 1.2440758293838863, "step": 525, "train/alpha": 0.2, "train/clean_ce_loss": 0.0703125, "train/clean_loss_scale": 13.411630630493164, "train/clean_loss_weight": 2.682326078414917, "train/combined_loss": 1.1316063404083252, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12440758293838862, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9430052638053894 }, { "epoch": 1.2440758293838863, "step": 525, "train/alpha": 0.2, "train/clean_ce_loss": 0.11572265625, "train/clean_loss_scale": 8.437020301818848, "train/clean_loss_weight": 1.6874040365219116, "train/combined_loss": 1.1716252565383911, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12440758293838862, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9763543605804443 }, { "epoch": 1.2440758293838863, "step": 525, "train/alpha": 0.2, "train/clean_ce_loss": 0.05859375, "train/clean_loss_scale": 9.332310676574707, "train/clean_loss_weight": 1.8664621114730835, "train/combined_loss": 0.6561781167984009, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12440758293838862, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5468150973320007 }, { "epoch": 1.2440758293838863, "step": 525, "train/alpha": 0.2, "train/clean_ce_loss": 0.01287841796875, "train/clean_loss_scale": 18.124814987182617, "train/clean_loss_weight": 3.6249630451202393, "train/combined_loss": 0.2801027297973633, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12440758293838862, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.23341894149780273 }, { "epoch": 1.2464454976303316, "step": 526, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 8.788941383361816, "train/clean_loss_weight": 1.7577883005142212, "train/combined_loss": 0.7518664598464966, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12464454976303317, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6265553832054138 }, { "epoch": 1.2464454976303316, "step": 526, "train/alpha": 0.2, "train/clean_ce_loss": 0.10009765625, "train/clean_loss_scale": 7.889380931854248, "train/clean_loss_weight": 1.5778762102127075, "train/combined_loss": 0.9476502537727356, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12464454976303317, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7897085547447205 }, { "epoch": 1.2464454976303316, "step": 526, "train/alpha": 0.2, "train/clean_ce_loss": 0.02294921875, "train/clean_loss_scale": 16.722335815429688, "train/clean_loss_weight": 3.3444671630859375, "train/combined_loss": 0.46051743626594543, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12464454976303317, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3837645351886749 }, { "epoch": 1.2464454976303316, "step": 526, "train/alpha": 0.2, "train/clean_ce_loss": 0.072265625, "train/clean_loss_scale": 6.6234002113342285, "train/clean_loss_weight": 1.3246800899505615, "train/combined_loss": 0.5743730068206787, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12464454976303317, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.47864416241645813 }, { "epoch": 1.2488151658767772, "step": 527, "train/alpha": 0.2, "train/clean_ce_loss": 0.0223388671875, "train/clean_loss_scale": 9.220844268798828, "train/clean_loss_weight": 1.8441689014434814, "train/combined_loss": 0.24717986583709717, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12488151658767772, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2059832215309143 }, { "epoch": 1.2488151658767772, "step": 527, "train/alpha": 0.2, "train/clean_ce_loss": 0.0498046875, "train/clean_loss_scale": 10.370431900024414, "train/clean_loss_weight": 2.0740864276885986, "train/combined_loss": 0.6197953224182129, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12488151658767772, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5164961218833923 }, { "epoch": 1.2488151658767772, "step": 527, "train/alpha": 0.2, "train/clean_ce_loss": 0.0299072265625, "train/clean_loss_scale": 8.008369445800781, "train/clean_loss_weight": 1.60167396068573, "train/combined_loss": 0.2874097228050232, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12488151658767772, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.23950810730457306 }, { "epoch": 1.2488151658767772, "step": 527, "train/alpha": 0.2, "train/clean_ce_loss": 0.047607421875, "train/clean_loss_scale": 8.05860710144043, "train/clean_loss_weight": 1.611721396446228, "train/combined_loss": 0.46037939190864563, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12488151658767772, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3836494982242584 }, { "epoch": 1.2511848341232228, "step": 528, "train/alpha": 0.2, "train/clean_ce_loss": 0.0771484375, "train/clean_loss_scale": 13.887774467468262, "train/clean_loss_weight": 2.777554988861084, "train/combined_loss": 1.2857041358947754, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12511848341232226, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0714200735092163 }, { "epoch": 1.2511848341232228, "step": 528, "train/alpha": 0.2, "train/clean_ce_loss": 0.173828125, "train/clean_loss_scale": 7.64864444732666, "train/clean_loss_weight": 1.529728889465332, "train/combined_loss": 1.5954594612121582, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12511848341232226, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3295495510101318 }, { "epoch": 1.2511848341232228, "step": 528, "train/alpha": 0.2, "train/clean_ce_loss": 0.05712890625, "train/clean_loss_scale": 8.663270950317383, "train/clean_loss_weight": 1.7326542139053345, "train/combined_loss": 0.5939078330993652, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12511848341232226, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4949232041835785 }, { "epoch": 1.2511848341232228, "step": 528, "train/alpha": 0.2, "train/clean_ce_loss": 0.0072021484375, "train/clean_loss_scale": 35.22002029418945, "train/clean_loss_weight": 7.044003963470459, "train/combined_loss": 0.3043917715549469, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12511848341232226, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.25365981459617615 }, { "epoch": 1.2535545023696684, "step": 529, "train/alpha": 0.2, "train/clean_ce_loss": 0.09130859375, "train/clean_loss_scale": 8.689091682434082, "train/clean_loss_weight": 1.7378183603286743, "train/combined_loss": 0.9520665407180786, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12535545023696681, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7933887839317322 }, { "epoch": 1.2535545023696684, "step": 529, "train/alpha": 0.2, "train/clean_ce_loss": 0.1611328125, "train/clean_loss_scale": 7.333209037780762, "train/clean_loss_weight": 1.4666417837142944, "train/combined_loss": 1.4179446697235107, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12535545023696681, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1816205978393555 }, { "epoch": 1.2535545023696684, "step": 529, "train/alpha": 0.2, "train/clean_ce_loss": 0.091796875, "train/clean_loss_scale": 7.119629383087158, "train/clean_loss_weight": 1.4239258766174316, "train/combined_loss": 0.7842717170715332, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12535545023696681, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6535597443580627 }, { "epoch": 1.2535545023696684, "step": 529, "train/alpha": 0.2, "train/clean_ce_loss": 0.12255859375, "train/clean_loss_scale": 7.4917311668396, "train/clean_loss_weight": 1.498346209526062, "train/combined_loss": 1.1018112897872925, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12535545023696681, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9181760549545288 }, { "epoch": 1.2559241706161137, "grad_norm": 12.073049545288086, "learning_rate": 2e-06, "loss": 3.6246, "step": 530 }, { "epoch": 1.2559241706161137, "step": 530, "train/alpha": 0.2, "train/clean_ce_loss": 0.0703125, "train/clean_loss_scale": 8.885886192321777, "train/clean_loss_weight": 1.7771772146224976, "train/combined_loss": 0.7497466802597046, "train/pos_loss/pos0": 0.27062270045280457, "train/pos_loss/pos1": 0.5585565567016602, "train/pos_loss/pos2": 0.8935233950614929, "train/pos_loss/pos3": 1.1241660118103027, "train/pos_loss/pos4": 1.5453107357025146, "train/pos_loss/pos5": 2.0167577266693115, "train/pos_loss/pos6": 2.258028030395508, "train/pos_loss/weight0": 2.7636923789978027, "train/pos_loss/weight1": 1.9279811382293701, "train/pos_loss/weight2": 1.2223658561706543, "train/pos_loss/weight3": 0.6762331128120422, "train/pos_loss/weight4": 0.2889823615550995, "train/pos_loss/weight5": 0.08702912926673889, "train/pos_loss/weight6": 0.033715471625328064, "train/pos_weight_progress": 0.12559241706161137, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6247888803482056 }, { "epoch": 1.2559241706161137, "step": 530, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 8.503166198730469, "train/clean_loss_weight": 1.7006332874298096, "train/combined_loss": 0.7423663139343262, "train/pos_loss/pos0": 0.2237457036972046, "train/pos_loss/pos1": 0.430087149143219, "train/pos_loss/pos2": 0.7052699327468872, "train/pos_loss/pos3": 1.1084010601043701, "train/pos_loss/pos4": 1.487862467765808, "train/pos_loss/pos5": 1.8331892490386963, "train/pos_loss/pos6": 2.4017388820648193, "train/pos_loss/weight0": 2.464817762374878, "train/pos_loss/weight1": 1.8063480854034424, "train/pos_loss/weight2": 1.2190629243850708, "train/pos_loss/weight3": 0.7505540251731873, "train/pos_loss/weight4": 0.42340224981307983, "train/pos_loss/weight5": 0.2186134308576584, "train/pos_loss/weight6": 0.11720096319913864, "train/pos_weight_progress": 0.12559241706161137, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6186385750770569 }, { "epoch": 1.2559241706161137, "step": 530, "train/alpha": 0.2, "train/clean_ce_loss": 0.1279296875, "train/clean_loss_scale": 8.713425636291504, "train/clean_loss_weight": 1.7426851987838745, "train/combined_loss": 1.3376469612121582, "train/pos_loss/pos0": 0.522824764251709, "train/pos_loss/pos1": 0.9223048686981201, "train/pos_loss/pos2": 1.1145747900009155, "train/pos_loss/pos3": 1.696199893951416, "train/pos_loss/pos4": 2.4365406036376953, "train/pos_loss/pos5": 3.1460397243499756, "train/pos_loss/pos6": 3.539217472076416, "train/pos_loss/weight0": 2.4292407035827637, "train/pos_loss/weight1": 1.7821669578552246, "train/pos_loss/weight2": 1.2301080226898193, "train/pos_loss/weight3": 0.7760773301124573, "train/pos_loss/weight4": 0.4305832087993622, "train/pos_loss/weight5": 0.2266400009393692, "train/pos_loss/weight6": 0.12518304586410522, "train/pos_weight_progress": 0.12559241706161137, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1147058010101318 }, { "epoch": 1.2559241706161137, "step": 530, "train/alpha": 0.2, "train/clean_ce_loss": 0.032958984375, "train/clean_loss_scale": 4.847559928894043, "train/clean_loss_weight": 0.9695119857788086, "train/combined_loss": 0.1917247772216797, "train/pos_loss/pos0": 0.03823556378483772, "train/pos_loss/pos1": 0.03645140677690506, "train/pos_loss/pos2": 0.13655412197113037, "train/pos_loss/pos3": 0.5755889415740967, "train/pos_loss/pos4": 1.3776452541351318, "train/pos_loss/pos5": 2.110032796859741, "train/pos_loss/pos6": 2.115250825881958, "train/pos_loss/weight0": 3.03118634223938, "train/pos_loss/weight1": 1.983664631843567, "train/pos_loss/weight2": 1.1499557495117188, "train/pos_loss/weight3": 0.5429292917251587, "train/pos_loss/weight4": 0.21331676840782166, "train/pos_loss/weight5": 0.06173970550298691, "train/pos_loss/weight6": 0.017207833006978035, "train/pos_weight_progress": 0.12559241706161137, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1597706526517868 }, { "epoch": 1.2582938388625593, "step": 531, "train/alpha": 0.2, "train/clean_ce_loss": 0.06396484375, "train/clean_loss_scale": 11.923201560974121, "train/clean_loss_weight": 2.3846404552459717, "train/combined_loss": 0.9151989221572876, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12582938388625592, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7626657485961914 }, { "epoch": 1.2582938388625593, "step": 531, "train/alpha": 0.2, "train/clean_ce_loss": 0.14453125, "train/clean_loss_scale": 8.77829360961914, "train/clean_loss_weight": 1.755658745765686, "train/combined_loss": 1.522485375404358, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12582938388625592, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.26873779296875 }, { "epoch": 1.2582938388625593, "step": 531, "train/alpha": 0.2, "train/clean_ce_loss": 0.02783203125, "train/clean_loss_scale": 22.314664840698242, "train/clean_loss_weight": 4.46293306350708, "train/combined_loss": 0.7452749609947205, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12582938388625592, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6210624575614929 }, { "epoch": 1.2582938388625593, "step": 531, "train/alpha": 0.2, "train/clean_ce_loss": 0.072265625, "train/clean_loss_scale": 12.674846649169922, "train/clean_loss_weight": 2.5349693298339844, "train/combined_loss": 1.099146842956543, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12582938388625592, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9159557223320007 }, { "epoch": 1.2606635071090047, "step": 532, "train/alpha": 0.2, "train/clean_ce_loss": 0.11767578125, "train/clean_loss_scale": 11.064825057983398, "train/clean_loss_weight": 2.2129650115966797, "train/combined_loss": 1.562474250793457, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12606635071090047, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.302061915397644 }, { "epoch": 1.2606635071090047, "step": 532, "train/alpha": 0.2, "train/clean_ce_loss": 0.1845703125, "train/clean_loss_scale": 9.112667083740234, "train/clean_loss_weight": 1.8225334882736206, "train/combined_loss": 2.018313407897949, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12606635071090047, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.6819278001785278 }, { "epoch": 1.2606635071090047, "step": 532, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 11.595865249633789, "train/clean_loss_weight": 2.3191730976104736, "train/combined_loss": 0.9104565978050232, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12606635071090047, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7587138414382935 }, { "epoch": 1.2606635071090047, "step": 532, "train/alpha": 0.2, "train/clean_ce_loss": 0.021240234375, "train/clean_loss_scale": 21.939510345458984, "train/clean_loss_weight": 4.38790225982666, "train/combined_loss": 0.5592004060745239, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12606635071090047, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46600034832954407 }, { "epoch": 1.2630331753554502, "step": 533, "train/alpha": 0.2, "train/clean_ce_loss": 0.07470703125, "train/clean_loss_scale": 10.70904541015625, "train/clean_loss_weight": 2.1418092250823975, "train/combined_loss": 0.9600492119789124, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12630331753554502, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8000410199165344 }, { "epoch": 1.2630331753554502, "step": 533, "train/alpha": 0.2, "train/clean_ce_loss": 0.015869140625, "train/clean_loss_scale": 23.69326400756836, "train/clean_loss_weight": 4.73865270614624, "train/combined_loss": 0.4511900842189789, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12630331753554502, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.37599173188209534 }, { "epoch": 1.2630331753554502, "step": 533, "train/alpha": 0.2, "train/clean_ce_loss": 0.05322265625, "train/clean_loss_scale": 13.519622802734375, "train/clean_loss_weight": 2.7039246559143066, "train/combined_loss": 0.8634603023529053, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12630331753554502, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7195502519607544 }, { "epoch": 1.2630331753554502, "step": 533, "train/alpha": 0.2, "train/clean_ce_loss": 0.025146484375, "train/clean_loss_scale": 16.179691314697266, "train/clean_loss_weight": 3.235938310623169, "train/combined_loss": 0.48823484778404236, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12630331753554502, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.40686237812042236 }, { "epoch": 1.2654028436018958, "step": 534, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 8.38790512084961, "train/clean_loss_weight": 1.6775810718536377, "train/combined_loss": 0.8404287695884705, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12654028436018958, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7003573179244995 }, { "epoch": 1.2654028436018958, "step": 534, "train/alpha": 0.2, "train/clean_ce_loss": 0.0177001953125, "train/clean_loss_scale": 26.445356369018555, "train/clean_loss_weight": 5.289071559906006, "train/combined_loss": 0.5617055892944336, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12654028436018958, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46808797121047974 }, { "epoch": 1.2654028436018958, "step": 534, "train/alpha": 0.2, "train/clean_ce_loss": 0.06982421875, "train/clean_loss_scale": 11.561385154724121, "train/clean_loss_weight": 2.31227707862854, "train/combined_loss": 0.968717634677887, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12654028436018958, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8072646856307983 }, { "epoch": 1.2654028436018958, "step": 534, "train/alpha": 0.2, "train/clean_ce_loss": 0.099609375, "train/clean_loss_scale": 15.680871963500977, "train/clean_loss_weight": 3.136174440383911, "train/combined_loss": 1.8743542432785034, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12654028436018958, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5619618892669678 }, { "epoch": 1.2677725118483412, "grad_norm": 17.946826934814453, "learning_rate": 2e-06, "loss": 3.449, "step": 535 }, { "epoch": 1.2677725118483412, "step": 535, "train/alpha": 0.2, "train/clean_ce_loss": 0.004913330078125, "train/clean_loss_scale": 106.9136734008789, "train/clean_loss_weight": 21.382734298706055, "train/combined_loss": 0.6303626298904419, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12677725118483413, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5253021717071533 }, { "epoch": 1.2677725118483412, "step": 535, "train/alpha": 0.2, "train/clean_ce_loss": 0.107421875, "train/clean_loss_scale": 6.676314353942871, "train/clean_loss_weight": 1.3352628946304321, "train/combined_loss": 0.8606186509132385, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12677725118483413, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7171822190284729 }, { "epoch": 1.2677725118483412, "step": 535, "train/alpha": 0.2, "train/clean_ce_loss": 0.06396484375, "train/clean_loss_scale": 17.8181209564209, "train/clean_loss_weight": 3.563624143600464, "train/combined_loss": 1.3676799535751343, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12677725118483413, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1397333145141602 }, { "epoch": 1.2677725118483412, "step": 535, "train/alpha": 0.2, "train/clean_ce_loss": 0.107421875, "train/clean_loss_scale": 9.851734161376953, "train/clean_loss_weight": 1.9703468084335327, "train/combined_loss": 1.2699501514434814, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12677725118483413, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0582917928695679 }, { "epoch": 1.2701421800947867, "step": 536, "train/alpha": 0.2, "train/clean_ce_loss": 0.009033203125, "train/clean_loss_scale": 51.18974685668945, "train/clean_loss_weight": 10.23794937133789, "train/combined_loss": 0.5548888444900513, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12701421800947868, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4624073803424835 }, { "epoch": 1.2701421800947867, "step": 536, "train/alpha": 0.2, "train/clean_ce_loss": 0.1171875, "train/clean_loss_scale": 8.253073692321777, "train/clean_loss_weight": 1.6506147384643555, "train/combined_loss": 1.1605885028839111, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12701421800947868, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9671570658683777 }, { "epoch": 1.2701421800947867, "step": 536, "train/alpha": 0.2, "train/clean_ce_loss": 0.09912109375, "train/clean_loss_scale": 7.157636642456055, "train/clean_loss_weight": 1.4315273761749268, "train/combined_loss": 0.8513673543930054, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12701421800947868, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7094727754592896 }, { "epoch": 1.2701421800947867, "step": 536, "train/alpha": 0.2, "train/clean_ce_loss": 0.042724609375, "train/clean_loss_scale": 9.372159004211426, "train/clean_loss_weight": 1.874431848526001, "train/combined_loss": 0.48050618171691895, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12701421800947868, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4004218280315399 }, { "epoch": 1.272511848341232, "step": 537, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 11.582060813903809, "train/clean_loss_weight": 2.3164122104644775, "train/combined_loss": 1.133322834968567, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12725118483412323, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9444356560707092 }, { "epoch": 1.272511848341232, "step": 537, "train/alpha": 0.2, "train/clean_ce_loss": 0.0169677734375, "train/clean_loss_scale": 23.93651008605957, "train/clean_loss_weight": 4.787302017211914, "train/combined_loss": 0.48737913370132446, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12725118483412323, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4061492681503296 }, { "epoch": 1.272511848341232, "step": 537, "train/alpha": 0.2, "train/clean_ce_loss": 0.06298828125, "train/clean_loss_scale": 11.844108581542969, "train/clean_loss_weight": 2.368821859359741, "train/combined_loss": 0.8952480554580688, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12725118483412323, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7460400462150574 }, { "epoch": 1.272511848341232, "step": 537, "train/alpha": 0.2, "train/clean_ce_loss": 0.0859375, "train/clean_loss_scale": 10.476544380187988, "train/clean_loss_weight": 2.095309019088745, "train/combined_loss": 1.080393671989441, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12725118483412323, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9003280401229858 }, { "epoch": 1.2748815165876777, "step": 538, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 10.301497459411621, "train/clean_loss_weight": 2.0602996349334717, "train/combined_loss": 0.8993690013885498, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12748815165876778, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7494741678237915 }, { "epoch": 1.2748815165876777, "step": 538, "train/alpha": 0.2, "train/clean_ce_loss": 0.1005859375, "train/clean_loss_scale": 12.429176330566406, "train/clean_loss_weight": 2.485835313796997, "train/combined_loss": 1.500240445137024, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12748815165876778, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2502003908157349 }, { "epoch": 1.2748815165876777, "step": 538, "train/alpha": 0.2, "train/clean_ce_loss": 0.07763671875, "train/clean_loss_scale": 10.959370613098145, "train/clean_loss_weight": 2.1918742656707764, "train/combined_loss": 1.021019458770752, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12748815165876778, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8508495688438416 }, { "epoch": 1.2748815165876777, "step": 538, "train/alpha": 0.2, "train/clean_ce_loss": 0.0947265625, "train/clean_loss_scale": 9.716866493225098, "train/clean_loss_weight": 1.9433733224868774, "train/combined_loss": 1.104534387588501, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12748815165876778, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9204453229904175 }, { "epoch": 1.2772511848341233, "step": 539, "train/alpha": 0.2, "train/clean_ce_loss": 0.0703125, "train/clean_loss_scale": 10.577669143676758, "train/clean_loss_weight": 2.1155338287353516, "train/combined_loss": 0.892490804195404, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.1277251184834123, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7437423467636108 }, { "epoch": 1.2772511848341233, "step": 539, "train/alpha": 0.2, "train/clean_ce_loss": 0.1015625, "train/clean_loss_scale": 9.120494842529297, "train/clean_loss_weight": 1.8240989446640015, "train/combined_loss": 1.1115602254867554, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.1277251184834123, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9263002276420593 }, { "epoch": 1.2772511848341233, "step": 539, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 10.982854843139648, "train/clean_loss_weight": 2.196571111679077, "train/combined_loss": 1.0682542324066162, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.1277251184834123, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8902118802070618 }, { "epoch": 1.2772511848341233, "step": 539, "train/alpha": 0.2, "train/clean_ce_loss": 0.01263427734375, "train/clean_loss_scale": 39.86916732788086, "train/clean_loss_weight": 7.9738335609436035, "train/combined_loss": 0.6044617891311646, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.1277251184834123, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5037181377410889 }, { "epoch": 1.2796208530805688, "grad_norm": 12.670347213745117, "learning_rate": 2e-06, "loss": 3.3756, "step": 540 }, { "epoch": 1.2796208530805688, "step": 540, "train/alpha": 0.2, "train/clean_ce_loss": 0.057861328125, "train/clean_loss_scale": 12.053025245666504, "train/clean_loss_weight": 2.4106051921844482, "train/combined_loss": 0.8368848562240601, "train/pos_loss/pos0": 0.15669240057468414, "train/pos_loss/pos1": 0.4236481189727783, "train/pos_loss/pos2": 0.7337342500686646, "train/pos_loss/pos3": 1.102932333946228, "train/pos_loss/pos4": 1.4707740545272827, "train/pos_loss/pos5": 1.9357470273971558, "train/pos_loss/pos6": 2.3436827659606934, "train/pos_loss/weight0": 2.14363431930542, "train/pos_loss/weight1": 1.6452405452728271, "train/pos_loss/weight2": 1.2073894739151, "train/pos_loss/weight3": 0.8840251564979553, "train/pos_loss/weight4": 0.5682445168495178, "train/pos_loss/weight5": 0.3412994146347046, "train/pos_loss/weight6": 0.21016690135002136, "train/pos_weight_progress": 0.12796208530805686, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6974040269851685 }, { "epoch": 1.2796208530805688, "step": 540, "train/alpha": 0.2, "train/clean_ce_loss": 0.1328125, "train/clean_loss_scale": 10.437886238098145, "train/clean_loss_weight": 2.0875773429870605, "train/combined_loss": 1.6635380983352661, "train/pos_loss/pos0": 0.4950101673603058, "train/pos_loss/pos1": 1.008190393447876, "train/pos_loss/pos2": 1.8288222551345825, "train/pos_loss/pos3": 2.457810163497925, "train/pos_loss/pos4": 2.9173965454101562, "train/pos_loss/pos5": 3.4052698612213135, "train/pos_loss/pos6": 3.9672951698303223, "train/pos_loss/weight0": 2.3922107219696045, "train/pos_loss/weight1": 1.8142515420913696, "train/pos_loss/weight2": 1.2310709953308105, "train/pos_loss/weight3": 0.7792156934738159, "train/pos_loss/weight4": 0.43208688497543335, "train/pos_loss/weight5": 0.22417011857032776, "train/pos_loss/weight6": 0.1269935667514801, "train/pos_weight_progress": 0.12796208530805686, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3862817287445068 }, { "epoch": 1.2796208530805688, "step": 540, "train/alpha": 0.2, "train/clean_ce_loss": 0.10107421875, "train/clean_loss_scale": 6.14592170715332, "train/clean_loss_weight": 1.2291843891143799, "train/combined_loss": 0.7454330921173096, "train/pos_loss/pos0": 0.31196174025535583, "train/pos_loss/pos1": 0.6072598695755005, "train/pos_loss/pos2": 0.9711626768112183, "train/pos_loss/pos3": 1.4156495332717896, "train/pos_loss/pos4": 1.8483238220214844, "train/pos_loss/pos5": 2.1886894702911377, "train/pos_loss/pos6": 2.584059476852417, "train/pos_loss/weight0": 3.1865203380584717, "train/pos_loss/weight1": 2.0316262245178223, "train/pos_loss/weight2": 1.1238738298416138, "train/pos_loss/weight3": 0.46953651309013367, "train/pos_loss/weight4": 0.14621368050575256, "train/pos_loss/weight5": 0.03332281857728958, "train/pos_loss/weight6": 0.008905831724405289, "train/pos_weight_progress": 0.12796208530805686, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6211942434310913 }, { "epoch": 1.2796208530805688, "step": 540, "train/alpha": 0.2, "train/clean_ce_loss": 0.07373046875, "train/clean_loss_scale": 9.21975326538086, "train/clean_loss_weight": 1.843950629234314, "train/combined_loss": 0.8157320618629456, "train/pos_loss/pos0": 0.32134154438972473, "train/pos_loss/pos1": 0.5486240386962891, "train/pos_loss/pos2": 0.9269509315490723, "train/pos_loss/pos3": 1.2321580648422241, "train/pos_loss/pos4": 1.5848619937896729, "train/pos_loss/pos5": 1.9783034324645996, "train/pos_loss/pos6": 2.38533353805542, "train/pos_loss/weight0": 2.714669704437256, "train/pos_loss/weight1": 1.9106053113937378, "train/pos_loss/weight2": 1.19978666305542, "train/pos_loss/weight3": 0.6660513877868652, "train/pos_loss/weight4": 0.31874385476112366, "train/pos_loss/weight5": 0.13023628294467926, "train/pos_loss/weight6": 0.059906862676143646, "train/pos_weight_progress": 0.12796208530805686, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6797767281532288 }, { "epoch": 1.2819905213270142, "step": 541, "train/alpha": 0.2, "train/clean_ce_loss": 0.07666015625, "train/clean_loss_scale": 9.98270320892334, "train/clean_loss_weight": 1.9965406656265259, "train/combined_loss": 0.918330729007721, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.1281990521327014, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7652755975723267 }, { "epoch": 1.2819905213270142, "step": 541, "train/alpha": 0.2, "train/clean_ce_loss": 0.1064453125, "train/clean_loss_scale": 11.668027877807617, "train/clean_loss_weight": 2.3336055278778076, "train/combined_loss": 1.4904083013534546, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.1281990521327014, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2420068979263306 }, { "epoch": 1.2819905213270142, "step": 541, "train/alpha": 0.2, "train/clean_ce_loss": 0.19140625, "train/clean_loss_scale": 7.055774211883545, "train/clean_loss_weight": 1.411154866218567, "train/combined_loss": 1.6206231117248535, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.1281990521327014, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3505192995071411 }, { "epoch": 1.2819905213270142, "step": 541, "train/alpha": 0.2, "train/clean_ce_loss": 0.130859375, "train/clean_loss_scale": 7.481525897979736, "train/clean_loss_weight": 1.496305227279663, "train/combined_loss": 1.1748334169387817, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.1281990521327014, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9790278077125549 }, { "epoch": 1.2843601895734598, "step": 542, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 8.467947959899902, "train/clean_loss_weight": 1.6935895681381226, "train/combined_loss": 0.8037934303283691, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12843601895734597, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6698278784751892 }, { "epoch": 1.2843601895734598, "step": 542, "train/alpha": 0.2, "train/clean_ce_loss": 0.04296875, "train/clean_loss_scale": 10.008271217346191, "train/clean_loss_weight": 2.0016543865203857, "train/combined_loss": 0.5160514712333679, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12843601895734597, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43004289269447327 }, { "epoch": 1.2843601895734598, "step": 542, "train/alpha": 0.2, "train/clean_ce_loss": 0.07470703125, "train/clean_loss_scale": 10.807764053344727, "train/clean_loss_weight": 2.161552906036377, "train/combined_loss": 0.968899130821228, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12843601895734597, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8074159622192383 }, { "epoch": 1.2843601895734598, "step": 542, "train/alpha": 0.2, "train/clean_ce_loss": 0.1748046875, "train/clean_loss_scale": 8.755456924438477, "train/clean_loss_weight": 1.7510913610458374, "train/combined_loss": 1.8365938663482666, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12843601895734597, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5304949283599854 }, { "epoch": 1.2867298578199051, "step": 543, "train/alpha": 0.2, "train/clean_ce_loss": 0.1328125, "train/clean_loss_scale": 11.424477577209473, "train/clean_loss_weight": 2.284895658493042, "train/combined_loss": 1.8207762241363525, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12867298578199052, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5173134803771973 }, { "epoch": 1.2867298578199051, "step": 543, "train/alpha": 0.2, "train/clean_ce_loss": 0.1259765625, "train/clean_loss_scale": 9.211779594421387, "train/clean_loss_weight": 1.8423559665679932, "train/combined_loss": 1.3925620317459106, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12867298578199052, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.160468339920044 }, { "epoch": 1.2867298578199051, "step": 543, "train/alpha": 0.2, "train/clean_ce_loss": 0.0869140625, "train/clean_loss_scale": 9.022492408752441, "train/clean_loss_weight": 1.804498553276062, "train/combined_loss": 0.941017746925354, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12867298578199052, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7841814756393433 }, { "epoch": 1.2867298578199051, "step": 543, "train/alpha": 0.2, "train/clean_ce_loss": 0.0311279296875, "train/clean_loss_scale": 17.06688690185547, "train/clean_loss_weight": 3.413377523422241, "train/combined_loss": 0.63750821352005, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12867298578199052, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5312568545341492 }, { "epoch": 1.2890995260663507, "step": 544, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 8.784512519836426, "train/clean_loss_weight": 1.7569025754928589, "train/combined_loss": 0.6897214651107788, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12890995260663507, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5747678875923157 }, { "epoch": 1.2890995260663507, "step": 544, "train/alpha": 0.2, "train/clean_ce_loss": 0.0771484375, "train/clean_loss_scale": 9.059093475341797, "train/clean_loss_weight": 1.8118187189102173, "train/combined_loss": 0.8386738896369934, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12890995260663507, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6988949179649353 }, { "epoch": 1.2890995260663507, "step": 544, "train/alpha": 0.2, "train/clean_ce_loss": 0.052978515625, "train/clean_loss_scale": 7.837505340576172, "train/clean_loss_weight": 1.5675010681152344, "train/combined_loss": 0.49826326966285706, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12890995260663507, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.41521939635276794 }, { "epoch": 1.2890995260663507, "step": 544, "train/alpha": 0.2, "train/clean_ce_loss": 0.1083984375, "train/clean_loss_scale": 8.167091369628906, "train/clean_loss_weight": 1.633418321609497, "train/combined_loss": 1.0623599290847778, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12890995260663507, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8852999210357666 }, { "epoch": 1.2914691943127963, "grad_norm": 16.349973678588867, "learning_rate": 2e-06, "loss": 3.4993, "step": 545 }, { "epoch": 1.2914691943127963, "step": 545, "train/alpha": 0.2, "train/clean_ce_loss": 0.1337890625, "train/clean_loss_scale": 10.308320045471191, "train/clean_loss_weight": 2.06166410446167, "train/combined_loss": 1.6549686193466187, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12914691943127962, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3791404962539673 }, { "epoch": 1.2914691943127963, "step": 545, "train/alpha": 0.2, "train/clean_ce_loss": 0.0059814453125, "train/clean_loss_scale": 55.364742279052734, "train/clean_loss_weight": 11.072948455810547, "train/combined_loss": 0.3973934054374695, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12914691943127962, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.33116117119789124 }, { "epoch": 1.2914691943127963, "step": 545, "train/alpha": 0.2, "train/clean_ce_loss": 0.103515625, "train/clean_loss_scale": 5.571753025054932, "train/clean_loss_weight": 1.11435067653656, "train/combined_loss": 0.6921162009239197, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12914691943127962, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5767635107040405 }, { "epoch": 1.2914691943127963, "step": 545, "train/alpha": 0.2, "train/clean_ce_loss": 0.057861328125, "train/clean_loss_scale": 10.240750312805176, "train/clean_loss_weight": 2.048150062561035, "train/combined_loss": 0.7110521197319031, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12914691943127962, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5925434231758118 }, { "epoch": 1.2938388625592416, "step": 546, "train/alpha": 0.2, "train/clean_ce_loss": 0.16796875, "train/clean_loss_scale": 7.407251358032227, "train/clean_loss_weight": 1.4814503192901611, "train/combined_loss": 1.4930241107940674, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12938388625592417, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2441867589950562 }, { "epoch": 1.2938388625592416, "step": 546, "train/alpha": 0.2, "train/clean_ce_loss": 0.06787109375, "train/clean_loss_scale": 8.36594009399414, "train/clean_loss_weight": 1.6731880903244019, "train/combined_loss": 0.6813666224479675, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12938388625592417, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5678055286407471 }, { "epoch": 1.2938388625592416, "step": 546, "train/alpha": 0.2, "train/clean_ce_loss": 0.138671875, "train/clean_loss_scale": 7.650498867034912, "train/clean_loss_weight": 1.5300997495651245, "train/combined_loss": 1.2730908393859863, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12938388625592417, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0609090328216553 }, { "epoch": 1.2938388625592416, "step": 546, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 13.574131965637207, "train/clean_loss_weight": 2.7148263454437256, "train/combined_loss": 1.3202964067459106, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12938388625592417, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1002470254898071 }, { "epoch": 1.2962085308056872, "step": 547, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 7.992560863494873, "train/clean_loss_weight": 1.5985121726989746, "train/combined_loss": 0.7118374705314636, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12962085308056873, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5931978821754456 }, { "epoch": 1.2962085308056872, "step": 547, "train/alpha": 0.2, "train/clean_ce_loss": 0.09716796875, "train/clean_loss_scale": 8.149235725402832, "train/clean_loss_weight": 1.6298471689224243, "train/combined_loss": 0.950213611125946, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12962085308056873, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7918446660041809 }, { "epoch": 1.2962085308056872, "step": 547, "train/alpha": 0.2, "train/clean_ce_loss": 0.08056640625, "train/clean_loss_scale": 10.194595336914062, "train/clean_loss_weight": 2.03891921043396, "train/combined_loss": 0.9856102466583252, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12962085308056873, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.821341872215271 }, { "epoch": 1.2962085308056872, "step": 547, "train/alpha": 0.2, "train/clean_ce_loss": 0.11376953125, "train/clean_loss_scale": 8.42536449432373, "train/clean_loss_weight": 1.685072898864746, "train/combined_loss": 1.1502597332000732, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12962085308056873, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9585497975349426 }, { "epoch": 1.2985781990521326, "step": 548, "train/alpha": 0.2, "train/clean_ce_loss": 0.11181640625, "train/clean_loss_scale": 8.813776016235352, "train/clean_loss_weight": 1.762755274772644, "train/combined_loss": 1.1826297044754028, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12985781990521328, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9855247735977173 }, { "epoch": 1.2985781990521326, "step": 548, "train/alpha": 0.2, "train/clean_ce_loss": 0.0673828125, "train/clean_loss_scale": 9.75385856628418, "train/clean_loss_weight": 1.950771689414978, "train/combined_loss": 0.7886909246444702, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12985781990521328, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6572424173355103 }, { "epoch": 1.2985781990521326, "step": 548, "train/alpha": 0.2, "train/clean_ce_loss": 0.06982421875, "train/clean_loss_scale": 8.917229652404785, "train/clean_loss_weight": 1.783445954322815, "train/combined_loss": 0.7471662759780884, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12985781990521328, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6226385831832886 }, { "epoch": 1.2985781990521326, "step": 548, "train/alpha": 0.2, "train/clean_ce_loss": 0.177734375, "train/clean_loss_scale": 7.0137128829956055, "train/clean_loss_weight": 1.402742624282837, "train/combined_loss": 1.4958934783935547, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.12985781990521328, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2465778589248657 }, { "epoch": 1.3009478672985781, "step": 549, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 9.583657264709473, "train/clean_loss_weight": 1.9167314767837524, "train/combined_loss": 0.7468514442443848, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.13009478672985783, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6223762035369873 }, { "epoch": 1.3009478672985781, "step": 549, "train/alpha": 0.2, "train/clean_ce_loss": 0.1162109375, "train/clean_loss_scale": 6.631356716156006, "train/clean_loss_weight": 1.326271414756775, "train/combined_loss": 0.9247634410858154, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.13009478672985783, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7706362009048462 }, { "epoch": 1.3009478672985781, "step": 549, "train/alpha": 0.2, "train/clean_ce_loss": 0.055908203125, "train/clean_loss_scale": 5.787202835083008, "train/clean_loss_weight": 1.1574405431747437, "train/combined_loss": 0.3882625102996826, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.13009478672985783, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.32355210185050964 }, { "epoch": 1.3009478672985781, "step": 549, "train/alpha": 0.2, "train/clean_ce_loss": 0.07373046875, "train/clean_loss_scale": 10.088172912597656, "train/clean_loss_weight": 2.017634630203247, "train/combined_loss": 0.8925668597221375, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.13009478672985783, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7438057065010071 }, { "epoch": 1.3033175355450237, "grad_norm": 13.12017822265625, "learning_rate": 2e-06, "loss": 3.3044, "step": 550 }, { "epoch": 1.3033175355450237, "step": 550, "train/alpha": 0.2, "train/clean_ce_loss": 0.10693359375, "train/clean_loss_scale": 8.73931884765625, "train/clean_loss_weight": 1.74786376953125, "train/combined_loss": 1.1214321851730347, "train/pos_loss/pos0": 0.33201783895492554, "train/pos_loss/pos1": 0.7251343727111816, "train/pos_loss/pos2": 1.1995004415512085, "train/pos_loss/pos3": 1.7160658836364746, "train/pos_loss/pos4": 2.16459321975708, "train/pos_loss/pos5": 2.468305826187134, "train/pos_loss/pos6": 2.8851311206817627, "train/pos_loss/weight0": 2.5166752338409424, "train/pos_loss/weight1": 1.8204761743545532, "train/pos_loss/weight2": 1.2245121002197266, "train/pos_loss/weight3": 0.7339890003204346, "train/pos_loss/weight4": 0.40054944157600403, "train/pos_loss/weight5": 0.19995561242103577, "train/pos_loss/weight6": 0.10384213924407959, "train/pos_weight_progress": 0.13033175355450238, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.934526801109314 }, { "epoch": 1.3033175355450237, "step": 550, "train/alpha": 0.2, "train/clean_ce_loss": 0.01318359375, "train/clean_loss_scale": 20.896638870239258, "train/clean_loss_weight": 4.179327964782715, "train/combined_loss": 0.3305913507938385, "train/pos_loss/pos0": 0.03264617919921875, "train/pos_loss/pos1": 0.12772038578987122, "train/pos_loss/pos2": 0.35207027196884155, "train/pos_loss/pos3": 0.8817917704582214, "train/pos_loss/pos4": 1.451917290687561, "train/pos_loss/pos5": 1.9673943519592285, "train/pos_loss/pos6": 2.379652976989746, "train/pos_loss/weight0": 2.8710544109344482, "train/pos_loss/weight1": 1.9635796546936035, "train/pos_loss/weight2": 1.1763606071472168, "train/pos_loss/weight3": 0.6059023141860962, "train/pos_loss/weight4": 0.2553899586200714, "train/pos_loss/weight5": 0.09280068427324295, "train/pos_loss/weight6": 0.03491266071796417, "train/pos_weight_progress": 0.13033175355450238, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.275492787361145 }, { "epoch": 1.3033175355450237, "step": 550, "train/alpha": 0.2, "train/clean_ce_loss": 0.126953125, "train/clean_loss_scale": 9.869756698608398, "train/clean_loss_weight": 1.9739513397216797, "train/combined_loss": 1.5035957098007202, "train/pos_loss/pos0": 0.4666435122489929, "train/pos_loss/pos1": 0.9483045339584351, "train/pos_loss/pos2": 1.3237388134002686, "train/pos_loss/pos3": 1.81437087059021, "train/pos_loss/pos4": 2.5406577587127686, "train/pos_loss/pos5": 3.1731767654418945, "train/pos_loss/pos6": 3.396474838256836, "train/pos_loss/weight0": 2.194319009780884, "train/pos_loss/weight1": 1.6808555126190186, "train/pos_loss/weight2": 1.23397958278656, "train/pos_loss/weight3": 0.8627514839172363, "train/pos_loss/weight4": 0.541475772857666, "train/pos_loss/weight5": 0.30105578899383545, "train/pos_loss/weight6": 0.1855631023645401, "train/pos_weight_progress": 0.13033175355450238, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2529964447021484 }, { "epoch": 1.3033175355450237, "step": 550, "train/alpha": 0.2, "train/clean_ce_loss": 0.01336669921875, "train/clean_loss_scale": 16.85730743408203, "train/clean_loss_weight": 3.3714616298675537, "train/combined_loss": 0.27039188146591187, "train/pos_loss/pos0": 0.012285152450203896, "train/pos_loss/pos1": 0.1023838222026825, "train/pos_loss/pos2": 0.3982667326927185, "train/pos_loss/pos3": 0.7033035159111023, "train/pos_loss/pos4": 1.4667598009109497, "train/pos_loss/pos5": 2.0238537788391113, "train/pos_loss/pos6": 2.257866382598877, "train/pos_loss/weight0": 2.9781742095947266, "train/pos_loss/weight1": 2.046635866165161, "train/pos_loss/weight2": 1.1220829486846924, "train/pos_loss/weight3": 0.5500873327255249, "train/pos_loss/weight4": 0.21512550115585327, "train/pos_loss/weight5": 0.06630997359752655, "train/pos_loss/weight6": 0.021584071218967438, "train/pos_weight_progress": 0.13033175355450238, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2253265529870987 }, { "epoch": 1.3056872037914693, "step": 551, "train/alpha": 0.2, "train/clean_ce_loss": 0.13671875, "train/clean_loss_scale": 9.884204864501953, "train/clean_loss_weight": 1.9768409729003906, "train/combined_loss": 1.6216273307800293, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.1305687203791469, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3513561487197876 }, { "epoch": 1.3056872037914693, "step": 551, "train/alpha": 0.2, "train/clean_ce_loss": 0.06689453125, "train/clean_loss_scale": 9.677092552185059, "train/clean_loss_weight": 1.9354184865951538, "train/combined_loss": 0.7768135070800781, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.1305687203791469, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6473445892333984 }, { "epoch": 1.3056872037914693, "step": 551, "train/alpha": 0.2, "train/clean_ce_loss": 0.16796875, "train/clean_loss_scale": 8.06942367553711, "train/clean_loss_weight": 1.6138848066329956, "train/combined_loss": 1.626493215560913, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.1305687203791469, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3554109334945679 }, { "epoch": 1.3056872037914693, "step": 551, "train/alpha": 0.2, "train/clean_ce_loss": 0.09423828125, "train/clean_loss_scale": 16.116445541381836, "train/clean_loss_weight": 3.2232892513275146, "train/combined_loss": 1.8225433826446533, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.1305687203791469, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5187861919403076 }, { "epoch": 1.3080568720379147, "step": 552, "train/alpha": 0.2, "train/clean_ce_loss": 0.044677734375, "train/clean_loss_scale": 13.677165985107422, "train/clean_loss_weight": 2.735433340072632, "train/combined_loss": 0.7332777380943298, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13080568720379146, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6110647916793823 }, { "epoch": 1.3080568720379147, "step": 552, "train/alpha": 0.2, "train/clean_ce_loss": 0.13671875, "train/clean_loss_scale": 7.24724006652832, "train/clean_loss_weight": 1.4494479894638062, "train/combined_loss": 1.1890002489089966, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13080568720379146, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.990833580493927 }, { "epoch": 1.3080568720379147, "step": 552, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 13.0459566116333, "train/clean_loss_weight": 2.609191417694092, "train/combined_loss": 1.0319554805755615, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13080568720379146, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8599629402160645 }, { "epoch": 1.3080568720379147, "step": 552, "train/alpha": 0.2, "train/clean_ce_loss": 0.0615234375, "train/clean_loss_scale": 10.905171394348145, "train/clean_loss_weight": 2.1810343265533447, "train/combined_loss": 0.805108368396759, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13080568720379146, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.67092365026474 }, { "epoch": 1.3104265402843602, "step": 553, "train/alpha": 0.2, "train/clean_ce_loss": 0.045166015625, "train/clean_loss_scale": 14.982353210449219, "train/clean_loss_weight": 2.9964706897735596, "train/combined_loss": 0.8120318651199341, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.131042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6766932010650635 }, { "epoch": 1.3104265402843602, "step": 553, "train/alpha": 0.2, "train/clean_ce_loss": 0.06298828125, "train/clean_loss_scale": 9.995796203613281, "train/clean_loss_weight": 1.9991592168807983, "train/combined_loss": 0.7555416822433472, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.131042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6296180486679077 }, { "epoch": 1.3104265402843602, "step": 553, "train/alpha": 0.2, "train/clean_ce_loss": 0.09912109375, "train/clean_loss_scale": 16.610414505004883, "train/clean_loss_weight": 3.322082996368408, "train/combined_loss": 1.9757310152053833, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.131042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.6464425325393677 }, { "epoch": 1.3104265402843602, "step": 553, "train/alpha": 0.2, "train/clean_ce_loss": 4.351139068603516e-06, "train/clean_loss_scale": 63768.45703125, "train/clean_loss_weight": 12753.69140625, "train/combined_loss": 0.33295851945877075, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.131042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.27746543288230896 }, { "epoch": 1.3127962085308056, "step": 554, "train/alpha": 0.2, "train/clean_ce_loss": 0.119140625, "train/clean_loss_scale": 10.396939277648926, "train/clean_loss_weight": 2.079387903213501, "train/combined_loss": 1.486437439918518, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13127962085308056, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.23869788646698 }, { "epoch": 1.3127962085308056, "step": 554, "train/alpha": 0.2, "train/clean_ce_loss": 0.044189453125, "train/clean_loss_scale": 13.839436531066895, "train/clean_loss_weight": 2.7678873538970947, "train/combined_loss": 0.7338685393333435, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13127962085308056, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.611557126045227 }, { "epoch": 1.3127962085308056, "step": 554, "train/alpha": 0.2, "train/clean_ce_loss": 0.0218505859375, "train/clean_loss_scale": 22.777523040771484, "train/clean_loss_weight": 4.55550479888916, "train/combined_loss": 0.5972426533699036, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13127962085308056, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4977022111415863 }, { "epoch": 1.3127962085308056, "step": 554, "train/alpha": 0.2, "train/clean_ce_loss": 0.15234375, "train/clean_loss_scale": 10.20217514038086, "train/clean_loss_weight": 2.0404350757598877, "train/combined_loss": 1.8650851249694824, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13127962085308056, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5542376041412354 }, { "epoch": 1.3151658767772512, "grad_norm": 13.089982986450195, "learning_rate": 2e-06, "loss": 3.4921, "step": 555 }, { "epoch": 1.3151658767772512, "step": 555, "train/alpha": 0.2, "train/clean_ce_loss": 0.1328125, "train/clean_loss_scale": 10.94367790222168, "train/clean_loss_weight": 2.1887357234954834, "train/combined_loss": 1.7441487312316895, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13151658767772512, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.453457236289978 }, { "epoch": 1.3151658767772512, "step": 555, "train/alpha": 0.2, "train/clean_ce_loss": 0.12890625, "train/clean_loss_scale": 10.085442543029785, "train/clean_loss_weight": 2.0170886516571045, "train/combined_loss": 1.5600919723510742, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13151658767772512, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3000766038894653 }, { "epoch": 1.3151658767772512, "step": 555, "train/alpha": 0.2, "train/clean_ce_loss": 0.1611328125, "train/clean_loss_scale": 10.400221824645996, "train/clean_loss_weight": 2.0800445079803467, "train/combined_loss": 2.0109803676605225, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13151658767772512, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.6758170127868652 }, { "epoch": 1.3151658767772512, "step": 555, "train/alpha": 0.2, "train/clean_ce_loss": 0.09228515625, "train/clean_loss_scale": 10.43494987487793, "train/clean_loss_weight": 2.0869901180267334, "train/combined_loss": 1.1555891036987305, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13151658767772512, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.962990939617157 }, { "epoch": 1.3175355450236967, "step": 556, "train/alpha": 0.2, "train/clean_ce_loss": 0.1201171875, "train/clean_loss_scale": 9.81202507019043, "train/clean_loss_weight": 1.9624050855636597, "train/combined_loss": 1.414311408996582, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13175355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1785928010940552 }, { "epoch": 1.3175355450236967, "step": 556, "train/alpha": 0.2, "train/clean_ce_loss": 0.0223388671875, "train/clean_loss_scale": 15.795363426208496, "train/clean_loss_weight": 3.1590726375579834, "train/combined_loss": 0.42342063784599304, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13175355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3528505265712738 }, { "epoch": 1.3175355450236967, "step": 556, "train/alpha": 0.2, "train/clean_ce_loss": 0.107421875, "train/clean_loss_scale": 10.724693298339844, "train/clean_loss_weight": 2.1449387073516846, "train/combined_loss": 1.3824800252914429, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13175355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.152066707611084 }, { "epoch": 1.3175355450236967, "step": 556, "train/alpha": 0.2, "train/clean_ce_loss": 0.10400390625, "train/clean_loss_scale": 11.820276260375977, "train/clean_loss_weight": 2.3640553951263428, "train/combined_loss": 1.475225806236267, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13175355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2293548583984375 }, { "epoch": 1.319905213270142, "step": 557, "train/alpha": 0.2, "train/clean_ce_loss": 0.169921875, "train/clean_loss_scale": 8.654006958007812, "train/clean_loss_weight": 1.7308014631271362, "train/combined_loss": 1.7646061182022095, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13199052132701422, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4705051183700562 }, { "epoch": 1.319905213270142, "step": 557, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 8.029255867004395, "train/clean_loss_weight": 1.605851173400879, "train/combined_loss": 0.780970573425293, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13199052132701422, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6508088111877441 }, { "epoch": 1.319905213270142, "step": 557, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 9.785173416137695, "train/clean_loss_weight": 1.957034707069397, "train/combined_loss": 1.089365005493164, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13199052132701422, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9078041911125183 }, { "epoch": 1.319905213270142, "step": 557, "train/alpha": 0.2, "train/clean_ce_loss": 0.10498046875, "train/clean_loss_scale": 13.131717681884766, "train/clean_loss_weight": 2.6263434886932373, "train/combined_loss": 1.654288649559021, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13199052132701422, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3785738945007324 }, { "epoch": 1.3222748815165877, "step": 558, "train/alpha": 0.2, "train/clean_ce_loss": 0.046875, "train/clean_loss_scale": 13.9248685836792, "train/clean_loss_weight": 2.7849738597869873, "train/combined_loss": 0.7832738161087036, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13222748815165877, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6527281999588013 }, { "epoch": 1.3222748815165877, "step": 558, "train/alpha": 0.2, "train/clean_ce_loss": 0.09375, "train/clean_loss_scale": 10.556144714355469, "train/clean_loss_weight": 2.1112289428710938, "train/combined_loss": 1.1875662803649902, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13222748815165877, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9896385669708252 }, { "epoch": 1.3222748815165877, "step": 558, "train/alpha": 0.2, "train/clean_ce_loss": 0.060546875, "train/clean_loss_scale": 8.67795181274414, "train/clean_loss_weight": 1.7355903387069702, "train/combined_loss": 0.6305074691772461, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13222748815165877, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5254228711128235 }, { "epoch": 1.3222748815165877, "step": 558, "train/alpha": 0.2, "train/clean_ce_loss": 0.08447265625, "train/clean_loss_scale": 9.596656799316406, "train/clean_loss_weight": 1.919331431388855, "train/combined_loss": 0.9727860689163208, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13222748815165877, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8106550574302673 }, { "epoch": 1.3246445497630333, "step": 559, "train/alpha": 0.2, "train/clean_ce_loss": 0.078125, "train/clean_loss_scale": 10.66010856628418, "train/clean_loss_weight": 2.13202166557312, "train/combined_loss": 0.9993852376937866, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13246445497630333, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8328210115432739 }, { "epoch": 1.3246445497630333, "step": 559, "train/alpha": 0.2, "train/clean_ce_loss": 0.043701171875, "train/clean_loss_scale": 10.741129875183105, "train/clean_loss_weight": 2.148226022720337, "train/combined_loss": 0.5632799863815308, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13246445497630333, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46939995884895325 }, { "epoch": 1.3246445497630333, "step": 559, "train/alpha": 0.2, "train/clean_ce_loss": 0.0771484375, "train/clean_loss_scale": 21.003761291503906, "train/clean_loss_weight": 4.200752258300781, "train/combined_loss": 1.944488763809204, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13246445497630333, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.6204073429107666 }, { "epoch": 1.3246445497630333, "step": 559, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 11.703845024108887, "train/clean_loss_weight": 2.340769052505493, "train/combined_loss": 0.9120769500732422, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13246445497630333, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7600641250610352 }, { "epoch": 1.3270142180094786, "grad_norm": 16.617630004882812, "learning_rate": 2e-06, "loss": 3.4638, "step": 560 }, { "epoch": 1.3270142180094786, "step": 560, "train/alpha": 0.2, "train/clean_ce_loss": 0.041259765625, "train/clean_loss_scale": 5.667245864868164, "train/clean_loss_weight": 1.1334491968154907, "train/combined_loss": 0.2805950939655304, "train/pos_loss/pos0": 0.001196022261865437, "train/pos_loss/pos1": 0.01094165537506342, "train/pos_loss/pos2": 0.060607582330703735, "train/pos_loss/pos3": 0.197299063205719, "train/pos_loss/pos4": 0.8527213335037231, "train/pos_loss/pos5": 1.5147565603256226, "train/pos_loss/pos6": 2.136200189590454, "train/pos_loss/weight0": 2.15285062789917, "train/pos_loss/weight1": 1.6665942668914795, "train/pos_loss/weight2": 1.2549386024475098, "train/pos_loss/weight3": 0.8534725904464722, "train/pos_loss/weight4": 0.558977484703064, "train/pos_loss/weight5": 0.3180006742477417, "train/pos_loss/weight6": 0.19516561925411224, "train/pos_weight_progress": 0.13270142180094788, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.23382924497127533 }, { "epoch": 1.3270142180094786, "step": 560, "train/alpha": 0.2, "train/clean_ce_loss": 0.087890625, "train/clean_loss_scale": 9.978754997253418, "train/clean_loss_weight": 1.9957510232925415, "train/combined_loss": 1.0524468421936035, "train/pos_loss/pos0": 0.12186149507761002, "train/pos_loss/pos1": 0.3432108759880066, "train/pos_loss/pos2": 0.7401127219200134, "train/pos_loss/pos3": 1.375599980354309, "train/pos_loss/pos4": 1.8456734418869019, "train/pos_loss/pos5": 2.235623598098755, "train/pos_loss/pos6": 2.6196210384368896, "train/pos_loss/weight0": 1.8860116004943848, "train/pos_loss/weight1": 1.5203989744186401, "train/pos_loss/weight2": 1.1844666004180908, "train/pos_loss/weight3": 0.8787375688552856, "train/pos_loss/weight4": 0.6651750206947327, "train/pos_loss/weight5": 0.49834558367729187, "train/pos_loss/weight6": 0.36686524748802185, "train/pos_weight_progress": 0.13270142180094788, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8770390152931213 }, { "epoch": 1.3270142180094786, "step": 560, "train/alpha": 0.2, "train/clean_ce_loss": 0.019287109375, "train/clean_loss_scale": 29.748828887939453, "train/clean_loss_weight": 5.949765682220459, "train/combined_loss": 0.6885226964950562, "train/pos_loss/pos0": 0.0661151260137558, "train/pos_loss/pos1": 0.30732375383377075, "train/pos_loss/pos2": 0.6210562586784363, "train/pos_loss/pos3": 1.2669541835784912, "train/pos_loss/pos4": 1.8783588409423828, "train/pos_loss/pos5": 2.253891706466675, "train/pos_loss/pos6": 2.880664825439453, "train/pos_loss/weight0": 2.441418170928955, "train/pos_loss/weight1": 1.7987096309661865, "train/pos_loss/weight2": 1.2540392875671387, "train/pos_loss/weight3": 0.7899720668792725, "train/pos_loss/weight4": 0.41113388538360596, "train/pos_loss/weight5": 0.1996016651391983, "train/pos_loss/weight6": 0.10512512177228928, "train/pos_weight_progress": 0.13270142180094788, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5737689137458801 }, { "epoch": 1.3270142180094786, "step": 560, "train/alpha": 0.2, "train/clean_ce_loss": 0.130859375, "train/clean_loss_scale": 9.912934303283691, "train/clean_loss_weight": 1.9825868606567383, "train/combined_loss": 1.5566405057907104, "train/pos_loss/pos0": 0.6234526634216309, "train/pos_loss/pos1": 1.115085244178772, "train/pos_loss/pos2": 1.5746735334396362, "train/pos_loss/pos3": 2.0504748821258545, "train/pos_loss/pos4": 2.544438362121582, "train/pos_loss/pos5": 2.9513158798217773, "train/pos_loss/pos6": 3.434269428253174, "train/pos_loss/weight0": 2.4380338191986084, "train/pos_loss/weight1": 1.8112952709197998, "train/pos_loss/weight2": 1.2523471117019653, "train/pos_loss/weight3": 0.8115166425704956, "train/pos_loss/weight4": 0.41125982999801636, "train/pos_loss/weight5": 0.18124239146709442, "train/pos_loss/weight6": 0.09430468827486038, "train/pos_weight_progress": 0.13270142180094788, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2972004413604736 }, { "epoch": 1.3293838862559242, "step": 561, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 6.937437534332275, "train/clean_loss_weight": 1.387487530708313, "train/combined_loss": 0.5528270602226257, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13293838862559243, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4606892168521881 }, { "epoch": 1.3293838862559242, "step": 561, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 6.757361888885498, "train/clean_loss_weight": 1.3514723777770996, "train/combined_loss": 0.7522844672203064, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13293838862559243, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6269037127494812 }, { "epoch": 1.3293838862559242, "step": 561, "train/alpha": 0.2, "train/clean_ce_loss": 0.09375, "train/clean_loss_scale": 6.718151092529297, "train/clean_loss_weight": 1.3436301946640015, "train/combined_loss": 0.7557920217514038, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13293838862559243, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6298266649246216 }, { "epoch": 1.3293838862559242, "step": 561, "train/alpha": 0.2, "train/clean_ce_loss": 0.08935546875, "train/clean_loss_scale": 6.326038360595703, "train/clean_loss_weight": 1.2652076482772827, "train/combined_loss": 0.6783193349838257, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13293838862559243, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5652661323547363 }, { "epoch": 1.3317535545023698, "step": 562, "train/alpha": 0.2, "train/clean_ce_loss": 0.041748046875, "train/clean_loss_scale": 6.7797346115112305, "train/clean_loss_weight": 1.3559468984603882, "train/combined_loss": 0.3396488130092621, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13317535545023695, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2830406725406647 }, { "epoch": 1.3317535545023698, "step": 562, "train/alpha": 0.2, "train/clean_ce_loss": 0.09375, "train/clean_loss_scale": 7.41494607925415, "train/clean_loss_weight": 1.4829891920089722, "train/combined_loss": 0.8341814279556274, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13317535545023695, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6951512098312378 }, { "epoch": 1.3317535545023698, "step": 562, "train/alpha": 0.2, "train/clean_ce_loss": 0.042724609375, "train/clean_loss_scale": 7.174100399017334, "train/clean_loss_weight": 1.4348200559616089, "train/combined_loss": 0.36781275272369385, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13317535545023695, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3065106272697449 }, { "epoch": 1.3317535545023698, "step": 562, "train/alpha": 0.2, "train/clean_ce_loss": 0.053466796875, "train/clean_loss_scale": 6.704455375671387, "train/clean_loss_weight": 1.3408911228179932, "train/combined_loss": 0.43015891313552856, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13317535545023695, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3584657609462738 }, { "epoch": 1.3341232227488151, "step": 563, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 8.41324234008789, "train/clean_loss_weight": 1.6826485395431519, "train/combined_loss": 0.6655006408691406, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.1334123222748815, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5545838475227356 }, { "epoch": 1.3341232227488151, "step": 563, "train/alpha": 0.2, "train/clean_ce_loss": 0.059326171875, "train/clean_loss_scale": 7.606186389923096, "train/clean_loss_weight": 1.5212372541427612, "train/combined_loss": 0.5414950847625732, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.1334123222748815, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4512459337711334 }, { "epoch": 1.3341232227488151, "step": 563, "train/alpha": 0.2, "train/clean_ce_loss": 0.045654296875, "train/clean_loss_scale": 5.505946159362793, "train/clean_loss_weight": 1.1011892557144165, "train/combined_loss": 0.30164411664009094, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.1334123222748815, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2513701021671295 }, { "epoch": 1.3341232227488151, "step": 563, "train/alpha": 0.2, "train/clean_ce_loss": 0.09423828125, "train/clean_loss_scale": 9.742233276367188, "train/clean_loss_weight": 1.9484466314315796, "train/combined_loss": 1.1017096042633057, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.1334123222748815, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.918091356754303 }, { "epoch": 1.3364928909952607, "step": 564, "train/alpha": 0.2, "train/clean_ce_loss": 0.00958251953125, "train/clean_loss_scale": 21.39850616455078, "train/clean_loss_weight": 4.279701232910156, "train/combined_loss": 0.24606192111968994, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13364928909952606, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.20505160093307495 }, { "epoch": 1.3364928909952607, "step": 564, "train/alpha": 0.2, "train/clean_ce_loss": 0.1123046875, "train/clean_loss_scale": 6.891483306884766, "train/clean_loss_weight": 1.3782967329025269, "train/combined_loss": 0.9287350177764893, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13364928909952606, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7739458680152893 }, { "epoch": 1.3364928909952607, "step": 564, "train/alpha": 0.2, "train/clean_ce_loss": 0.08544921875, "train/clean_loss_scale": 5.368635177612305, "train/clean_loss_weight": 1.073727011680603, "train/combined_loss": 0.5504947900772095, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13364928909952606, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45874568819999695 }, { "epoch": 1.3364928909952607, "step": 564, "train/alpha": 0.2, "train/clean_ce_loss": 0.047119140625, "train/clean_loss_scale": 10.156269073486328, "train/clean_loss_weight": 2.0312538146972656, "train/combined_loss": 0.5742655992507935, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13364928909952606, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4785546660423279 }, { "epoch": 1.338862559241706, "grad_norm": 22.380937576293945, "learning_rate": 2e-06, "loss": 3.3251, "step": 565 }, { "epoch": 1.338862559241706, "step": 565, "train/alpha": 0.2, "train/clean_ce_loss": 0.01409912109375, "train/clean_loss_scale": 18.045209884643555, "train/clean_loss_weight": 3.609041929244995, "train/combined_loss": 0.3053058981895447, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.1338862559241706, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.254421591758728 }, { "epoch": 1.338862559241706, "step": 565, "train/alpha": 0.2, "train/clean_ce_loss": 0.126953125, "train/clean_loss_scale": 9.221196174621582, "train/clean_loss_weight": 1.8442392349243164, "train/combined_loss": 1.4047915935516357, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.1338862559241706, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1706596612930298 }, { "epoch": 1.338862559241706, "step": 565, "train/alpha": 0.2, "train/clean_ce_loss": 0.040771484375, "train/clean_loss_scale": 8.699819564819336, "train/clean_loss_weight": 1.7399638891220093, "train/combined_loss": 0.42564547061920166, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.1338862559241706, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3547045588493347 }, { "epoch": 1.338862559241706, "step": 565, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 6.302840709686279, "train/clean_loss_weight": 1.2605681419372559, "train/combined_loss": 0.491178423166275, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.1338862559241706, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4093153476715088 }, { "epoch": 1.3412322274881516, "step": 566, "train/alpha": 0.2, "train/clean_ce_loss": 0.0299072265625, "train/clean_loss_scale": 10.359545707702637, "train/clean_loss_weight": 2.071909189224243, "train/combined_loss": 0.3717903196811676, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13412322274881516, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.30982527136802673 }, { "epoch": 1.3412322274881516, "step": 566, "train/alpha": 0.2, "train/clean_ce_loss": 0.0830078125, "train/clean_loss_scale": 5.615772724151611, "train/clean_loss_weight": 1.1231545209884644, "train/combined_loss": 0.5593835711479187, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13412322274881516, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46615299582481384 }, { "epoch": 1.3412322274881516, "step": 566, "train/alpha": 0.2, "train/clean_ce_loss": 0.06005859375, "train/clean_loss_scale": 7.329276084899902, "train/clean_loss_weight": 1.4658552408218384, "train/combined_loss": 0.5282232165336609, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13412322274881516, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4401860237121582 }, { "epoch": 1.3412322274881516, "step": 566, "train/alpha": 0.2, "train/clean_ce_loss": 0.05224609375, "train/clean_loss_scale": 6.258282661437988, "train/clean_loss_weight": 1.2516565322875977, "train/combined_loss": 0.3923649787902832, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13412322274881516, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.32697081565856934 }, { "epoch": 1.3436018957345972, "step": 567, "train/alpha": 0.2, "train/clean_ce_loss": 0.16015625, "train/clean_loss_scale": 5.5344085693359375, "train/clean_loss_weight": 1.1068817377090454, "train/combined_loss": 1.0636441707611084, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13436018957345972, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8863701224327087 }, { "epoch": 1.3436018957345972, "step": 567, "train/alpha": 0.2, "train/clean_ce_loss": 0.038818359375, "train/clean_loss_scale": 9.442218780517578, "train/clean_loss_weight": 1.8884438276290894, "train/combined_loss": 0.4398377239704132, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13436018957345972, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3665314316749573 }, { "epoch": 1.3436018957345972, "step": 567, "train/alpha": 0.2, "train/clean_ce_loss": 0.10888671875, "train/clean_loss_scale": 5.562560081481934, "train/clean_loss_weight": 1.1125119924545288, "train/combined_loss": 0.7268267273902893, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13436018957345972, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6056889295578003 }, { "epoch": 1.3436018957345972, "step": 567, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 7.2562408447265625, "train/clean_loss_weight": 1.4512481689453125, "train/combined_loss": 0.7100344896316528, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13436018957345972, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5916954278945923 }, { "epoch": 1.3459715639810428, "step": 568, "train/alpha": 0.2, "train/clean_ce_loss": 0.053466796875, "train/clean_loss_scale": 5.918437957763672, "train/clean_loss_weight": 1.1836875677108765, "train/combined_loss": 0.37972790002822876, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13459715639810427, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3164399266242981 }, { "epoch": 1.3459715639810428, "step": 568, "train/alpha": 0.2, "train/clean_ce_loss": 0.189453125, "train/clean_loss_scale": 5.7746901512146, "train/clean_loss_weight": 1.1549381017684937, "train/combined_loss": 1.3128397464752197, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13459715639810427, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.094033122062683 }, { "epoch": 1.3459715639810428, "step": 568, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 6.525264263153076, "train/clean_loss_weight": 1.3050528764724731, "train/combined_loss": 0.5696861743927002, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13459715639810427, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4747384786605835 }, { "epoch": 1.3459715639810428, "step": 568, "train/alpha": 0.2, "train/clean_ce_loss": 0.10791015625, "train/clean_loss_scale": 8.313485145568848, "train/clean_loss_weight": 1.6626970767974854, "train/combined_loss": 1.0765314102172852, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13459715639810427, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8971095085144043 }, { "epoch": 1.3483412322274881, "step": 569, "train/alpha": 0.2, "train/clean_ce_loss": 0.051513671875, "train/clean_loss_scale": 9.034236907958984, "train/clean_loss_weight": 1.8068474531173706, "train/combined_loss": 0.5584640502929688, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13483412322274882, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4653867185115814 }, { "epoch": 1.3483412322274881, "step": 569, "train/alpha": 0.2, "train/clean_ce_loss": 0.00787353515625, "train/clean_loss_scale": 37.906822204589844, "train/clean_loss_weight": 7.581364631652832, "train/combined_loss": 0.358152836561203, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13483412322274882, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2984606921672821 }, { "epoch": 1.3483412322274881, "step": 569, "train/alpha": 0.2, "train/clean_ce_loss": 0.060546875, "train/clean_loss_scale": 7.549919605255127, "train/clean_loss_weight": 1.5099838972091675, "train/combined_loss": 0.5485488176345825, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13483412322274882, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4571240246295929 }, { "epoch": 1.3483412322274881, "step": 569, "train/alpha": 0.2, "train/clean_ce_loss": 0.154296875, "train/clean_loss_scale": 7.7656474113464355, "train/clean_loss_weight": 1.5531295537948608, "train/combined_loss": 1.4378581047058105, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13483412322274882, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.198215126991272 }, { "epoch": 1.3507109004739337, "grad_norm": 15.608816146850586, "learning_rate": 2e-06, "loss": 3.2781, "step": 570 }, { "epoch": 1.3507109004739337, "step": 570, "train/alpha": 0.2, "train/clean_ce_loss": 0.0179443359375, "train/clean_loss_scale": 15.3021879196167, "train/clean_loss_weight": 3.0604376792907715, "train/combined_loss": 0.3295051157474518, "train/pos_loss/pos0": 0.08698540925979614, "train/pos_loss/pos1": 0.21934570372104645, "train/pos_loss/pos2": 0.43797850608825684, "train/pos_loss/pos3": 0.8234294056892395, "train/pos_loss/pos4": 1.5163003206253052, "train/pos_loss/pos5": 1.9323689937591553, "train/pos_loss/pos6": 2.2483725547790527, "train/pos_loss/weight0": 3.2145073413848877, "train/pos_loss/weight1": 2.0418381690979004, "train/pos_loss/weight2": 1.072675108909607, "train/pos_loss/weight3": 0.4583519697189331, "train/pos_loss/weight4": 0.15697208046913147, "train/pos_loss/weight5": 0.04346858337521553, "train/pos_loss/weight6": 0.01218663901090622, "train/pos_weight_progress": 0.13507109004739337, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.27458760142326355 }, { "epoch": 1.3507109004739337, "step": 570, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 9.963860511779785, "train/clean_loss_weight": 1.992772102355957, "train/combined_loss": 0.8698917627334595, "train/pos_loss/pos0": 0.15131330490112305, "train/pos_loss/pos1": 0.42174866795539856, "train/pos_loss/pos2": 0.7190470695495605, "train/pos_loss/pos3": 1.2149771451950073, "train/pos_loss/pos4": 1.5790570974349976, "train/pos_loss/pos5": 2.009369373321533, "train/pos_loss/pos6": 2.4612624645233154, "train/pos_loss/weight0": 2.1077589988708496, "train/pos_loss/weight1": 1.6695338487625122, "train/pos_loss/weight2": 1.2375288009643555, "train/pos_loss/weight3": 0.8552260398864746, "train/pos_loss/weight4": 0.5728985667228699, "train/pos_loss/weight5": 0.3378199338912964, "train/pos_loss/weight6": 0.21923384070396423, "train/pos_weight_progress": 0.13507109004739337, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.724909782409668 }, { "epoch": 1.3507109004739337, "step": 570, "train/alpha": 0.2, "train/clean_ce_loss": 0.01483154296875, "train/clean_loss_scale": 37.304622650146484, "train/clean_loss_weight": 7.4609246253967285, "train/combined_loss": 0.6639421582221985, "train/pos_loss/pos0": 0.030289649963378906, "train/pos_loss/pos1": 0.23734945058822632, "train/pos_loss/pos2": 0.6546072959899902, "train/pos_loss/pos3": 1.23686945438385, "train/pos_loss/pos4": 1.9693291187286377, "train/pos_loss/pos5": 2.622270107269287, "train/pos_loss/pos6": 3.102029323577881, "train/pos_loss/weight0": 2.476639747619629, "train/pos_loss/weight1": 1.821374535560608, "train/pos_loss/weight2": 1.242199182510376, "train/pos_loss/weight3": 0.7647495269775391, "train/pos_loss/weight4": 0.4006209671497345, "train/pos_loss/weight5": 0.19353771209716797, "train/pos_loss/weight6": 0.10087816417217255, "train/pos_weight_progress": 0.13507109004739337, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5532851219177246 }, { "epoch": 1.3507109004739337, "step": 570, "train/alpha": 0.2, "train/clean_ce_loss": 0.1396484375, "train/clean_loss_scale": 9.684930801391602, "train/clean_loss_weight": 1.9369862079620361, "train/combined_loss": 1.6229825019836426, "train/pos_loss/pos0": 0.4953957796096802, "train/pos_loss/pos1": 1.1986358165740967, "train/pos_loss/pos2": 1.7291076183319092, "train/pos_loss/pos3": 2.3762338161468506, "train/pos_loss/pos4": 2.7961223125457764, "train/pos_loss/pos5": 3.425536870956421, "train/pos_loss/pos6": 3.6761634349823, "train/pos_loss/weight0": 2.4517531394958496, "train/pos_loss/weight1": 1.8382346630096436, "train/pos_loss/weight2": 1.289405345916748, "train/pos_loss/weight3": 0.77931147813797, "train/pos_loss/weight4": 0.3929053246974945, "train/pos_loss/weight5": 0.169064462184906, "train/pos_loss/weight6": 0.07932567596435547, "train/pos_weight_progress": 0.13507109004739337, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3524854183197021 }, { "epoch": 1.353080568720379, "step": 571, "train/alpha": 0.2, "train/clean_ce_loss": 0.13671875, "train/clean_loss_scale": 6.952914714813232, "train/clean_loss_weight": 1.3905829191207886, "train/combined_loss": 1.1407126188278198, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13530805687203792, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9505938291549683 }, { "epoch": 1.353080568720379, "step": 571, "train/alpha": 0.2, "train/clean_ce_loss": 0.1328125, "train/clean_loss_scale": 6.2033209800720215, "train/clean_loss_weight": 1.2406642436981201, "train/combined_loss": 0.9886543154716492, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13530805687203792, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8238785862922668 }, { "epoch": 1.353080568720379, "step": 571, "train/alpha": 0.2, "train/clean_ce_loss": 0.12890625, "train/clean_loss_scale": 7.135725021362305, "train/clean_loss_weight": 1.427145004272461, "train/combined_loss": 1.1038074493408203, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13530805687203792, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9198395609855652 }, { "epoch": 1.353080568720379, "step": 571, "train/alpha": 0.2, "train/clean_ce_loss": 0.1357421875, "train/clean_loss_scale": 7.500927448272705, "train/clean_loss_weight": 1.500185489654541, "train/combined_loss": 1.2218307256698608, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13530805687203792, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0181922912597656 }, { "epoch": 1.3554502369668247, "step": 572, "train/alpha": 0.2, "train/clean_ce_loss": 0.0947265625, "train/clean_loss_scale": 6.4793291091918945, "train/clean_loss_weight": 1.2958658933639526, "train/combined_loss": 0.7365174889564514, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13554502369668248, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6137645840644836 }, { "epoch": 1.3554502369668247, "step": 572, "train/alpha": 0.2, "train/clean_ce_loss": 0.10693359375, "train/clean_loss_scale": 8.197592735290527, "train/clean_loss_weight": 1.6395186185836792, "train/combined_loss": 1.0519176721572876, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13554502369668248, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.876598060131073 }, { "epoch": 1.3554502369668247, "step": 572, "train/alpha": 0.2, "train/clean_ce_loss": 0.1826171875, "train/clean_loss_scale": 5.180296421051025, "train/clean_loss_weight": 1.0360592603683472, "train/combined_loss": 1.1352133750915527, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13554502369668248, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9460111856460571 }, { "epoch": 1.3554502369668247, "step": 572, "train/alpha": 0.2, "train/clean_ce_loss": 0.01519775390625, "train/clean_loss_scale": 5.837484359741211, "train/clean_loss_weight": 1.167496919631958, "train/combined_loss": 0.10645997524261475, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13554502369668248, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.08871664851903915 }, { "epoch": 1.3578199052132702, "step": 573, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 6.1512770652771, "train/clean_loss_weight": 1.2302554845809937, "train/combined_loss": 0.547848105430603, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.135781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4565400779247284 }, { "epoch": 1.3578199052132702, "step": 573, "train/alpha": 0.2, "train/clean_ce_loss": 0.10205078125, "train/clean_loss_scale": 6.310979843139648, "train/clean_loss_weight": 1.2621959447860718, "train/combined_loss": 0.7728484869003296, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.135781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6440404057502747 }, { "epoch": 1.3578199052132702, "step": 573, "train/alpha": 0.2, "train/clean_ce_loss": 0.0947265625, "train/clean_loss_scale": 5.980671405792236, "train/clean_loss_weight": 1.196134328842163, "train/combined_loss": 0.6798341274261475, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.135781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5665284395217896 }, { "epoch": 1.3578199052132702, "step": 573, "train/alpha": 0.2, "train/clean_ce_loss": 0.134765625, "train/clean_loss_scale": 5.341639518737793, "train/clean_loss_weight": 1.0683279037475586, "train/combined_loss": 0.8638432621955872, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.135781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7198693752288818 }, { "epoch": 1.3601895734597156, "step": 574, "train/alpha": 0.2, "train/clean_ce_loss": 0.10498046875, "train/clean_loss_scale": 8.236979484558105, "train/clean_loss_weight": 1.6473959684371948, "train/combined_loss": 1.0376663208007812, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13601895734597155, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.864721953868866 }, { "epoch": 1.3601895734597156, "step": 574, "train/alpha": 0.2, "train/clean_ce_loss": 0.0625, "train/clean_loss_scale": 6.493710041046143, "train/clean_loss_weight": 1.2987420558929443, "train/combined_loss": 0.48702824115753174, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13601895734597155, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4058568775653839 }, { "epoch": 1.3601895734597156, "step": 574, "train/alpha": 0.2, "train/clean_ce_loss": 0.00933837890625, "train/clean_loss_scale": 7.604335784912109, "train/clean_loss_weight": 1.5208672285079956, "train/combined_loss": 0.08521459996700287, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13601895734597155, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.07101216912269592 }, { "epoch": 1.3601895734597156, "step": 574, "train/alpha": 0.2, "train/clean_ce_loss": 0.1376953125, "train/clean_loss_scale": 5.710973262786865, "train/clean_loss_weight": 1.1421946287155151, "train/combined_loss": 0.9436491131782532, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13601895734597155, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7863742709159851 }, { "epoch": 1.3625592417061612, "grad_norm": 26.653676986694336, "learning_rate": 2e-06, "loss": 3.1923, "step": 575 }, { "epoch": 1.3625592417061612, "step": 575, "train/alpha": 0.2, "train/clean_ce_loss": 0.07373046875, "train/clean_loss_scale": 6.069429874420166, "train/clean_loss_weight": 1.213886022567749, "train/combined_loss": 0.5370022654533386, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.1362559241706161, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.44750189781188965 }, { "epoch": 1.3625592417061612, "step": 575, "train/alpha": 0.2, "train/clean_ce_loss": 0.1435546875, "train/clean_loss_scale": 5.96488094329834, "train/clean_loss_weight": 1.1929762363433838, "train/combined_loss": 1.0275440216064453, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.1362559241706161, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8562866449356079 }, { "epoch": 1.3625592417061612, "step": 575, "train/alpha": 0.2, "train/clean_ce_loss": 0.150390625, "train/clean_loss_scale": 5.152132511138916, "train/clean_loss_weight": 1.0304265022277832, "train/combined_loss": 0.9297989010810852, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.1362559241706161, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7748324275016785 }, { "epoch": 1.3625592417061612, "step": 575, "train/alpha": 0.2, "train/clean_ce_loss": 0.1318359375, "train/clean_loss_scale": 8.02989673614502, "train/clean_loss_weight": 1.605979323387146, "train/combined_loss": 1.2703547477722168, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.1362559241706161, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0586289167404175 }, { "epoch": 1.3649289099526065, "step": 576, "train/alpha": 0.2, "train/clean_ce_loss": 0.09130859375, "train/clean_loss_scale": 4.909139633178711, "train/clean_loss_weight": 0.9818279147148132, "train/combined_loss": 0.5378959774971008, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13649289099526066, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.44824662804603577 }, { "epoch": 1.3649289099526065, "step": 576, "train/alpha": 0.2, "train/clean_ce_loss": 0.0098876953125, "train/clean_loss_scale": 13.508514404296875, "train/clean_loss_weight": 2.701702833175659, "train/combined_loss": 0.1602816879749298, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13649289099526066, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.13356807827949524 }, { "epoch": 1.3649289099526065, "step": 576, "train/alpha": 0.2, "train/clean_ce_loss": 0.0303955078125, "train/clean_loss_scale": 6.276907444000244, "train/clean_loss_weight": 1.255381464958191, "train/combined_loss": 0.2289477437734604, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13649289099526066, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.19078978896141052 }, { "epoch": 1.3649289099526065, "step": 576, "train/alpha": 0.2, "train/clean_ce_loss": 0.047607421875, "train/clean_loss_scale": 5.786168575286865, "train/clean_loss_weight": 1.157233715057373, "train/combined_loss": 0.3305574655532837, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13649289099526066, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.27546456456184387 }, { "epoch": 1.367298578199052, "step": 577, "train/alpha": 0.2, "train/clean_ce_loss": 0.047119140625, "train/clean_loss_scale": 7.784289836883545, "train/clean_loss_weight": 1.556857943534851, "train/combined_loss": 0.44014686346054077, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.1367298578199052, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3667890429496765 }, { "epoch": 1.367298578199052, "step": 577, "train/alpha": 0.2, "train/clean_ce_loss": 0.058837890625, "train/clean_loss_scale": 5.773889541625977, "train/clean_loss_weight": 1.1547778844833374, "train/combined_loss": 0.40766817331314087, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.1367298578199052, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.33972346782684326 }, { "epoch": 1.367298578199052, "step": 577, "train/alpha": 0.2, "train/clean_ce_loss": 0.12060546875, "train/clean_loss_scale": 8.42241096496582, "train/clean_loss_weight": 1.684482216835022, "train/combined_loss": 1.2189465761184692, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.1367298578199052, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0157887935638428 }, { "epoch": 1.367298578199052, "step": 577, "train/alpha": 0.2, "train/clean_ce_loss": 0.02783203125, "train/clean_loss_scale": 5.230702877044678, "train/clean_loss_weight": 1.0461405515670776, "train/combined_loss": 0.17469729483127594, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.1367298578199052, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.14558108150959015 }, { "epoch": 1.3696682464454977, "step": 578, "train/alpha": 0.2, "train/clean_ce_loss": 0.055908203125, "train/clean_loss_scale": 7.440318584442139, "train/clean_loss_weight": 1.4880636930465698, "train/combined_loss": 0.49916982650756836, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13696682464454976, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.41597485542297363 }, { "epoch": 1.3696682464454977, "step": 578, "train/alpha": 0.2, "train/clean_ce_loss": 0.03369140625, "train/clean_loss_scale": 5.814918518066406, "train/clean_loss_weight": 1.162983775138855, "train/combined_loss": 0.2350953370332718, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13696682464454976, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.19591277837753296 }, { "epoch": 1.3696682464454977, "step": 578, "train/alpha": 0.2, "train/clean_ce_loss": 0.1455078125, "train/clean_loss_scale": 6.7694292068481445, "train/clean_loss_weight": 1.3538858890533447, "train/combined_loss": 1.1820058822631836, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13696682464454976, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9850048422813416 }, { "epoch": 1.3696682464454977, "step": 578, "train/alpha": 0.2, "train/clean_ce_loss": 0.07177734375, "train/clean_loss_scale": 6.307515621185303, "train/clean_loss_weight": 1.2615031003952026, "train/combined_loss": 0.5432840585708618, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13696682464454976, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4527367055416107 }, { "epoch": 1.3720379146919433, "step": 579, "train/alpha": 0.2, "train/clean_ce_loss": 0.169921875, "train/clean_loss_scale": 5.4904465675354, "train/clean_loss_weight": 1.098089337348938, "train/combined_loss": 1.1195363998413086, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13720379146919431, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9329469799995422 }, { "epoch": 1.3720379146919433, "step": 579, "train/alpha": 0.2, "train/clean_ce_loss": 0.140625, "train/clean_loss_scale": 6.701192855834961, "train/clean_loss_weight": 1.3402385711669922, "train/combined_loss": 1.1308262348175049, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13720379146919431, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.942355215549469 }, { "epoch": 1.3720379146919433, "step": 579, "train/alpha": 0.2, "train/clean_ce_loss": 0.0130615234375, "train/clean_loss_scale": 14.596075057983398, "train/clean_loss_weight": 2.919214963912964, "train/combined_loss": 0.22877636551856995, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13720379146919431, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1906469762325287 }, { "epoch": 1.3720379146919433, "step": 579, "train/alpha": 0.2, "train/clean_ce_loss": 0.0025634765625, "train/clean_loss_scale": 54.39183807373047, "train/clean_loss_weight": 10.87836742401123, "train/combined_loss": 0.16731864213943481, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13720379146919431, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.13943220674991608 }, { "epoch": 1.3744075829383886, "grad_norm": 31.453990936279297, "learning_rate": 2e-06, "loss": 3.4694, "step": 580 }, { "epoch": 1.3744075829383886, "step": 580, "train/alpha": 0.2, "train/clean_ce_loss": 0.006500244140625, "train/clean_loss_scale": 25.194913864135742, "train/clean_loss_weight": 5.03898286819458, "train/combined_loss": 0.19652770459651947, "train/pos_loss/pos0": 0.006787188351154327, "train/pos_loss/pos1": 0.04644171521067619, "train/pos_loss/pos2": 0.33957475423812866, "train/pos_loss/pos3": 0.939710795879364, "train/pos_loss/pos4": 1.4894827604293823, "train/pos_loss/pos5": 2.144531011581421, "train/pos_loss/pos6": 2.740591526031494, "train/pos_loss/weight0": 3.2547948360443115, "train/pos_loss/weight1": 2.0939688682556152, "train/pos_loss/weight2": 1.0669742822647095, "train/pos_loss/weight3": 0.4207506477832794, "train/pos_loss/weight4": 0.1277400404214859, "train/pos_loss/weight5": 0.02954097092151642, "train/pos_loss/weight6": 0.006230560131371021, "train/pos_weight_progress": 0.13744075829383887, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1637730896472931 }, { "epoch": 1.3744075829383886, "step": 580, "train/alpha": 0.2, "train/clean_ce_loss": 0.091796875, "train/clean_loss_scale": 9.251354217529297, "train/clean_loss_weight": 1.8502708673477173, "train/combined_loss": 1.019094467163086, "train/pos_loss/pos0": 0.28820550441741943, "train/pos_loss/pos1": 0.5751712322235107, "train/pos_loss/pos2": 0.9457605481147766, "train/pos_loss/pos3": 1.4004441499710083, "train/pos_loss/pos4": 1.8552204370498657, "train/pos_loss/pos5": 2.1477108001708984, "train/pos_loss/pos6": 2.5050876140594482, "train/pos_loss/weight0": 2.245171070098877, "train/pos_loss/weight1": 1.6955076456069946, "train/pos_loss/weight2": 1.2496275901794434, "train/pos_loss/weight3": 0.870529294013977, "train/pos_loss/weight4": 0.5146651268005371, "train/pos_loss/weight5": 0.2687875032424927, "train/pos_loss/weight6": 0.15571185946464539, "train/pos_weight_progress": 0.13744075829383887, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8492453694343567 }, { "epoch": 1.3744075829383886, "step": 580, "train/alpha": 0.2, "train/clean_ce_loss": 0.0634765625, "train/clean_loss_scale": 9.983990669250488, "train/clean_loss_weight": 1.9967981576919556, "train/combined_loss": 0.7604992985725403, "train/pos_loss/pos0": 0.20280729234218597, "train/pos_loss/pos1": 0.41254425048828125, "train/pos_loss/pos2": 0.8582987189292908, "train/pos_loss/pos3": 1.4139683246612549, "train/pos_loss/pos4": 1.7251551151275635, "train/pos_loss/pos5": 2.099912643432617, "train/pos_loss/pos6": 2.450881242752075, "train/pos_loss/weight0": 2.6324288845062256, "train/pos_loss/weight1": 1.8755638599395752, "train/pos_loss/weight2": 1.2187873125076294, "train/pos_loss/weight3": 0.7005466818809509, "train/pos_loss/weight4": 0.34648388624191284, "train/pos_loss/weight5": 0.15213847160339355, "train/pos_loss/weight6": 0.07405102998018265, "train/pos_weight_progress": 0.13744075829383887, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6337494254112244 }, { "epoch": 1.3744075829383886, "step": 580, "train/alpha": 0.2, "train/clean_ce_loss": 0.0255126953125, "train/clean_loss_scale": 22.38360023498535, "train/clean_loss_weight": 4.476720333099365, "train/combined_loss": 0.685279130935669, "train/pos_loss/pos0": 0.07579673826694489, "train/pos_loss/pos1": 0.272787481546402, "train/pos_loss/pos2": 0.6718305945396423, "train/pos_loss/pos3": 1.235242247581482, "train/pos_loss/pos4": 1.9907512664794922, "train/pos_loss/pos5": 2.340394973754883, "train/pos_loss/pos6": 2.904914379119873, "train/pos_loss/weight0": 2.48128342628479, "train/pos_loss/weight1": 1.824270486831665, "train/pos_loss/weight2": 1.2470314502716064, "train/pos_loss/weight3": 0.7486764192581177, "train/pos_loss/weight4": 0.4034506678581238, "train/pos_loss/weight5": 0.1929844617843628, "train/pos_loss/weight6": 0.10230362415313721, "train/pos_weight_progress": 0.13744075829383887, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5710659623146057 }, { "epoch": 1.3767772511848342, "step": 581, "train/alpha": 0.2, "train/clean_ce_loss": 0.06298828125, "train/clean_loss_scale": 8.100852012634277, "train/clean_loss_weight": 1.6201704740524292, "train/combined_loss": 0.6123104691505432, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13767772511848342, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5102587342262268 }, { "epoch": 1.3767772511848342, "step": 581, "train/alpha": 0.2, "train/clean_ce_loss": 0.06884765625, "train/clean_loss_scale": 8.716370582580566, "train/clean_loss_weight": 1.7432740926742554, "train/combined_loss": 0.7201220393180847, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13767772511848342, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6001017093658447 }, { "epoch": 1.3767772511848342, "step": 581, "train/alpha": 0.2, "train/clean_ce_loss": 0.000949859619140625, "train/clean_loss_scale": 260.16357421875, "train/clean_loss_weight": 52.03271484375, "train/combined_loss": 0.2965426445007324, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13767772511848342, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.24711886048316956 }, { "epoch": 1.3767772511848342, "step": 581, "train/alpha": 0.2, "train/clean_ce_loss": 0.1572265625, "train/clean_loss_scale": 7.202874660491943, "train/clean_loss_weight": 1.4405750036239624, "train/combined_loss": 1.3589799404144287, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13767772511848342, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1324832439422607 }, { "epoch": 1.3791469194312795, "step": 582, "train/alpha": 0.2, "train/clean_ce_loss": 0.1640625, "train/clean_loss_scale": 7.933191776275635, "train/clean_loss_weight": 1.586638331413269, "train/combined_loss": 1.56184720993042, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13791469194312797, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3015393018722534 }, { "epoch": 1.3791469194312795, "step": 582, "train/alpha": 0.2, "train/clean_ce_loss": 0.06982421875, "train/clean_loss_scale": 7.351502895355225, "train/clean_loss_weight": 1.470300555229187, "train/combined_loss": 0.6159754991531372, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13791469194312797, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5133129358291626 }, { "epoch": 1.3791469194312795, "step": 582, "train/alpha": 0.2, "train/clean_ce_loss": 0.1572265625, "train/clean_loss_scale": 8.65750503540039, "train/clean_loss_weight": 1.7315009832382202, "train/combined_loss": 1.633427619934082, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13791469194312797, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3611897230148315 }, { "epoch": 1.3791469194312795, "step": 582, "train/alpha": 0.2, "train/clean_ce_loss": 0.1298828125, "train/clean_loss_scale": 11.052929878234863, "train/clean_loss_weight": 2.2105860710144043, "train/combined_loss": 1.7227027416229248, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13791469194312797, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.435585618019104 }, { "epoch": 1.3815165876777251, "step": 583, "train/alpha": 0.2, "train/clean_ce_loss": 0.0030517578125, "train/clean_loss_scale": 42.22575378417969, "train/clean_loss_weight": 8.445151329040527, "train/combined_loss": 0.15463532507419586, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13815165876777252, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.12886276841163635 }, { "epoch": 1.3815165876777251, "step": 583, "train/alpha": 0.2, "train/clean_ce_loss": 0.03759765625, "train/clean_loss_scale": 9.97179126739502, "train/clean_loss_weight": 1.9943583011627197, "train/combined_loss": 0.44989919662475586, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13815165876777252, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3749159872531891 }, { "epoch": 1.3815165876777251, "step": 583, "train/alpha": 0.2, "train/clean_ce_loss": 0.05908203125, "train/clean_loss_scale": 18.68802261352539, "train/clean_loss_weight": 3.7376046180725098, "train/combined_loss": 1.3249516487121582, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13815165876777252, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1041263341903687 }, { "epoch": 1.3815165876777251, "step": 583, "train/alpha": 0.2, "train/clean_ce_loss": 0.11279296875, "train/clean_loss_scale": 9.39603328704834, "train/clean_loss_weight": 1.879206657409668, "train/combined_loss": 1.2717677354812622, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13815165876777252, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0598064661026 }, { "epoch": 1.3838862559241707, "step": 584, "train/alpha": 0.2, "train/clean_ce_loss": 0.0693359375, "train/clean_loss_scale": 10.36107063293457, "train/clean_loss_weight": 2.072214126586914, "train/combined_loss": 0.8620734810829163, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13838862559241707, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7183945775032043 }, { "epoch": 1.3838862559241707, "step": 584, "train/alpha": 0.2, "train/clean_ce_loss": 0.08984375, "train/clean_loss_scale": 8.23725414276123, "train/clean_loss_weight": 1.6474508047103882, "train/combined_loss": 0.8880789875984192, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13838862559241707, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7400658130645752 }, { "epoch": 1.3838862559241707, "step": 584, "train/alpha": 0.2, "train/clean_ce_loss": 0.111328125, "train/clean_loss_scale": 7.697615146636963, "train/clean_loss_weight": 1.5395230054855347, "train/combined_loss": 1.028353214263916, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13838862559241707, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8569610714912415 }, { "epoch": 1.3838862559241707, "step": 584, "train/alpha": 0.2, "train/clean_ce_loss": 0.10205078125, "train/clean_loss_scale": 7.349565029144287, "train/clean_loss_weight": 1.4699130058288574, "train/combined_loss": 0.9000346064567566, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13838862559241707, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7500288486480713 }, { "epoch": 1.386255924170616, "grad_norm": 14.979560852050781, "learning_rate": 2e-06, "loss": 3.2452, "step": 585 }, { "epoch": 1.386255924170616, "step": 585, "train/alpha": 0.2, "train/clean_ce_loss": 0.14453125, "train/clean_loss_scale": 7.630733966827393, "train/clean_loss_weight": 1.5261467695236206, "train/combined_loss": 1.3234554529190063, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.1386255924170616, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.102879524230957 }, { "epoch": 1.386255924170616, "step": 585, "train/alpha": 0.2, "train/clean_ce_loss": 0.0849609375, "train/clean_loss_scale": 7.727585315704346, "train/clean_loss_weight": 1.545517086982727, "train/combined_loss": 0.7878514528274536, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.1386255924170616, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6565428972244263 }, { "epoch": 1.386255924170616, "step": 585, "train/alpha": 0.2, "train/clean_ce_loss": 0.01092529296875, "train/clean_loss_scale": 26.788352966308594, "train/clean_loss_weight": 5.357670783996582, "train/combined_loss": 0.3512047231197357, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.1386255924170616, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2926706075668335 }, { "epoch": 1.386255924170616, "step": 585, "train/alpha": 0.2, "train/clean_ce_loss": 0.05859375, "train/clean_loss_scale": 8.952241897583008, "train/clean_loss_weight": 1.7904484272003174, "train/combined_loss": 0.629454493522644, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.1386255924170616, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.524545431137085 }, { "epoch": 1.3886255924170616, "step": 586, "train/alpha": 0.2, "train/clean_ce_loss": 0.052978515625, "train/clean_loss_scale": 13.153095245361328, "train/clean_loss_weight": 2.6306190490722656, "train/combined_loss": 0.8361977338790894, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13886255924170615, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.696831464767456 }, { "epoch": 1.3886255924170616, "step": 586, "train/alpha": 0.2, "train/clean_ce_loss": 0.047119140625, "train/clean_loss_scale": 12.496020317077637, "train/clean_loss_weight": 2.499204158782959, "train/combined_loss": 0.7065621018409729, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13886255924170615, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5888017416000366 }, { "epoch": 1.3886255924170616, "step": 586, "train/alpha": 0.2, "train/clean_ce_loss": 0.17578125, "train/clean_loss_scale": 6.602348327636719, "train/clean_loss_weight": 1.3204697370529175, "train/combined_loss": 1.3926829099655151, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13886255924170615, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1605690717697144 }, { "epoch": 1.3886255924170616, "step": 586, "train/alpha": 0.2, "train/clean_ce_loss": 0.03125, "train/clean_loss_scale": 11.307740211486816, "train/clean_loss_weight": 2.2615480422973633, "train/combined_loss": 0.4240402579307556, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13886255924170615, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.353366881608963 }, { "epoch": 1.390995260663507, "step": 587, "train/alpha": 0.2, "train/clean_ce_loss": 0.10302734375, "train/clean_loss_scale": 8.568888664245605, "train/clean_loss_weight": 1.713777780532837, "train/combined_loss": 1.0593957901000977, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.1390995260663507, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8828298449516296 }, { "epoch": 1.390995260663507, "step": 587, "train/alpha": 0.2, "train/clean_ce_loss": 0.1259765625, "train/clean_loss_scale": 10.88835620880127, "train/clean_loss_weight": 2.177671194076538, "train/combined_loss": 1.6460131406784058, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.1390995260663507, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3716776371002197 }, { "epoch": 1.390995260663507, "step": 587, "train/alpha": 0.2, "train/clean_ce_loss": 0.053466796875, "train/clean_loss_scale": 22.048797607421875, "train/clean_loss_weight": 4.409759521484375, "train/combined_loss": 1.41465425491333, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.1390995260663507, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1788785457611084 }, { "epoch": 1.390995260663507, "step": 587, "train/alpha": 0.2, "train/clean_ce_loss": 0.1845703125, "train/clean_loss_scale": 6.848836421966553, "train/clean_loss_weight": 1.3697673082351685, "train/combined_loss": 1.516910195350647, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.1390995260663507, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2640918493270874 }, { "epoch": 1.3933649289099526, "step": 588, "train/alpha": 0.2, "train/clean_ce_loss": 0.12890625, "train/clean_loss_scale": 7.159319877624512, "train/clean_loss_weight": 1.4318640232086182, "train/combined_loss": 1.1074572801589966, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13933649289099526, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9228810667991638 }, { "epoch": 1.3933649289099526, "step": 588, "train/alpha": 0.2, "train/clean_ce_loss": 0.12451171875, "train/clean_loss_scale": 9.852509498596191, "train/clean_loss_weight": 1.9705018997192383, "train/combined_loss": 1.472103476524353, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13933649289099526, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2267528772354126 }, { "epoch": 1.3933649289099526, "step": 588, "train/alpha": 0.2, "train/clean_ce_loss": 0.056640625, "train/clean_loss_scale": 7.8926472663879395, "train/clean_loss_weight": 1.5785294771194458, "train/combined_loss": 0.536453366279602, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13933649289099526, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4470444619655609 }, { "epoch": 1.3933649289099526, "step": 588, "train/alpha": 0.2, "train/clean_ce_loss": 0.025634765625, "train/clean_loss_scale": 20.68729591369629, "train/clean_loss_weight": 4.1374592781066895, "train/combined_loss": 0.6363767385482788, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13933649289099526, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5303139686584473 }, { "epoch": 1.3957345971563981, "step": 589, "train/alpha": 0.2, "train/clean_ce_loss": 0.1689453125, "train/clean_loss_scale": 6.480576515197754, "train/clean_loss_weight": 1.2961152791976929, "train/combined_loss": 1.313835620880127, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.1395734597156398, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0948630571365356 }, { "epoch": 1.3957345971563981, "step": 589, "train/alpha": 0.2, "train/clean_ce_loss": 0.0260009765625, "train/clean_loss_scale": 12.075773239135742, "train/clean_loss_weight": 2.4151546955108643, "train/combined_loss": 0.3767782747745514, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.1395734597156398, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.31398189067840576 }, { "epoch": 1.3957345971563981, "step": 589, "train/alpha": 0.2, "train/clean_ce_loss": 0.01226806640625, "train/clean_loss_scale": 28.470617294311523, "train/clean_loss_weight": 5.6941237449646, "train/combined_loss": 0.41913533210754395, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.1395734597156398, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3492794334888458 }, { "epoch": 1.3957345971563981, "step": 589, "train/alpha": 0.2, "train/clean_ce_loss": 0.0771484375, "train/clean_loss_scale": 8.944147109985352, "train/clean_loss_weight": 1.7888294458389282, "train/combined_loss": 0.828032374382019, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.1395734597156398, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6900269985198975 }, { "epoch": 1.3981042654028437, "grad_norm": 32.64021682739258, "learning_rate": 2e-06, "loss": 3.4304, "step": 590 }, { "epoch": 1.3981042654028437, "step": 590, "train/alpha": 0.2, "train/clean_ce_loss": 0.14453125, "train/clean_loss_scale": 7.939741611480713, "train/clean_loss_weight": 1.5879483222961426, "train/combined_loss": 1.3770489692687988, "train/pos_loss/pos0": 0.4086136519908905, "train/pos_loss/pos1": 0.9274575710296631, "train/pos_loss/pos2": 1.5367482900619507, "train/pos_loss/pos3": 2.1581368446350098, "train/pos_loss/pos4": 2.543322801589966, "train/pos_loss/pos5": 2.8557019233703613, "train/pos_loss/pos6": 3.377009630203247, "train/pos_loss/weight0": 2.4923095703125, "train/pos_loss/weight1": 1.8411895036697388, "train/pos_loss/weight2": 1.2773164510726929, "train/pos_loss/weight3": 0.7631439566612244, "train/pos_loss/weight4": 0.38869011402130127, "train/pos_loss/weight5": 0.15768808126449585, "train/pos_loss/weight6": 0.07966256886720657, "train/pos_weight_progress": 0.13981042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.147540807723999 }, { "epoch": 1.3981042654028437, "step": 590, "train/alpha": 0.2, "train/clean_ce_loss": 0.004425048828125, "train/clean_loss_scale": 39.208595275878906, "train/clean_loss_weight": 7.841719150543213, "train/combined_loss": 0.20819994807243347, "train/pos_loss/pos0": 0.014689525589346886, "train/pos_loss/pos1": 0.04566394165158272, "train/pos_loss/pos2": 0.3338209390640259, "train/pos_loss/pos3": 0.7695422172546387, "train/pos_loss/pos4": 1.5241904258728027, "train/pos_loss/pos5": 2.0829458236694336, "train/pos_loss/pos6": 2.5533173084259033, "train/pos_loss/weight0": 3.1636292934417725, "train/pos_loss/weight1": 2.0577595233917236, "train/pos_loss/weight2": 1.1066843271255493, "train/pos_loss/weight3": 0.4672347903251648, "train/pos_loss/weight4": 0.1557106077671051, "train/pos_loss/weight5": 0.03802874684333801, "train/pos_loss/weight6": 0.010953349992632866, "train/pos_weight_progress": 0.13981042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1734999567270279 }, { "epoch": 1.3981042654028437, "step": 590, "train/alpha": 0.2, "train/clean_ce_loss": 0.0380859375, "train/clean_loss_scale": 14.255949974060059, "train/clean_loss_weight": 2.8511900901794434, "train/combined_loss": 0.6515414714813232, "train/pos_loss/pos0": 0.051135506480932236, "train/pos_loss/pos1": 0.21401754021644592, "train/pos_loss/pos2": 0.561500072479248, "train/pos_loss/pos3": 1.0061753988265991, "train/pos_loss/pos4": 1.431538462638855, "train/pos_loss/pos5": 1.977620005607605, "train/pos_loss/pos6": 2.661790609359741, "train/pos_loss/weight0": 2.2307610511779785, "train/pos_loss/weight1": 1.6744502782821655, "train/pos_loss/weight2": 1.2511252164840698, "train/pos_loss/weight3": 0.8652478456497192, "train/pos_loss/weight4": 0.5374264717102051, "train/pos_loss/weight5": 0.27476534247398376, "train/pos_loss/weight6": 0.16622363030910492, "train/pos_weight_progress": 0.13981042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.542951226234436 }, { "epoch": 1.3981042654028437, "step": 590, "train/alpha": 0.2, "train/clean_ce_loss": 0.1591796875, "train/clean_loss_scale": 8.221999168395996, "train/clean_loss_weight": 1.644399881362915, "train/combined_loss": 1.5705301761627197, "train/pos_loss/pos0": 0.4546390175819397, "train/pos_loss/pos1": 1.0018203258514404, "train/pos_loss/pos2": 1.5720194578170776, "train/pos_loss/pos3": 2.073415994644165, "train/pos_loss/pos4": 2.593832015991211, "train/pos_loss/pos5": 3.0887813568115234, "train/pos_loss/pos6": 3.322597026824951, "train/pos_loss/weight0": 2.2365522384643555, "train/pos_loss/weight1": 1.7280099391937256, "train/pos_loss/weight2": 1.24619722366333, "train/pos_loss/weight3": 0.8466566801071167, "train/pos_loss/weight4": 0.4960379898548126, "train/pos_loss/weight5": 0.2805103361606598, "train/pos_loss/weight6": 0.1660352200269699, "train/pos_weight_progress": 0.13981042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3087751865386963 }, { "epoch": 1.400473933649289, "step": 591, "train/alpha": 0.2, "train/clean_ce_loss": 0.060791015625, "train/clean_loss_scale": 6.606433868408203, "train/clean_loss_weight": 1.3212867975234985, "train/combined_loss": 0.48193418979644775, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.1400473933649289, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4016118347644806 }, { "epoch": 1.400473933649289, "step": 591, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 6.587322235107422, "train/clean_loss_weight": 1.3174644708633423, "train/combined_loss": 0.6252809762954712, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.1400473933649289, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5210675001144409 }, { "epoch": 1.400473933649289, "step": 591, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 6.562692642211914, "train/clean_loss_weight": 1.312538504600525, "train/combined_loss": 0.7306122779846191, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.1400473933649289, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6088435649871826 }, { "epoch": 1.400473933649289, "step": 591, "train/alpha": 0.2, "train/clean_ce_loss": 0.1611328125, "train/clean_loss_scale": 5.702715873718262, "train/clean_loss_weight": 1.1405432224273682, "train/combined_loss": 1.1026735305786133, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.1400473933649289, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9188946485519409 }, { "epoch": 1.4028436018957346, "step": 592, "train/alpha": 0.2, "train/clean_ce_loss": 0.0184326171875, "train/clean_loss_scale": 14.451873779296875, "train/clean_loss_weight": 2.8903748989105225, "train/combined_loss": 0.31966301798820496, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14028436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.26638585329055786 }, { "epoch": 1.4028436018957346, "step": 592, "train/alpha": 0.2, "train/clean_ce_loss": 0.0159912109375, "train/clean_loss_scale": 17.9732723236084, "train/clean_loss_weight": 3.5946545600891113, "train/combined_loss": 0.3448972702026367, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14028436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.28741440176963806 }, { "epoch": 1.4028436018957346, "step": 592, "train/alpha": 0.2, "train/clean_ce_loss": 0.130859375, "train/clean_loss_scale": 6.271546840667725, "train/clean_loss_weight": 1.2543094158172607, "train/combined_loss": 0.9848288297653198, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14028436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8206906914710999 }, { "epoch": 1.4028436018957346, "step": 592, "train/alpha": 0.2, "train/clean_ce_loss": 0.1982421875, "train/clean_loss_scale": 5.940616607666016, "train/clean_loss_weight": 1.188123345375061, "train/combined_loss": 1.4132170677185059, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14028436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1776808500289917 }, { "epoch": 1.40521327014218, "step": 593, "train/alpha": 0.2, "train/clean_ce_loss": 0.054443359375, "train/clean_loss_scale": 12.960783004760742, "train/clean_loss_weight": 2.5921566486358643, "train/combined_loss": 0.8467543125152588, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14052132701421802, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7056285738945007 }, { "epoch": 1.40521327014218, "step": 593, "train/alpha": 0.2, "train/clean_ce_loss": 0.158203125, "train/clean_loss_scale": 7.661300182342529, "train/clean_loss_weight": 1.5322600603103638, "train/combined_loss": 1.4544498920440674, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14052132701421802, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2120416164398193 }, { "epoch": 1.40521327014218, "step": 593, "train/alpha": 0.2, "train/clean_ce_loss": 0.1337890625, "train/clean_loss_scale": 7.512524127960205, "train/clean_loss_weight": 1.502504825592041, "train/combined_loss": 1.206112265586853, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14052132701421802, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0050935745239258 }, { "epoch": 1.40521327014218, "step": 593, "train/alpha": 0.2, "train/clean_ce_loss": 0.1064453125, "train/clean_loss_scale": 9.007993698120117, "train/clean_loss_weight": 1.8015987873077393, "train/combined_loss": 1.1506304740905762, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14052132701421802, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9588587284088135 }, { "epoch": 1.4075829383886256, "step": 594, "train/alpha": 0.2, "train/clean_ce_loss": 0.07666015625, "train/clean_loss_scale": 6.924286842346191, "train/clean_loss_weight": 1.384857416152954, "train/combined_loss": 0.6369802951812744, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14075829383886257, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.530816912651062 }, { "epoch": 1.4075829383886256, "step": 594, "train/alpha": 0.2, "train/clean_ce_loss": 0.08935546875, "train/clean_loss_scale": 7.266866207122803, "train/clean_loss_weight": 1.4533733129501343, "train/combined_loss": 0.779201090335846, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14075829383886257, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6493342518806458 }, { "epoch": 1.4075829383886256, "step": 594, "train/alpha": 0.2, "train/clean_ce_loss": 0.015625, "train/clean_loss_scale": 19.523170471191406, "train/clean_loss_weight": 3.9046342372894287, "train/combined_loss": 0.36605945229530334, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14075829383886257, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3050495386123657 }, { "epoch": 1.4075829383886256, "step": 594, "train/alpha": 0.2, "train/clean_ce_loss": 0.0703125, "train/clean_loss_scale": 13.817839622497559, "train/clean_loss_weight": 2.7635679244995117, "train/combined_loss": 1.1658802032470703, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14075829383886257, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9715668559074402 }, { "epoch": 1.4099526066350712, "grad_norm": 15.847481727600098, "learning_rate": 2e-06, "loss": 3.3829, "step": 595 }, { "epoch": 1.4099526066350712, "step": 595, "train/alpha": 0.2, "train/clean_ce_loss": 0.08642578125, "train/clean_loss_scale": 8.025339126586914, "train/clean_loss_weight": 1.6050678491592407, "train/combined_loss": 0.8323154449462891, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14099526066350712, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6935961842536926 }, { "epoch": 1.4099526066350712, "step": 595, "train/alpha": 0.2, "train/clean_ce_loss": 0.041748046875, "train/clean_loss_scale": 5.899890899658203, "train/clean_loss_weight": 1.1799782514572144, "train/combined_loss": 0.2955707013607025, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14099526066350712, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.24630892276763916 }, { "epoch": 1.4099526066350712, "step": 595, "train/alpha": 0.2, "train/clean_ce_loss": 0.0859375, "train/clean_loss_scale": 7.571251392364502, "train/clean_loss_weight": 1.5142502784729004, "train/combined_loss": 0.780785322189331, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14099526066350712, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6506544351577759 }, { "epoch": 1.4099526066350712, "step": 595, "train/alpha": 0.2, "train/clean_ce_loss": 0.041748046875, "train/clean_loss_scale": 5.899890899658203, "train/clean_loss_weight": 1.1799782514572144, "train/combined_loss": 0.2955707013607025, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14099526066350712, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.24630892276763916 }, { "epoch": 1.4123222748815165, "step": 596, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 9.150315284729004, "train/clean_loss_weight": 1.8300631046295166, "train/combined_loss": 0.7077197432518005, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14123222748815165, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5897664427757263 }, { "epoch": 1.4123222748815165, "step": 596, "train/alpha": 0.2, "train/clean_ce_loss": 0.11376953125, "train/clean_loss_scale": 7.969581604003906, "train/clean_loss_weight": 1.5939162969589233, "train/combined_loss": 1.0880346298217773, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14123222748815165, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.906695544719696 }, { "epoch": 1.4123222748815165, "step": 596, "train/alpha": 0.2, "train/clean_ce_loss": 0.06884765625, "train/clean_loss_scale": 6.824905872344971, "train/clean_loss_weight": 1.3649811744689941, "train/combined_loss": 0.5638545155525208, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14123222748815165, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46987876296043396 }, { "epoch": 1.4123222748815165, "step": 596, "train/alpha": 0.2, "train/clean_ce_loss": 0.0081787109375, "train/clean_loss_scale": 22.570995330810547, "train/clean_loss_weight": 4.514199256896973, "train/combined_loss": 0.22152197360992432, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14123222748815165, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.18460164964199066 }, { "epoch": 1.414691943127962, "step": 597, "train/alpha": 0.2, "train/clean_ce_loss": 0.050537109375, "train/clean_loss_scale": 7.8036580085754395, "train/clean_loss_weight": 1.5607316493988037, "train/combined_loss": 0.4732491672039032, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.1414691943127962, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3943743109703064 }, { "epoch": 1.414691943127962, "step": 597, "train/alpha": 0.2, "train/clean_ce_loss": 0.03564453125, "train/clean_loss_scale": 7.991901397705078, "train/clean_loss_weight": 1.5983803272247314, "train/combined_loss": 0.34184110164642334, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.1414691943127962, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2848675847053528 }, { "epoch": 1.414691943127962, "step": 597, "train/alpha": 0.2, "train/clean_ce_loss": 0.0036773681640625, "train/clean_loss_scale": 55.38365936279297, "train/clean_loss_weight": 11.07673168182373, "train/combined_loss": 0.2443993240594864, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.1414691943127962, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2036661058664322 }, { "epoch": 1.414691943127962, "step": 597, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 8.1859712600708, "train/clean_loss_weight": 1.637194275856018, "train/combined_loss": 0.801010012626648, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.1414691943127962, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6675083637237549 }, { "epoch": 1.4170616113744074, "step": 598, "train/alpha": 0.2, "train/clean_ce_loss": 0.00225830078125, "train/clean_loss_scale": 84.3320083618164, "train/clean_loss_weight": 16.86640167236328, "train/combined_loss": 0.2285364419221878, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14170616113744075, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.19044703245162964 }, { "epoch": 1.4170616113744074, "step": 598, "train/alpha": 0.2, "train/clean_ce_loss": 0.103515625, "train/clean_loss_scale": 8.554414749145508, "train/clean_loss_weight": 1.7108830213546753, "train/combined_loss": 1.0626187324523926, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14170616113744075, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.885515570640564 }, { "epoch": 1.4170616113744074, "step": 598, "train/alpha": 0.2, "train/clean_ce_loss": 0.10595703125, "train/clean_loss_scale": 5.505550384521484, "train/clean_loss_weight": 1.1011101007461548, "train/combined_loss": 0.7000221610069275, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14170616113744075, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5833517909049988 }, { "epoch": 1.4170616113744074, "step": 598, "train/alpha": 0.2, "train/clean_ce_loss": 0.061767578125, "train/clean_loss_scale": 7.544496536254883, "train/clean_loss_weight": 1.5088993310928345, "train/combined_loss": 0.5592063069343567, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14170616113744075, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46600526571273804 }, { "epoch": 1.419431279620853, "step": 599, "train/alpha": 0.2, "train/clean_ce_loss": 0.095703125, "train/clean_loss_scale": 7.977075576782227, "train/clean_loss_weight": 1.5954151153564453, "train/combined_loss": 0.9161173105239868, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.1419431279620853, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7634310722351074 }, { "epoch": 1.419431279620853, "step": 599, "train/alpha": 0.2, "train/clean_ce_loss": 0.00762939453125, "train/clean_loss_scale": 24.40700340270996, "train/clean_loss_weight": 4.8814005851745605, "train/combined_loss": 0.22345279157161713, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.1419431279620853, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.18621066212654114 }, { "epoch": 1.419431279620853, "step": 599, "train/alpha": 0.2, "train/clean_ce_loss": 0.07861328125, "train/clean_loss_scale": 8.480247497558594, "train/clean_loss_weight": 1.6960495710372925, "train/combined_loss": 0.7999920845031738, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.1419431279620853, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6666600704193115 }, { "epoch": 1.419431279620853, "step": 599, "train/alpha": 0.2, "train/clean_ce_loss": 0.0830078125, "train/clean_loss_scale": 5.690300464630127, "train/clean_loss_weight": 1.1380600929260254, "train/combined_loss": 0.5668072700500488, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.1419431279620853, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.472339391708374 }, { "epoch": 1.4218009478672986, "grad_norm": 14.021217346191406, "learning_rate": 2e-06, "loss": 3.119, "step": 600 }, { "epoch": 1.4218009478672986, "step": 600, "train/alpha": 0.2, "train/clean_ce_loss": 0.09619140625, "train/clean_loss_scale": 7.158542633056641, "train/clean_loss_weight": 1.431708574295044, "train/combined_loss": 0.8263083696365356, "train/pos_loss/pos0": 0.32644185423851013, "train/pos_loss/pos1": 0.6693258881568909, "train/pos_loss/pos2": 0.9845273494720459, "train/pos_loss/pos3": 1.5734461545944214, "train/pos_loss/pos4": 1.9273899793624878, "train/pos_loss/pos5": 2.630220890045166, "train/pos_loss/pos6": 2.802990674972534, "train/pos_loss/weight0": 3.083913564682007, "train/pos_loss/weight1": 2.010641574859619, "train/pos_loss/weight2": 1.1409791707992554, "train/pos_loss/weight3": 0.5035961270332336, "train/pos_loss/weight4": 0.18969468772411346, "train/pos_loss/weight5": 0.05494847893714905, "train/pos_loss/weight6": 0.016225989907979965, "train/pos_weight_progress": 0.14218009478672985, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6885902881622314 }, { "epoch": 1.4218009478672986, "step": 600, "train/alpha": 0.2, "train/clean_ce_loss": 0.08544921875, "train/clean_loss_scale": 11.747551918029785, "train/clean_loss_weight": 2.349510431289673, "train/combined_loss": 1.204582929611206, "train/pos_loss/pos0": 0.3098156154155731, "train/pos_loss/pos1": 0.7339969277381897, "train/pos_loss/pos2": 1.4804415702819824, "train/pos_loss/pos3": 2.1515910625457764, "train/pos_loss/pos4": 2.98626446723938, "train/pos_loss/pos5": 3.5020411014556885, "train/pos_loss/pos6": 3.840935707092285, "train/pos_loss/weight0": 2.733504056930542, "train/pos_loss/weight1": 1.8975939750671387, "train/pos_loss/weight2": 1.2191452980041504, "train/pos_loss/weight3": 0.665472686290741, "train/pos_loss/weight4": 0.3067653179168701, "train/pos_loss/weight5": 0.12657703459262848, "train/pos_loss/weight6": 0.05094214528799057, "train/pos_weight_progress": 0.14218009478672985, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0038191080093384 }, { "epoch": 1.4218009478672986, "step": 600, "train/alpha": 0.2, "train/clean_ce_loss": 0.10009765625, "train/clean_loss_scale": 6.505290508270264, "train/clean_loss_weight": 1.3010581731796265, "train/combined_loss": 0.7813972234725952, "train/pos_loss/pos0": 0.3355507552623749, "train/pos_loss/pos1": 0.6712016463279724, "train/pos_loss/pos2": 0.9504821300506592, "train/pos_loss/pos3": 1.3472470045089722, "train/pos_loss/pos4": 1.5876132249832153, "train/pos_loss/pos5": 1.9022411108016968, "train/pos_loss/pos6": 2.231718063354492, "train/pos_loss/weight0": 3.0531249046325684, "train/pos_loss/weight1": 2.0514822006225586, "train/pos_loss/weight2": 1.1646015644073486, "train/pos_loss/weight3": 0.5224928855895996, "train/pos_loss/weight4": 0.1661905199289322, "train/pos_loss/weight5": 0.03374912589788437, "train/pos_loss/weight6": 0.008359109051525593, "train/pos_weight_progress": 0.14218009478672985, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6511643528938293 }, { "epoch": 1.4218009478672986, "step": 600, "train/alpha": 0.2, "train/clean_ce_loss": 0.0703125, "train/clean_loss_scale": 8.491259574890137, "train/clean_loss_weight": 1.6982519626617432, "train/combined_loss": 0.7164499759674072, "train/pos_loss/pos0": 0.21072129905223846, "train/pos_loss/pos1": 0.47776955366134644, "train/pos_loss/pos2": 0.7007802724838257, "train/pos_loss/pos3": 1.0241107940673828, "train/pos_loss/pos4": 1.414336919784546, "train/pos_loss/pos5": 1.680373191833496, "train/pos_loss/pos6": 2.0755207538604736, "train/pos_loss/weight0": 2.4937891960144043, "train/pos_loss/weight1": 1.8151218891143799, "train/pos_loss/weight2": 1.2124913930892944, "train/pos_loss/weight3": 0.7406747341156006, "train/pos_loss/weight4": 0.4045822322368622, "train/pos_loss/weight5": 0.21542273461818695, "train/pos_loss/weight6": 0.11791786551475525, "train/pos_weight_progress": 0.14218009478672985, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.597041666507721 }, { "epoch": 1.4241706161137442, "step": 601, "train/alpha": 0.2, "train/clean_ce_loss": 0.00762939453125, "train/clean_loss_scale": 40.2019157409668, "train/clean_loss_weight": 8.040383338928223, "train/combined_loss": 0.36805951595306396, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.1424170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.30671626329421997 }, { "epoch": 1.4241706161137442, "step": 601, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 16.871118545532227, "train/clean_loss_weight": 3.3742237091064453, "train/combined_loss": 1.3246464729309082, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.1424170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1038720607757568 }, { "epoch": 1.4241706161137442, "step": 601, "train/alpha": 0.2, "train/clean_ce_loss": 0.09326171875, "train/clean_loss_scale": 7.481657028198242, "train/clean_loss_weight": 1.4963314533233643, "train/combined_loss": 0.8373026251792908, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.1424170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6977521777153015 }, { "epoch": 1.4241706161137442, "step": 601, "train/alpha": 0.2, "train/clean_ce_loss": 0.1142578125, "train/clean_loss_scale": 14.01302433013916, "train/clean_loss_weight": 2.802604913711548, "train/combined_loss": 1.9213169813156128, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.1424170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.6010974645614624 }, { "epoch": 1.4265402843601895, "step": 602, "train/alpha": 0.2, "train/clean_ce_loss": 0.115234375, "train/clean_loss_scale": 9.521156311035156, "train/clean_loss_weight": 1.904231309890747, "train/combined_loss": 1.3165974617004395, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.14265402843601896, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.097164511680603 }, { "epoch": 1.4265402843601895, "step": 602, "train/alpha": 0.2, "train/clean_ce_loss": 0.1396484375, "train/clean_loss_scale": 10.300517082214355, "train/clean_loss_weight": 2.060103416442871, "train/combined_loss": 1.7261414527893066, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.14265402843601896, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4384511709213257 }, { "epoch": 1.4265402843601895, "step": 602, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 8.828277587890625, "train/clean_loss_weight": 1.765655517578125, "train/combined_loss": 0.8379967212677002, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.14265402843601896, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6983305811882019 }, { "epoch": 1.4265402843601895, "step": 602, "train/alpha": 0.2, "train/clean_ce_loss": 0.002716064453125, "train/clean_loss_scale": 160.79388427734375, "train/clean_loss_weight": 32.15877914428711, "train/combined_loss": 0.5240718722343445, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.14265402843601896, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43672657012939453 }, { "epoch": 1.4289099526066351, "step": 603, "train/alpha": 0.2, "train/clean_ce_loss": 0.0849609375, "train/clean_loss_scale": 9.869305610656738, "train/clean_loss_weight": 1.9738610982894897, "train/combined_loss": 1.0062065124511719, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.1428909952606635, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8385054469108582 }, { "epoch": 1.4289099526066351, "step": 603, "train/alpha": 0.2, "train/clean_ce_loss": 0.014404296875, "train/clean_loss_scale": 22.216793060302734, "train/clean_loss_weight": 4.443358898162842, "train/combined_loss": 0.38402074575424194, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.1428909952606635, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3200172781944275 }, { "epoch": 1.4289099526066351, "step": 603, "train/alpha": 0.2, "train/clean_ce_loss": 0.040283203125, "train/clean_loss_scale": 13.50031852722168, "train/clean_loss_weight": 2.700063705444336, "train/combined_loss": 0.652603268623352, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.1428909952606635, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5438360571861267 }, { "epoch": 1.4289099526066351, "step": 603, "train/alpha": 0.2, "train/clean_ce_loss": 0.05810546875, "train/clean_loss_scale": 7.3759989738464355, "train/clean_loss_weight": 1.475199818611145, "train/combined_loss": 0.5143030881881714, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.1428909952606635, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.42858588695526123 }, { "epoch": 1.4312796208530805, "step": 604, "train/alpha": 0.2, "train/clean_ce_loss": 0.083984375, "train/clean_loss_scale": 8.764305114746094, "train/clean_loss_weight": 1.7528610229492188, "train/combined_loss": 0.8832775950431824, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.14312796208530806, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7360646724700928 }, { "epoch": 1.4312796208530805, "step": 604, "train/alpha": 0.2, "train/clean_ce_loss": 0.12353515625, "train/clean_loss_scale": 9.702683448791504, "train/clean_loss_weight": 1.9405367374420166, "train/combined_loss": 1.4383469820022583, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.14312796208530806, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.198622465133667 }, { "epoch": 1.4312796208530805, "step": 604, "train/alpha": 0.2, "train/clean_ce_loss": 0.091796875, "train/clean_loss_scale": 13.372984886169434, "train/clean_loss_weight": 2.6745970249176025, "train/combined_loss": 1.4731178283691406, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.14312796208530806, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2275981903076172 }, { "epoch": 1.4312796208530805, "step": 604, "train/alpha": 0.2, "train/clean_ce_loss": 0.0157470703125, "train/clean_loss_scale": 22.898212432861328, "train/clean_loss_weight": 4.5796427726745605, "train/combined_loss": 0.4326957166194916, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.14312796208530806, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3605797588825226 }, { "epoch": 1.433649289099526, "grad_norm": 18.00126838684082, "learning_rate": 2e-06, "loss": 3.4104, "step": 605 }, { "epoch": 1.433649289099526, "step": 605, "train/alpha": 0.2, "train/clean_ce_loss": 0.04931640625, "train/clean_loss_scale": 9.542311668395996, "train/clean_loss_weight": 1.908462405204773, "train/combined_loss": 0.5647110342979431, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.14336492890995262, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.47059252858161926 }, { "epoch": 1.433649289099526, "step": 605, "train/alpha": 0.2, "train/clean_ce_loss": 0.09423828125, "train/clean_loss_scale": 9.614757537841797, "train/clean_loss_weight": 1.922951579093933, "train/combined_loss": 1.0872938632965088, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.14336492890995262, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9060782194137573 }, { "epoch": 1.433649289099526, "step": 605, "train/alpha": 0.2, "train/clean_ce_loss": 0.053955078125, "train/clean_loss_scale": 10.467514038085938, "train/clean_loss_weight": 2.0935027599334717, "train/combined_loss": 0.6777306199073792, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.14336492890995262, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5647755265235901 }, { "epoch": 1.433649289099526, "step": 605, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 7.9662556648254395, "train/clean_loss_weight": 1.59325110912323, "train/combined_loss": 0.6768205165863037, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.14336492890995262, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.564017117023468 }, { "epoch": 1.4360189573459716, "step": 606, "train/alpha": 0.2, "train/clean_ce_loss": 0.0947265625, "train/clean_loss_scale": 13.915472030639648, "train/clean_loss_weight": 2.7830944061279297, "train/combined_loss": 1.5817978382110596, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.14360189573459717, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3181648254394531 }, { "epoch": 1.4360189573459716, "step": 606, "train/alpha": 0.2, "train/clean_ce_loss": 0.10986328125, "train/clean_loss_scale": 8.133530616760254, "train/clean_loss_weight": 1.6267061233520508, "train/combined_loss": 1.072291612625122, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.14360189573459717, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8935763239860535 }, { "epoch": 1.4360189573459716, "step": 606, "train/alpha": 0.2, "train/clean_ce_loss": 0.10498046875, "train/clean_loss_scale": 10.52538013458252, "train/clean_loss_weight": 2.1050760746002197, "train/combined_loss": 1.3259512186050415, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.14360189573459717, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1049593687057495 }, { "epoch": 1.4360189573459716, "step": 606, "train/alpha": 0.2, "train/clean_ce_loss": 0.059814453125, "train/clean_loss_scale": 18.25387954711914, "train/clean_loss_weight": 3.650775909423828, "train/combined_loss": 1.3102149963378906, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.14360189573459717, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0918458700180054 }, { "epoch": 1.438388625592417, "step": 607, "train/alpha": 0.2, "train/clean_ce_loss": 0.007049560546875, "train/clean_loss_scale": 59.06989288330078, "train/clean_loss_weight": 11.813979148864746, "train/combined_loss": 0.4997001588344574, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.1438388625592417, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.41641679406166077 }, { "epoch": 1.438388625592417, "step": 607, "train/alpha": 0.2, "train/clean_ce_loss": 0.10888671875, "train/clean_loss_scale": 9.657330513000488, "train/clean_loss_weight": 1.9314661026000977, "train/combined_loss": 1.2618660926818848, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.1438388625592417, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0515550374984741 }, { "epoch": 1.438388625592417, "step": 607, "train/alpha": 0.2, "train/clean_ce_loss": 0.07470703125, "train/clean_loss_scale": 10.288125038146973, "train/clean_loss_weight": 2.0576250553131104, "train/combined_loss": 0.9223143458366394, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.1438388625592417, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.768595278263092 }, { "epoch": 1.438388625592417, "step": 607, "train/alpha": 0.2, "train/clean_ce_loss": 0.060791015625, "train/clean_loss_scale": 10.221406936645508, "train/clean_loss_weight": 2.044281482696533, "train/combined_loss": 0.745643675327301, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.1438388625592417, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6213697195053101 }, { "epoch": 1.4407582938388626, "step": 608, "train/alpha": 0.2, "train/clean_ce_loss": 0.1513671875, "train/clean_loss_scale": 7.5420708656311035, "train/clean_loss_weight": 1.5084141492843628, "train/combined_loss": 1.3699464797973633, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.14407582938388624, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1416220664978027 }, { "epoch": 1.4407582938388626, "step": 608, "train/alpha": 0.2, "train/clean_ce_loss": 0.115234375, "train/clean_loss_scale": 9.54210090637207, "train/clean_loss_weight": 1.908420205116272, "train/combined_loss": 1.3194936513900757, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.14407582938388624, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0995780229568481 }, { "epoch": 1.4407582938388626, "step": 608, "train/alpha": 0.2, "train/clean_ce_loss": 0.05029296875, "train/clean_loss_scale": 17.072114944458008, "train/clean_loss_weight": 3.4144229888916016, "train/combined_loss": 1.0303288698196411, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.14407582938388624, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8586073517799377 }, { "epoch": 1.4407582938388626, "step": 608, "train/alpha": 0.2, "train/clean_ce_loss": 0.0380859375, "train/clean_loss_scale": 11.032320022583008, "train/clean_loss_weight": 2.2064640522003174, "train/combined_loss": 0.5042114853858948, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.14407582938388624, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.420176237821579 }, { "epoch": 1.4431279620853081, "step": 609, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 9.184237480163574, "train/clean_loss_weight": 1.8368475437164307, "train/combined_loss": 0.893310546875, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.1443127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7444254755973816 }, { "epoch": 1.4431279620853081, "step": 609, "train/alpha": 0.2, "train/clean_ce_loss": 0.060302734375, "train/clean_loss_scale": 11.696370124816895, "train/clean_loss_weight": 2.3392741680145264, "train/combined_loss": 0.8463877439498901, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.1443127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7053231000900269 }, { "epoch": 1.4431279620853081, "step": 609, "train/alpha": 0.2, "train/clean_ce_loss": 0.0908203125, "train/clean_loss_scale": 12.768289566040039, "train/clean_loss_weight": 2.5536580085754395, "train/combined_loss": 1.3915441036224365, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.1443127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1596200466156006 }, { "epoch": 1.4431279620853081, "step": 609, "train/alpha": 0.2, "train/clean_ce_loss": 0.00174713134765625, "train/clean_loss_scale": 220.3670196533203, "train/clean_loss_weight": 44.07340621948242, "train/combined_loss": 0.4620121419429779, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.1443127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.38501012325286865 }, { "epoch": 1.4454976303317535, "grad_norm": 14.245183944702148, "learning_rate": 2e-06, "loss": 3.7265, "step": 610 }, { "epoch": 1.4454976303317535, "step": 610, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 9.390841484069824, "train/clean_loss_weight": 1.8781683444976807, "train/combined_loss": 0.8968987464904785, "train/pos_loss/pos0": 0.23289082944393158, "train/pos_loss/pos1": 0.46790334582328796, "train/pos_loss/pos2": 0.8532657027244568, "train/pos_loss/pos3": 1.2239187955856323, "train/pos_loss/pos4": 1.8573449850082397, "train/pos_loss/pos5": 2.2549617290496826, "train/pos_loss/pos6": 2.7210586071014404, "train/pos_loss/weight0": 2.3540589809417725, "train/pos_loss/weight1": 1.7631080150604248, "train/pos_loss/weight2": 1.2240567207336426, "train/pos_loss/weight3": 0.7833295464515686, "train/pos_loss/weight4": 0.4694300889968872, "train/pos_loss/weight5": 0.25603294372558594, "train/pos_loss/weight6": 0.14998359978199005, "train/pos_weight_progress": 0.14454976303317535, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7474156022071838 }, { "epoch": 1.4454976303317535, "step": 610, "train/alpha": 0.2, "train/clean_ce_loss": 0.1513671875, "train/clean_loss_scale": 8.428218841552734, "train/clean_loss_weight": 1.6856437921524048, "train/combined_loss": 1.5309069156646729, "train/pos_loss/pos0": 0.4295729696750641, "train/pos_loss/pos1": 0.9871864318847656, "train/pos_loss/pos2": 1.6711643934249878, "train/pos_loss/pos3": 2.5002992153167725, "train/pos_loss/pos4": 2.987589120864868, "train/pos_loss/pos5": 3.43481183052063, "train/pos_loss/pos6": 4.121903419494629, "train/pos_loss/weight0": 2.5421254634857178, "train/pos_loss/weight1": 1.8341906070709229, "train/pos_loss/weight2": 1.2460272312164307, "train/pos_loss/weight3": 0.7221988439559937, "train/pos_loss/weight4": 0.3839210569858551, "train/pos_loss/weight5": 0.1796104609966278, "train/pos_loss/weight6": 0.09192650765180588, "train/pos_weight_progress": 0.14454976303317535, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.275755763053894 }, { "epoch": 1.4454976303317535, "step": 610, "train/alpha": 0.2, "train/clean_ce_loss": 0.061279296875, "train/clean_loss_scale": 7.034754753112793, "train/clean_loss_weight": 1.4069509506225586, "train/combined_loss": 0.5173017978668213, "train/pos_loss/pos0": 0.18684989213943481, "train/pos_loss/pos1": 0.44001105427742004, "train/pos_loss/pos2": 0.7308266162872314, "train/pos_loss/pos3": 0.9576494693756104, "train/pos_loss/pos4": 1.2261580228805542, "train/pos_loss/pos5": 1.6679962873458862, "train/pos_loss/pos6": 2.0523509979248047, "train/pos_loss/weight0": 3.2110636234283447, "train/pos_loss/weight1": 1.988487958908081, "train/pos_loss/weight2": 1.1315137147903442, "train/pos_loss/weight3": 0.4754559397697449, "train/pos_loss/weight4": 0.1490529328584671, "train/pos_loss/weight5": 0.03515976294875145, "train/pos_loss/weight6": 0.009266051463782787, "train/pos_weight_progress": 0.14454976303317535, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4310848116874695 }, { "epoch": 1.4454976303317535, "step": 610, "train/alpha": 0.2, "train/clean_ce_loss": 0.053466796875, "train/clean_loss_scale": 12.635749816894531, "train/clean_loss_weight": 2.5271499156951904, "train/combined_loss": 0.810711681842804, "train/pos_loss/pos0": 0.2032988965511322, "train/pos_loss/pos1": 0.4236106872558594, "train/pos_loss/pos2": 0.7519242167472839, "train/pos_loss/pos3": 1.1374255418777466, "train/pos_loss/pos4": 1.6064985990524292, "train/pos_loss/pos5": 1.8386197090148926, "train/pos_loss/pos6": 2.2216341495513916, "train/pos_loss/weight0": 2.2716360092163086, "train/pos_loss/weight1": 1.7391276359558105, "train/pos_loss/weight2": 1.2405238151550293, "train/pos_loss/weight3": 0.8077101707458496, "train/pos_loss/weight4": 0.4864460825920105, "train/pos_loss/weight5": 0.2842985987663269, "train/pos_loss/weight6": 0.1702573299407959, "train/pos_weight_progress": 0.14454976303317535, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6755930781364441 }, { "epoch": 1.447867298578199, "step": 611, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 11.766160011291504, "train/clean_loss_weight": 2.3532321453094482, "train/combined_loss": 0.9996640086174011, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.1447867298578199, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8330533504486084 }, { "epoch": 1.447867298578199, "step": 611, "train/alpha": 0.2, "train/clean_ce_loss": 0.022216796875, "train/clean_loss_scale": 11.576016426086426, "train/clean_loss_weight": 2.3152034282684326, "train/combined_loss": 0.3086183965206146, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.1447867298578199, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2571820020675659 }, { "epoch": 1.447867298578199, "step": 611, "train/alpha": 0.2, "train/clean_ce_loss": 0.0693359375, "train/clean_loss_scale": 8.49278450012207, "train/clean_loss_weight": 1.698556900024414, "train/combined_loss": 0.7066261768341064, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.1447867298578199, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5888551473617554 }, { "epoch": 1.447867298578199, "step": 611, "train/alpha": 0.2, "train/clean_ce_loss": 0.057861328125, "train/clean_loss_scale": 6.200615406036377, "train/clean_loss_weight": 1.2401231527328491, "train/combined_loss": 0.43053102493286133, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.1447867298578199, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3587758541107178 }, { "epoch": 1.4502369668246446, "step": 612, "train/alpha": 0.2, "train/clean_ce_loss": 0.169921875, "train/clean_loss_scale": 8.054049491882324, "train/clean_loss_weight": 1.6108099222183228, "train/combined_loss": 1.6422710418701172, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.14502369668246445, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3685592412948608 }, { "epoch": 1.4502369668246446, "step": 612, "train/alpha": 0.2, "train/clean_ce_loss": 0.0673828125, "train/clean_loss_scale": 8.757294654846191, "train/clean_loss_weight": 1.751459002494812, "train/combined_loss": 0.7081093788146973, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.14502369668246445, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5900911688804626 }, { "epoch": 1.4502369668246446, "step": 612, "train/alpha": 0.2, "train/clean_ce_loss": 0.06103515625, "train/clean_loss_scale": 6.233482837677002, "train/clean_loss_weight": 1.2466965913772583, "train/combined_loss": 0.45655393600463867, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.14502369668246445, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.38046160340309143 }, { "epoch": 1.4502369668246446, "step": 612, "train/alpha": 0.2, "train/clean_ce_loss": 0.1474609375, "train/clean_loss_scale": 8.824286460876465, "train/clean_loss_weight": 1.764857292175293, "train/combined_loss": 1.5614850521087646, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.14502369668246445, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3012375831604004 }, { "epoch": 1.45260663507109, "step": 613, "train/alpha": 0.2, "train/clean_ce_loss": 0.007720947265625, "train/clean_loss_scale": 51.486507415771484, "train/clean_loss_weight": 10.297301292419434, "train/combined_loss": 0.4770295023918152, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.145260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3975245952606201 }, { "epoch": 1.45260663507109, "step": 613, "train/alpha": 0.2, "train/clean_ce_loss": 0.134765625, "train/clean_loss_scale": 6.999145030975342, "train/clean_loss_weight": 1.3998290300369263, "train/combined_loss": 1.1318930387496948, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.145260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9432441592216492 }, { "epoch": 1.45260663507109, "step": 613, "train/alpha": 0.2, "train/clean_ce_loss": 0.061279296875, "train/clean_loss_scale": 8.833415985107422, "train/clean_loss_weight": 1.7666832208633423, "train/combined_loss": 0.649566650390625, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.145260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5413055419921875 }, { "epoch": 1.45260663507109, "step": 613, "train/alpha": 0.2, "train/clean_ce_loss": 0.177734375, "train/clean_loss_scale": 8.48127555847168, "train/clean_loss_weight": 1.696255087852478, "train/combined_loss": 1.8088970184326172, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.145260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5074142217636108 }, { "epoch": 1.4549763033175356, "step": 614, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 10.436911582946777, "train/clean_loss_weight": 2.0873823165893555, "train/combined_loss": 0.9295374155044556, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.14549763033175356, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7746145129203796 }, { "epoch": 1.4549763033175356, "step": 614, "train/alpha": 0.2, "train/clean_ce_loss": 0.0235595703125, "train/clean_loss_scale": 14.985074996948242, "train/clean_loss_weight": 2.9970149993896484, "train/combined_loss": 0.4236502945423126, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.14549763033175356, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3530419170856476 }, { "epoch": 1.4549763033175356, "step": 614, "train/alpha": 0.2, "train/clean_ce_loss": 0.047119140625, "train/clean_loss_scale": 5.04058313369751, "train/clean_loss_weight": 1.008116602897644, "train/combined_loss": 0.2850095331668854, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.14549763033175356, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.23750795423984528 }, { "epoch": 1.4549763033175356, "step": 614, "train/alpha": 0.2, "train/clean_ce_loss": 0.1279296875, "train/clean_loss_scale": 7.784291744232178, "train/clean_loss_weight": 1.5568584203720093, "train/combined_loss": 1.1950104236602783, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.14549763033175356, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9958420395851135 }, { "epoch": 1.457345971563981, "grad_norm": 16.42146110534668, "learning_rate": 2e-06, "loss": 3.5728, "step": 615 }, { "epoch": 1.457345971563981, "step": 615, "train/alpha": 0.2, "train/clean_ce_loss": 0.03857421875, "train/clean_loss_scale": 11.131349563598633, "train/clean_loss_weight": 2.2262699604034424, "train/combined_loss": 0.5152597427368164, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.1457345971563981, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4293830990791321 }, { "epoch": 1.457345971563981, "step": 615, "train/alpha": 0.2, "train/clean_ce_loss": 0.0908203125, "train/clean_loss_scale": 8.806777954101562, "train/clean_loss_weight": 1.7613556385040283, "train/combined_loss": 0.959801197052002, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.1457345971563981, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7998343110084534 }, { "epoch": 1.457345971563981, "step": 615, "train/alpha": 0.2, "train/clean_ce_loss": 0.07470703125, "train/clean_loss_scale": 9.12292194366455, "train/clean_loss_weight": 1.8245843648910522, "train/combined_loss": 0.817855715751648, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.1457345971563981, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6815464496612549 }, { "epoch": 1.457345971563981, "step": 615, "train/alpha": 0.2, "train/clean_ce_loss": 0.103515625, "train/clean_loss_scale": 11.882497787475586, "train/clean_loss_weight": 2.376499652862549, "train/combined_loss": 1.4760290384292603, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.1457345971563981, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2300242185592651 }, { "epoch": 1.4597156398104265, "step": 616, "train/alpha": 0.2, "train/clean_ce_loss": 0.1064453125, "train/clean_loss_scale": 9.497673988342285, "train/clean_loss_weight": 1.899534821510315, "train/combined_loss": 1.2131794691085815, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.14597156398104266, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.010982871055603 }, { "epoch": 1.4597156398104265, "step": 616, "train/alpha": 0.2, "train/clean_ce_loss": 0.08642578125, "train/clean_loss_scale": 8.791362762451172, "train/clean_loss_weight": 1.7582725286483765, "train/combined_loss": 0.9117605090141296, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.14597156398104266, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7598004341125488 }, { "epoch": 1.4597156398104265, "step": 616, "train/alpha": 0.2, "train/clean_ce_loss": 0.1279296875, "train/clean_loss_scale": 10.152254104614258, "train/clean_loss_weight": 2.0304508209228516, "train/combined_loss": 1.5585296154022217, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.14597156398104266, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2987747192382812 }, { "epoch": 1.4597156398104265, "step": 616, "train/alpha": 0.2, "train/clean_ce_loss": 0.09326171875, "train/clean_loss_scale": 8.091301918029785, "train/clean_loss_weight": 1.618260383605957, "train/combined_loss": 0.9055304527282715, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.14597156398104266, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.754608690738678 }, { "epoch": 1.462085308056872, "step": 617, "train/alpha": 0.2, "train/clean_ce_loss": 0.12158203125, "train/clean_loss_scale": 9.067829132080078, "train/clean_loss_weight": 1.8135658502578735, "train/combined_loss": 1.3229820728302002, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.14620853080568721, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1024850606918335 }, { "epoch": 1.462085308056872, "step": 617, "train/alpha": 0.2, "train/clean_ce_loss": 0.04296875, "train/clean_loss_scale": 22.91067123413086, "train/clean_loss_weight": 4.582134246826172, "train/combined_loss": 1.1813315153121948, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.14620853080568721, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9844428896903992 }, { "epoch": 1.462085308056872, "step": 617, "train/alpha": 0.2, "train/clean_ce_loss": 0.0294189453125, "train/clean_loss_scale": 12.226551055908203, "train/clean_loss_weight": 2.445310354232788, "train/combined_loss": 0.4316307008266449, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.14620853080568721, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.359692245721817 }, { "epoch": 1.462085308056872, "step": 617, "train/alpha": 0.2, "train/clean_ce_loss": 0.1220703125, "train/clean_loss_scale": 10.300594329833984, "train/clean_loss_weight": 2.0601189136505127, "train/combined_loss": 1.5088762044906616, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.14620853080568721, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2573968172073364 }, { "epoch": 1.4644549763033177, "step": 618, "train/alpha": 0.2, "train/clean_ce_loss": 0.0517578125, "train/clean_loss_scale": 9.88771915435791, "train/clean_loss_weight": 1.977543830871582, "train/combined_loss": 0.6141200661659241, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.14644549763033177, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5117667317390442 }, { "epoch": 1.4644549763033177, "step": 618, "train/alpha": 0.2, "train/clean_ce_loss": 0.0054931640625, "train/clean_loss_scale": 50.00856018066406, "train/clean_loss_weight": 10.00171184539795, "train/combined_loss": 0.3296462893486023, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.14644549763033177, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2747052311897278 }, { "epoch": 1.4644549763033177, "step": 618, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 9.293142318725586, "train/clean_loss_weight": 1.858628511428833, "train/combined_loss": 0.8276704549789429, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.14644549763033177, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.689725399017334 }, { "epoch": 1.4644549763033177, "step": 618, "train/alpha": 0.2, "train/clean_ce_loss": 0.1416015625, "train/clean_loss_scale": 11.998822212219238, "train/clean_loss_weight": 2.3997645378112793, "train/combined_loss": 2.038862466812134, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.14644549763033177, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.6990519762039185 }, { "epoch": 1.466824644549763, "step": 619, "train/alpha": 0.2, "train/clean_ce_loss": 0.03125, "train/clean_loss_scale": 11.60456657409668, "train/clean_loss_weight": 2.320913314819336, "train/combined_loss": 0.4351712465286255, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.1466824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.36264270544052124 }, { "epoch": 1.466824644549763, "step": 619, "train/alpha": 0.2, "train/clean_ce_loss": 0.078125, "train/clean_loss_scale": 6.611031532287598, "train/clean_loss_weight": 1.3222063779830933, "train/combined_loss": 0.6197841763496399, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.1466824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5164868235588074 }, { "epoch": 1.466824644549763, "step": 619, "train/alpha": 0.2, "train/clean_ce_loss": 0.0098876953125, "train/clean_loss_scale": 31.213130950927734, "train/clean_loss_weight": 6.242626190185547, "train/combined_loss": 0.3703511357307434, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.1466824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3086259365081787 }, { "epoch": 1.466824644549763, "step": 619, "train/alpha": 0.2, "train/clean_ce_loss": 0.072265625, "train/clean_loss_scale": 9.576045989990234, "train/clean_loss_weight": 1.915209174156189, "train/combined_loss": 0.8304226994514465, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.1466824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6920189261436462 }, { "epoch": 1.4691943127962086, "grad_norm": 22.66509437561035, "learning_rate": 2e-06, "loss": 3.6875, "step": 620 }, { "epoch": 1.4691943127962086, "step": 620, "train/alpha": 0.2, "train/clean_ce_loss": 0.058349609375, "train/clean_loss_scale": 11.12249755859375, "train/clean_loss_weight": 2.224499464035034, "train/combined_loss": 0.7787920236587524, "train/pos_loss/pos0": 0.23391857743263245, "train/pos_loss/pos1": 0.47136160731315613, "train/pos_loss/pos2": 0.749871015548706, "train/pos_loss/pos3": 1.1705595254898071, "train/pos_loss/pos4": 1.4207545518875122, "train/pos_loss/pos5": 1.9171011447906494, "train/pos_loss/pos6": 2.347064256668091, "train/pos_loss/weight0": 2.426628828048706, "train/pos_loss/weight1": 1.8130543231964111, "train/pos_loss/weight2": 1.22214937210083, "train/pos_loss/weight3": 0.777302086353302, "train/pos_loss/weight4": 0.43016326427459717, "train/pos_loss/weight5": 0.21238628029823303, "train/pos_loss/weight6": 0.11831621825695038, "train/pos_weight_progress": 0.14691943127962084, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6489933729171753 }, { "epoch": 1.4691943127962086, "step": 620, "train/alpha": 0.2, "train/clean_ce_loss": 0.0181884765625, "train/clean_loss_scale": 17.899600982666016, "train/clean_loss_weight": 3.5799202919006348, "train/combined_loss": 0.3906797766685486, "train/pos_loss/pos0": 0.045063264667987823, "train/pos_loss/pos1": 0.08595982939004898, "train/pos_loss/pos2": 0.21308334171772003, "train/pos_loss/pos3": 0.6538828611373901, "train/pos_loss/pos4": 1.3015209436416626, "train/pos_loss/pos5": 2.1038131713867188, "train/pos_loss/pos6": 2.475813150405884, "train/pos_loss/weight0": 2.4720046520233154, "train/pos_loss/weight1": 1.806592583656311, "train/pos_loss/weight2": 1.2303274869918823, "train/pos_loss/weight3": 0.7577862739562988, "train/pos_loss/weight4": 0.41027799248695374, "train/pos_loss/weight5": 0.21129818260669708, "train/pos_loss/weight6": 0.11171233654022217, "train/pos_weight_progress": 0.14691943127962084, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.32556647062301636 }, { "epoch": 1.4691943127962086, "step": 620, "train/alpha": 0.2, "train/clean_ce_loss": 0.056640625, "train/clean_loss_scale": 13.164926528930664, "train/clean_loss_weight": 2.6329853534698486, "train/combined_loss": 0.8948035836219788, "train/pos_loss/pos0": 0.17938721179962158, "train/pos_loss/pos1": 0.4199554920196533, "train/pos_loss/pos2": 0.7577406764030457, "train/pos_loss/pos3": 1.1979148387908936, "train/pos_loss/pos4": 1.7077966928482056, "train/pos_loss/pos5": 2.1522510051727295, "train/pos_loss/pos6": 2.546870708465576, "train/pos_loss/weight0": 2.1689116954803467, "train/pos_loss/weight1": 1.655623197555542, "train/pos_loss/weight2": 1.2119022607803345, "train/pos_loss/weight3": 0.8790492415428162, "train/pos_loss/weight4": 0.5590095520019531, "train/pos_loss/weight5": 0.3264065384864807, "train/pos_loss/weight6": 0.19909724593162537, "train/pos_weight_progress": 0.14691943127962084, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7456696629524231 }, { "epoch": 1.4691943127962086, "step": 620, "train/alpha": 0.2, "train/clean_ce_loss": 0.09521484375, "train/clean_loss_scale": 9.054936408996582, "train/clean_loss_weight": 1.8109873533248901, "train/combined_loss": 1.0345971584320068, "train/pos_loss/pos0": 0.35551002621650696, "train/pos_loss/pos1": 0.6702782511711121, "train/pos_loss/pos2": 1.1275043487548828, "train/pos_loss/pos3": 1.4974740743637085, "train/pos_loss/pos4": 1.9712961912155151, "train/pos_loss/pos5": 2.4431450366973877, "train/pos_loss/pos6": 2.8237788677215576, "train/pos_loss/weight0": 2.5413308143615723, "train/pos_loss/weight1": 1.8527429103851318, "train/pos_loss/weight2": 1.2437710762023926, "train/pos_loss/weight3": 0.7325071692466736, "train/pos_loss/weight4": 0.3786434531211853, "train/pos_loss/weight5": 0.16854523122310638, "train/pos_loss/weight6": 0.0824589878320694, "train/pos_weight_progress": 0.14691943127962084, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.862164318561554 }, { "epoch": 1.471563981042654, "step": 621, "train/alpha": 0.2, "train/clean_ce_loss": 0.154296875, "train/clean_loss_scale": 7.264358043670654, "train/clean_loss_weight": 1.4528716802597046, "train/combined_loss": 1.345041275024414, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.1471563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1208677291870117 }, { "epoch": 1.471563981042654, "step": 621, "train/alpha": 0.2, "train/clean_ce_loss": 0.12060546875, "train/clean_loss_scale": 11.135946273803711, "train/clean_loss_weight": 2.227189302444458, "train/combined_loss": 1.6116671562194824, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.1471563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3430559635162354 }, { "epoch": 1.471563981042654, "step": 621, "train/alpha": 0.2, "train/clean_ce_loss": 0.08544921875, "train/clean_loss_scale": 8.997633934020996, "train/clean_loss_weight": 1.7995268106460571, "train/combined_loss": 0.9226089715957642, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.1471563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7688407897949219 }, { "epoch": 1.471563981042654, "step": 621, "train/alpha": 0.2, "train/clean_ce_loss": 0.09228515625, "train/clean_loss_scale": 7.198509693145752, "train/clean_loss_weight": 1.4397019147872925, "train/combined_loss": 0.7971786856651306, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.1471563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6643155813217163 }, { "epoch": 1.4739336492890995, "step": 622, "train/alpha": 0.2, "train/clean_ce_loss": 0.1064453125, "train/clean_loss_scale": 7.446202754974365, "train/clean_loss_weight": 1.4892405271530151, "train/combined_loss": 0.95113605260849, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.14739336492890995, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7926133871078491 }, { "epoch": 1.4739336492890995, "step": 622, "train/alpha": 0.2, "train/clean_ce_loss": 0.0029449462890625, "train/clean_loss_scale": 66.3076400756836, "train/clean_loss_weight": 13.261528015136719, "train/combined_loss": 0.23432691395282745, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.14739336492890995, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.19527243077754974 }, { "epoch": 1.4739336492890995, "step": 622, "train/alpha": 0.2, "train/clean_ce_loss": 0.095703125, "train/clean_loss_scale": 9.270567893981934, "train/clean_loss_weight": 1.8541135787963867, "train/combined_loss": 1.064666748046875, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.14739336492890995, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8872223496437073 }, { "epoch": 1.4739336492890995, "step": 622, "train/alpha": 0.2, "train/clean_ce_loss": 0.050048828125, "train/clean_loss_scale": 12.71019458770752, "train/clean_loss_weight": 2.542038917541504, "train/combined_loss": 0.7633563876152039, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.14739336492890995, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6361303329467773 }, { "epoch": 1.4763033175355451, "step": 623, "train/alpha": 0.2, "train/clean_ce_loss": 0.08056640625, "train/clean_loss_scale": 5.194462299346924, "train/clean_loss_weight": 1.0388925075531006, "train/combined_loss": 0.5021989941596985, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.1476303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4184991419315338 }, { "epoch": 1.4763033175355451, "step": 623, "train/alpha": 0.2, "train/clean_ce_loss": 0.1533203125, "train/clean_loss_scale": 6.988547325134277, "train/clean_loss_weight": 1.3977094888687134, "train/combined_loss": 1.2857835292816162, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.1476303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.071486234664917 }, { "epoch": 1.4763033175355451, "step": 623, "train/alpha": 0.2, "train/clean_ce_loss": 0.107421875, "train/clean_loss_scale": 9.652888298034668, "train/clean_loss_weight": 1.9305776357650757, "train/combined_loss": 1.244317650794983, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.1476303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0369313955307007 }, { "epoch": 1.4763033175355451, "step": 623, "train/alpha": 0.2, "train/clean_ce_loss": 0.060791015625, "train/clean_loss_scale": 7.399406909942627, "train/clean_loss_weight": 1.4798814058303833, "train/combined_loss": 0.5397809743881226, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.1476303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4498174488544464 }, { "epoch": 1.4786729857819905, "step": 624, "train/alpha": 0.2, "train/clean_ce_loss": 0.1220703125, "train/clean_loss_scale": 7.600526809692383, "train/clean_loss_weight": 1.5201053619384766, "train/combined_loss": 1.1133583784103394, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.14786729857819905, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9277986884117126 }, { "epoch": 1.4786729857819905, "step": 624, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 13.84588623046875, "train/clean_loss_weight": 2.769177198410034, "train/combined_loss": 1.1844723224639893, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.14786729857819905, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9870602488517761 }, { "epoch": 1.4786729857819905, "step": 624, "train/alpha": 0.2, "train/clean_ce_loss": 0.1328125, "train/clean_loss_scale": 6.836630821228027, "train/clean_loss_weight": 1.3673261404037476, "train/combined_loss": 1.0895880460739136, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.14786729857819905, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.907990038394928 }, { "epoch": 1.4786729857819905, "step": 624, "train/alpha": 0.2, "train/clean_ce_loss": 0.06689453125, "train/clean_loss_scale": 7.965531826019287, "train/clean_loss_weight": 1.5931063890457153, "train/combined_loss": 0.6394206285476685, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.14786729857819905, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5328505039215088 }, { "epoch": 1.481042654028436, "grad_norm": 23.044504165649414, "learning_rate": 2e-06, "loss": 3.2375, "step": 625 }, { "epoch": 1.481042654028436, "step": 625, "train/alpha": 0.2, "train/clean_ce_loss": 0.03271484375, "train/clean_loss_scale": 12.52994155883789, "train/clean_loss_weight": 2.505988359451294, "train/combined_loss": 0.4918981194496155, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.1481042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.40991508960723877 }, { "epoch": 1.481042654028436, "step": 625, "train/alpha": 0.2, "train/clean_ce_loss": 0.0208740234375, "train/clean_loss_scale": 7.764091491699219, "train/clean_loss_weight": 1.5528182983398438, "train/combined_loss": 0.19448140263557434, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.1481042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.16206783056259155 }, { "epoch": 1.481042654028436, "step": 625, "train/alpha": 0.2, "train/clean_ce_loss": 0.055908203125, "train/clean_loss_scale": 8.076900482177734, "train/clean_loss_weight": 1.6153801679611206, "train/combined_loss": 0.5418779850006104, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.1481042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4515649676322937 }, { "epoch": 1.481042654028436, "step": 625, "train/alpha": 0.2, "train/clean_ce_loss": 0.09033203125, "train/clean_loss_scale": 8.279560089111328, "train/clean_loss_weight": 1.6559120416641235, "train/combined_loss": 0.8974913954734802, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.1481042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7479094862937927 }, { "epoch": 1.4834123222748814, "step": 626, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 8.17663860321045, "train/clean_loss_weight": 1.635327696800232, "train/combined_loss": 0.776141881942749, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.14834123222748816, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6467849016189575 }, { "epoch": 1.4834123222748814, "step": 626, "train/alpha": 0.2, "train/clean_ce_loss": 0.08251953125, "train/clean_loss_scale": 9.20269775390625, "train/clean_loss_weight": 1.840539574623108, "train/combined_loss": 0.9112827181816101, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.14834123222748816, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7594022750854492 }, { "epoch": 1.4834123222748814, "step": 626, "train/alpha": 0.2, "train/clean_ce_loss": 0.11474609375, "train/clean_loss_scale": 8.161856651306152, "train/clean_loss_weight": 1.6323713064193726, "train/combined_loss": 1.1238493919372559, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.14834123222748816, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9365411400794983 }, { "epoch": 1.4834123222748814, "step": 626, "train/alpha": 0.2, "train/clean_ce_loss": 0.11376953125, "train/clean_loss_scale": 7.1191301345825195, "train/clean_loss_weight": 1.4238260984420776, "train/combined_loss": 0.9719281196594238, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.14834123222748816, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8099400997161865 }, { "epoch": 1.485781990521327, "step": 627, "train/alpha": 0.2, "train/clean_ce_loss": 0.003265380859375, "train/clean_loss_scale": 90.96171569824219, "train/clean_loss_weight": 18.19234275817871, "train/combined_loss": 0.3564295768737793, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.1485781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2970246374607086 }, { "epoch": 1.485781990521327, "step": 627, "train/alpha": 0.2, "train/clean_ce_loss": 0.05029296875, "train/clean_loss_scale": 21.36503791809082, "train/clean_loss_weight": 4.273007869720459, "train/combined_loss": 1.2894134521484375, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.1485781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.074511170387268 }, { "epoch": 1.485781990521327, "step": 627, "train/alpha": 0.2, "train/clean_ce_loss": 0.09619140625, "train/clean_loss_scale": 8.086440086364746, "train/clean_loss_weight": 1.6172879934310913, "train/combined_loss": 0.9334152340888977, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.1485781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7778460383415222 }, { "epoch": 1.485781990521327, "step": 627, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 7.065013408660889, "train/clean_loss_weight": 1.4130027294158936, "train/combined_loss": 0.7078812122344971, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.1485781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5899010300636292 }, { "epoch": 1.4881516587677726, "step": 628, "train/alpha": 0.2, "train/clean_ce_loss": 0.146484375, "train/clean_loss_scale": 6.699742317199707, "train/clean_loss_weight": 1.3399485349655151, "train/combined_loss": 1.1776890754699707, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.14881516587677726, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9814075827598572 }, { "epoch": 1.4881516587677726, "step": 628, "train/alpha": 0.2, "train/clean_ce_loss": 0.1767578125, "train/clean_loss_scale": 7.34493350982666, "train/clean_loss_weight": 1.4689867496490479, "train/combined_loss": 1.557929277420044, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.14881516587677726, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2982743978500366 }, { "epoch": 1.4881516587677726, "step": 628, "train/alpha": 0.2, "train/clean_ce_loss": 0.002349853515625, "train/clean_loss_scale": 65.15323638916016, "train/clean_loss_weight": 13.030647277832031, "train/combined_loss": 0.1837206780910492, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.14881516587677726, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.15310056507587433 }, { "epoch": 1.4881516587677726, "step": 628, "train/alpha": 0.2, "train/clean_ce_loss": 0.1083984375, "train/clean_loss_scale": 8.228525161743164, "train/clean_loss_weight": 1.6457051038742065, "train/combined_loss": 1.0703511238098145, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.14881516587677726, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8919593095779419 }, { "epoch": 1.4905213270142181, "step": 629, "train/alpha": 0.2, "train/clean_ce_loss": 0.1435546875, "train/clean_loss_scale": 8.670051574707031, "train/clean_loss_weight": 1.7340103387832642, "train/combined_loss": 1.4935518503189087, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.1490521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.244626522064209 }, { "epoch": 1.4905213270142181, "step": 629, "train/alpha": 0.2, "train/clean_ce_loss": 0.05029296875, "train/clean_loss_scale": 13.927129745483398, "train/clean_loss_weight": 2.785425901412964, "train/combined_loss": 0.8405240774154663, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.1490521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7004367113113403 }, { "epoch": 1.4905213270142181, "step": 629, "train/alpha": 0.2, "train/clean_ce_loss": 0.07861328125, "train/clean_loss_scale": 13.65268611907959, "train/clean_loss_weight": 2.730537176132202, "train/combined_loss": 1.2879389524459839, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.1490521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0732824802398682 }, { "epoch": 1.4905213270142181, "step": 629, "train/alpha": 0.2, "train/clean_ce_loss": 0.08203125, "train/clean_loss_scale": 11.982258796691895, "train/clean_loss_weight": 2.396451711654663, "train/combined_loss": 1.1795035600662231, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.1490521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9829196333885193 }, { "epoch": 1.4928909952606635, "grad_norm": 18.553308486938477, "learning_rate": 2e-06, "loss": 3.2962, "step": 630 }, { "epoch": 1.4928909952606635, "step": 630, "train/alpha": 0.2, "train/clean_ce_loss": 0.1337890625, "train/clean_loss_scale": 7.663974285125732, "train/clean_loss_weight": 1.5327948331832886, "train/combined_loss": 1.2304271459579468, "train/pos_loss/pos0": 0.4712628424167633, "train/pos_loss/pos1": 0.8055604100227356, "train/pos_loss/pos2": 1.4089577198028564, "train/pos_loss/pos3": 2.1657347679138184, "train/pos_loss/pos4": 2.415567398071289, "train/pos_loss/pos5": 3.0752413272857666, "train/pos_loss/pos6": 3.249931573867798, "train/pos_loss/weight0": 2.7770230770111084, "train/pos_loss/weight1": 1.906909704208374, "train/pos_loss/weight2": 1.206133484840393, "train/pos_loss/weight3": 0.6448495388031006, "train/pos_loss/weight4": 0.2963249087333679, "train/pos_loss/weight5": 0.1176268681883812, "train/pos_loss/weight6": 0.05113256722688675, "train/pos_weight_progress": 0.14928909952606634, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0253559350967407 }, { "epoch": 1.4928909952606635, "step": 630, "train/alpha": 0.2, "train/clean_ce_loss": 0.08544921875, "train/clean_loss_scale": 10.503788948059082, "train/clean_loss_weight": 2.1007578372955322, "train/combined_loss": 1.077048659324646, "train/pos_loss/pos0": 0.34842246770858765, "train/pos_loss/pos1": 0.7882850170135498, "train/pos_loss/pos2": 1.4603331089019775, "train/pos_loss/pos3": 2.1388163566589355, "train/pos_loss/pos4": 2.739450454711914, "train/pos_loss/pos5": 3.3316657543182373, "train/pos_loss/pos6": 3.7127535343170166, "train/pos_loss/weight0": 3.051403284072876, "train/pos_loss/weight1": 1.963034749031067, "train/pos_loss/weight2": 1.1534324884414673, "train/pos_loss/weight3": 0.5555276274681091, "train/pos_loss/weight4": 0.19920144975185394, "train/pos_loss/weight5": 0.0593666210770607, "train/pos_loss/weight6": 0.018034126609563828, "train/pos_weight_progress": 0.14928909952606634, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8975405693054199 }, { "epoch": 1.4928909952606635, "step": 630, "train/alpha": 0.2, "train/clean_ce_loss": 0.08447265625, "train/clean_loss_scale": 6.52515983581543, "train/clean_loss_weight": 1.3050320148468018, "train/combined_loss": 0.6614370942115784, "train/pos_loss/pos0": 0.2529011368751526, "train/pos_loss/pos1": 0.477357417345047, "train/pos_loss/pos2": 0.9244658350944519, "train/pos_loss/pos3": 1.4185494184494019, "train/pos_loss/pos4": 1.888409972190857, "train/pos_loss/pos5": 2.3186707496643066, "train/pos_loss/pos6": 2.7235777378082275, "train/pos_loss/weight0": 3.163667678833008, "train/pos_loss/weight1": 2.045527935028076, "train/pos_loss/weight2": 1.1325151920318604, "train/pos_loss/weight3": 0.4798792004585266, "train/pos_loss/weight4": 0.14145125448703766, "train/pos_loss/weight5": 0.030417833477258682, "train/pos_loss/weight6": 0.006541045382618904, "train/pos_weight_progress": 0.14928909952606634, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5511975884437561 }, { "epoch": 1.4928909952606635, "step": 630, "train/alpha": 0.2, "train/clean_ce_loss": 0.004547119140625, "train/clean_loss_scale": 72.3372573852539, "train/clean_loss_weight": 14.467452049255371, "train/combined_loss": 0.39471134543418884, "train/pos_loss/pos0": 0.015621233731508255, "train/pos_loss/pos1": 0.07233879715204239, "train/pos_loss/pos2": 0.31262296438217163, "train/pos_loss/pos3": 0.9104171991348267, "train/pos_loss/pos4": 1.5746608972549438, "train/pos_loss/pos5": 2.200449228286743, "train/pos_loss/pos6": 2.9964115619659424, "train/pos_loss/weight0": 2.6050865650177, "train/pos_loss/weight1": 1.8678414821624756, "train/pos_loss/weight2": 1.2434386014938354, "train/pos_loss/weight3": 0.713290274143219, "train/pos_loss/weight4": 0.3524759113788605, "train/pos_loss/weight5": 0.14898532629013062, "train/pos_loss/weight6": 0.06888217478990555, "train/pos_weight_progress": 0.14928909952606634, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.32892611622810364 }, { "epoch": 1.495260663507109, "step": 631, "train/alpha": 0.2, "train/clean_ce_loss": 0.005859375, "train/clean_loss_scale": 49.69828414916992, "train/clean_loss_weight": 9.939657211303711, "train/combined_loss": 0.3494410514831543, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.1495260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2912008762359619 }, { "epoch": 1.495260663507109, "step": 631, "train/alpha": 0.2, "train/clean_ce_loss": 0.068359375, "train/clean_loss_scale": 10.040390968322754, "train/clean_loss_weight": 2.0080783367156982, "train/combined_loss": 0.8236258625984192, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.1495260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6863548755645752 }, { "epoch": 1.495260663507109, "step": 631, "train/alpha": 0.2, "train/clean_ce_loss": 0.03271484375, "train/clean_loss_scale": 13.414145469665527, "train/clean_loss_weight": 2.6828291416168213, "train/combined_loss": 0.5266100168228149, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.1495260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43884167075157166 }, { "epoch": 1.495260663507109, "step": 631, "train/alpha": 0.2, "train/clean_ce_loss": 0.10400390625, "train/clean_loss_scale": 10.790094375610352, "train/clean_loss_weight": 2.1580188274383545, "train/combined_loss": 1.3466542959213257, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.1495260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1222119331359863 }, { "epoch": 1.4976303317535544, "step": 632, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 9.846357345581055, "train/clean_loss_weight": 1.9692715406417847, "train/combined_loss": 0.9634814262390137, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.14976303317535544, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8029012084007263 }, { "epoch": 1.4976303317535544, "step": 632, "train/alpha": 0.2, "train/clean_ce_loss": 0.125, "train/clean_loss_scale": 8.850690841674805, "train/clean_loss_weight": 1.770138144493103, "train/combined_loss": 1.3276035785675049, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.14976303317535544, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1063363552093506 }, { "epoch": 1.4976303317535544, "step": 632, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 13.236379623413086, "train/clean_loss_weight": 2.647275924682617, "train/combined_loss": 1.1788650751113892, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.14976303317535544, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9823875427246094 }, { "epoch": 1.4976303317535544, "step": 632, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 8.355327606201172, "train/clean_loss_weight": 1.6710655689239502, "train/combined_loss": 0.6658151745796204, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.14976303317535544, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5548459887504578 }, { "epoch": 1.5, "step": 633, "train/alpha": 0.2, "train/clean_ce_loss": 0.10400390625, "train/clean_loss_scale": 10.181260108947754, "train/clean_loss_weight": 2.036252021789551, "train/combined_loss": 1.2706689834594727, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.15, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0588908195495605 }, { "epoch": 1.5, "step": 633, "train/alpha": 0.2, "train/clean_ce_loss": 0.09326171875, "train/clean_loss_scale": 7.803549766540527, "train/clean_loss_weight": 1.5607099533081055, "train/combined_loss": 0.8733269572257996, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.15, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7277724742889404 }, { "epoch": 1.5, "step": 633, "train/alpha": 0.2, "train/clean_ce_loss": 0.05517578125, "train/clean_loss_scale": 10.193048477172852, "train/clean_loss_weight": 2.038609743118286, "train/combined_loss": 0.6748912930488586, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.15, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5624094009399414 }, { "epoch": 1.5, "step": 633, "train/alpha": 0.2, "train/clean_ce_loss": 0.0546875, "train/clean_loss_scale": 15.252569198608398, "train/clean_loss_weight": 3.050513982772827, "train/combined_loss": 1.0009498596191406, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.15, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8341248631477356 }, { "epoch": 1.5023696682464456, "step": 634, "train/alpha": 0.2, "train/clean_ce_loss": 0.12451171875, "train/clean_loss_scale": 9.312713623046875, "train/clean_loss_weight": 1.862542748451233, "train/combined_loss": 1.3914504051208496, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.15023696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1595419645309448 }, { "epoch": 1.5023696682464456, "step": 634, "train/alpha": 0.2, "train/clean_ce_loss": 0.10400390625, "train/clean_loss_scale": 8.7909574508667, "train/clean_loss_weight": 1.758191466331482, "train/combined_loss": 1.0971527099609375, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.15023696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9142938852310181 }, { "epoch": 1.5023696682464456, "step": 634, "train/alpha": 0.2, "train/clean_ce_loss": 0.11279296875, "train/clean_loss_scale": 8.142208099365234, "train/clean_loss_weight": 1.6284416913986206, "train/combined_loss": 1.1020605564117432, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.15023696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.918383777141571 }, { "epoch": 1.5023696682464456, "step": 634, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 10.026481628417969, "train/clean_loss_weight": 2.005296468734741, "train/combined_loss": 0.7989853024482727, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.15023696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6658210754394531 }, { "epoch": 1.5047393364928912, "grad_norm": 28.247314453125, "learning_rate": 2e-06, "loss": 3.4139, "step": 635 }, { "epoch": 1.5047393364928912, "step": 635, "train/alpha": 0.2, "train/clean_ce_loss": 0.1455078125, "train/clean_loss_scale": 8.026413917541504, "train/clean_loss_weight": 1.6052827835083008, "train/combined_loss": 1.401487112045288, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.1504739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1679059267044067 }, { "epoch": 1.5047393364928912, "step": 635, "train/alpha": 0.2, "train/clean_ce_loss": 0.04443359375, "train/clean_loss_scale": 14.539081573486328, "train/clean_loss_weight": 2.9078164100646973, "train/combined_loss": 0.7752283811569214, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.1504739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6460236310958862 }, { "epoch": 1.5047393364928912, "step": 635, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 9.137555122375488, "train/clean_loss_weight": 1.8275110721588135, "train/combined_loss": 0.7067327499389648, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.1504739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5889439582824707 }, { "epoch": 1.5047393364928912, "step": 635, "train/alpha": 0.2, "train/clean_ce_loss": 0.048828125, "train/clean_loss_scale": 11.293451309204102, "train/clean_loss_weight": 2.258690357208252, "train/combined_loss": 0.661725640296936, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.1504739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.55143803358078 }, { "epoch": 1.5071090047393365, "step": 636, "train/alpha": 0.2, "train/clean_ce_loss": 0.0673828125, "train/clean_loss_scale": 12.619416236877441, "train/clean_loss_weight": 2.52388334274292, "train/combined_loss": 1.0203981399536133, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.15071090047393365, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8503317832946777 }, { "epoch": 1.5071090047393365, "step": 636, "train/alpha": 0.2, "train/clean_ce_loss": 0.1513671875, "train/clean_loss_scale": 8.619702339172363, "train/clean_loss_weight": 1.7239404916763306, "train/combined_loss": 1.565688133239746, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.15071090047393365, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.304740071296692 }, { "epoch": 1.5071090047393365, "step": 636, "train/alpha": 0.2, "train/clean_ce_loss": 0.08984375, "train/clean_loss_scale": 10.295746803283691, "train/clean_loss_weight": 2.0591495037078857, "train/combined_loss": 1.1100102663040161, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.15071090047393365, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9250085353851318 }, { "epoch": 1.5071090047393365, "step": 636, "train/alpha": 0.2, "train/clean_ce_loss": 0.1416015625, "train/clean_loss_scale": 8.820219993591309, "train/clean_loss_weight": 1.7640440464019775, "train/combined_loss": 1.4987483024597168, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.15071090047393365, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2489569187164307 }, { "epoch": 1.5094786729857819, "step": 637, "train/alpha": 0.2, "train/clean_ce_loss": 0.10888671875, "train/clean_loss_scale": 11.207331657409668, "train/clean_loss_weight": 2.2414662837982178, "train/combined_loss": 1.4643954038619995, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.1509478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2203295230865479 }, { "epoch": 1.5094786729857819, "step": 637, "train/alpha": 0.2, "train/clean_ce_loss": 0.087890625, "train/clean_loss_scale": 9.146830558776855, "train/clean_loss_weight": 1.8293660879135132, "train/combined_loss": 0.9647047519683838, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.1509478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8039206266403198 }, { "epoch": 1.5094786729857819, "step": 637, "train/alpha": 0.2, "train/clean_ce_loss": 0.083984375, "train/clean_loss_scale": 7.569821357727051, "train/clean_loss_weight": 1.513964295387268, "train/combined_loss": 0.7628960609436035, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.1509478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6357467174530029 }, { "epoch": 1.5094786729857819, "step": 637, "train/alpha": 0.2, "train/clean_ce_loss": 0.04931640625, "train/clean_loss_scale": 9.41362190246582, "train/clean_loss_weight": 1.882724404335022, "train/combined_loss": 0.5570952296257019, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.1509478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46424600481987 }, { "epoch": 1.5118483412322274, "step": 638, "train/alpha": 0.2, "train/clean_ce_loss": 0.08642578125, "train/clean_loss_scale": 9.103361129760742, "train/clean_loss_weight": 1.8206722736358643, "train/combined_loss": 0.9441181421279907, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.15118483412322276, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7867650985717773 }, { "epoch": 1.5118483412322274, "step": 638, "train/alpha": 0.2, "train/clean_ce_loss": 0.03369140625, "train/clean_loss_scale": 17.710264205932617, "train/clean_loss_weight": 3.542052984237671, "train/combined_loss": 0.716020405292511, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.15118483412322276, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5966836810112 }, { "epoch": 1.5118483412322274, "step": 638, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 16.13249969482422, "train/clean_loss_weight": 3.2265000343322754, "train/combined_loss": 1.5313270092010498, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.15118483412322276, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2761058807373047 }, { "epoch": 1.5118483412322274, "step": 638, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 9.895101547241211, "train/clean_loss_weight": 1.979020357131958, "train/combined_loss": 0.9450594782829285, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.15118483412322276, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7875495553016663 }, { "epoch": 1.514218009478673, "step": 639, "train/alpha": 0.2, "train/clean_ce_loss": 0.078125, "train/clean_loss_scale": 14.36487865447998, "train/clean_loss_weight": 2.8729758262634277, "train/combined_loss": 1.3467073440551758, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.1514218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1222561597824097 }, { "epoch": 1.514218009478673, "step": 639, "train/alpha": 0.2, "train/clean_ce_loss": 0.003448486328125, "train/clean_loss_scale": 77.55902862548828, "train/clean_loss_weight": 15.511805534362793, "train/combined_loss": 0.32095348834991455, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.1514218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.26746124029159546 }, { "epoch": 1.514218009478673, "step": 639, "train/alpha": 0.2, "train/clean_ce_loss": 0.134765625, "train/clean_loss_scale": 10.80383586883545, "train/clean_loss_weight": 2.1607673168182373, "train/combined_loss": 1.747182846069336, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.1514218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.45598566532135 }, { "epoch": 1.514218009478673, "step": 639, "train/alpha": 0.2, "train/clean_ce_loss": 0.003082275390625, "train/clean_loss_scale": 92.32646179199219, "train/clean_loss_weight": 18.46529197692871, "train/combined_loss": 0.3414906859397888, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.1514218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.28457558155059814 }, { "epoch": 1.5165876777251186, "grad_norm": 21.17659568786621, "learning_rate": 2e-06, "loss": 3.2628, "step": 640 }, { "epoch": 1.5165876777251186, "step": 640, "train/alpha": 0.2, "train/clean_ce_loss": 0.0517578125, "train/clean_loss_scale": 11.575169563293457, "train/clean_loss_weight": 2.3150339126586914, "train/combined_loss": 0.7189265489578247, "train/pos_loss/pos0": 0.15463028848171234, "train/pos_loss/pos1": 0.35813477635383606, "train/pos_loss/pos2": 0.7305387854576111, "train/pos_loss/pos3": 0.9939493536949158, "train/pos_loss/pos4": 1.3950448036193848, "train/pos_loss/pos5": 1.8463853597640991, "train/pos_loss/pos6": 2.21051025390625, "train/pos_loss/weight0": 2.288882255554199, "train/pos_loss/weight1": 1.7180349826812744, "train/pos_loss/weight2": 1.2529716491699219, "train/pos_loss/weight3": 0.8621255159378052, "train/pos_loss/weight4": 0.48888498544692993, "train/pos_loss/weight5": 0.24482713639736176, "train/pos_loss/weight6": 0.14427344501018524, "train/pos_weight_progress": 0.15165876777251186, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5991054773330688 }, { "epoch": 1.5165876777251186, "step": 640, "train/alpha": 0.2, "train/clean_ce_loss": 0.11767578125, "train/clean_loss_scale": 11.425726890563965, "train/clean_loss_weight": 2.2851455211639404, "train/combined_loss": 1.6134376525878906, "train/pos_loss/pos0": 0.26893550157546997, "train/pos_loss/pos1": 1.1772990226745605, "train/pos_loss/pos2": 1.6637295484542847, "train/pos_loss/pos3": 2.313076972961426, "train/pos_loss/pos4": 2.8531737327575684, "train/pos_loss/pos5": 3.6455466747283936, "train/pos_loss/pos6": 3.8155477046966553, "train/pos_loss/weight0": 2.3339431285858154, "train/pos_loss/weight1": 1.7648236751556396, "train/pos_loss/weight2": 1.2477012872695923, "train/pos_loss/weight3": 0.7864101529121399, "train/pos_loss/weight4": 0.46573057770729065, "train/pos_loss/weight5": 0.2529586851596832, "train/pos_loss/weight6": 0.14843270182609558, "train/pos_weight_progress": 0.15165876777251186, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3445312976837158 }, { "epoch": 1.5165876777251186, "step": 640, "train/alpha": 0.2, "train/clean_ce_loss": 0.10498046875, "train/clean_loss_scale": 7.644495010375977, "train/clean_loss_weight": 1.528899073600769, "train/combined_loss": 0.9630271792411804, "train/pos_loss/pos0": 0.34460094571113586, "train/pos_loss/pos1": 0.7884670495986938, "train/pos_loss/pos2": 1.3405933380126953, "train/pos_loss/pos3": 1.7406330108642578, "train/pos_loss/pos4": 2.3384342193603516, "train/pos_loss/pos5": 2.8818063735961914, "train/pos_loss/pos6": 3.2255618572235107, "train/pos_loss/weight0": 3.0420103073120117, "train/pos_loss/weight1": 2.127838134765625, "train/pos_loss/weight2": 1.1298496723175049, "train/pos_loss/weight3": 0.4854661524295807, "train/pos_loss/weight4": 0.16645526885986328, "train/pos_loss/weight5": 0.03864492475986481, "train/pos_loss/weight6": 0.009736277163028717, "train/pos_weight_progress": 0.15165876777251186, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8025226593017578 }, { "epoch": 1.5165876777251186, "step": 640, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 6.96315336227417, "train/clean_loss_weight": 1.392630696296692, "train/combined_loss": 0.697675347328186, "train/pos_loss/pos0": 0.25624752044677734, "train/pos_loss/pos1": 0.5133180022239685, "train/pos_loss/pos2": 0.8738393783569336, "train/pos_loss/pos3": 1.2761247158050537, "train/pos_loss/pos4": 1.6987941265106201, "train/pos_loss/pos5": 2.0420446395874023, "train/pos_loss/pos6": 2.3706858158111572, "train/pos_loss/weight0": 2.9500114917755127, "train/pos_loss/weight1": 2.0256948471069336, "train/pos_loss/weight2": 1.165476679801941, "train/pos_loss/weight3": 0.5615454316139221, "train/pos_loss/weight4": 0.2114890217781067, "train/pos_loss/weight5": 0.06605149805545807, "train/pos_loss/weight6": 0.019730892032384872, "train/pos_weight_progress": 0.15165876777251186, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5813961029052734 }, { "epoch": 1.518957345971564, "step": 641, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 8.598381042480469, "train/clean_loss_weight": 1.7196762561798096, "train/combined_loss": 0.735564649105072, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.15189573459715638, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6129705309867859 }, { "epoch": 1.518957345971564, "step": 641, "train/alpha": 0.2, "train/clean_ce_loss": 0.1015625, "train/clean_loss_scale": 10.885589599609375, "train/clean_loss_weight": 2.1771180629730225, "train/combined_loss": 1.3266812562942505, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.15189573459715638, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1055676937103271 }, { "epoch": 1.518957345971564, "step": 641, "train/alpha": 0.2, "train/clean_ce_loss": 0.11767578125, "train/clean_loss_scale": 10.270251274108887, "train/clean_loss_weight": 2.0540502071380615, "train/combined_loss": 1.4502718448638916, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.15189573459715638, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2085598707199097 }, { "epoch": 1.518957345971564, "step": 641, "train/alpha": 0.2, "train/clean_ce_loss": 0.01904296875, "train/clean_loss_scale": 14.090583801269531, "train/clean_loss_weight": 2.8181169033050537, "train/combined_loss": 0.32199186086654663, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.15189573459715638, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2683265507221222 }, { "epoch": 1.5213270142180095, "step": 642, "train/alpha": 0.2, "train/clean_ce_loss": 0.024169921875, "train/clean_loss_scale": 19.79620361328125, "train/clean_loss_weight": 3.959240674972534, "train/combined_loss": 0.5741672515869141, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.15213270142180094, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4784727096557617 }, { "epoch": 1.5213270142180095, "step": 642, "train/alpha": 0.2, "train/clean_ce_loss": 0.03662109375, "train/clean_loss_scale": 13.832303047180176, "train/clean_loss_weight": 2.766460657119751, "train/combined_loss": 0.6078648567199707, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.15213270142180094, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5065540671348572 }, { "epoch": 1.5213270142180095, "step": 642, "train/alpha": 0.2, "train/clean_ce_loss": 0.06689453125, "train/clean_loss_scale": 9.691421508789062, "train/clean_loss_weight": 1.9382842779159546, "train/combined_loss": 0.7779636979103088, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.15213270142180094, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6483030915260315 }, { "epoch": 1.5213270142180095, "step": 642, "train/alpha": 0.2, "train/clean_ce_loss": 0.06787109375, "train/clean_loss_scale": 9.472810745239258, "train/clean_loss_weight": 1.8945621252059937, "train/combined_loss": 0.7715160250663757, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.15213270142180094, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6429300308227539 }, { "epoch": 1.5236966824644549, "step": 643, "train/alpha": 0.2, "train/clean_ce_loss": 0.0069580078125, "train/clean_loss_scale": 48.949363708496094, "train/clean_loss_weight": 9.789873123168945, "train/combined_loss": 0.4087080657482147, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.1523696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.34059005975723267 }, { "epoch": 1.5236966824644549, "step": 643, "train/alpha": 0.2, "train/clean_ce_loss": 0.11474609375, "train/clean_loss_scale": 8.325892448425293, "train/clean_loss_weight": 1.6651785373687744, "train/combined_loss": 1.146436333656311, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.1523696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9553636312484741 }, { "epoch": 1.5236966824644549, "step": 643, "train/alpha": 0.2, "train/clean_ce_loss": 0.0830078125, "train/clean_loss_scale": 8.299177169799805, "train/clean_loss_weight": 1.6598354578018188, "train/combined_loss": 0.8266758322715759, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.1523696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6888965368270874 }, { "epoch": 1.5236966824644549, "step": 643, "train/alpha": 0.2, "train/clean_ce_loss": 0.02001953125, "train/clean_loss_scale": 21.254438400268555, "train/clean_loss_weight": 4.250887870788574, "train/combined_loss": 0.5106046795845032, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.1523696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4255038797855377 }, { "epoch": 1.5260663507109005, "step": 644, "train/alpha": 0.2, "train/clean_ce_loss": 0.07763671875, "train/clean_loss_scale": 9.766255378723145, "train/clean_loss_weight": 1.9532511234283447, "train/combined_loss": 0.9098640084266663, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.15260663507109004, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7582200169563293 }, { "epoch": 1.5260663507109005, "step": 644, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 9.086472511291504, "train/clean_loss_weight": 1.8172944784164429, "train/combined_loss": 0.9530147314071655, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.15260663507109004, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7941789627075195 }, { "epoch": 1.5260663507109005, "step": 644, "train/alpha": 0.2, "train/clean_ce_loss": 0.08251953125, "train/clean_loss_scale": 7.032686710357666, "train/clean_loss_weight": 1.406537413597107, "train/combined_loss": 0.6964008212089539, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.15260663507109004, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5803340077400208 }, { "epoch": 1.5260663507109005, "step": 644, "train/alpha": 0.2, "train/clean_ce_loss": 0.0869140625, "train/clean_loss_scale": 9.361950874328613, "train/clean_loss_weight": 1.8723901510238647, "train/combined_loss": 0.9764221906661987, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.15260663507109004, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8136851787567139 }, { "epoch": 1.528436018957346, "grad_norm": 29.475830078125, "learning_rate": 2e-06, "loss": 3.5768, "step": 645 }, { "epoch": 1.528436018957346, "step": 645, "train/alpha": 0.2, "train/clean_ce_loss": 0.00933837890625, "train/clean_loss_scale": 29.176374435424805, "train/clean_loss_weight": 5.835275173187256, "train/combined_loss": 0.3269520401954651, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.1528436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.27246004343032837 }, { "epoch": 1.528436018957346, "step": 645, "train/alpha": 0.2, "train/clean_ce_loss": 0.0185546875, "train/clean_loss_scale": 15.976473808288574, "train/clean_loss_weight": 3.1952948570251465, "train/combined_loss": 0.3557261824607849, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.1528436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2964384853839874 }, { "epoch": 1.528436018957346, "step": 645, "train/alpha": 0.2, "train/clean_ce_loss": 0.0859375, "train/clean_loss_scale": 7.400089740753174, "train/clean_loss_weight": 1.4800180196762085, "train/combined_loss": 0.763134241104126, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.1528436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.635945200920105 }, { "epoch": 1.528436018957346, "step": 645, "train/alpha": 0.2, "train/clean_ce_loss": 0.01312255859375, "train/clean_loss_scale": 21.04816246032715, "train/clean_loss_weight": 4.209632396697998, "train/combined_loss": 0.33144688606262207, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.1528436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.27620574831962585 }, { "epoch": 1.5308056872037916, "step": 646, "train/alpha": 0.2, "train/clean_ce_loss": 0.02099609375, "train/clean_loss_scale": 12.355572700500488, "train/clean_loss_weight": 2.4711146354675293, "train/combined_loss": 0.31130251288414, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.15308056872037915, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2594187557697296 }, { "epoch": 1.5308056872037916, "step": 646, "train/alpha": 0.2, "train/clean_ce_loss": 0.059814453125, "train/clean_loss_scale": 10.708943367004395, "train/clean_loss_weight": 2.1417887210845947, "train/combined_loss": 0.7686595320701599, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.15308056872037915, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6405496001243591 }, { "epoch": 1.5308056872037916, "step": 646, "train/alpha": 0.2, "train/clean_ce_loss": 0.06396484375, "train/clean_loss_scale": 10.283559799194336, "train/clean_loss_weight": 2.0567119121551514, "train/combined_loss": 0.7893435955047607, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.15308056872037915, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6577863097190857 }, { "epoch": 1.5308056872037916, "step": 646, "train/alpha": 0.2, "train/clean_ce_loss": 0.03271484375, "train/clean_loss_scale": 14.255328178405762, "train/clean_loss_weight": 2.8510656356811523, "train/combined_loss": 0.5596330165863037, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.15308056872037915, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46636083722114563 }, { "epoch": 1.533175355450237, "step": 647, "train/alpha": 0.2, "train/clean_ce_loss": 0.052978515625, "train/clean_loss_scale": 9.185341835021973, "train/clean_loss_weight": 1.8370684385299683, "train/combined_loss": 0.583950936794281, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.1533175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4866257905960083 }, { "epoch": 1.533175355450237, "step": 647, "train/alpha": 0.2, "train/clean_ce_loss": 0.055908203125, "train/clean_loss_scale": 10.314372062683105, "train/clean_loss_weight": 2.0628745555877686, "train/combined_loss": 0.6919896006584167, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.1533175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5766580104827881 }, { "epoch": 1.533175355450237, "step": 647, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 15.397058486938477, "train/clean_loss_weight": 3.079411745071411, "train/combined_loss": 1.4705393314361572, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.1533175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2254494428634644 }, { "epoch": 1.533175355450237, "step": 647, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 10.353819847106934, "train/clean_loss_weight": 2.0707640647888184, "train/combined_loss": 0.9039369821548462, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.1533175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7532808184623718 }, { "epoch": 1.5355450236966823, "step": 648, "train/alpha": 0.2, "train/clean_ce_loss": 0.111328125, "train/clean_loss_scale": 7.772326469421387, "train/clean_loss_weight": 1.5544652938842773, "train/combined_loss": 1.0383342504501343, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.15355450236966825, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8652785420417786 }, { "epoch": 1.5355450236966823, "step": 648, "train/alpha": 0.2, "train/clean_ce_loss": 0.041015625, "train/clean_loss_scale": 9.389570236206055, "train/clean_loss_weight": 1.8779140710830688, "train/combined_loss": 0.4621428847312927, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.15355450236966825, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.38511908054351807 }, { "epoch": 1.5355450236966823, "step": 648, "train/alpha": 0.2, "train/clean_ce_loss": 0.09814453125, "train/clean_loss_scale": 8.35693645477295, "train/clean_loss_weight": 1.6713873147964478, "train/combined_loss": 0.9842250943183899, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.15355450236966825, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8201875686645508 }, { "epoch": 1.5355450236966823, "step": 648, "train/alpha": 0.2, "train/clean_ce_loss": 0.1337890625, "train/clean_loss_scale": 7.8631815910339355, "train/clean_loss_weight": 1.572636365890503, "train/combined_loss": 1.2624092102050781, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.15355450236966825, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0520076751708984 }, { "epoch": 1.537914691943128, "step": 649, "train/alpha": 0.2, "train/clean_ce_loss": 0.04150390625, "train/clean_loss_scale": 13.229548454284668, "train/clean_loss_weight": 2.6459097862243652, "train/combined_loss": 0.6588935256004333, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.1537914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5490779280662537 }, { "epoch": 1.537914691943128, "step": 649, "train/alpha": 0.2, "train/clean_ce_loss": 0.00799560546875, "train/clean_loss_scale": 64.47064208984375, "train/clean_loss_weight": 12.894128799438477, "train/combined_loss": 0.6185781955718994, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.1537914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5154818296432495 }, { "epoch": 1.537914691943128, "step": 649, "train/alpha": 0.2, "train/clean_ce_loss": 0.002899169921875, "train/clean_loss_scale": 120.29108428955078, "train/clean_loss_weight": 24.058218002319336, "train/combined_loss": 0.4184931516647339, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.1537914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.34874430298805237 }, { "epoch": 1.537914691943128, "step": 649, "train/alpha": 0.2, "train/clean_ce_loss": 0.09326171875, "train/clean_loss_scale": 8.462162971496582, "train/clean_loss_weight": 1.6924326419830322, "train/combined_loss": 0.947035014629364, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.1537914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7891958355903625 }, { "epoch": 1.5402843601895735, "grad_norm": 26.844295501708984, "learning_rate": 2e-06, "loss": 3.2611, "step": 650 }, { "epoch": 1.5402843601895735, "step": 650, "train/alpha": 0.2, "train/clean_ce_loss": 0.061767578125, "train/clean_loss_scale": 10.10055160522461, "train/clean_loss_weight": 2.0201103687286377, "train/combined_loss": 0.7486639022827148, "train/pos_loss/pos0": 0.2149120420217514, "train/pos_loss/pos1": 0.4354352653026581, "train/pos_loss/pos2": 0.7869887948036194, "train/pos_loss/pos3": 1.105970025062561, "train/pos_loss/pos4": 1.6662788391113281, "train/pos_loss/pos5": 1.9534614086151123, "train/pos_loss/pos6": 2.412533760070801, "train/pos_loss/weight0": 2.520801544189453, "train/pos_loss/weight1": 1.8426754474639893, "train/pos_loss/weight2": 1.2280082702636719, "train/pos_loss/weight3": 0.7398027777671814, "train/pos_loss/weight4": 0.3868807554244995, "train/pos_loss/weight5": 0.18620380759239197, "train/pos_loss/weight6": 0.09562775492668152, "train/pos_weight_progress": 0.15402843601895735, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6238865852355957 }, { "epoch": 1.5402843601895735, "step": 650, "train/alpha": 0.2, "train/clean_ce_loss": 0.099609375, "train/clean_loss_scale": 7.850640296936035, "train/clean_loss_weight": 1.570128083229065, "train/combined_loss": 0.9383968710899353, "train/pos_loss/pos0": 0.3010401129722595, "train/pos_loss/pos1": 0.57672518491745, "train/pos_loss/pos2": 0.9482024908065796, "train/pos_loss/pos3": 1.4031528234481812, "train/pos_loss/pos4": 1.8424814939498901, "train/pos_loss/pos5": 2.250504970550537, "train/pos_loss/pos6": 2.553234815597534, "train/pos_loss/weight0": 2.487957000732422, "train/pos_loss/weight1": 1.8256089687347412, "train/pos_loss/weight2": 1.2279748916625977, "train/pos_loss/weight3": 0.753400981426239, "train/pos_loss/weight4": 0.40702757239341736, "train/pos_loss/weight5": 0.1938772052526474, "train/pos_loss/weight6": 0.10415301471948624, "train/pos_weight_progress": 0.15402843601895735, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7819973826408386 }, { "epoch": 1.5402843601895735, "step": 650, "train/alpha": 0.2, "train/clean_ce_loss": 0.1435546875, "train/clean_loss_scale": 7.5979790687561035, "train/clean_loss_weight": 1.5195958614349365, "train/combined_loss": 1.3088706731796265, "train/pos_loss/pos0": 0.4274950623512268, "train/pos_loss/pos1": 0.8691059947013855, "train/pos_loss/pos2": 1.5063507556915283, "train/pos_loss/pos3": 2.0983800888061523, "train/pos_loss/pos4": 2.5863475799560547, "train/pos_loss/pos5": 3.026538372039795, "train/pos_loss/pos6": 3.4591784477233887, "train/pos_loss/weight0": 2.6427860260009766, "train/pos_loss/weight1": 1.8773068189620972, "train/pos_loss/weight2": 1.226373314857483, "train/pos_loss/weight3": 0.6971743702888489, "train/pos_loss/weight4": 0.3391731083393097, "train/pos_loss/weight5": 0.14893074333667755, "train/pos_loss/weight6": 0.06825514137744904, "train/pos_weight_progress": 0.15402843601895735, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0907255411148071 }, { "epoch": 1.5402843601895735, "step": 650, "train/alpha": 0.2, "train/clean_ce_loss": 0.061767578125, "train/clean_loss_scale": 7.585780143737793, "train/clean_loss_weight": 1.5171560049057007, "train/combined_loss": 0.5622663497924805, "train/pos_loss/pos0": 0.22465163469314575, "train/pos_loss/pos1": 0.40771880745887756, "train/pos_loss/pos2": 0.7230474948883057, "train/pos_loss/pos3": 1.0152318477630615, "train/pos_loss/pos4": 1.443656325340271, "train/pos_loss/pos5": 1.8843175172805786, "train/pos_loss/pos6": 2.2380318641662598, "train/pos_loss/weight0": 3.0749716758728027, "train/pos_loss/weight1": 1.9991856813430786, "train/pos_loss/weight2": 1.1389274597167969, "train/pos_loss/weight3": 0.5234773755073547, "train/pos_loss/weight4": 0.18830564618110657, "train/pos_loss/weight5": 0.05752558261156082, "train/pos_loss/weight6": 0.01760701835155487, "train/pos_weight_progress": 0.15402843601895735, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4685552716255188 }, { "epoch": 1.542654028436019, "step": 651, "train/alpha": 0.2, "train/clean_ce_loss": 0.091796875, "train/clean_loss_scale": 11.671513557434082, "train/clean_loss_weight": 2.3343026638031006, "train/combined_loss": 1.285690188407898, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.1542654028436019, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0714085102081299 }, { "epoch": 1.542654028436019, "step": 651, "train/alpha": 0.2, "train/clean_ce_loss": 0.1416015625, "train/clean_loss_scale": 9.62541389465332, "train/clean_loss_weight": 1.925082802772522, "train/combined_loss": 1.635568380355835, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.1542654028436019, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.362973690032959 }, { "epoch": 1.542654028436019, "step": 651, "train/alpha": 0.2, "train/clean_ce_loss": 0.087890625, "train/clean_loss_scale": 8.812172889709473, "train/clean_loss_weight": 1.7624346017837524, "train/combined_loss": 0.9294087886810303, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.1542654028436019, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7745073437690735 }, { "epoch": 1.542654028436019, "step": 651, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 8.717595100402832, "train/clean_loss_weight": 1.7435190677642822, "train/combined_loss": 0.6895754337310791, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.1542654028436019, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5746461749076843 }, { "epoch": 1.5450236966824644, "step": 652, "train/alpha": 0.2, "train/clean_ce_loss": 0.12353515625, "train/clean_loss_scale": 8.491971969604492, "train/clean_loss_weight": 1.6983944177627563, "train/combined_loss": 1.2588685750961304, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.15450236966824646, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.049057126045227 }, { "epoch": 1.5450236966824644, "step": 652, "train/alpha": 0.2, "train/clean_ce_loss": 0.1376953125, "train/clean_loss_scale": 8.657766342163086, "train/clean_loss_weight": 1.731553316116333, "train/combined_loss": 1.4305607080459595, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.15450236966824646, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.192133903503418 }, { "epoch": 1.5450236966824644, "step": 652, "train/alpha": 0.2, "train/clean_ce_loss": 0.00994873046875, "train/clean_loss_scale": 41.79568862915039, "train/clean_loss_weight": 8.359137535095215, "train/combined_loss": 0.49897685647010803, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.15450236966824646, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.41581404209136963 }, { "epoch": 1.5450236966824644, "step": 652, "train/alpha": 0.2, "train/clean_ce_loss": 0.10791015625, "train/clean_loss_scale": 9.72226619720459, "train/clean_loss_weight": 1.944453239440918, "train/combined_loss": 1.2589575052261353, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.15450236966824646, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0491312742233276 }, { "epoch": 1.54739336492891, "step": 653, "train/alpha": 0.2, "train/clean_ce_loss": 0.06982421875, "train/clean_loss_scale": 10.436388969421387, "train/clean_loss_weight": 2.087277889251709, "train/combined_loss": 0.8744552731513977, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.15473933649289098, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7287127375602722 }, { "epoch": 1.54739336492891, "step": 653, "train/alpha": 0.2, "train/clean_ce_loss": 0.01251220703125, "train/clean_loss_scale": 29.822099685668945, "train/clean_loss_weight": 5.964419841766357, "train/combined_loss": 0.44776833057403564, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.15473933649289098, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.37314027547836304 }, { "epoch": 1.54739336492891, "step": 653, "train/alpha": 0.2, "train/clean_ce_loss": 0.1220703125, "train/clean_loss_scale": 6.866238594055176, "train/clean_loss_weight": 1.373247742652893, "train/combined_loss": 1.0057966709136963, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.15473933649289098, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8381639122962952 }, { "epoch": 1.54739336492891, "step": 653, "train/alpha": 0.2, "train/clean_ce_loss": 0.05078125, "train/clean_loss_scale": 10.189383506774902, "train/clean_loss_weight": 2.037876844406128, "train/combined_loss": 0.6209155917167664, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.15473933649289098, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5174296498298645 }, { "epoch": 1.5497630331753554, "step": 654, "train/alpha": 0.2, "train/clean_ce_loss": 0.15625, "train/clean_loss_scale": 9.212179183959961, "train/clean_loss_weight": 1.8424358367919922, "train/combined_loss": 1.7272834777832031, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.15497630331753554, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4394029378890991 }, { "epoch": 1.5497630331753554, "step": 654, "train/alpha": 0.2, "train/clean_ce_loss": 0.09228515625, "train/clean_loss_scale": 9.7930269241333, "train/clean_loss_weight": 1.958605408668518, "train/combined_loss": 1.0845012664794922, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.15497630331753554, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.903751015663147 }, { "epoch": 1.5497630331753554, "step": 654, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 15.776995658874512, "train/clean_loss_weight": 3.1553990840911865, "train/combined_loss": 1.2202520370483398, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.15497630331753554, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0168766975402832 }, { "epoch": 1.5497630331753554, "step": 654, "train/alpha": 0.2, "train/clean_ce_loss": 0.02978515625, "train/clean_loss_scale": 30.1214656829834, "train/clean_loss_weight": 6.024293422698975, "train/combined_loss": 1.07660710811615, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.15497630331753554, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8971725702285767 }, { "epoch": 1.552132701421801, "grad_norm": 15.625990867614746, "learning_rate": 2e-06, "loss": 3.594, "step": 655 }, { "epoch": 1.552132701421801, "step": 655, "train/alpha": 0.2, "train/clean_ce_loss": 0.02734375, "train/clean_loss_scale": 19.471155166625977, "train/clean_loss_weight": 3.894231081008911, "train/combined_loss": 0.6388972401618958, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.1552132701421801, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5324143767356873 }, { "epoch": 1.552132701421801, "step": 655, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 15.399849891662598, "train/clean_loss_weight": 3.079970121383667, "train/combined_loss": 1.3444790840148926, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.1552132701421801, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1203992366790771 }, { "epoch": 1.552132701421801, "step": 655, "train/alpha": 0.2, "train/clean_ce_loss": 0.1611328125, "train/clean_loss_scale": 8.050230979919434, "train/clean_loss_weight": 1.6100462675094604, "train/combined_loss": 1.55658757686615, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.1552132701421801, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2971563339233398 }, { "epoch": 1.552132701421801, "step": 655, "train/alpha": 0.2, "train/clean_ce_loss": 0.1123046875, "train/clean_loss_scale": 7.6081085205078125, "train/clean_loss_weight": 1.5216217041015625, "train/combined_loss": 1.0253114700317383, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.1552132701421801, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8544262647628784 }, { "epoch": 1.5545023696682465, "step": 656, "train/alpha": 0.2, "train/clean_ce_loss": 0.01373291015625, "train/clean_loss_scale": 28.925518035888672, "train/clean_loss_weight": 5.785103797912598, "train/combined_loss": 0.47667786478996277, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.15545023696682464, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3972315490245819 }, { "epoch": 1.5545023696682465, "step": 656, "train/alpha": 0.2, "train/clean_ce_loss": 0.06787109375, "train/clean_loss_scale": 10.566876411437988, "train/clean_loss_weight": 2.113375425338745, "train/combined_loss": 0.8606225252151489, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.15545023696682464, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7171854376792908 }, { "epoch": 1.5545023696682465, "step": 656, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 16.79767417907715, "train/clean_loss_weight": 3.359534978866577, "train/combined_loss": 1.3287222385406494, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.15545023696682464, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1072685718536377 }, { "epoch": 1.5545023696682465, "step": 656, "train/alpha": 0.2, "train/clean_ce_loss": 0.04638671875, "train/clean_loss_scale": 21.423660278320312, "train/clean_loss_weight": 4.284732341766357, "train/combined_loss": 1.1925280094146729, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.15545023696682464, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9937732815742493 }, { "epoch": 1.556872037914692, "step": 657, "train/alpha": 0.2, "train/clean_ce_loss": 0.02880859375, "train/clean_loss_scale": 15.859213829040527, "train/clean_loss_weight": 3.1718428134918213, "train/combined_loss": 0.5482579469680786, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.1556872037914692, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45688164234161377 }, { "epoch": 1.556872037914692, "step": 657, "train/alpha": 0.2, "train/clean_ce_loss": 0.08251953125, "train/clean_loss_scale": 12.642437934875488, "train/clean_loss_weight": 2.5284876823425293, "train/combined_loss": 1.2518976926803589, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.1556872037914692, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0432480573654175 }, { "epoch": 1.556872037914692, "step": 657, "train/alpha": 0.2, "train/clean_ce_loss": 0.10693359375, "train/clean_loss_scale": 8.751191139221191, "train/clean_loss_weight": 1.750238299369812, "train/combined_loss": 1.122955560684204, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.1556872037914692, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.935796320438385 }, { "epoch": 1.556872037914692, "step": 657, "train/alpha": 0.2, "train/clean_ce_loss": 0.0277099609375, "train/clean_loss_scale": 14.187909126281738, "train/clean_loss_weight": 2.8375818729400635, "train/combined_loss": 0.47177568078041077, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.1556872037914692, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3931463956832886 }, { "epoch": 1.5592417061611374, "step": 658, "train/alpha": 0.2, "train/clean_ce_loss": 0.1748046875, "train/clean_loss_scale": 7.630547523498535, "train/clean_loss_weight": 1.5261095762252808, "train/combined_loss": 1.6006265878677368, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.15592417061611374, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3338555097579956 }, { "epoch": 1.5592417061611374, "step": 658, "train/alpha": 0.2, "train/clean_ce_loss": 0.119140625, "train/clean_loss_scale": 10.755614280700684, "train/clean_loss_weight": 2.151122808456421, "train/combined_loss": 1.5377167463302612, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.15592417061611374, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2814306020736694 }, { "epoch": 1.5592417061611374, "step": 658, "train/alpha": 0.2, "train/clean_ce_loss": 0.087890625, "train/clean_loss_scale": 8.029352188110352, "train/clean_loss_weight": 1.6058704853057861, "train/combined_loss": 0.8468457460403442, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.15592417061611374, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7057047486305237 }, { "epoch": 1.5592417061611374, "step": 658, "train/alpha": 0.2, "train/clean_ce_loss": 0.0830078125, "train/clean_loss_scale": 8.190359115600586, "train/clean_loss_weight": 1.638071894645691, "train/combined_loss": 0.8158365488052368, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.15592417061611374, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6798638105392456 }, { "epoch": 1.5616113744075828, "step": 659, "train/alpha": 0.2, "train/clean_ce_loss": 0.10302734375, "train/clean_loss_scale": 10.469780921936035, "train/clean_loss_weight": 2.093956232070923, "train/combined_loss": 1.2944084405899048, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.1561611374407583, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0786737203598022 }, { "epoch": 1.5616113744075828, "step": 659, "train/alpha": 0.2, "train/clean_ce_loss": 0.08251953125, "train/clean_loss_scale": 12.170486450195312, "train/clean_loss_weight": 2.4340972900390625, "train/combined_loss": 1.2051634788513184, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.1561611374407583, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0043028593063354 }, { "epoch": 1.5616113744075828, "step": 659, "train/alpha": 0.2, "train/clean_ce_loss": 0.1123046875, "train/clean_loss_scale": 11.49555492401123, "train/clean_loss_weight": 2.2991111278533936, "train/combined_loss": 1.5492055416107178, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.1561611374407583, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2910046577453613 }, { "epoch": 1.5616113744075828, "step": 659, "train/alpha": 0.2, "train/clean_ce_loss": 0.095703125, "train/clean_loss_scale": 9.577102661132812, "train/clean_loss_weight": 1.9154205322265625, "train/combined_loss": 1.0998704433441162, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.1561611374407583, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9165586829185486 }, { "epoch": 1.5639810426540284, "grad_norm": 19.642744064331055, "learning_rate": 2e-06, "loss": 3.2429, "step": 660 }, { "epoch": 1.5639810426540284, "step": 660, "train/alpha": 0.2, "train/clean_ce_loss": 0.06689453125, "train/clean_loss_scale": 16.947349548339844, "train/clean_loss_weight": 3.389469861984253, "train/combined_loss": 1.3604220151901245, "train/pos_loss/pos0": 0.3490205407142639, "train/pos_loss/pos1": 0.8111486434936523, "train/pos_loss/pos2": 1.3452521562576294, "train/pos_loss/pos3": 1.972713828086853, "train/pos_loss/pos4": 2.6652042865753174, "train/pos_loss/pos5": 2.9020533561706543, "train/pos_loss/pos6": 3.6236510276794434, "train/pos_loss/weight0": 2.3637728691101074, "train/pos_loss/weight1": 1.7591646909713745, "train/pos_loss/weight2": 1.2352349758148193, "train/pos_loss/weight3": 0.7880960702896118, "train/pos_loss/weight4": 0.45860612392425537, "train/pos_loss/weight5": 0.25189974904060364, "train/pos_loss/weight6": 0.14322540163993835, "train/pos_weight_progress": 0.15639810426540285, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1336849927902222 }, { "epoch": 1.5639810426540284, "step": 660, "train/alpha": 0.2, "train/clean_ce_loss": 0.10498046875, "train/clean_loss_scale": 5.740818500518799, "train/clean_loss_weight": 1.1481636762619019, "train/combined_loss": 0.7232086062431335, "train/pos_loss/pos0": 0.23267357051372528, "train/pos_loss/pos1": 0.5510014295578003, "train/pos_loss/pos2": 0.9499244093894958, "train/pos_loss/pos3": 1.4629400968551636, "train/pos_loss/pos4": 1.954463243484497, "train/pos_loss/pos5": 2.361898422241211, "train/pos_loss/pos6": 2.741527557373047, "train/pos_loss/weight0": 2.997734308242798, "train/pos_loss/weight1": 2.061619520187378, "train/pos_loss/weight2": 1.1759071350097656, "train/pos_loss/weight3": 0.5178667306900024, "train/pos_loss/weight4": 0.1845533698797226, "train/pos_loss/weight5": 0.049663182348012924, "train/pos_loss/weight6": 0.01265615876764059, "train/pos_weight_progress": 0.15639810426540285, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6026738286018372 }, { "epoch": 1.5639810426540284, "step": 660, "train/alpha": 0.2, "train/clean_ce_loss": 0.1826171875, "train/clean_loss_scale": 7.175307750701904, "train/clean_loss_weight": 1.4350615739822388, "train/combined_loss": 1.572401523590088, "train/pos_loss/pos0": 0.580802857875824, "train/pos_loss/pos1": 1.1389806270599365, "train/pos_loss/pos2": 1.7971769571304321, "train/pos_loss/pos3": 2.3510687351226807, "train/pos_loss/pos4": 2.9621317386627197, "train/pos_loss/pos5": 3.4673163890838623, "train/pos_loss/pos6": 3.7714884281158447, "train/pos_loss/weight0": 2.6580464839935303, "train/pos_loss/weight1": 1.9117668867111206, "train/pos_loss/weight2": 1.2304478883743286, "train/pos_loss/weight3": 0.6899358034133911, "train/pos_loss/weight4": 0.32401198148727417, "train/pos_loss/weight5": 0.12883931398391724, "train/pos_loss/weight6": 0.05695183947682381, "train/pos_weight_progress": 0.15639810426540285, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.31033456325531 }, { "epoch": 1.5639810426540284, "step": 660, "train/alpha": 0.2, "train/clean_ce_loss": 0.09765625, "train/clean_loss_scale": 5.695390701293945, "train/clean_loss_weight": 1.139078140258789, "train/combined_loss": 0.6674286127090454, "train/pos_loss/pos0": 0.2936876118183136, "train/pos_loss/pos1": 0.5921950936317444, "train/pos_loss/pos2": 0.9202165603637695, "train/pos_loss/pos3": 1.3380873203277588, "train/pos_loss/pos4": 1.8127501010894775, "train/pos_loss/pos5": 2.302534818649292, "train/pos_loss/pos6": 2.478008508682251, "train/pos_loss/weight0": 3.4302523136138916, "train/pos_loss/weight1": 2.073606014251709, "train/pos_loss/weight2": 0.9942930340766907, "train/pos_loss/weight3": 0.3745286464691162, "train/pos_loss/weight4": 0.10405804216861725, "train/pos_loss/weight5": 0.01988055743277073, "train/pos_loss/weight6": 0.003381924470886588, "train/pos_weight_progress": 0.15639810426540285, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5561904907226562 }, { "epoch": 1.566350710900474, "step": 661, "train/alpha": 0.2, "train/clean_ce_loss": 0.08203125, "train/clean_loss_scale": 8.398185729980469, "train/clean_loss_weight": 1.6796371936798096, "train/combined_loss": 0.8266963958740234, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.1566350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6889136433601379 }, { "epoch": 1.566350710900474, "step": 661, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 6.404473781585693, "train/clean_loss_weight": 1.2808947563171387, "train/combined_loss": 0.4953460395336151, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.1566350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.41278836131095886 }, { "epoch": 1.566350710900474, "step": 661, "train/alpha": 0.2, "train/clean_ce_loss": 0.09130859375, "train/clean_loss_scale": 11.855348587036133, "train/clean_loss_weight": 2.3710696697235107, "train/combined_loss": 1.2989943027496338, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.1566350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0824952125549316 }, { "epoch": 1.566350710900474, "step": 661, "train/alpha": 0.2, "train/clean_ce_loss": 0.051513671875, "train/clean_loss_scale": 17.010671615600586, "train/clean_loss_weight": 3.402134418487549, "train/combined_loss": 1.0515385866165161, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.1566350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8762821555137634 }, { "epoch": 1.5687203791469195, "step": 662, "train/alpha": 0.2, "train/clean_ce_loss": 0.049560546875, "train/clean_loss_scale": 16.600862503051758, "train/clean_loss_weight": 3.3201725482940674, "train/combined_loss": 0.9872974157333374, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.15687203791469195, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8227478265762329 }, { "epoch": 1.5687203791469195, "step": 662, "train/alpha": 0.2, "train/clean_ce_loss": 0.06787109375, "train/clean_loss_scale": 7.782768249511719, "train/clean_loss_weight": 1.5565537214279175, "train/combined_loss": 0.633870005607605, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.15687203791469195, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5282250046730042 }, { "epoch": 1.5687203791469195, "step": 662, "train/alpha": 0.2, "train/clean_ce_loss": 0.123046875, "train/clean_loss_scale": 7.270725250244141, "train/clean_loss_weight": 1.454145073890686, "train/combined_loss": 1.0735679864883423, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.15687203791469195, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8946400284767151 }, { "epoch": 1.5687203791469195, "step": 662, "train/alpha": 0.2, "train/clean_ce_loss": 0.11376953125, "train/clean_loss_scale": 6.891319274902344, "train/clean_loss_weight": 1.3782638311386108, "train/combined_loss": 0.9408265948295593, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.15687203791469195, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7840221524238586 }, { "epoch": 1.5710900473933649, "step": 663, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 6.55811882019043, "train/clean_loss_weight": 1.3116238117218018, "train/combined_loss": 0.730103075504303, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.1571090047393365, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6084192395210266 }, { "epoch": 1.5710900473933649, "step": 663, "train/alpha": 0.2, "train/clean_ce_loss": 0.09326171875, "train/clean_loss_scale": 7.120792865753174, "train/clean_loss_weight": 1.4241585731506348, "train/combined_loss": 0.7969168424606323, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.1571090047393365, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6640973687171936 }, { "epoch": 1.5710900473933649, "step": 663, "train/alpha": 0.2, "train/clean_ce_loss": 0.0966796875, "train/clean_loss_scale": 9.089131355285645, "train/clean_loss_weight": 1.817826271057129, "train/combined_loss": 1.0544812679290771, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.1571090047393365, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8787344098091125 }, { "epoch": 1.5710900473933649, "step": 663, "train/alpha": 0.2, "train/clean_ce_loss": 0.1572265625, "train/clean_loss_scale": 6.970402717590332, "train/clean_loss_weight": 1.3940805196762085, "train/combined_loss": 1.3151190280914307, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.1571090047393365, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0959324836730957 }, { "epoch": 1.5734597156398105, "step": 664, "train/alpha": 0.2, "train/clean_ce_loss": 0.16015625, "train/clean_loss_scale": 8.873732566833496, "train/clean_loss_weight": 1.7747465372085571, "train/combined_loss": 1.7054204940795898, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.15734597156398103, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.421183705329895 }, { "epoch": 1.5734597156398105, "step": 664, "train/alpha": 0.2, "train/clean_ce_loss": 0.0595703125, "train/clean_loss_scale": 10.02884578704834, "train/clean_loss_weight": 2.0057692527770996, "train/combined_loss": 0.7169057726860046, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.15734597156398103, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5974214673042297 }, { "epoch": 1.5734597156398105, "step": 664, "train/alpha": 0.2, "train/clean_ce_loss": 0.11181640625, "train/clean_loss_scale": 6.627654552459717, "train/clean_loss_weight": 1.3255308866500854, "train/combined_loss": 0.8892966508865356, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.15734597156398103, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7410805225372314 }, { "epoch": 1.5734597156398105, "step": 664, "train/alpha": 0.2, "train/clean_ce_loss": 0.10693359375, "train/clean_loss_scale": 10.217309951782227, "train/clean_loss_weight": 2.043462038040161, "train/combined_loss": 1.3110883235931396, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.15734597156398103, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.092573642730713 }, { "epoch": 1.5758293838862558, "grad_norm": 15.263248443603516, "learning_rate": 2e-06, "loss": 3.3581, "step": 665 }, { "epoch": 1.5758293838862558, "step": 665, "train/alpha": 0.2, "train/clean_ce_loss": 0.00372314453125, "train/clean_loss_scale": 62.54399871826172, "train/clean_loss_weight": 12.50879955291748, "train/combined_loss": 0.27943241596221924, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.15758293838862558, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.23286034166812897 }, { "epoch": 1.5758293838862558, "step": 665, "train/alpha": 0.2, "train/clean_ce_loss": 0.12255859375, "train/clean_loss_scale": 9.249985694885254, "train/clean_loss_weight": 1.8499971628189087, "train/combined_loss": 1.360398292541504, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.15758293838862558, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1336652040481567 }, { "epoch": 1.5758293838862558, "step": 665, "train/alpha": 0.2, "train/clean_ce_loss": 0.05029296875, "train/clean_loss_scale": 19.815317153930664, "train/clean_loss_weight": 3.9630634784698486, "train/combined_loss": 1.1958853006362915, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.15758293838862558, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9965711236000061 }, { "epoch": 1.5758293838862558, "step": 665, "train/alpha": 0.2, "train/clean_ce_loss": 0.1630859375, "train/clean_loss_scale": 7.611677646636963, "train/clean_loss_weight": 1.5223355293273926, "train/combined_loss": 1.4896290302276611, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.15758293838862558, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2413575649261475 }, { "epoch": 1.5781990521327014, "step": 666, "train/alpha": 0.2, "train/clean_ce_loss": 0.01318359375, "train/clean_loss_scale": 21.269216537475586, "train/clean_loss_weight": 4.253843307495117, "train/combined_loss": 0.3364856541156769, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.15781990521327013, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2804047167301178 }, { "epoch": 1.5781990521327014, "step": 666, "train/alpha": 0.2, "train/clean_ce_loss": 0.032958984375, "train/clean_loss_scale": 16.337547302246094, "train/clean_loss_weight": 3.2675094604492188, "train/combined_loss": 0.646162748336792, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.15781990521327013, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5384689569473267 }, { "epoch": 1.5781990521327014, "step": 666, "train/alpha": 0.2, "train/clean_ce_loss": 0.1455078125, "train/clean_loss_scale": 8.298340797424316, "train/clean_loss_weight": 1.659668207168579, "train/combined_loss": 1.4489680528640747, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.15781990521327013, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2074733972549438 }, { "epoch": 1.5781990521327014, "step": 666, "train/alpha": 0.2, "train/clean_ce_loss": 0.1669921875, "train/clean_loss_scale": 7.771694183349609, "train/clean_loss_weight": 1.554338812828064, "train/combined_loss": 1.5573747158050537, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.15781990521327013, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2978122234344482 }, { "epoch": 1.580568720379147, "step": 667, "train/alpha": 0.2, "train/clean_ce_loss": 0.08837890625, "train/clean_loss_scale": 10.92300796508789, "train/clean_loss_weight": 2.1846015453338623, "train/combined_loss": 1.1584361791610718, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.1580568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9653635025024414 }, { "epoch": 1.580568720379147, "step": 667, "train/alpha": 0.2, "train/clean_ce_loss": 0.05078125, "train/clean_loss_scale": 17.330663681030273, "train/clean_loss_weight": 3.466132879257202, "train/combined_loss": 1.0560873746871948, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.1580568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8800727725028992 }, { "epoch": 1.580568720379147, "step": 667, "train/alpha": 0.2, "train/clean_ce_loss": 0.00360107421875, "train/clean_loss_scale": 75.24285125732422, "train/clean_loss_weight": 15.04857063293457, "train/combined_loss": 0.3251461088657379, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.1580568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.27095508575439453 }, { "epoch": 1.580568720379147, "step": 667, "train/alpha": 0.2, "train/clean_ce_loss": 4.172325134277344e-05, "train/clean_loss_scale": 3492.098876953125, "train/clean_loss_weight": 698.4197998046875, "train/combined_loss": 0.17484205961227417, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.1580568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.14570172131061554 }, { "epoch": 1.5829383886255926, "step": 668, "train/alpha": 0.2, "train/clean_ce_loss": 0.002655029296875, "train/clean_loss_scale": 98.77843475341797, "train/clean_loss_weight": 19.755687713623047, "train/combined_loss": 0.3147115707397461, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.15829383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2622596323490143 }, { "epoch": 1.5829383886255926, "step": 668, "train/alpha": 0.2, "train/clean_ce_loss": 0.08837890625, "train/clean_loss_scale": 12.941802978515625, "train/clean_loss_weight": 2.588360548019409, "train/combined_loss": 1.3725388050079346, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.15829383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.143782377243042 }, { "epoch": 1.5829383886255926, "step": 668, "train/alpha": 0.2, "train/clean_ce_loss": 0.1103515625, "train/clean_loss_scale": 8.855393409729004, "train/clean_loss_weight": 1.7710787057876587, "train/combined_loss": 1.1726477146148682, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.15829383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9772064685821533 }, { "epoch": 1.5829383886255926, "step": 668, "train/alpha": 0.2, "train/clean_ce_loss": 0.0869140625, "train/clean_loss_scale": 7.881575107574463, "train/clean_loss_weight": 1.5763150453567505, "train/combined_loss": 0.8220236897468567, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.15829383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6850197315216064 }, { "epoch": 1.585308056872038, "step": 669, "train/alpha": 0.2, "train/clean_ce_loss": 0.1337890625, "train/clean_loss_scale": 9.607843399047852, "train/clean_loss_weight": 1.921568751335144, "train/combined_loss": 1.5425091981887817, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.1585308056872038, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2854243516921997 }, { "epoch": 1.585308056872038, "step": 669, "train/alpha": 0.2, "train/clean_ce_loss": 0.10498046875, "train/clean_loss_scale": 7.689517498016357, "train/clean_loss_weight": 1.5379035472869873, "train/combined_loss": 0.9686989784240723, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.1585308056872038, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.807249128818512 }, { "epoch": 1.585308056872038, "step": 669, "train/alpha": 0.2, "train/clean_ce_loss": 0.00058746337890625, "train/clean_loss_scale": 518.9761352539062, "train/clean_loss_weight": 103.79522705078125, "train/combined_loss": 0.36585533618927, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.1585308056872038, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.30487945675849915 }, { "epoch": 1.585308056872038, "step": 669, "train/alpha": 0.2, "train/clean_ce_loss": 0.08203125, "train/clean_loss_scale": 10.79713249206543, "train/clean_loss_weight": 2.15942645072937, "train/combined_loss": 1.0628427267074585, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.1585308056872038, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8857022523880005 }, { "epoch": 1.5876777251184833, "grad_norm": 14.640969276428223, "learning_rate": 2e-06, "loss": 3.34, "step": 670 }, { "epoch": 1.5876777251184833, "step": 670, "train/alpha": 0.2, "train/clean_ce_loss": 0.01324462890625, "train/clean_loss_scale": 11.584402084350586, "train/clean_loss_weight": 2.316880464553833, "train/combined_loss": 0.18411731719970703, "train/pos_loss/pos0": 0.0014978839317336679, "train/pos_loss/pos1": 0.010639647953212261, "train/pos_loss/pos2": 0.130869522690773, "train/pos_loss/pos3": 0.3211691379547119, "train/pos_loss/pos4": 0.783924400806427, "train/pos_loss/pos5": 1.2870315313339233, "train/pos_loss/pos6": 2.119126319885254, "train/pos_loss/weight0": 2.6180481910705566, "train/pos_loss/weight1": 1.8582167625427246, "train/pos_loss/weight2": 1.2160612344741821, "train/pos_loss/weight3": 0.6992285847663879, "train/pos_loss/weight4": 0.3639608323574066, "train/pos_loss/weight5": 0.16558727622032166, "train/pos_loss/weight6": 0.07889716327190399, "train/pos_weight_progress": 0.15876777251184834, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.15343110263347626 }, { "epoch": 1.5876777251184833, "step": 670, "train/alpha": 0.2, "train/clean_ce_loss": 0.04052734375, "train/clean_loss_scale": 12.832810401916504, "train/clean_loss_weight": 2.5665621757507324, "train/combined_loss": 0.6240956783294678, "train/pos_loss/pos0": 0.055905215442180634, "train/pos_loss/pos1": 0.1074909120798111, "train/pos_loss/pos2": 0.3393357992172241, "train/pos_loss/pos3": 0.6973661184310913, "train/pos_loss/pos4": 1.0779744386672974, "train/pos_loss/pos5": 1.6545459032058716, "train/pos_loss/pos6": 2.195739507675171, "train/pos_loss/weight0": 1.8730307817459106, "train/pos_loss/weight1": 1.5008583068847656, "train/pos_loss/weight2": 1.197251558303833, "train/pos_loss/weight3": 0.9002190828323364, "train/pos_loss/weight4": 0.6785385012626648, "train/pos_loss/weight5": 0.47450587153434753, "train/pos_loss/weight6": 0.3755960762500763, "train/pos_weight_progress": 0.15876777251184834, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5200797319412231 }, { "epoch": 1.5876777251184833, "step": 670, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 11.639182090759277, "train/clean_loss_weight": 2.327836513519287, "train/combined_loss": 1.1116328239440918, "train/pos_loss/pos0": 0.2603442966938019, "train/pos_loss/pos1": 0.6282193064689636, "train/pos_loss/pos2": 0.9547361731529236, "train/pos_loss/pos3": 1.487733244895935, "train/pos_loss/pos4": 1.8426644802093506, "train/pos_loss/pos5": 2.3747270107269287, "train/pos_loss/pos6": 2.7021567821502686, "train/pos_loss/weight0": 2.139254331588745, "train/pos_loss/weight1": 1.6407727003097534, "train/pos_loss/weight2": 1.234287142753601, "train/pos_loss/weight3": 0.8693735003471375, "train/pos_loss/weight4": 0.5574414134025574, "train/pos_loss/weight5": 0.3402012288570404, "train/pos_loss/weight6": 0.21866925060749054, "train/pos_weight_progress": 0.15876777251184834, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9263606667518616 }, { "epoch": 1.5876777251184833, "step": 670, "train/alpha": 0.2, "train/clean_ce_loss": 0.1201171875, "train/clean_loss_scale": 9.72464656829834, "train/clean_loss_weight": 1.9449293613433838, "train/combined_loss": 1.4017165899276733, "train/pos_loss/pos0": 0.5076285600662231, "train/pos_loss/pos1": 0.9064760804176331, "train/pos_loss/pos2": 1.5547250509262085, "train/pos_loss/pos3": 2.265880823135376, "train/pos_loss/pos4": 2.982551336288452, "train/pos_loss/pos5": 3.3948519229888916, "train/pos_loss/pos6": 4.015497207641602, "train/pos_loss/weight0": 2.6863677501678467, "train/pos_loss/weight1": 1.9084023237228394, "train/pos_loss/weight2": 1.2099895477294922, "train/pos_loss/weight3": 0.6729090809822083, "train/pos_loss/weight4": 0.3205670118331909, "train/pos_loss/weight5": 0.13950134813785553, "train/pos_loss/weight6": 0.0622626356780529, "train/pos_weight_progress": 0.15876777251184834, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1680971384048462 }, { "epoch": 1.5900473933649288, "step": 671, "train/alpha": 0.2, "train/clean_ce_loss": 0.087890625, "train/clean_loss_scale": 12.89895248413086, "train/clean_loss_weight": 2.5797905921936035, "train/combined_loss": 1.3604364395141602, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.1590047393364929, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1336970329284668 }, { "epoch": 1.5900473933649288, "step": 671, "train/alpha": 0.2, "train/clean_ce_loss": 0.09521484375, "train/clean_loss_scale": 6.502351760864258, "train/clean_loss_weight": 1.3004703521728516, "train/combined_loss": 0.7429444789886475, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.1590047393364929, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6191204190254211 }, { "epoch": 1.5900473933649288, "step": 671, "train/alpha": 0.2, "train/clean_ce_loss": 0.039794921875, "train/clean_loss_scale": 7.345738410949707, "train/clean_loss_weight": 1.4691476821899414, "train/combined_loss": 0.3507876992225647, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.1590047393364929, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2923230826854706 }, { "epoch": 1.5900473933649288, "step": 671, "train/alpha": 0.2, "train/clean_ce_loss": 0.024658203125, "train/clean_loss_scale": 9.950708389282227, "train/clean_loss_weight": 1.990141749382019, "train/combined_loss": 0.2944399118423462, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.1590047393364929, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.24536658823490143 }, { "epoch": 1.5924170616113744, "step": 672, "train/alpha": 0.2, "train/clean_ce_loss": 0.0693359375, "train/clean_loss_scale": 14.50243091583252, "train/clean_loss_weight": 2.9004862308502197, "train/combined_loss": 1.2066476345062256, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.15924170616113745, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0055396556854248 }, { "epoch": 1.5924170616113744, "step": 672, "train/alpha": 0.2, "train/clean_ce_loss": 0.0084228515625, "train/clean_loss_scale": 19.10283851623535, "train/clean_loss_weight": 3.8205678462982178, "train/combined_loss": 0.19308044016361237, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.15924170616113745, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.16090036928653717 }, { "epoch": 1.5924170616113744, "step": 672, "train/alpha": 0.2, "train/clean_ce_loss": 0.011474609375, "train/clean_loss_scale": 28.486562728881836, "train/clean_loss_weight": 5.697312831878662, "train/combined_loss": 0.3922466039657593, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.15924170616113745, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.32687216997146606 }, { "epoch": 1.5924170616113744, "step": 672, "train/alpha": 0.2, "train/clean_ce_loss": 0.1416015625, "train/clean_loss_scale": 5.9297075271606445, "train/clean_loss_weight": 1.1859415769577026, "train/combined_loss": 1.0075870752334595, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.15924170616113745, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.839655876159668 }, { "epoch": 1.59478672985782, "step": 673, "train/alpha": 0.2, "train/clean_ce_loss": 0.09765625, "train/clean_loss_scale": 7.966111660003662, "train/clean_loss_weight": 1.5932223796844482, "train/combined_loss": 0.9335286617279053, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.159478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.777940571308136 }, { "epoch": 1.59478672985782, "step": 673, "train/alpha": 0.2, "train/clean_ce_loss": 0.12451171875, "train/clean_loss_scale": 6.359133243560791, "train/clean_loss_weight": 1.2718266248703003, "train/combined_loss": 0.9501439332962036, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.159478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7917866110801697 }, { "epoch": 1.59478672985782, "step": 673, "train/alpha": 0.2, "train/clean_ce_loss": 0.1318359375, "train/clean_loss_scale": 7.970085620880127, "train/clean_loss_weight": 1.5940171480178833, "train/combined_loss": 1.260892391204834, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.159478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0507436990737915 }, { "epoch": 1.59478672985782, "step": 673, "train/alpha": 0.2, "train/clean_ce_loss": 0.0142822265625, "train/clean_loss_scale": 15.269956588745117, "train/clean_loss_weight": 3.0539913177490234, "train/combined_loss": 0.26170676946640015, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.159478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.21808898448944092 }, { "epoch": 1.5971563981042654, "step": 674, "train/alpha": 0.2, "train/clean_ce_loss": 0.051513671875, "train/clean_loss_scale": 5.87401008605957, "train/clean_loss_weight": 1.1748020648956299, "train/combined_loss": 0.36311018466949463, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.15971563981042655, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.30259183049201965 }, { "epoch": 1.5971563981042654, "step": 674, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 8.615693092346191, "train/clean_loss_weight": 1.723138689994812, "train/combined_loss": 0.8632521033287048, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.15971563981042655, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7193767428398132 }, { "epoch": 1.5971563981042654, "step": 674, "train/alpha": 0.2, "train/clean_ce_loss": 0.051513671875, "train/clean_loss_scale": 7.641528606414795, "train/clean_loss_weight": 1.5283057689666748, "train/combined_loss": 0.4723718464374542, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.15971563981042655, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.39364320039749146 }, { "epoch": 1.5971563981042654, "step": 674, "train/alpha": 0.2, "train/clean_ce_loss": 0.019775390625, "train/clean_loss_scale": 9.636835098266602, "train/clean_loss_weight": 1.927367091178894, "train/combined_loss": 0.2286866307258606, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.15971563981042655, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.19057218730449677 }, { "epoch": 1.599526066350711, "grad_norm": 20.03014373779297, "learning_rate": 2e-06, "loss": 3.6275, "step": 675 }, { "epoch": 1.599526066350711, "step": 675, "train/alpha": 0.2, "train/clean_ce_loss": 4.5299530029296875e-05, "train/clean_loss_scale": 1665.2645263671875, "train/clean_loss_weight": 333.05291748046875, "train/combined_loss": 0.09052284061908722, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.15995260663507108, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.07543569803237915 }, { "epoch": 1.599526066350711, "step": 675, "train/alpha": 0.2, "train/clean_ce_loss": 0.06884765625, "train/clean_loss_scale": 9.230531692504883, "train/clean_loss_weight": 1.8461064100265503, "train/combined_loss": 0.7626006007194519, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.15995260663507108, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6355004906654358 }, { "epoch": 1.599526066350711, "step": 675, "train/alpha": 0.2, "train/clean_ce_loss": 0.1025390625, "train/clean_loss_scale": 5.555586814880371, "train/clean_loss_weight": 1.1111173629760742, "train/combined_loss": 0.6835975646972656, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.15995260663507108, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5696646571159363 }, { "epoch": 1.599526066350711, "step": 675, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 8.618143081665039, "train/clean_loss_weight": 1.7236286401748657, "train/combined_loss": 0.6716091632843018, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.15995260663507108, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5596743226051331 }, { "epoch": 1.6018957345971563, "step": 676, "train/alpha": 0.2, "train/clean_ce_loss": 0.1181640625, "train/clean_loss_scale": 7.6873087882995605, "train/clean_loss_weight": 1.537461757659912, "train/combined_loss": 1.090036392211914, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.16018957345971563, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9083636403083801 }, { "epoch": 1.6018957345971563, "step": 676, "train/alpha": 0.2, "train/clean_ce_loss": 0.035400390625, "train/clean_loss_scale": 10.818727493286133, "train/clean_loss_weight": 2.163745641708374, "train/combined_loss": 0.4595845937728882, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.16018957345971563, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3829871714115143 }, { "epoch": 1.6018957345971563, "step": 676, "train/alpha": 0.2, "train/clean_ce_loss": 0.0167236328125, "train/clean_loss_scale": 10.6106538772583, "train/clean_loss_weight": 2.122130870819092, "train/combined_loss": 0.21293841302394867, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.16018957345971563, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.17744867503643036 }, { "epoch": 1.6018957345971563, "step": 676, "train/alpha": 0.2, "train/clean_ce_loss": 0.05908203125, "train/clean_loss_scale": 6.195187091827393, "train/clean_loss_weight": 1.2390373945236206, "train/combined_loss": 0.4392290711402893, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.16018957345971563, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3660242259502411 }, { "epoch": 1.6042654028436019, "step": 677, "train/alpha": 0.2, "train/clean_ce_loss": 0.07763671875, "train/clean_loss_scale": 6.360973358154297, "train/clean_loss_weight": 1.272194743156433, "train/combined_loss": 0.5926141142845154, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.16042654028436018, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4938451051712036 }, { "epoch": 1.6042654028436019, "step": 677, "train/alpha": 0.2, "train/clean_ce_loss": 0.06396484375, "train/clean_loss_scale": 7.72807502746582, "train/clean_loss_weight": 1.5456150770187378, "train/combined_loss": 0.5931901335716248, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.16042654028436018, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.49432510137557983 }, { "epoch": 1.6042654028436019, "step": 677, "train/alpha": 0.2, "train/clean_ce_loss": 0.0018310546875, "train/clean_loss_scale": 102.73545837402344, "train/clean_loss_weight": 20.54709243774414, "train/combined_loss": 0.2257370948791504, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.16042654028436018, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1881142407655716 }, { "epoch": 1.6042654028436019, "step": 677, "train/alpha": 0.2, "train/clean_ce_loss": 0.0167236328125, "train/clean_loss_scale": 16.41063117980957, "train/clean_loss_weight": 3.2821261882781982, "train/combined_loss": 0.32933443784713745, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.16042654028436018, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2744453549385071 }, { "epoch": 1.6066350710900474, "step": 678, "train/alpha": 0.2, "train/clean_ce_loss": 0.19140625, "train/clean_loss_scale": 6.04440975189209, "train/clean_loss_weight": 1.2088819742202759, "train/combined_loss": 1.3883254528045654, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.16066350710900473, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.156937837600708 }, { "epoch": 1.6066350710900474, "step": 678, "train/alpha": 0.2, "train/clean_ce_loss": 0.044921875, "train/clean_loss_scale": 8.959543228149414, "train/clean_loss_weight": 1.791908621788025, "train/combined_loss": 0.48297539353370667, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.16066350710900473, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.40247949957847595 }, { "epoch": 1.6066350710900474, "step": 678, "train/alpha": 0.2, "train/clean_ce_loss": 0.123046875, "train/clean_loss_scale": 7.203921794891357, "train/clean_loss_weight": 1.4407843351364136, "train/combined_loss": 1.0637041330337524, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.16066350710900473, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8864200711250305 }, { "epoch": 1.6066350710900474, "step": 678, "train/alpha": 0.2, "train/clean_ce_loss": 0.06005859375, "train/clean_loss_scale": 10.49032211303711, "train/clean_loss_weight": 2.098064422607422, "train/combined_loss": 0.7560407519340515, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.16066350710900473, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6300339698791504 }, { "epoch": 1.609004739336493, "step": 679, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 8.700248718261719, "train/clean_loss_weight": 1.7400497198104858, "train/combined_loss": 0.9125065207481384, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.16090047393364929, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7604221105575562 }, { "epoch": 1.609004739336493, "step": 679, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 7.495380401611328, "train/clean_loss_weight": 1.4990761280059814, "train/combined_loss": 0.6543818116188049, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.16090047393364929, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5453181862831116 }, { "epoch": 1.609004739336493, "step": 679, "train/alpha": 0.2, "train/clean_ce_loss": 0.054931640625, "train/clean_loss_scale": 8.783303260803223, "train/clean_loss_weight": 1.7566606998443604, "train/combined_loss": 0.5789775252342224, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.16090047393364929, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4824812710285187 }, { "epoch": 1.609004739336493, "step": 679, "train/alpha": 0.2, "train/clean_ce_loss": 0.057373046875, "train/clean_loss_scale": 8.274225234985352, "train/clean_loss_weight": 1.654845118522644, "train/combined_loss": 0.569661021232605, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.16090047393364929, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.47471749782562256 }, { "epoch": 1.6113744075829384, "grad_norm": 19.417943954467773, "learning_rate": 2e-06, "loss": 3.5907, "step": 680 }, { "epoch": 1.6113744075829384, "step": 680, "train/alpha": 0.2, "train/clean_ce_loss": 0.001007080078125, "train/clean_loss_scale": 171.90518188476562, "train/clean_loss_weight": 34.381038665771484, "train/combined_loss": 0.2077467441558838, "train/pos_loss/pos0": 0.012949416413903236, "train/pos_loss/pos1": 0.03961480036377907, "train/pos_loss/pos2": 0.23531626164913177, "train/pos_loss/pos3": 0.7777060866355896, "train/pos_loss/pos4": 1.4430344104766846, "train/pos_loss/pos5": 2.128923177719116, "train/pos_loss/pos6": 2.9202654361724854, "train/pos_loss/weight0": 3.0608255863189697, "train/pos_loss/weight1": 2.0344314575195312, "train/pos_loss/weight2": 1.1342109441757202, "train/pos_loss/weight3": 0.5147420167922974, "train/pos_loss/weight4": 0.18984295427799225, "train/pos_loss/weight5": 0.05206301435828209, "train/pos_loss/weight6": 0.013883877545595169, "train/pos_weight_progress": 0.16113744075829384, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.17312228679656982 }, { "epoch": 1.6113744075829384, "step": 680, "train/alpha": 0.2, "train/clean_ce_loss": 0.033447265625, "train/clean_loss_scale": 18.951263427734375, "train/clean_loss_weight": 3.790252685546875, "train/combined_loss": 0.7606415152549744, "train/pos_loss/pos0": 0.07434941083192825, "train/pos_loss/pos1": 0.24193476140499115, "train/pos_loss/pos2": 0.5848494172096252, "train/pos_loss/pos3": 1.1550309658050537, "train/pos_loss/pos4": 1.8012146949768066, "train/pos_loss/pos5": 2.4576492309570312, "train/pos_loss/pos6": 2.8269195556640625, "train/pos_loss/weight0": 2.224337577819824, "train/pos_loss/weight1": 1.6794147491455078, "train/pos_loss/weight2": 1.241615891456604, "train/pos_loss/weight3": 0.8740609884262085, "train/pos_loss/weight4": 0.5262963771820068, "train/pos_loss/weight5": 0.2792448103427887, "train/pos_loss/weight6": 0.1750296652317047, "train/pos_weight_progress": 0.16113744075829384, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6338679194450378 }, { "epoch": 1.6113744075829384, "step": 680, "train/alpha": 0.2, "train/clean_ce_loss": 0.16796875, "train/clean_loss_scale": 7.898744583129883, "train/clean_loss_weight": 1.5797489881515503, "train/combined_loss": 1.5920907258987427, "train/pos_loss/pos0": 0.5809371471405029, "train/pos_loss/pos1": 1.0745952129364014, "train/pos_loss/pos2": 1.6327767372131348, "train/pos_loss/pos3": 2.328329086303711, "train/pos_loss/pos4": 2.6422207355499268, "train/pos_loss/pos5": 3.359175443649292, "train/pos_loss/pos6": 3.8271644115448, "train/pos_loss/weight0": 2.454716682434082, "train/pos_loss/weight1": 1.826915979385376, "train/pos_loss/weight2": 1.256923794746399, "train/pos_loss/weight3": 0.7875078320503235, "train/pos_loss/weight4": 0.4061846137046814, "train/pos_loss/weight5": 0.17976759374141693, "train/pos_loss/weight6": 0.08798382431268692, "train/pos_weight_progress": 0.16113744075829384, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3267422914505005 }, { "epoch": 1.6113744075829384, "step": 680, "train/alpha": 0.2, "train/clean_ce_loss": 0.0859375, "train/clean_loss_scale": 6.2418131828308105, "train/clean_loss_weight": 1.24836266040802, "train/combined_loss": 0.6436869502067566, "train/pos_loss/pos0": 0.2315179705619812, "train/pos_loss/pos1": 0.5143098831176758, "train/pos_loss/pos2": 0.9093356728553772, "train/pos_loss/pos3": 1.3638358116149902, "train/pos_loss/pos4": 1.8578461408615112, "train/pos_loss/pos5": 2.2019195556640625, "train/pos_loss/pos6": 2.543233871459961, "train/pos_loss/weight0": 3.2234394550323486, "train/pos_loss/weight1": 2.0778021812438965, "train/pos_loss/weight2": 1.0613811016082764, "train/pos_loss/weight3": 0.4572376012802124, "train/pos_loss/weight4": 0.1388893723487854, "train/pos_loss/weight5": 0.03330995514988899, "train/pos_loss/weight6": 0.007940668612718582, "train/pos_weight_progress": 0.16113744075829384, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5364058017730713 }, { "epoch": 1.6137440758293837, "step": 681, "train/alpha": 0.2, "train/clean_ce_loss": 0.09326171875, "train/clean_loss_scale": 8.943015098571777, "train/clean_loss_weight": 1.7886030673980713, "train/combined_loss": 1.0008491277694702, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.1613744075829384, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8340409398078918 }, { "epoch": 1.6137440758293837, "step": 681, "train/alpha": 0.2, "train/clean_ce_loss": 0.10302734375, "train/clean_loss_scale": 7.596297264099121, "train/clean_loss_weight": 1.5192594528198242, "train/combined_loss": 0.9391515851020813, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.1613744075829384, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7826263308525085 }, { "epoch": 1.6137440758293837, "step": 681, "train/alpha": 0.2, "train/clean_ce_loss": 0.10498046875, "train/clean_loss_scale": 7.7869720458984375, "train/clean_loss_weight": 1.5573943853378296, "train/combined_loss": 0.9809759855270386, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.1613744075829384, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8174799680709839 }, { "epoch": 1.6137440758293837, "step": 681, "train/alpha": 0.2, "train/clean_ce_loss": 0.055419921875, "train/clean_loss_scale": 17.48550033569336, "train/clean_loss_weight": 3.4971001148223877, "train/combined_loss": 1.1628540754318237, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.1613744075829384, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.969045102596283 }, { "epoch": 1.6161137440758293, "step": 682, "train/alpha": 0.2, "train/clean_ce_loss": 0.126953125, "train/clean_loss_scale": 10.411490440368652, "train/clean_loss_weight": 2.0822980403900146, "train/combined_loss": 1.5861254930496216, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16161137440758294, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.321771264076233 }, { "epoch": 1.6161137440758293, "step": 682, "train/alpha": 0.2, "train/clean_ce_loss": 0.04052734375, "train/clean_loss_scale": 10.121249198913574, "train/clean_loss_weight": 2.024249792098999, "train/combined_loss": 0.4922248125076294, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16161137440758294, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.41018733382225037 }, { "epoch": 1.6161137440758293, "step": 682, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 8.578973770141602, "train/clean_loss_weight": 1.7157948017120361, "train/combined_loss": 0.7339043617248535, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16161137440758294, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6115869879722595 }, { "epoch": 1.6161137440758293, "step": 682, "train/alpha": 0.2, "train/clean_ce_loss": 0.10546875, "train/clean_loss_scale": 8.910086631774902, "train/clean_loss_weight": 1.7820173501968384, "train/combined_loss": 1.1276828050613403, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16161137440758294, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9397357106208801 }, { "epoch": 1.6184834123222749, "step": 683, "train/alpha": 0.2, "train/clean_ce_loss": 0.0400390625, "train/clean_loss_scale": 11.368328094482422, "train/clean_loss_weight": 2.2736656665802, "train/combined_loss": 0.5462126731872559, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.1618483412322275, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4551772177219391 }, { "epoch": 1.6184834123222749, "step": 683, "train/alpha": 0.2, "train/clean_ce_loss": 8.821487426757812e-05, "train/clean_loss_scale": 1825.8746337890625, "train/clean_loss_weight": 365.1749267578125, "train/combined_loss": 0.19328317046165466, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.1618483412322275, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.16106930375099182 }, { "epoch": 1.6184834123222749, "step": 683, "train/alpha": 0.2, "train/clean_ce_loss": 0.1357421875, "train/clean_loss_scale": 7.367933750152588, "train/clean_loss_weight": 1.4735867977142334, "train/combined_loss": 1.2001674175262451, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.1618483412322275, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0001394748687744 }, { "epoch": 1.6184834123222749, "step": 683, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 7.978842735290527, "train/clean_loss_weight": 1.5957685708999634, "train/combined_loss": 0.7573667168617249, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.1618483412322275, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6311389207839966 }, { "epoch": 1.6208530805687205, "step": 684, "train/alpha": 0.2, "train/clean_ce_loss": 0.0299072265625, "train/clean_loss_scale": 25.470224380493164, "train/clean_loss_weight": 5.094045162200928, "train/combined_loss": 0.9140925407409668, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16208530805687205, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7617437839508057 }, { "epoch": 1.6208530805687205, "step": 684, "train/alpha": 0.2, "train/clean_ce_loss": 0.0771484375, "train/clean_loss_scale": 8.41988754272461, "train/clean_loss_weight": 1.683977484703064, "train/combined_loss": 0.7794973850250244, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16208530805687205, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6495811343193054 }, { "epoch": 1.6208530805687205, "step": 684, "train/alpha": 0.2, "train/clean_ce_loss": 0.01483154296875, "train/clean_loss_scale": 23.3713321685791, "train/clean_loss_weight": 4.674266338348389, "train/combined_loss": 0.41595950722694397, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16208530805687205, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3466329276561737 }, { "epoch": 1.6208530805687205, "step": 684, "train/alpha": 0.2, "train/clean_ce_loss": 0.130859375, "train/clean_loss_scale": 8.061838150024414, "train/clean_loss_weight": 1.6123676300048828, "train/combined_loss": 1.265960454940796, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16208530805687205, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0549670457839966 }, { "epoch": 1.623222748815166, "grad_norm": 20.629974365234375, "learning_rate": 2e-06, "loss": 3.5529, "step": 685 }, { "epoch": 1.623222748815166, "step": 685, "train/alpha": 0.2, "train/clean_ce_loss": 0.06884765625, "train/clean_loss_scale": 5.743035793304443, "train/clean_loss_weight": 1.1486071348190308, "train/combined_loss": 0.4744734764099121, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.1623222748815166, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.39539456367492676 }, { "epoch": 1.623222748815166, "step": 685, "train/alpha": 0.2, "train/clean_ce_loss": 0.05810546875, "train/clean_loss_scale": 8.378959655761719, "train/clean_loss_weight": 1.6757919788360596, "train/combined_loss": 0.5842360854148865, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.1623222748815166, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4868634045124054 }, { "epoch": 1.623222748815166, "step": 685, "train/alpha": 0.2, "train/clean_ce_loss": 0.0908203125, "train/clean_loss_scale": 7.650729179382324, "train/clean_loss_weight": 1.5301458835601807, "train/combined_loss": 0.8338099718093872, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.1623222748815166, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6948416233062744 }, { "epoch": 1.623222748815166, "step": 685, "train/alpha": 0.2, "train/clean_ce_loss": 0.09375, "train/clean_loss_scale": 10.603439331054688, "train/clean_loss_weight": 2.120687961578369, "train/combined_loss": 1.1928869485855103, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.1623222748815166, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.994072437286377 }, { "epoch": 1.6255924170616114, "step": 686, "train/alpha": 0.2, "train/clean_ce_loss": 0.01611328125, "train/clean_loss_scale": 23.1407470703125, "train/clean_loss_weight": 4.628149509429932, "train/combined_loss": 0.4474480450153351, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16255924170616115, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.37287336587905884 }, { "epoch": 1.6255924170616114, "step": 686, "train/alpha": 0.2, "train/clean_ce_loss": 0.0291748046875, "train/clean_loss_scale": 15.203751564025879, "train/clean_loss_weight": 3.04075026512146, "train/combined_loss": 0.5322797894477844, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16255924170616115, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4435664713382721 }, { "epoch": 1.6255924170616114, "step": 686, "train/alpha": 0.2, "train/clean_ce_loss": 0.01025390625, "train/clean_loss_scale": 18.299640655517578, "train/clean_loss_weight": 3.6599280834198, "train/combined_loss": 0.22517135739326477, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16255924170616115, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.18764279782772064 }, { "epoch": 1.6255924170616114, "step": 686, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 8.173203468322754, "train/clean_loss_weight": 1.6346406936645508, "train/combined_loss": 0.7758157849311829, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16255924170616115, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6465131640434265 }, { "epoch": 1.6279620853080567, "step": 687, "train/alpha": 0.2, "train/clean_ce_loss": 0.103515625, "train/clean_loss_scale": 9.581147193908691, "train/clean_loss_weight": 1.916229486465454, "train/combined_loss": 1.1901581287384033, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16279620853080567, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9917984008789062 }, { "epoch": 1.6279620853080567, "step": 687, "train/alpha": 0.2, "train/clean_ce_loss": 0.1025390625, "train/clean_loss_scale": 7.343913555145264, "train/clean_loss_weight": 1.4687827825546265, "train/combined_loss": 0.9036456346511841, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16279620853080567, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7530379891395569 }, { "epoch": 1.6279620853080567, "step": 687, "train/alpha": 0.2, "train/clean_ce_loss": 0.09765625, "train/clean_loss_scale": 10.628061294555664, "train/clean_loss_weight": 2.125612258911133, "train/combined_loss": 1.2454760074615479, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16279620853080567, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0378966331481934 }, { "epoch": 1.6279620853080567, "step": 687, "train/alpha": 0.2, "train/clean_ce_loss": 0.05322265625, "train/clean_loss_scale": 8.415841102600098, "train/clean_loss_weight": 1.6831682920455933, "train/combined_loss": 0.5374961495399475, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16279620853080567, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.44791343808174133 }, { "epoch": 1.6303317535545023, "step": 688, "train/alpha": 0.2, "train/clean_ce_loss": 0.1494140625, "train/clean_loss_scale": 8.394379615783691, "train/clean_loss_weight": 1.6788759231567383, "train/combined_loss": 1.505086064338684, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16303317535545023, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2542383670806885 }, { "epoch": 1.6303317535545023, "step": 688, "train/alpha": 0.2, "train/clean_ce_loss": 0.0751953125, "train/clean_loss_scale": 9.807737350463867, "train/clean_loss_weight": 1.9615474939346313, "train/combined_loss": 0.8849949836730957, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16303317535545023, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7374958395957947 }, { "epoch": 1.6303317535545023, "step": 688, "train/alpha": 0.2, "train/clean_ce_loss": 0.0301513671875, "train/clean_loss_scale": 14.267402648925781, "train/clean_loss_weight": 2.853480577468872, "train/combined_loss": 0.5162180066108704, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16303317535545023, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43018168210983276 }, { "epoch": 1.6303317535545023, "step": 688, "train/alpha": 0.2, "train/clean_ce_loss": 0.09521484375, "train/clean_loss_scale": 7.738379001617432, "train/clean_loss_weight": 1.5476758480072021, "train/combined_loss": 0.8841702342033386, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16303317535545023, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7368085384368896 }, { "epoch": 1.632701421800948, "step": 689, "train/alpha": 0.2, "train/clean_ce_loss": 0.1181640625, "train/clean_loss_scale": 8.282303810119629, "train/clean_loss_weight": 1.6564607620239258, "train/combined_loss": 1.1744048595428467, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16327014218009478, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9786707162857056 }, { "epoch": 1.632701421800948, "step": 689, "train/alpha": 0.2, "train/clean_ce_loss": 0.138671875, "train/clean_loss_scale": 7.323147773742676, "train/clean_loss_weight": 1.4646295309066772, "train/combined_loss": 1.218617558479309, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16327014218009478, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.015514612197876 }, { "epoch": 1.632701421800948, "step": 689, "train/alpha": 0.2, "train/clean_ce_loss": 0.053955078125, "train/clean_loss_scale": 9.709122657775879, "train/clean_loss_weight": 1.9418245553970337, "train/combined_loss": 0.6286277770996094, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16327014218009478, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5238564610481262 }, { "epoch": 1.632701421800948, "step": 689, "train/alpha": 0.2, "train/clean_ce_loss": 0.162109375, "train/clean_loss_scale": 7.568530082702637, "train/clean_loss_weight": 1.513706088066101, "train/combined_loss": 1.4723155498504639, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16327014218009478, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2269296646118164 }, { "epoch": 1.6350710900473935, "grad_norm": 36.687355041503906, "learning_rate": 2e-06, "loss": 3.744, "step": 690 }, { "epoch": 1.6350710900473935, "step": 690, "train/alpha": 0.2, "train/clean_ce_loss": 0.050048828125, "train/clean_loss_scale": 10.192106246948242, "train/clean_loss_weight": 2.038421392440796, "train/combined_loss": 0.6121236085891724, "train/pos_loss/pos0": 0.16845417022705078, "train/pos_loss/pos1": 0.3608589768409729, "train/pos_loss/pos2": 0.6582629680633545, "train/pos_loss/pos3": 0.9795231223106384, "train/pos_loss/pos4": 1.3409903049468994, "train/pos_loss/pos5": 1.761756181716919, "train/pos_loss/pos6": 2.1537773609161377, "train/pos_loss/weight0": 2.574782133102417, "train/pos_loss/weight1": 1.8696578741073608, "train/pos_loss/weight2": 1.2281601428985596, "train/pos_loss/weight3": 0.7159639596939087, "train/pos_loss/weight4": 0.367995947599411, "train/pos_loss/weight5": 0.16387887299060822, "train/pos_loss/weight6": 0.07956057786941528, "train/pos_weight_progress": 0.16350710900473933, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5101029872894287 }, { "epoch": 1.6350710900473935, "step": 690, "train/alpha": 0.2, "train/clean_ce_loss": 0.1484375, "train/clean_loss_scale": 9.416818618774414, "train/clean_loss_weight": 1.8833637237548828, "train/combined_loss": 1.6773707866668701, "train/pos_loss/pos0": 0.4965878427028656, "train/pos_loss/pos1": 1.0276474952697754, "train/pos_loss/pos2": 1.873619794845581, "train/pos_loss/pos3": 2.1729230880737305, "train/pos_loss/pos4": 2.573383331298828, "train/pos_loss/pos5": 3.341057300567627, "train/pos_loss/pos6": 4.148159503936768, "train/pos_loss/weight0": 2.304955005645752, "train/pos_loss/weight1": 1.7472199201583862, "train/pos_loss/weight2": 1.2316381931304932, "train/pos_loss/weight3": 0.803946316242218, "train/pos_loss/weight4": 0.48061978816986084, "train/pos_loss/weight5": 0.2710454761981964, "train/pos_loss/weight6": 0.16057510673999786, "train/pos_weight_progress": 0.16350710900473933, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3978090286254883 }, { "epoch": 1.6350710900473935, "step": 690, "train/alpha": 0.2, "train/clean_ce_loss": 0.134765625, "train/clean_loss_scale": 6.813772678375244, "train/clean_loss_weight": 1.3627545833587646, "train/combined_loss": 1.101914882659912, "train/pos_loss/pos0": 0.4911620318889618, "train/pos_loss/pos1": 0.9754999876022339, "train/pos_loss/pos2": 1.4488428831100464, "train/pos_loss/pos3": 1.9689218997955322, "train/pos_loss/pos4": 2.3672196865081787, "train/pos_loss/pos5": 2.782144546508789, "train/pos_loss/pos6": 3.0848658084869385, "train/pos_loss/weight0": 3.26285982131958, "train/pos_loss/weight1": 2.0708253383636475, "train/pos_loss/weight2": 1.0740416049957275, "train/pos_loss/weight3": 0.4260927140712738, "train/pos_loss/weight4": 0.12613095343112946, "train/pos_loss/weight5": 0.032347217202186584, "train/pos_loss/weight6": 0.007701893802732229, "train/pos_weight_progress": 0.16350710900473933, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9182623624801636 }, { "epoch": 1.6350710900473935, "step": 690, "train/alpha": 0.2, "train/clean_ce_loss": 0.08251953125, "train/clean_loss_scale": 7.8504838943481445, "train/clean_loss_weight": 1.5700968503952026, "train/combined_loss": 0.7773818969726562, "train/pos_loss/pos0": 0.29640936851501465, "train/pos_loss/pos1": 0.5655834674835205, "train/pos_loss/pos2": 0.9482586979866028, "train/pos_loss/pos3": 1.3947608470916748, "train/pos_loss/pos4": 1.84482741355896, "train/pos_loss/pos5": 2.320594549179077, "train/pos_loss/pos6": 2.646010637283325, "train/pos_loss/weight0": 3.008704900741577, "train/pos_loss/weight1": 1.9590299129486084, "train/pos_loss/weight2": 1.1408518552780151, "train/pos_loss/weight3": 0.5562006831169128, "train/pos_loss/weight4": 0.2266210913658142, "train/pos_loss/weight5": 0.07966066896915436, "train/pos_loss/weight6": 0.028931036591529846, "train/pos_weight_progress": 0.16350710900473933, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6478182673454285 }, { "epoch": 1.6374407582938388, "step": 691, "train/alpha": 0.2, "train/clean_ce_loss": 0.1025390625, "train/clean_loss_scale": 7.058792591094971, "train/clean_loss_weight": 1.411758542060852, "train/combined_loss": 0.8685623407363892, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16374407582938388, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7238019704818726 }, { "epoch": 1.6374407582938388, "step": 691, "train/alpha": 0.2, "train/clean_ce_loss": 0.053466796875, "train/clean_loss_scale": 10.111756324768066, "train/clean_loss_weight": 2.0223512649536133, "train/combined_loss": 0.6487718820571899, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16374407582938388, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5406432151794434 }, { "epoch": 1.6374407582938388, "step": 691, "train/alpha": 0.2, "train/clean_ce_loss": 0.091796875, "train/clean_loss_scale": 8.881346702575684, "train/clean_loss_weight": 1.7762693166732788, "train/combined_loss": 0.9783357977867126, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16374407582938388, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8152798414230347 }, { "epoch": 1.6374407582938388, "step": 691, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 8.221233367919922, "train/clean_loss_weight": 1.6442466974258423, "train/combined_loss": 0.6454952955245972, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16374407582938388, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5379127264022827 }, { "epoch": 1.6398104265402842, "step": 692, "train/alpha": 0.2, "train/clean_ce_loss": 0.1630859375, "train/clean_loss_scale": 8.36106014251709, "train/clean_loss_weight": 1.67221200466156, "train/combined_loss": 1.6362855434417725, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16398104265402844, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.363571286201477 }, { "epoch": 1.6398104265402842, "step": 692, "train/alpha": 0.2, "train/clean_ce_loss": 0.103515625, "train/clean_loss_scale": 9.525508880615234, "train/clean_loss_weight": 1.9051017761230469, "train/combined_loss": 1.1832468509674072, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16398104265402844, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9860390424728394 }, { "epoch": 1.6398104265402842, "step": 692, "train/alpha": 0.2, "train/clean_ce_loss": 0.06298828125, "train/clean_loss_scale": 10.581888198852539, "train/clean_loss_weight": 2.116377592086792, "train/combined_loss": 0.7998419404029846, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16398104265402844, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.666534960269928 }, { "epoch": 1.6398104265402842, "step": 692, "train/alpha": 0.2, "train/clean_ce_loss": 0.1318359375, "train/clean_loss_scale": 9.013021469116211, "train/clean_loss_weight": 1.8026043176651, "train/combined_loss": 1.425888180732727, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16398104265402844, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1882401704788208 }, { "epoch": 1.6421800947867298, "step": 693, "train/alpha": 0.2, "train/clean_ce_loss": 0.037841796875, "train/clean_loss_scale": 19.222434997558594, "train/clean_loss_weight": 3.844486951828003, "train/combined_loss": 0.872893750667572, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.164218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7274114489555359 }, { "epoch": 1.6421800947867298, "step": 693, "train/alpha": 0.2, "train/clean_ce_loss": 0.109375, "train/clean_loss_scale": 10.121342658996582, "train/clean_loss_weight": 2.024268627166748, "train/combined_loss": 1.3284261226654053, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.164218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1070218086242676 }, { "epoch": 1.6421800947867298, "step": 693, "train/alpha": 0.2, "train/clean_ce_loss": 0.019287109375, "train/clean_loss_scale": 19.064119338989258, "train/clean_loss_weight": 3.812824010848999, "train/combined_loss": 0.4412301182746887, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.164218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3676917552947998 }, { "epoch": 1.6421800947867298, "step": 693, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 9.942230224609375, "train/clean_loss_weight": 1.9884461164474487, "train/combined_loss": 1.1068497896194458, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.164218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9223748445510864 }, { "epoch": 1.6445497630331753, "step": 694, "train/alpha": 0.2, "train/clean_ce_loss": 0.0625, "train/clean_loss_scale": 12.905189514160156, "train/clean_loss_weight": 2.581037998199463, "train/combined_loss": 0.9678891897201538, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16445497630331754, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8065743446350098 }, { "epoch": 1.6445497630331753, "step": 694, "train/alpha": 0.2, "train/clean_ce_loss": 1.609325408935547e-05, "train/clean_loss_scale": 8003.13037109375, "train/clean_loss_weight": 1600.6260986328125, "train/combined_loss": 0.15455569326877594, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16445497630331754, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.12879641354084015 }, { "epoch": 1.6445497630331753, "step": 694, "train/alpha": 0.2, "train/clean_ce_loss": 0.181640625, "train/clean_loss_scale": 7.116262912750244, "train/clean_loss_weight": 1.4232525825500488, "train/combined_loss": 1.5511229038238525, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16445497630331754, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2926024198532104 }, { "epoch": 1.6445497630331753, "step": 694, "train/alpha": 0.2, "train/clean_ce_loss": 0.046630859375, "train/clean_loss_scale": 8.655938148498535, "train/clean_loss_weight": 1.7311877012252808, "train/combined_loss": 0.48436060547828674, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16445497630331754, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.40363383293151855 }, { "epoch": 1.646919431279621, "grad_norm": 15.54315185546875, "learning_rate": 2e-06, "loss": 3.2384, "step": 695 }, { "epoch": 1.646919431279621, "step": 695, "train/alpha": 0.2, "train/clean_ce_loss": 0.10986328125, "train/clean_loss_scale": 9.15283203125, "train/clean_loss_weight": 1.83056640625, "train/combined_loss": 1.206672191619873, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.1646919431279621, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0055601596832275 }, { "epoch": 1.646919431279621, "step": 695, "train/alpha": 0.2, "train/clean_ce_loss": 0.06884765625, "train/clean_loss_scale": 7.065757751464844, "train/clean_loss_weight": 1.4131516218185425, "train/combined_loss": 0.5837530493736267, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.1646919431279621, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4864608645439148 }, { "epoch": 1.646919431279621, "step": 695, "train/alpha": 0.2, "train/clean_ce_loss": 0.10888671875, "train/clean_loss_scale": 9.950804710388184, "train/clean_loss_weight": 1.9901609420776367, "train/combined_loss": 1.3002126216888428, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.1646919431279621, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0835105180740356 }, { "epoch": 1.646919431279621, "step": 695, "train/alpha": 0.2, "train/clean_ce_loss": 0.1630859375, "train/clean_loss_scale": 7.414791107177734, "train/clean_loss_weight": 1.482958197593689, "train/combined_loss": 1.451097846031189, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.1646919431279621, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2092481851577759 }, { "epoch": 1.6492890995260665, "step": 696, "train/alpha": 0.2, "train/clean_ce_loss": 0.00054931640625, "train/clean_loss_scale": 533.740966796875, "train/clean_loss_weight": 106.7481918334961, "train/combined_loss": 0.35183119773864746, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16492890995260664, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.29319265484809875 }, { "epoch": 1.6492890995260665, "step": 696, "train/alpha": 0.2, "train/clean_ce_loss": 0.072265625, "train/clean_loss_scale": 8.762039184570312, "train/clean_loss_weight": 1.7524079084396362, "train/combined_loss": 0.7598330974578857, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16492890995260664, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6331942677497864 }, { "epoch": 1.6492890995260665, "step": 696, "train/alpha": 0.2, "train/clean_ce_loss": 0.01129150390625, "train/clean_loss_scale": 22.032068252563477, "train/clean_loss_weight": 4.406413555145264, "train/combined_loss": 0.2985302209854126, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16492890995260664, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2487751841545105 }, { "epoch": 1.6492890995260665, "step": 696, "train/alpha": 0.2, "train/clean_ce_loss": 0.006500244140625, "train/clean_loss_scale": 53.97951126098633, "train/clean_loss_weight": 10.795902252197266, "train/combined_loss": 0.42105600237846375, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16492890995260664, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3508799970149994 }, { "epoch": 1.6516587677725119, "step": 697, "train/alpha": 0.2, "train/clean_ce_loss": 0.08837890625, "train/clean_loss_scale": 7.495351314544678, "train/clean_loss_weight": 1.4990702867507935, "train/combined_loss": 0.794917106628418, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.1651658767772512, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6624309420585632 }, { "epoch": 1.6516587677725119, "step": 697, "train/alpha": 0.2, "train/clean_ce_loss": 0.08203125, "train/clean_loss_scale": 11.28670597076416, "train/clean_loss_weight": 2.257341146469116, "train/combined_loss": 1.1110351085662842, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.1651658767772512, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9258626103401184 }, { "epoch": 1.6516587677725119, "step": 697, "train/alpha": 0.2, "train/clean_ce_loss": 0.16015625, "train/clean_loss_scale": 6.191557884216309, "train/clean_loss_weight": 1.2383116483688354, "train/combined_loss": 1.1899399757385254, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.1651658767772512, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9916166663169861 }, { "epoch": 1.6516587677725119, "step": 697, "train/alpha": 0.2, "train/clean_ce_loss": 0.09423828125, "train/clean_loss_scale": 11.505393028259277, "train/clean_loss_weight": 2.3010785579681396, "train/combined_loss": 1.301098108291626, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.1651658767772512, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.084248423576355 }, { "epoch": 1.6540284360189572, "step": 698, "train/alpha": 0.2, "train/clean_ce_loss": 0.08544921875, "train/clean_loss_scale": 7.2833027839660645, "train/clean_loss_weight": 1.4566606283187866, "train/combined_loss": 0.7468230724334717, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16540284360189572, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6223525404930115 }, { "epoch": 1.6540284360189572, "step": 698, "train/alpha": 0.2, "train/clean_ce_loss": 0.0137939453125, "train/clean_loss_scale": 15.943931579589844, "train/clean_loss_weight": 3.188786268234253, "train/combined_loss": 0.26391565799713135, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16540284360189572, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.21992972493171692 }, { "epoch": 1.6540284360189572, "step": 698, "train/alpha": 0.2, "train/clean_ce_loss": 0.166015625, "train/clean_loss_scale": 7.962030410766602, "train/clean_loss_weight": 1.592406153678894, "train/combined_loss": 1.5861856937408447, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16540284360189572, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3218214511871338 }, { "epoch": 1.6540284360189572, "step": 698, "train/alpha": 0.2, "train/clean_ce_loss": 0.119140625, "train/clean_loss_scale": 9.542037963867188, "train/clean_loss_weight": 1.9084075689315796, "train/combined_loss": 1.364213228225708, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16540284360189572, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1368443965911865 }, { "epoch": 1.6563981042654028, "step": 699, "train/alpha": 0.2, "train/clean_ce_loss": 0.1044921875, "train/clean_loss_scale": 8.499723434448242, "train/clean_loss_weight": 1.6999447345733643, "train/combined_loss": 1.0657856464385986, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16563981042654027, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8881546854972839 }, { "epoch": 1.6563981042654028, "step": 699, "train/alpha": 0.2, "train/clean_ce_loss": 0.041259765625, "train/clean_loss_scale": 4.964054107666016, "train/clean_loss_weight": 0.992810845375061, "train/combined_loss": 0.2457788586616516, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16563981042654027, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.20481571555137634 }, { "epoch": 1.6563981042654028, "step": 699, "train/alpha": 0.2, "train/clean_ce_loss": 0.1533203125, "train/clean_loss_scale": 9.189009666442871, "train/clean_loss_weight": 1.8378019332885742, "train/combined_loss": 1.6906342506408691, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16563981042654027, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4088618755340576 }, { "epoch": 1.6563981042654028, "step": 699, "train/alpha": 0.2, "train/clean_ce_loss": 0.06884765625, "train/clean_loss_scale": 8.918177604675293, "train/clean_loss_weight": 1.7836354970932007, "train/combined_loss": 0.7367947101593018, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16563981042654027, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6139956116676331 }, { "epoch": 1.6587677725118484, "grad_norm": 12.826157569885254, "learning_rate": 2e-06, "loss": 3.2279, "step": 700 }, { "epoch": 1.6587677725118484, "step": 700, "train/alpha": 0.2, "train/clean_ce_loss": 0.07861328125, "train/clean_loss_scale": 9.920778274536133, "train/clean_loss_weight": 1.9841556549072266, "train/combined_loss": 0.9358859658241272, "train/pos_loss/pos0": 0.310634046792984, "train/pos_loss/pos1": 0.665048360824585, "train/pos_loss/pos2": 1.0856488943099976, "train/pos_loss/pos3": 1.4581962823867798, "train/pos_loss/pos4": 1.835162878036499, "train/pos_loss/pos5": 2.3594810962677, "train/pos_loss/pos6": 2.6593525409698486, "train/pos_loss/weight0": 2.685364007949829, "train/pos_loss/weight1": 1.9076348543167114, "train/pos_loss/weight2": 1.2224960327148438, "train/pos_loss/weight3": 0.6809221506118774, "train/pos_loss/weight4": 0.3169640302658081, "train/pos_loss/weight5": 0.13101467490196228, "train/pos_loss/weight6": 0.05560450628399849, "train/pos_weight_progress": 0.16587677725118483, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7799049615859985 }, { "epoch": 1.6587677725118484, "step": 700, "train/alpha": 0.2, "train/clean_ce_loss": 0.06396484375, "train/clean_loss_scale": 9.209835052490234, "train/clean_loss_weight": 1.841966986656189, "train/combined_loss": 0.7069268226623535, "train/pos_loss/pos0": 0.23416496813297272, "train/pos_loss/pos1": 0.47557833790779114, "train/pos_loss/pos2": 0.7891860604286194, "train/pos_loss/pos3": 1.1455161571502686, "train/pos_loss/pos4": 1.5459572076797485, "train/pos_loss/pos5": 1.7967150211334229, "train/pos_loss/pos6": 2.2439088821411133, "train/pos_loss/weight0": 2.694018602371216, "train/pos_loss/weight1": 1.9165033102035522, "train/pos_loss/weight2": 1.2143769264221191, "train/pos_loss/weight3": 0.6665444374084473, "train/pos_loss/weight4": 0.31519365310668945, "train/pos_loss/weight5": 0.13372384011745453, "train/pos_loss/weight6": 0.05963928624987602, "train/pos_weight_progress": 0.16587677725118483, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5891056656837463 }, { "epoch": 1.6587677725118484, "step": 700, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 6.368195533752441, "train/clean_loss_weight": 1.2736390829086304, "train/combined_loss": 0.6679142713546753, "train/pos_loss/pos0": 0.16193081438541412, "train/pos_loss/pos1": 0.3196967840194702, "train/pos_loss/pos2": 0.6756035089492798, "train/pos_loss/pos3": 1.0212273597717285, "train/pos_loss/pos4": 1.518807291984558, "train/pos_loss/pos5": 1.8383387327194214, "train/pos_loss/pos6": 2.343402624130249, "train/pos_loss/weight0": 2.453152894973755, "train/pos_loss/weight1": 1.794851303100586, "train/pos_loss/weight2": 1.234336018562317, "train/pos_loss/weight3": 0.7630161643028259, "train/pos_loss/weight4": 0.420951783657074, "train/pos_loss/weight5": 0.21552197635173798, "train/pos_loss/weight6": 0.11817005276679993, "train/pos_weight_progress": 0.16587677725118483, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5565952062606812 }, { "epoch": 1.6587677725118484, "step": 700, "train/alpha": 0.2, "train/clean_ce_loss": 0.0142822265625, "train/clean_loss_scale": 27.955793380737305, "train/clean_loss_weight": 5.591158866882324, "train/combined_loss": 0.47912517189979553, "train/pos_loss/pos0": 0.010929697193205357, "train/pos_loss/pos1": 0.08106068521738052, "train/pos_loss/pos2": 0.3230077028274536, "train/pos_loss/pos3": 0.9375709891319275, "train/pos_loss/pos4": 1.3145930767059326, "train/pos_loss/pos5": 2.143826961517334, "train/pos_loss/pos6": 2.896965980529785, "train/pos_loss/weight0": 2.382591724395752, "train/pos_loss/weight1": 1.7536994218826294, "train/pos_loss/weight2": 1.249742031097412, "train/pos_loss/weight3": 0.7942719459533691, "train/pos_loss/weight4": 0.45426106452941895, "train/pos_loss/weight5": 0.23244096338748932, "train/pos_loss/weight6": 0.13299281895160675, "train/pos_weight_progress": 0.16587677725118483, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3992709815502167 }, { "epoch": 1.661137440758294, "step": 701, "train/alpha": 0.2, "train/clean_ce_loss": 0.1103515625, "train/clean_loss_scale": 11.036046028137207, "train/clean_loss_weight": 2.207209348678589, "train/combined_loss": 1.4614139795303345, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16611374407582938, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2178449630737305 }, { "epoch": 1.661137440758294, "step": 701, "train/alpha": 0.2, "train/clean_ce_loss": 0.162109375, "train/clean_loss_scale": 10.264662742614746, "train/clean_loss_weight": 2.0529325008392334, "train/combined_loss": 1.996797800064087, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16611374407582938, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.6639981269836426 }, { "epoch": 1.661137440758294, "step": 701, "train/alpha": 0.2, "train/clean_ce_loss": 0.0458984375, "train/clean_loss_scale": 8.987306594848633, "train/clean_loss_weight": 1.7974613904953003, "train/combined_loss": 0.49500399827957153, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16611374407582938, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.41250333189964294 }, { "epoch": 1.661137440758294, "step": 701, "train/alpha": 0.2, "train/clean_ce_loss": 0.056884765625, "train/clean_loss_scale": 9.95564079284668, "train/clean_loss_weight": 1.9911282062530518, "train/combined_loss": 0.6795891523361206, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16611374407582938, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5663242936134338 }, { "epoch": 1.6635071090047393, "step": 702, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 9.785592079162598, "train/clean_loss_weight": 1.9571183919906616, "train/combined_loss": 0.7568543553352356, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16635071090047393, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6307119727134705 }, { "epoch": 1.6635071090047393, "step": 702, "train/alpha": 0.2, "train/clean_ce_loss": 0.1396484375, "train/clean_loss_scale": 11.096758842468262, "train/clean_loss_weight": 2.2193517684936523, "train/combined_loss": 1.8595740795135498, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16635071090047393, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5496450662612915 }, { "epoch": 1.6635071090047393, "step": 702, "train/alpha": 0.2, "train/clean_ce_loss": 0.1064453125, "train/clean_loss_scale": 10.416454315185547, "train/clean_loss_weight": 2.0832908153533936, "train/combined_loss": 1.330539345741272, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16635071090047393, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1087827682495117 }, { "epoch": 1.6635071090047393, "step": 702, "train/alpha": 0.2, "train/clean_ce_loss": 0.033203125, "train/clean_loss_scale": 12.77865219116211, "train/clean_loss_weight": 2.5557305812835693, "train/combined_loss": 0.509149432182312, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16635071090047393, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.42429119348526 }, { "epoch": 1.6658767772511849, "step": 703, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 11.888550758361816, "train/clean_loss_weight": 2.3777101039886475, "train/combined_loss": 1.1911770105361938, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16658767772511848, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9926475286483765 }, { "epoch": 1.6658767772511849, "step": 703, "train/alpha": 0.2, "train/clean_ce_loss": 0.111328125, "train/clean_loss_scale": 8.365348815917969, "train/clean_loss_weight": 1.6730698347091675, "train/combined_loss": 1.117558240890503, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16658767772511848, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.931298553943634 }, { "epoch": 1.6658767772511849, "step": 703, "train/alpha": 0.2, "train/clean_ce_loss": 0.10302734375, "train/clean_loss_scale": 9.272760391235352, "train/clean_loss_weight": 1.854552149772644, "train/combined_loss": 1.146417498588562, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16658767772511848, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9553478956222534 }, { "epoch": 1.6658767772511849, "step": 703, "train/alpha": 0.2, "train/clean_ce_loss": 0.053955078125, "train/clean_loss_scale": 10.039411544799805, "train/clean_loss_weight": 2.0078823566436768, "train/combined_loss": 0.6500126719474792, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16658767772511848, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5416772365570068 }, { "epoch": 1.6682464454976302, "step": 704, "train/alpha": 0.2, "train/clean_ce_loss": 0.030517578125, "train/clean_loss_scale": 18.712125778198242, "train/clean_loss_weight": 3.7424252033233643, "train/combined_loss": 0.6852585077285767, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16682464454976303, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5710487365722656 }, { "epoch": 1.6682464454976302, "step": 704, "train/alpha": 0.2, "train/clean_ce_loss": 0.0947265625, "train/clean_loss_scale": 13.167027473449707, "train/clean_loss_weight": 2.6334054470062256, "train/combined_loss": 1.4967206716537476, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16682464454976303, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.247267246246338 }, { "epoch": 1.6682464454976302, "step": 704, "train/alpha": 0.2, "train/clean_ce_loss": 0.05908203125, "train/clean_loss_scale": 10.873960494995117, "train/clean_loss_weight": 2.1747920513153076, "train/combined_loss": 0.7709468603134155, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16682464454976303, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6424556970596313 }, { "epoch": 1.6682464454976302, "step": 704, "train/alpha": 0.2, "train/clean_ce_loss": 0.1328125, "train/clean_loss_scale": 9.050307273864746, "train/clean_loss_weight": 1.8100614547729492, "train/combined_loss": 1.4423927068710327, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16682464454976303, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2019939422607422 }, { "epoch": 1.6706161137440758, "grad_norm": 22.241544723510742, "learning_rate": 2e-06, "loss": 3.3918, "step": 705 }, { "epoch": 1.6706161137440758, "step": 705, "train/alpha": 0.2, "train/clean_ce_loss": 0.138671875, "train/clean_loss_scale": 7.566248893737793, "train/clean_loss_weight": 1.5132497549057007, "train/combined_loss": 1.2590711116790771, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.1670616113744076, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.049225926399231 }, { "epoch": 1.6706161137440758, "step": 705, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 11.891173362731934, "train/clean_loss_weight": 2.378234624862671, "train/combined_loss": 1.156602382659912, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.1670616113744076, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9638353586196899 }, { "epoch": 1.6706161137440758, "step": 705, "train/alpha": 0.2, "train/clean_ce_loss": 0.10693359375, "train/clean_loss_scale": 10.298934936523438, "train/clean_loss_weight": 2.0597870349884033, "train/combined_loss": 1.3215625286102295, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.1670616113744076, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.101302146911621 }, { "epoch": 1.6706161137440758, "step": 705, "train/alpha": 0.2, "train/clean_ce_loss": 0.10888671875, "train/clean_loss_scale": 13.058990478515625, "train/clean_loss_weight": 2.611798048019409, "train/combined_loss": 1.7063406705856323, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.1670616113744076, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4219505786895752 }, { "epoch": 1.6729857819905214, "step": 706, "train/alpha": 0.2, "train/clean_ce_loss": 0.13671875, "train/clean_loss_scale": 9.217519760131836, "train/clean_loss_weight": 1.8435039520263672, "train/combined_loss": 1.5122493505477905, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16729857819905214, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2602077722549438 }, { "epoch": 1.6729857819905214, "step": 706, "train/alpha": 0.2, "train/clean_ce_loss": 0.08056640625, "train/clean_loss_scale": 7.667296409606934, "train/clean_loss_weight": 1.5334593057632446, "train/combined_loss": 0.7412717938423157, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16729857819905214, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6177265048027039 }, { "epoch": 1.6729857819905214, "step": 706, "train/alpha": 0.2, "train/clean_ce_loss": 0.171875, "train/clean_loss_scale": 7.339939594268799, "train/clean_loss_weight": 1.4679878950119019, "train/combined_loss": 1.5138624906539917, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16729857819905214, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.261552095413208 }, { "epoch": 1.6729857819905214, "step": 706, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 10.158706665039062, "train/clean_loss_weight": 2.0317413806915283, "train/combined_loss": 0.9761881828308105, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16729857819905214, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8134901523590088 }, { "epoch": 1.675355450236967, "step": 707, "train/alpha": 0.2, "train/clean_ce_loss": 0.04296875, "train/clean_loss_scale": 30.656299591064453, "train/clean_loss_weight": 6.131259918212891, "train/combined_loss": 1.5807154178619385, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.1675355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.317262887954712 }, { "epoch": 1.675355450236967, "step": 707, "train/alpha": 0.2, "train/clean_ce_loss": 0.08984375, "train/clean_loss_scale": 10.454997062683105, "train/clean_loss_weight": 2.0909993648529053, "train/combined_loss": 1.1271793842315674, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.1675355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9393161535263062 }, { "epoch": 1.675355450236967, "step": 707, "train/alpha": 0.2, "train/clean_ce_loss": 0.00787353515625, "train/clean_loss_scale": 36.893863677978516, "train/clean_loss_weight": 7.378772735595703, "train/combined_loss": 0.3485821783542633, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.1675355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.290485143661499 }, { "epoch": 1.675355450236967, "step": 707, "train/alpha": 0.2, "train/clean_ce_loss": 0.047607421875, "train/clean_loss_scale": 14.919342041015625, "train/clean_loss_weight": 2.983868360519409, "train/combined_loss": 0.8523256778717041, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.1675355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.710271418094635 }, { "epoch": 1.6777251184834123, "step": 708, "train/alpha": 0.2, "train/clean_ce_loss": 0.032958984375, "train/clean_loss_scale": 21.96517562866211, "train/clean_loss_weight": 4.393035411834717, "train/combined_loss": 0.868739902973175, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16777251184834124, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7239499092102051 }, { "epoch": 1.6777251184834123, "step": 708, "train/alpha": 0.2, "train/clean_ce_loss": 0.08837890625, "train/clean_loss_scale": 10.243817329406738, "train/clean_loss_weight": 2.0487635135650635, "train/combined_loss": 1.086404800415039, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16777251184834124, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9053373336791992 }, { "epoch": 1.6777251184834123, "step": 708, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 9.8527193069458, "train/clean_loss_weight": 1.9705438613891602, "train/combined_loss": 0.7620462775230408, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16777251184834124, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6350385546684265 }, { "epoch": 1.6777251184834123, "step": 708, "train/alpha": 0.2, "train/clean_ce_loss": 0.033203125, "train/clean_loss_scale": 13.527641296386719, "train/clean_loss_weight": 2.7055282592773438, "train/combined_loss": 0.5389919877052307, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16777251184834124, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.44915997982025146 }, { "epoch": 1.6800947867298577, "step": 709, "train/alpha": 0.2, "train/clean_ce_loss": 0.032958984375, "train/clean_loss_scale": 17.396665573120117, "train/clean_loss_weight": 3.4793331623077393, "train/combined_loss": 0.688051700592041, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.1680094786729858, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5733764171600342 }, { "epoch": 1.6800947867298577, "step": 709, "train/alpha": 0.2, "train/clean_ce_loss": 0.05908203125, "train/clean_loss_scale": 9.322935104370117, "train/clean_loss_weight": 1.8645870685577393, "train/combined_loss": 0.6609815359115601, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.1680094786729858, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5508179664611816 }, { "epoch": 1.6800947867298577, "step": 709, "train/alpha": 0.2, "train/clean_ce_loss": 0.0013580322265625, "train/clean_loss_scale": 137.84173583984375, "train/clean_loss_weight": 27.568347930908203, "train/combined_loss": 0.22463223338127136, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.1680094786729858, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.18719352781772614 }, { "epoch": 1.6800947867298577, "step": 709, "train/alpha": 0.2, "train/clean_ce_loss": 0.021728515625, "train/clean_loss_scale": 17.050642013549805, "train/clean_loss_weight": 3.410128355026245, "train/combined_loss": 0.44458216428756714, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.1680094786729858, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3704851269721985 }, { "epoch": 1.6824644549763033, "grad_norm": 16.61313819885254, "learning_rate": 2e-06, "loss": 3.1113, "step": 710 }, { "epoch": 1.6824644549763033, "step": 710, "train/alpha": 0.2, "train/clean_ce_loss": 0.0150146484375, "train/clean_loss_scale": 25.69637107849121, "train/clean_loss_weight": 5.1392741203308105, "train/combined_loss": 0.4629863500595093, "train/pos_loss/pos0": 0.016439097002148628, "train/pos_loss/pos1": 0.12460540980100632, "train/pos_loss/pos2": 0.2560836672782898, "train/pos_loss/pos3": 0.6792536377906799, "train/pos_loss/pos4": 1.248022437095642, "train/pos_loss/pos5": 1.9392858743667603, "train/pos_loss/pos6": 2.3385794162750244, "train/pos_loss/weight0": 2.232121706008911, "train/pos_loss/weight1": 1.7096779346466064, "train/pos_loss/weight2": 1.2490352392196655, "train/pos_loss/weight3": 0.8691143989562988, "train/pos_loss/weight4": 0.5001435279846191, "train/pos_loss/weight5": 0.2804758548736572, "train/pos_loss/weight6": 0.15943105518817902, "train/pos_weight_progress": 0.16824644549763032, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.38582196831703186 }, { "epoch": 1.6824644549763033, "step": 710, "train/alpha": 0.2, "train/clean_ce_loss": 0.049072265625, "train/clean_loss_scale": 25.025392532348633, "train/clean_loss_weight": 5.0050787925720215, "train/combined_loss": 1.473663330078125, "train/pos_loss/pos0": 0.2053776979446411, "train/pos_loss/pos1": 0.599375307559967, "train/pos_loss/pos2": 1.4125490188598633, "train/pos_loss/pos3": 2.3230724334716797, "train/pos_loss/pos4": 2.682227611541748, "train/pos_loss/pos5": 3.381563186645508, "train/pos_loss/pos6": 3.749739646911621, "train/pos_loss/weight0": 2.134654998779297, "train/pos_loss/weight1": 1.6634079217910767, "train/pos_loss/weight2": 1.2241116762161255, "train/pos_loss/weight3": 0.8774441480636597, "train/pos_loss/weight4": 0.5579608082771301, "train/pos_loss/weight5": 0.32534193992614746, "train/pos_loss/weight6": 0.21707837283611298, "train/pos_weight_progress": 0.16824644549763032, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2280527353286743 }, { "epoch": 1.6824644549763033, "step": 710, "train/alpha": 0.2, "train/clean_ce_loss": 0.1337890625, "train/clean_loss_scale": 7.06768274307251, "train/clean_loss_weight": 1.413536548614502, "train/combined_loss": 1.1346943378448486, "train/pos_loss/pos0": 0.4497188925743103, "train/pos_loss/pos1": 0.9061763286590576, "train/pos_loss/pos2": 1.3274469375610352, "train/pos_loss/pos3": 1.9735803604125977, "train/pos_loss/pos4": 2.3629262447357178, "train/pos_loss/pos5": 2.9103565216064453, "train/pos_loss/pos6": 3.380566120147705, "train/pos_loss/weight0": 2.912395477294922, "train/pos_loss/weight1": 1.978874921798706, "train/pos_loss/weight2": 1.2552344799041748, "train/pos_loss/weight3": 0.5656540989875793, "train/pos_loss/weight4": 0.20711176097393036, "train/pos_loss/weight5": 0.06096271425485611, "train/pos_loss/weight6": 0.019766097888350487, "train/pos_weight_progress": 0.16824644549763032, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9455786347389221 }, { "epoch": 1.6824644549763033, "step": 710, "train/alpha": 0.2, "train/clean_ce_loss": 0.0172119140625, "train/clean_loss_scale": 11.885473251342773, "train/clean_loss_weight": 2.3770947456359863, "train/combined_loss": 0.2454860806465149, "train/pos_loss/pos0": 0.024976948276162148, "train/pos_loss/pos1": 0.14569886028766632, "train/pos_loss/pos2": 0.33896321058273315, "train/pos_loss/pos3": 0.7579336762428284, "train/pos_loss/pos4": 1.0720034837722778, "train/pos_loss/pos5": 1.6513177156448364, "train/pos_loss/pos6": 2.1271817684173584, "train/pos_loss/weight0": 3.135862350463867, "train/pos_loss/weight1": 2.012639045715332, "train/pos_loss/weight2": 1.1057398319244385, "train/pos_loss/weight3": 0.5063788294792175, "train/pos_loss/weight4": 0.173769012093544, "train/pos_loss/weight5": 0.05021188035607338, "train/pos_loss/weight6": 0.015398883260786533, "train/pos_weight_progress": 0.16824644549763032, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.20457173883914948 }, { "epoch": 1.6848341232227488, "step": 711, "train/alpha": 0.2, "train/clean_ce_loss": 0.14453125, "train/clean_loss_scale": 8.214531898498535, "train/clean_loss_weight": 1.6429064273834229, "train/combined_loss": 1.4247078895568848, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16848341232227487, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1872565746307373 }, { "epoch": 1.6848341232227488, "step": 711, "train/alpha": 0.2, "train/clean_ce_loss": 0.021728515625, "train/clean_loss_scale": 17.19426155090332, "train/clean_loss_weight": 3.438852310180664, "train/combined_loss": 0.4483269453048706, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16848341232227487, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.37360578775405884 }, { "epoch": 1.6848341232227488, "step": 711, "train/alpha": 0.2, "train/clean_ce_loss": 0.10888671875, "train/clean_loss_scale": 11.070508003234863, "train/clean_loss_weight": 2.214101552963257, "train/combined_loss": 1.4465175867080688, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16848341232227487, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.205431342124939 }, { "epoch": 1.6848341232227488, "step": 711, "train/alpha": 0.2, "train/clean_ce_loss": 0.056396484375, "train/clean_loss_scale": 17.163484573364258, "train/clean_loss_weight": 3.432697057723999, "train/combined_loss": 1.1615521907806396, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16848341232227487, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9679601788520813 }, { "epoch": 1.6872037914691944, "step": 712, "train/alpha": 0.2, "train/clean_ce_loss": 0.0771484375, "train/clean_loss_scale": 13.337278366088867, "train/clean_loss_weight": 2.6674556732177734, "train/combined_loss": 1.2347402572631836, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16872037914691942, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0289502143859863 }, { "epoch": 1.6872037914691944, "step": 712, "train/alpha": 0.2, "train/clean_ce_loss": 0.0291748046875, "train/clean_loss_scale": 18.033143997192383, "train/clean_loss_weight": 3.606628894805908, "train/combined_loss": 0.6313361525535583, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16872037914691942, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5261134505271912 }, { "epoch": 1.6872037914691944, "step": 712, "train/alpha": 0.2, "train/clean_ce_loss": 0.007659912109375, "train/clean_loss_scale": 79.94078063964844, "train/clean_loss_weight": 15.98815631866455, "train/combined_loss": 0.7348072528839111, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16872037914691942, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6123393774032593 }, { "epoch": 1.6872037914691944, "step": 712, "train/alpha": 0.2, "train/clean_ce_loss": 0.166015625, "train/clean_loss_scale": 9.255857467651367, "train/clean_loss_weight": 1.8511714935302734, "train/combined_loss": 1.843940258026123, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16872037914691942, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5366169214248657 }, { "epoch": 1.6895734597156398, "step": 713, "train/alpha": 0.2, "train/clean_ce_loss": 0.04150390625, "train/clean_loss_scale": 10.412386894226074, "train/clean_loss_weight": 2.082477331161499, "train/combined_loss": 0.5185856819152832, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16895734597156398, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43215471506118774 }, { "epoch": 1.6895734597156398, "step": 713, "train/alpha": 0.2, "train/clean_ce_loss": 0.046142578125, "train/clean_loss_scale": 12.975172996520996, "train/clean_loss_weight": 2.595034599304199, "train/combined_loss": 0.7184494733810425, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16895734597156398, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.598707914352417 }, { "epoch": 1.6895734597156398, "step": 713, "train/alpha": 0.2, "train/clean_ce_loss": 0.061279296875, "train/clean_loss_scale": 17.292415618896484, "train/clean_loss_weight": 3.4584832191467285, "train/combined_loss": 1.2716004848480225, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16895734597156398, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0596671104431152 }, { "epoch": 1.6895734597156398, "step": 713, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 14.132658004760742, "train/clean_loss_weight": 2.8265316486358643, "train/combined_loss": 1.3580601215362549, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16895734597156398, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1317167282104492 }, { "epoch": 1.6919431279620853, "step": 714, "train/alpha": 0.2, "train/clean_ce_loss": 0.1171875, "train/clean_loss_scale": 11.637114524841309, "train/clean_loss_weight": 2.327422857284546, "train/combined_loss": 1.6364692449569702, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16919431279620853, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3637243509292603 }, { "epoch": 1.6919431279620853, "step": 714, "train/alpha": 0.2, "train/clean_ce_loss": 0.0615234375, "train/clean_loss_scale": 12.371383666992188, "train/clean_loss_weight": 2.4742767810821533, "train/combined_loss": 0.9133560657501221, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16919431279620853, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7611300349235535 }, { "epoch": 1.6919431279620853, "step": 714, "train/alpha": 0.2, "train/clean_ce_loss": 0.044677734375, "train/clean_loss_scale": 11.848014831542969, "train/clean_loss_weight": 2.369602918624878, "train/combined_loss": 0.6352109909057617, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16919431279620853, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5293424725532532 }, { "epoch": 1.6919431279620853, "step": 714, "train/alpha": 0.2, "train/clean_ce_loss": 0.006927490234375, "train/clean_loss_scale": 69.6891098022461, "train/clean_loss_weight": 13.937822341918945, "train/combined_loss": 0.5793247818946838, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16919431279620853, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4827706515789032 }, { "epoch": 1.6943127962085307, "grad_norm": 19.535253524780273, "learning_rate": 2e-06, "loss": 3.3798, "step": 715 }, { "epoch": 1.6943127962085307, "step": 715, "train/alpha": 0.2, "train/clean_ce_loss": 0.0272216796875, "train/clean_loss_scale": 15.804071426391602, "train/clean_loss_weight": 3.1608142852783203, "train/combined_loss": 0.5162560343742371, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16943127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4302133619785309 }, { "epoch": 1.6943127962085307, "step": 715, "train/alpha": 0.2, "train/clean_ce_loss": 0.12890625, "train/clean_loss_scale": 11.163447380065918, "train/clean_loss_weight": 2.232689619064331, "train/combined_loss": 1.7268457412719727, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16943127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4390381574630737 }, { "epoch": 1.6943127962085307, "step": 715, "train/alpha": 0.2, "train/clean_ce_loss": 0.06396484375, "train/clean_loss_scale": 13.24881649017334, "train/clean_loss_weight": 2.649763345718384, "train/combined_loss": 1.0169501304626465, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16943127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8474584817886353 }, { "epoch": 1.6943127962085307, "step": 715, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 14.624162673950195, "train/clean_loss_weight": 2.924832582473755, "train/combined_loss": 1.1567940711975098, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16943127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.963995099067688 }, { "epoch": 1.6966824644549763, "step": 716, "train/alpha": 0.2, "train/clean_ce_loss": 0.007720947265625, "train/clean_loss_scale": 66.17658996582031, "train/clean_loss_weight": 13.235318183898926, "train/combined_loss": 0.6131351590156555, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16966824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5109459757804871 }, { "epoch": 1.6966824644549763, "step": 716, "train/alpha": 0.2, "train/clean_ce_loss": 0.0615234375, "train/clean_loss_scale": 14.500664710998535, "train/clean_loss_weight": 2.900132894515991, "train/combined_loss": 1.070556879043579, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16966824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8921307325363159 }, { "epoch": 1.6966824644549763, "step": 716, "train/alpha": 0.2, "train/clean_ce_loss": 0.0751953125, "train/clean_loss_scale": 9.931461334228516, "train/clean_loss_weight": 1.9862922430038452, "train/combined_loss": 0.8961592316627502, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16966824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7467993497848511 }, { "epoch": 1.6966824644549763, "step": 716, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 12.696806907653809, "train/clean_loss_weight": 2.5393614768981934, "train/combined_loss": 1.4135117530822754, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16966824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1779264211654663 }, { "epoch": 1.6990521327014219, "step": 717, "train/alpha": 0.2, "train/clean_ce_loss": 0.08642578125, "train/clean_loss_scale": 17.6523494720459, "train/clean_loss_weight": 3.5304698944091797, "train/combined_loss": 1.8307416439056396, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16990521327014219, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.525618076324463 }, { "epoch": 1.6990521327014219, "step": 717, "train/alpha": 0.2, "train/clean_ce_loss": 0.08251953125, "train/clean_loss_scale": 9.942237854003906, "train/clean_loss_weight": 1.9884475469589233, "train/combined_loss": 0.9845145344734192, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16990521327014219, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8204287886619568 }, { "epoch": 1.6990521327014219, "step": 717, "train/alpha": 0.2, "train/clean_ce_loss": 0.015869140625, "train/clean_loss_scale": 26.751386642456055, "train/clean_loss_weight": 5.350277423858643, "train/combined_loss": 0.5094258189201355, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16990521327014219, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4245215058326721 }, { "epoch": 1.6990521327014219, "step": 717, "train/alpha": 0.2, "train/clean_ce_loss": 0.057861328125, "train/clean_loss_scale": 11.360933303833008, "train/clean_loss_weight": 2.272186756134033, "train/combined_loss": 0.7888304591178894, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.16990521327014219, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.657358705997467 }, { "epoch": 1.7014218009478674, "step": 718, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 11.709152221679688, "train/clean_loss_weight": 2.3418304920196533, "train/combined_loss": 0.9193513989448547, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.17014218009478674, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7661261558532715 }, { "epoch": 1.7014218009478674, "step": 718, "train/alpha": 0.2, "train/clean_ce_loss": 0.0908203125, "train/clean_loss_scale": 8.249166488647461, "train/clean_loss_weight": 1.64983332157135, "train/combined_loss": 0.8990302681922913, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.17014218009478674, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7491918802261353 }, { "epoch": 1.7014218009478674, "step": 718, "train/alpha": 0.2, "train/clean_ce_loss": 0.103515625, "train/clean_loss_scale": 10.285454750061035, "train/clean_loss_weight": 2.057090997695923, "train/combined_loss": 1.2776463031768799, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.17014218009478674, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0647052526474 }, { "epoch": 1.7014218009478674, "step": 718, "train/alpha": 0.2, "train/clean_ce_loss": 0.0458984375, "train/clean_loss_scale": 10.815450668334961, "train/clean_loss_weight": 2.163090229034424, "train/combined_loss": 0.5956947803497314, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.17014218009478674, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4964123070240021 }, { "epoch": 1.7037914691943128, "step": 719, "train/alpha": 0.2, "train/clean_ce_loss": 0.10205078125, "train/clean_loss_scale": 9.388919830322266, "train/clean_loss_weight": 1.877784013748169, "train/combined_loss": 1.1497759819030762, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.1703791469194313, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9581466317176819 }, { "epoch": 1.7037914691943128, "step": 719, "train/alpha": 0.2, "train/clean_ce_loss": 0.0029754638671875, "train/clean_loss_scale": 99.68096923828125, "train/clean_loss_weight": 19.936193466186523, "train/combined_loss": 0.3559165596961975, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.1703791469194313, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.29659712314605713 }, { "epoch": 1.7037914691943128, "step": 719, "train/alpha": 0.2, "train/clean_ce_loss": 0.053955078125, "train/clean_loss_scale": 22.77064323425293, "train/clean_loss_weight": 4.554128646850586, "train/combined_loss": 1.4743101596832275, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.1703791469194313, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.228591799736023 }, { "epoch": 1.7037914691943128, "step": 719, "train/alpha": 0.2, "train/clean_ce_loss": 0.041015625, "train/clean_loss_scale": 15.302803993225098, "train/clean_loss_weight": 3.060560941696167, "train/combined_loss": 0.7531849145889282, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.1703791469194313, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6276540756225586 }, { "epoch": 1.7061611374407581, "grad_norm": 27.122018814086914, "learning_rate": 2e-06, "loss": 3.428, "step": 720 }, { "epoch": 1.7061611374407581, "step": 720, "train/alpha": 0.2, "train/clean_ce_loss": 0.0189208984375, "train/clean_loss_scale": 25.742717742919922, "train/clean_loss_weight": 5.148543834686279, "train/combined_loss": 0.5844904184341431, "train/pos_loss/pos0": 0.07921815663576126, "train/pos_loss/pos1": 0.10737527906894684, "train/pos_loss/pos2": 0.3383953869342804, "train/pos_loss/pos3": 0.7378689646720886, "train/pos_loss/pos4": 1.2076853513717651, "train/pos_loss/pos5": 1.806860327720642, "train/pos_loss/pos6": 2.5646424293518066, "train/pos_loss/weight0": 2.061455726623535, "train/pos_loss/weight1": 1.6269153356552124, "train/pos_loss/weight2": 1.2365000247955322, "train/pos_loss/weight3": 0.8826724886894226, "train/pos_loss/weight4": 0.572197675704956, "train/pos_loss/weight5": 0.36986005306243896, "train/pos_loss/weight6": 0.2503986358642578, "train/pos_weight_progress": 0.17061611374407584, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4870753586292267 }, { "epoch": 1.7061611374407581, "step": 720, "train/alpha": 0.2, "train/clean_ce_loss": 0.0213623046875, "train/clean_loss_scale": 18.43931770324707, "train/clean_loss_weight": 3.68786358833313, "train/combined_loss": 0.47268760204315186, "train/pos_loss/pos0": 0.055671852082014084, "train/pos_loss/pos1": 0.1786995530128479, "train/pos_loss/pos2": 0.30361393094062805, "train/pos_loss/pos3": 0.6038338541984558, "train/pos_loss/pos4": 1.1387277841567993, "train/pos_loss/pos5": 1.4551984071731567, "train/pos_loss/pos6": 1.8202732801437378, "train/pos_loss/weight0": 2.176285982131958, "train/pos_loss/weight1": 1.6623033285140991, "train/pos_loss/weight2": 1.2343645095825195, "train/pos_loss/weight3": 0.8729231357574463, "train/pos_loss/weight4": 0.5373142957687378, "train/pos_loss/weight5": 0.3149428367614746, "train/pos_loss/weight6": 0.20186573266983032, "train/pos_weight_progress": 0.17061611374407584, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3939063251018524 }, { "epoch": 1.7061611374407581, "step": 720, "train/alpha": 0.2, "train/clean_ce_loss": 0.041748046875, "train/clean_loss_scale": 13.417218208312988, "train/clean_loss_weight": 2.683443784713745, "train/combined_loss": 0.6721711754798889, "train/pos_loss/pos0": 0.08953400701284409, "train/pos_loss/pos1": 0.2679377496242523, "train/pos_loss/pos2": 0.4223972260951996, "train/pos_loss/pos3": 0.8366395831108093, "train/pos_loss/pos4": 1.2671840190887451, "train/pos_loss/pos5": 1.8766140937805176, "train/pos_loss/pos6": 2.0803470611572266, "train/pos_loss/weight0": 2.0280237197875977, "train/pos_loss/weight1": 1.607017993927002, "train/pos_loss/weight2": 1.2148630619049072, "train/pos_loss/weight3": 0.8834865093231201, "train/pos_loss/weight4": 0.6120595932006836, "train/pos_loss/weight5": 0.39079174399375916, "train/pos_loss/weight6": 0.2637574076652527, "train/pos_weight_progress": 0.17061611374407584, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5601426362991333 }, { "epoch": 1.7061611374407581, "step": 720, "train/alpha": 0.2, "train/clean_ce_loss": 0.1357421875, "train/clean_loss_scale": 9.616829872131348, "train/clean_loss_weight": 1.9233659505844116, "train/combined_loss": 1.5664914846420288, "train/pos_loss/pos0": 0.4578695297241211, "train/pos_loss/pos1": 0.927388608455658, "train/pos_loss/pos2": 1.3768118619918823, "train/pos_loss/pos3": 2.037421226501465, "train/pos_loss/pos4": 2.61313796043396, "train/pos_loss/pos5": 3.1207244396209717, "train/pos_loss/pos6": 3.813945770263672, "train/pos_loss/weight0": 2.18208384513855, "train/pos_loss/weight1": 1.6816869974136353, "train/pos_loss/weight2": 1.2299988269805908, "train/pos_loss/weight3": 0.8533729314804077, "train/pos_loss/weight4": 0.5396796464920044, "train/pos_loss/weight5": 0.31399810314178467, "train/pos_loss/weight6": 0.19917960464954376, "train/pos_weight_progress": 0.17061611374407584, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.305409550666809 }, { "epoch": 1.7085308056872037, "step": 721, "train/alpha": 0.2, "train/clean_ce_loss": 0.142578125, "train/clean_loss_scale": 6.866701126098633, "train/clean_loss_weight": 1.3733402490615845, "train/combined_loss": 1.1748496294021606, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17085308056872037, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.979041337966919 }, { "epoch": 1.7085308056872037, "step": 721, "train/alpha": 0.2, "train/clean_ce_loss": 0.0126953125, "train/clean_loss_scale": 11.087657928466797, "train/clean_loss_weight": 2.217531681060791, "train/combined_loss": 0.16891354322433472, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17085308056872037, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.14076128602027893 }, { "epoch": 1.7085308056872037, "step": 721, "train/alpha": 0.2, "train/clean_ce_loss": 0.0458984375, "train/clean_loss_scale": 15.841348648071289, "train/clean_loss_weight": 3.1682698726654053, "train/combined_loss": 0.8725118041038513, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17085308056872037, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7270931601524353 }, { "epoch": 1.7085308056872037, "step": 721, "train/alpha": 0.2, "train/clean_ce_loss": 0.0703125, "train/clean_loss_scale": 8.878482818603516, "train/clean_loss_weight": 1.7756966352462769, "train/combined_loss": 0.7491220235824585, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17085308056872037, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6242683529853821 }, { "epoch": 1.7109004739336493, "step": 722, "train/alpha": 0.2, "train/clean_ce_loss": 0.09765625, "train/clean_loss_scale": 5.802281379699707, "train/clean_loss_weight": 1.1604562997817993, "train/combined_loss": 0.6799548864364624, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17109004739336492, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5666290521621704 }, { "epoch": 1.7109004739336493, "step": 722, "train/alpha": 0.2, "train/clean_ce_loss": 0.1357421875, "train/clean_loss_scale": 8.046817779541016, "train/clean_loss_weight": 1.6093635559082031, "train/combined_loss": 1.31075119972229, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17109004739336492, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0922926664352417 }, { "epoch": 1.7109004739336493, "step": 722, "train/alpha": 0.2, "train/clean_ce_loss": 0.04931640625, "train/clean_loss_scale": 8.348029136657715, "train/clean_loss_weight": 1.6696058511734009, "train/combined_loss": 0.4940337538719177, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17109004739336492, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.41169479489326477 }, { "epoch": 1.7109004739336493, "step": 722, "train/alpha": 0.2, "train/clean_ce_loss": 0.1337890625, "train/clean_loss_scale": 8.56623649597168, "train/clean_loss_weight": 1.713247299194336, "train/combined_loss": 1.3752825260162354, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17109004739336492, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.146068811416626 }, { "epoch": 1.7132701421800949, "step": 723, "train/alpha": 0.2, "train/clean_ce_loss": 0.03759765625, "train/clean_loss_scale": 9.158190727233887, "train/clean_loss_weight": 1.831638216972351, "train/combined_loss": 0.4131917953491211, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17132701421800947, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3443264961242676 }, { "epoch": 1.7132701421800949, "step": 723, "train/alpha": 0.2, "train/clean_ce_loss": 0.08447265625, "train/clean_loss_scale": 8.12495231628418, "train/clean_loss_weight": 1.624990463256836, "train/combined_loss": 0.8236035108566284, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17132701421800947, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6863362789154053 }, { "epoch": 1.7132701421800949, "step": 723, "train/alpha": 0.2, "train/clean_ce_loss": 0.046142578125, "train/clean_loss_scale": 7.611093521118164, "train/clean_loss_weight": 1.5222187042236328, "train/combined_loss": 0.42143458127975464, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17132701421800947, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.35119548439979553 }, { "epoch": 1.7132701421800949, "step": 723, "train/alpha": 0.2, "train/clean_ce_loss": 0.1318359375, "train/clean_loss_scale": 6.705546855926514, "train/clean_loss_weight": 1.3411093950271606, "train/combined_loss": 1.0608384609222412, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17132701421800947, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8840320706367493 }, { "epoch": 1.7156398104265402, "step": 724, "train/alpha": 0.2, "train/clean_ce_loss": 0.0830078125, "train/clean_loss_scale": 7.529605865478516, "train/clean_loss_weight": 1.5059212446212769, "train/combined_loss": 0.7500193119049072, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17156398104265402, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6250160932540894 }, { "epoch": 1.7156398104265402, "step": 724, "train/alpha": 0.2, "train/clean_ce_loss": 0.09228515625, "train/clean_loss_scale": 6.31802225112915, "train/clean_loss_weight": 1.2636045217514038, "train/combined_loss": 0.6996716260910034, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17156398104265402, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5830596685409546 }, { "epoch": 1.7156398104265402, "step": 724, "train/alpha": 0.2, "train/clean_ce_loss": 0.036376953125, "train/clean_loss_scale": 8.09101676940918, "train/clean_loss_weight": 1.6182034015655518, "train/combined_loss": 0.3531918525695801, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17156398104265402, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2943265438079834 }, { "epoch": 1.7156398104265402, "step": 724, "train/alpha": 0.2, "train/clean_ce_loss": 0.049560546875, "train/clean_loss_scale": 13.827232360839844, "train/clean_loss_weight": 2.765446424484253, "train/combined_loss": 0.8223422765731812, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17156398104265402, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.685285210609436 }, { "epoch": 1.7180094786729858, "grad_norm": 16.67449378967285, "learning_rate": 2e-06, "loss": 3.6725, "step": 725 }, { "epoch": 1.7180094786729858, "step": 725, "train/alpha": 0.2, "train/clean_ce_loss": 0.10791015625, "train/clean_loss_scale": 7.810417175292969, "train/clean_loss_weight": 1.5620834827423096, "train/combined_loss": 1.0113880634307861, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17180094786729858, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.842823326587677 }, { "epoch": 1.7180094786729858, "step": 725, "train/alpha": 0.2, "train/clean_ce_loss": 0.095703125, "train/clean_loss_scale": 7.911308288574219, "train/clean_loss_weight": 1.5822616815567017, "train/combined_loss": 0.9085643291473389, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17180094786729858, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7571369409561157 }, { "epoch": 1.7180094786729858, "step": 725, "train/alpha": 0.2, "train/clean_ce_loss": 0.095703125, "train/clean_loss_scale": 7.712500095367432, "train/clean_loss_weight": 1.5425000190734863, "train/combined_loss": 0.8857324123382568, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17180094786729858, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.738110363483429 }, { "epoch": 1.7180094786729858, "step": 725, "train/alpha": 0.2, "train/clean_ce_loss": 0.060302734375, "train/clean_loss_scale": 2.8330142498016357, "train/clean_loss_weight": 0.566602885723114, "train/combined_loss": 0.2050062119960785, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17180094786729858, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.17083850502967834 }, { "epoch": 1.7203791469194312, "step": 726, "train/alpha": 0.2, "train/clean_ce_loss": 0.1787109375, "train/clean_loss_scale": 5.5169782638549805, "train/clean_loss_weight": 1.103395700454712, "train/combined_loss": 1.1831332445144653, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17203791469194313, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9859443306922913 }, { "epoch": 1.7203791469194312, "step": 726, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 5.978992938995361, "train/clean_loss_weight": 1.195798635482788, "train/combined_loss": 0.5325040817260742, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17203791469194313, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4437533915042877 }, { "epoch": 1.7203791469194312, "step": 726, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 7.235395908355713, "train/clean_loss_weight": 1.4470791816711426, "train/combined_loss": 0.7588686943054199, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17203791469194313, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.632390558719635 }, { "epoch": 1.7203791469194312, "step": 726, "train/alpha": 0.2, "train/clean_ce_loss": 0.009765625, "train/clean_loss_scale": 18.47421646118164, "train/clean_loss_weight": 3.694843292236328, "train/combined_loss": 0.21649473905563354, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17203791469194313, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.18041227757930756 }, { "epoch": 1.7227488151658767, "step": 727, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 6.531686305999756, "train/clean_loss_weight": 1.3063372373580933, "train/combined_loss": 0.6353085041046143, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17227488151658768, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5294237732887268 }, { "epoch": 1.7227488151658767, "step": 727, "train/alpha": 0.2, "train/clean_ce_loss": 0.091796875, "train/clean_loss_scale": 5.671412467956543, "train/clean_loss_weight": 1.1342824697494507, "train/combined_loss": 0.6247415542602539, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17227488151658768, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5206179618835449 }, { "epoch": 1.7227488151658767, "step": 727, "train/alpha": 0.2, "train/clean_ce_loss": 0.078125, "train/clean_loss_scale": 6.745296478271484, "train/clean_loss_weight": 1.3490593433380127, "train/combined_loss": 0.6323715448379517, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17227488151658768, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5269762873649597 }, { "epoch": 1.7227488151658767, "step": 727, "train/alpha": 0.2, "train/clean_ce_loss": 0.01190185546875, "train/clean_loss_scale": 13.87975025177002, "train/clean_loss_weight": 2.7759501934051514, "train/combined_loss": 0.1982337385416031, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17227488151658768, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1651947796344757 }, { "epoch": 1.7251184834123223, "step": 728, "train/alpha": 0.2, "train/clean_ce_loss": 0.12255859375, "train/clean_loss_scale": 11.313819885253906, "train/clean_loss_weight": 2.2627639770507812, "train/combined_loss": 1.6639270782470703, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17251184834123223, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3866058588027954 }, { "epoch": 1.7251184834123223, "step": 728, "train/alpha": 0.2, "train/clean_ce_loss": 0.09033203125, "train/clean_loss_scale": 9.545166969299316, "train/clean_loss_weight": 1.9090334177017212, "train/combined_loss": 1.0346812009811401, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17251184834123223, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.862234354019165 }, { "epoch": 1.7251184834123223, "step": 728, "train/alpha": 0.2, "train/clean_ce_loss": 0.04248046875, "train/clean_loss_scale": 6.220672130584717, "train/clean_loss_weight": 1.2441344261169434, "train/combined_loss": 0.31710848212242126, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17251184834123223, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2642570734024048 }, { "epoch": 1.7251184834123223, "step": 728, "train/alpha": 0.2, "train/clean_ce_loss": 0.06689453125, "train/clean_loss_scale": 6.256336212158203, "train/clean_loss_weight": 1.2512673139572144, "train/combined_loss": 0.5022175908088684, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17251184834123223, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4185146689414978 }, { "epoch": 1.727488151658768, "step": 729, "train/alpha": 0.2, "train/clean_ce_loss": 0.060302734375, "train/clean_loss_scale": 14.47942066192627, "train/clean_loss_weight": 2.8958842754364014, "train/combined_loss": 1.0477783679962158, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17274881516587678, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8731486797332764 }, { "epoch": 1.727488151658768, "step": 729, "train/alpha": 0.2, "train/clean_ce_loss": 0.0303955078125, "train/clean_loss_scale": 23.262407302856445, "train/clean_loss_weight": 4.652481555938721, "train/combined_loss": 0.8484871983528137, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17274881516587678, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7070726752281189 }, { "epoch": 1.727488151658768, "step": 729, "train/alpha": 0.2, "train/clean_ce_loss": 0.07763671875, "train/clean_loss_scale": 6.307367324829102, "train/clean_loss_weight": 1.261473536491394, "train/combined_loss": 0.587619960308075, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17274881516587678, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4896833002567291 }, { "epoch": 1.727488151658768, "step": 729, "train/alpha": 0.2, "train/clean_ce_loss": 0.0023040771484375, "train/clean_loss_scale": 73.55389404296875, "train/clean_loss_weight": 14.710779190063477, "train/combined_loss": 0.20336860418319702, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17274881516587678, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.16947384178638458 }, { "epoch": 1.7298578199052133, "grad_norm": 15.21699333190918, "learning_rate": 2e-06, "loss": 3.6935, "step": 730 }, { "epoch": 1.7298578199052133, "step": 730, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 7.5880889892578125, "train/clean_loss_weight": 1.5176178216934204, "train/combined_loss": 0.7380602359771729, "train/pos_loss/pos0": 0.31374606490135193, "train/pos_loss/pos1": 0.5802839398384094, "train/pos_loss/pos2": 0.9546489715576172, "train/pos_loss/pos3": 1.3931392431259155, "train/pos_loss/pos4": 1.7597533464431763, "train/pos_loss/pos5": 2.1564652919769287, "train/pos_loss/pos6": 2.5823628902435303, "train/pos_loss/weight0": 3.153332233428955, "train/pos_loss/weight1": 2.038524866104126, "train/pos_loss/weight2": 1.1052184104919434, "train/pos_loss/weight3": 0.49754148721694946, "train/pos_loss/weight4": 0.1546640694141388, "train/pos_loss/weight5": 0.04090070351958275, "train/pos_loss/weight6": 0.009818526916205883, "train/pos_weight_progress": 0.17298578199052134, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.615050196647644 }, { "epoch": 1.7298578199052133, "step": 730, "train/alpha": 0.2, "train/clean_ce_loss": 0.06982421875, "train/clean_loss_scale": 10.09762191772461, "train/clean_loss_weight": 2.019524335861206, "train/combined_loss": 0.8460702896118164, "train/pos_loss/pos0": 0.21301008760929108, "train/pos_loss/pos1": 0.4960315227508545, "train/pos_loss/pos2": 0.9171483516693115, "train/pos_loss/pos3": 1.4780381917953491, "train/pos_loss/pos4": 2.1415398120880127, "train/pos_loss/pos5": 2.5040993690490723, "train/pos_loss/pos6": 2.6515684127807617, "train/pos_loss/weight0": 2.6714272499084473, "train/pos_loss/weight1": 1.8533637523651123, "train/pos_loss/weight2": 1.211317539215088, "train/pos_loss/weight3": 0.6937049031257629, "train/pos_loss/weight4": 0.34202998876571655, "train/pos_loss/weight5": 0.15477167069911957, "train/pos_loss/weight6": 0.07338469475507736, "train/pos_weight_progress": 0.17298578199052134, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7050585746765137 }, { "epoch": 1.7298578199052133, "step": 730, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 6.612227439880371, "train/clean_loss_weight": 1.3224455118179321, "train/combined_loss": 0.6470167636871338, "train/pos_loss/pos0": 0.23247507214546204, "train/pos_loss/pos1": 0.4705846309661865, "train/pos_loss/pos2": 0.7034820318222046, "train/pos_loss/pos3": 0.9804840087890625, "train/pos_loss/pos4": 1.2965469360351562, "train/pos_loss/pos5": 1.680168867111206, "train/pos_loss/pos6": 1.9810175895690918, "train/pos_loss/weight0": 2.6857845783233643, "train/pos_loss/weight1": 1.9061496257781982, "train/pos_loss/weight2": 1.2338941097259521, "train/pos_loss/weight3": 0.6895889639854431, "train/pos_loss/weight4": 0.3108471632003784, "train/pos_loss/weight5": 0.12402507662773132, "train/pos_loss/weight6": 0.04971044138073921, "train/pos_weight_progress": 0.17298578199052134, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5391806364059448 }, { "epoch": 1.7298578199052133, "step": 730, "train/alpha": 0.2, "train/clean_ce_loss": 0.04443359375, "train/clean_loss_scale": 10.198675155639648, "train/clean_loss_weight": 2.0397350788116455, "train/combined_loss": 0.5437965393066406, "train/pos_loss/pos0": 0.1331244558095932, "train/pos_loss/pos1": 0.2722889184951782, "train/pos_loss/pos2": 0.5405065417289734, "train/pos_loss/pos3": 0.7434396147727966, "train/pos_loss/pos4": 0.9883671998977661, "train/pos_loss/pos5": 1.3751394748687744, "train/pos_loss/pos6": 1.6492177248001099, "train/pos_loss/weight0": 2.310483694076538, "train/pos_loss/weight1": 1.7384281158447266, "train/pos_loss/weight2": 1.222214698791504, "train/pos_loss/weight3": 0.8036004304885864, "train/pos_loss/weight4": 0.47709599137306213, "train/pos_loss/weight5": 0.2810533046722412, "train/pos_loss/weight6": 0.16712434589862823, "train/pos_weight_progress": 0.17298578199052134, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4531638026237488 }, { "epoch": 1.7322274881516586, "step": 731, "train/alpha": 0.2, "train/clean_ce_loss": 0.095703125, "train/clean_loss_scale": 11.441019058227539, "train/clean_loss_weight": 2.2882039546966553, "train/combined_loss": 1.313929557800293, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.1732227488151659, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.094941258430481 }, { "epoch": 1.7322274881516586, "step": 731, "train/alpha": 0.2, "train/clean_ce_loss": 0.04296875, "train/clean_loss_scale": 11.88472843170166, "train/clean_loss_weight": 2.376945734024048, "train/combined_loss": 0.6128063201904297, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.1732227488151659, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5106719136238098 }, { "epoch": 1.7322274881516586, "step": 731, "train/alpha": 0.2, "train/clean_ce_loss": 0.0023193359375, "train/clean_loss_scale": 119.84857177734375, "train/clean_loss_weight": 23.969715118408203, "train/combined_loss": 0.3335629105567932, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.1732227488151659, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.277969092130661 }, { "epoch": 1.7322274881516586, "step": 731, "train/alpha": 0.2, "train/clean_ce_loss": 0.00482177734375, "train/clean_loss_scale": 124.93321990966797, "train/clean_loss_weight": 24.986644744873047, "train/combined_loss": 0.7228802442550659, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.1732227488151659, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6024001836776733 }, { "epoch": 1.7345971563981042, "step": 732, "train/alpha": 0.2, "train/clean_ce_loss": 0.0174560546875, "train/clean_loss_scale": 33.464412689208984, "train/clean_loss_weight": 6.692882537841797, "train/combined_loss": 0.7009879350662231, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.1734597156398104, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5841566324234009 }, { "epoch": 1.7345971563981042, "step": 732, "train/alpha": 0.2, "train/clean_ce_loss": 0.04248046875, "train/clean_loss_scale": 14.20095157623291, "train/clean_loss_weight": 2.8401904106140137, "train/combined_loss": 0.723915696144104, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.1734597156398104, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6032630801200867 }, { "epoch": 1.7345971563981042, "step": 732, "train/alpha": 0.2, "train/clean_ce_loss": 0.07568359375, "train/clean_loss_scale": 11.50563907623291, "train/clean_loss_weight": 2.3011279106140137, "train/combined_loss": 1.0449457168579102, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.1734597156398104, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8707880973815918 }, { "epoch": 1.7345971563981042, "step": 732, "train/alpha": 0.2, "train/clean_ce_loss": 0.10302734375, "train/clean_loss_scale": 9.512874603271484, "train/clean_loss_weight": 1.902574896812439, "train/combined_loss": 1.1761034727096558, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.1734597156398104, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9800862073898315 }, { "epoch": 1.7369668246445498, "step": 733, "train/alpha": 0.2, "train/clean_ce_loss": 0.134765625, "train/clean_loss_scale": 9.660927772521973, "train/clean_loss_weight": 1.9321855306625366, "train/combined_loss": 1.5623531341552734, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17369668246445497, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3019609451293945 }, { "epoch": 1.7369668246445498, "step": 733, "train/alpha": 0.2, "train/clean_ce_loss": 0.083984375, "train/clean_loss_scale": 10.042757987976074, "train/clean_loss_weight": 2.008551597595215, "train/combined_loss": 1.0121216773986816, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17369668246445497, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.843434751033783 }, { "epoch": 1.7369668246445498, "step": 733, "train/alpha": 0.2, "train/clean_ce_loss": 0.09716796875, "train/clean_loss_scale": 9.77678394317627, "train/clean_loss_weight": 1.9553568363189697, "train/combined_loss": 1.1399883031845093, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17369668246445497, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9499902129173279 }, { "epoch": 1.7369668246445498, "step": 733, "train/alpha": 0.2, "train/clean_ce_loss": 0.076171875, "train/clean_loss_scale": 12.83552360534668, "train/clean_loss_weight": 2.5671048164367676, "train/combined_loss": 1.1732470989227295, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17369668246445497, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9777058959007263 }, { "epoch": 1.7393364928909953, "step": 734, "train/alpha": 0.2, "train/clean_ce_loss": 0.024658203125, "train/clean_loss_scale": 22.533567428588867, "train/clean_loss_weight": 4.506713390350342, "train/combined_loss": 0.6667647361755371, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17393364928909952, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5556373000144958 }, { "epoch": 1.7393364928909953, "step": 734, "train/alpha": 0.2, "train/clean_ce_loss": 0.004638671875, "train/clean_loss_scale": 91.00775146484375, "train/clean_loss_weight": 18.20155143737793, "train/combined_loss": 0.5065860748291016, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17393364928909952, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.42215508222579956 }, { "epoch": 1.7393364928909953, "step": 734, "train/alpha": 0.2, "train/clean_ce_loss": 0.1044921875, "train/clean_loss_scale": 12.106621742248535, "train/clean_loss_weight": 2.4213244915008545, "train/combined_loss": 1.518056869506836, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17393364928909952, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2650474309921265 }, { "epoch": 1.7393364928909953, "step": 734, "train/alpha": 0.2, "train/clean_ce_loss": 0.09765625, "train/clean_loss_scale": 11.087985038757324, "train/clean_loss_weight": 2.217597007751465, "train/combined_loss": 1.2993731498718262, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17393364928909952, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0828109979629517 }, { "epoch": 1.741706161137441, "grad_norm": 24.079906463623047, "learning_rate": 2e-06, "loss": 3.5772, "step": 735 }, { "epoch": 1.741706161137441, "step": 735, "train/alpha": 0.2, "train/clean_ce_loss": 0.0458984375, "train/clean_loss_scale": 17.553110122680664, "train/clean_loss_weight": 3.510622024536133, "train/combined_loss": 0.9667923450469971, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17417061611374407, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8056603074073792 }, { "epoch": 1.741706161137441, "step": 735, "train/alpha": 0.2, "train/clean_ce_loss": 0.06982421875, "train/clean_loss_scale": 10.795469284057617, "train/clean_loss_weight": 2.1590938568115234, "train/combined_loss": 0.9045422077178955, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17417061611374407, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7537851929664612 }, { "epoch": 1.741706161137441, "step": 735, "train/alpha": 0.2, "train/clean_ce_loss": 0.0556640625, "train/clean_loss_scale": 11.843769073486328, "train/clean_loss_weight": 2.3687539100646973, "train/combined_loss": 0.7911267876625061, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17417061611374407, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.659272313117981 }, { "epoch": 1.741706161137441, "step": 735, "train/alpha": 0.2, "train/clean_ce_loss": 0.07666015625, "train/clean_loss_scale": 11.043359756469727, "train/clean_loss_weight": 2.208672046661377, "train/combined_loss": 1.0159028768539429, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17417061611374407, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8465856909751892 }, { "epoch": 1.7440758293838863, "step": 736, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 10.042956352233887, "train/clean_loss_weight": 2.008591413497925, "train/combined_loss": 1.1180634498596191, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17440758293838862, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9317195415496826 }, { "epoch": 1.7440758293838863, "step": 736, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 10.575040817260742, "train/clean_loss_weight": 2.1150081157684326, "train/combined_loss": 0.9232506155967712, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17440758293838862, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7693755030632019 }, { "epoch": 1.7440758293838863, "step": 736, "train/alpha": 0.2, "train/clean_ce_loss": 0.05419921875, "train/clean_loss_scale": 12.364450454711914, "train/clean_loss_weight": 2.4728901386260986, "train/combined_loss": 0.8041722774505615, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17440758293838862, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6701435446739197 }, { "epoch": 1.7440758293838863, "step": 736, "train/alpha": 0.2, "train/clean_ce_loss": 0.1083984375, "train/clean_loss_scale": 9.562849998474121, "train/clean_loss_weight": 1.9125699996948242, "train/combined_loss": 1.2439175844192505, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17440758293838862, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0365979671478271 }, { "epoch": 1.7464454976303316, "step": 737, "train/alpha": 0.2, "train/clean_ce_loss": 0.00634765625, "train/clean_loss_scale": 63.414127349853516, "train/clean_loss_weight": 12.682826042175293, "train/combined_loss": 0.4830372929573059, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17464454976303317, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.40253108739852905 }, { "epoch": 1.7464454976303316, "step": 737, "train/alpha": 0.2, "train/clean_ce_loss": 0.040283203125, "train/clean_loss_scale": 16.979976654052734, "train/clean_loss_weight": 3.3959953784942627, "train/combined_loss": 0.8208094835281372, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17464454976303317, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6840078830718994 }, { "epoch": 1.7464454976303316, "step": 737, "train/alpha": 0.2, "train/clean_ce_loss": 0.09326171875, "train/clean_loss_scale": 9.547232627868652, "train/clean_loss_weight": 1.9094465970993042, "train/combined_loss": 1.0684696435928345, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17464454976303317, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8903913497924805 }, { "epoch": 1.7464454976303316, "step": 737, "train/alpha": 0.2, "train/clean_ce_loss": 0.004547119140625, "train/clean_loss_scale": 83.12589263916016, "train/clean_loss_weight": 16.625179290771484, "train/combined_loss": 0.45357999205589294, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17464454976303317, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.37798333168029785 }, { "epoch": 1.7488151658767772, "step": 738, "train/alpha": 0.2, "train/clean_ce_loss": 0.048583984375, "train/clean_loss_scale": 12.752150535583496, "train/clean_loss_weight": 2.5504300594329834, "train/combined_loss": 0.7434603571891785, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17488151658767773, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6195502877235413 }, { "epoch": 1.7488151658767772, "step": 738, "train/alpha": 0.2, "train/clean_ce_loss": 0.020263671875, "train/clean_loss_scale": 31.490219116210938, "train/clean_loss_weight": 6.298043727874756, "train/combined_loss": 0.7657289505004883, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17488151658767773, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6381074786186218 }, { "epoch": 1.7488151658767772, "step": 738, "train/alpha": 0.2, "train/clean_ce_loss": 0.0257568359375, "train/clean_loss_scale": 20.024621963500977, "train/clean_loss_weight": 4.004924297332764, "train/combined_loss": 0.6189250946044922, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17488151658767773, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5157709121704102 }, { "epoch": 1.7488151658767772, "step": 738, "train/alpha": 0.2, "train/clean_ce_loss": 0.0947265625, "train/clean_loss_scale": 14.027542114257812, "train/clean_loss_weight": 2.8055083751678467, "train/combined_loss": 1.5945370197296143, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17488151658767773, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3287808895111084 }, { "epoch": 1.7511848341232228, "step": 739, "train/alpha": 0.2, "train/clean_ce_loss": 0.09326171875, "train/clean_loss_scale": 14.1459321975708, "train/clean_loss_weight": 2.82918643951416, "train/combined_loss": 1.5831286907196045, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17511848341232228, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3192739486694336 }, { "epoch": 1.7511848341232228, "step": 739, "train/alpha": 0.2, "train/clean_ce_loss": 0.05224609375, "train/clean_loss_scale": 14.970671653747559, "train/clean_loss_weight": 2.9941344261169434, "train/combined_loss": 0.9385908842086792, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17511848341232228, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7821590900421143 }, { "epoch": 1.7511848341232228, "step": 739, "train/alpha": 0.2, "train/clean_ce_loss": 0.0625, "train/clean_loss_scale": 18.267398834228516, "train/clean_loss_weight": 3.653479814529419, "train/combined_loss": 1.3700549602508545, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17511848341232228, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1417124271392822 }, { "epoch": 1.7511848341232228, "step": 739, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 11.181400299072266, "train/clean_loss_weight": 2.2362802028656006, "train/combined_loss": 1.2448043823242188, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17511848341232228, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0373369455337524 }, { "epoch": 1.7535545023696684, "grad_norm": 14.12901782989502, "learning_rate": 2e-06, "loss": 3.6689, "step": 740 }, { "epoch": 1.7535545023696684, "step": 740, "train/alpha": 0.2, "train/clean_ce_loss": 0.0078125, "train/clean_loss_scale": 67.9332275390625, "train/clean_loss_weight": 13.58664608001709, "train/combined_loss": 0.6368740200996399, "train/pos_loss/pos0": 0.015690557658672333, "train/pos_loss/pos1": 0.12192220985889435, "train/pos_loss/pos2": 0.4401761293411255, "train/pos_loss/pos3": 1.010733962059021, "train/pos_loss/pos4": 1.4697444438934326, "train/pos_loss/pos5": 1.747571587562561, "train/pos_loss/pos6": 2.245198965072632, "train/pos_loss/weight0": 2.0784661769866943, "train/pos_loss/weight1": 1.6305359601974487, "train/pos_loss/weight2": 1.2244595289230347, "train/pos_loss/weight3": 0.8600327968597412, "train/pos_loss/weight4": 0.5784854888916016, "train/pos_loss/weight5": 0.37372884154319763, "train/pos_loss/weight6": 0.25429144501686096, "train/pos_weight_progress": 0.17535545023696683, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5307283401489258 }, { "epoch": 1.7535545023696684, "step": 740, "train/alpha": 0.2, "train/clean_ce_loss": 0.1044921875, "train/clean_loss_scale": 9.7951078414917, "train/clean_loss_weight": 1.9590215682983398, "train/combined_loss": 1.2282147407531738, "train/pos_loss/pos0": 0.36179086565971375, "train/pos_loss/pos1": 0.6932928562164307, "train/pos_loss/pos2": 1.0917332172393799, "train/pos_loss/pos3": 1.6390957832336426, "train/pos_loss/pos4": 2.0990946292877197, "train/pos_loss/pos5": 2.487190008163452, "train/pos_loss/pos6": 2.748645305633545, "train/pos_loss/weight0": 2.1609389781951904, "train/pos_loss/weight1": 1.6764047145843506, "train/pos_loss/weight2": 1.2503184080123901, "train/pos_loss/weight3": 0.8747031688690186, "train/pos_loss/weight4": 0.5375268459320068, "train/pos_loss/weight5": 0.3067297339439392, "train/pos_loss/weight6": 0.19337792694568634, "train/pos_weight_progress": 0.17535545023696683, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0235122442245483 }, { "epoch": 1.7535545023696684, "step": 740, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 6.269959926605225, "train/clean_loss_weight": 1.253991961479187, "train/combined_loss": 0.4996374249458313, "train/pos_loss/pos0": 0.08441779017448425, "train/pos_loss/pos1": 0.3108890652656555, "train/pos_loss/pos2": 0.545358419418335, "train/pos_loss/pos3": 0.9533835649490356, "train/pos_loss/pos4": 1.3704665899276733, "train/pos_loss/pos5": 1.7858622074127197, "train/pos_loss/pos6": 2.2010281085968018, "train/pos_loss/weight0": 2.724999189376831, "train/pos_loss/weight1": 1.9054572582244873, "train/pos_loss/weight2": 1.1929463148117065, "train/pos_loss/weight3": 0.6645851135253906, "train/pos_loss/weight4": 0.3149067759513855, "train/pos_loss/weight5": 0.13631772994995117, "train/pos_loss/weight6": 0.06078759580850601, "train/pos_weight_progress": 0.17535545023696683, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.41636452078819275 }, { "epoch": 1.7535545023696684, "step": 740, "train/alpha": 0.2, "train/clean_ce_loss": 0.026123046875, "train/clean_loss_scale": 19.522809982299805, "train/clean_loss_weight": 3.904561996459961, "train/combined_loss": 0.6119943261146545, "train/pos_loss/pos0": 0.04706201329827309, "train/pos_loss/pos1": 0.18807344138622284, "train/pos_loss/pos2": 0.44128966331481934, "train/pos_loss/pos3": 0.9853708148002625, "train/pos_loss/pos4": 1.5085241794586182, "train/pos_loss/pos5": 1.9509981870651245, "train/pos_loss/pos6": 2.50581955909729, "train/pos_loss/weight0": 2.2265219688415527, "train/pos_loss/weight1": 1.7295938730239868, "train/pos_loss/weight2": 1.2356994152069092, "train/pos_loss/weight3": 0.8339368104934692, "train/pos_loss/weight4": 0.5089530944824219, "train/pos_loss/weight5": 0.28771883249282837, "train/pos_loss/weight6": 0.17757591605186462, "train/pos_weight_progress": 0.17535545023696683, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5099952816963196 }, { "epoch": 1.7559241706161137, "step": 741, "train/alpha": 0.2, "train/clean_ce_loss": 0.0174560546875, "train/clean_loss_scale": 32.48338317871094, "train/clean_loss_weight": 6.496676921844482, "train/combined_loss": 0.6804381012916565, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17559241706161138, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.567031741142273 }, { "epoch": 1.7559241706161137, "step": 741, "train/alpha": 0.2, "train/clean_ce_loss": 0.057373046875, "train/clean_loss_scale": 12.75906753540039, "train/clean_loss_weight": 2.5518136024475098, "train/combined_loss": 0.8784319162368774, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17559241706161138, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7320265769958496 }, { "epoch": 1.7559241706161137, "step": 741, "train/alpha": 0.2, "train/clean_ce_loss": 0.0615234375, "train/clean_loss_scale": 10.892898559570312, "train/clean_loss_weight": 2.178579807281494, "train/combined_loss": 0.8042023181915283, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17559241706161138, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6701685786247253 }, { "epoch": 1.7559241706161137, "step": 741, "train/alpha": 0.2, "train/clean_ce_loss": 0.0859375, "train/clean_loss_scale": 10.477395057678223, "train/clean_loss_weight": 2.0954790115356445, "train/combined_loss": 1.0804812908172607, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17559241706161138, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9004011154174805 }, { "epoch": 1.758293838862559, "step": 742, "train/alpha": 0.2, "train/clean_ce_loss": 0.07861328125, "train/clean_loss_scale": 10.45759391784668, "train/clean_loss_weight": 2.0915188789367676, "train/combined_loss": 0.9865269064903259, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17582938388625594, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8221057653427124 }, { "epoch": 1.758293838862559, "step": 742, "train/alpha": 0.2, "train/clean_ce_loss": 0.04052734375, "train/clean_loss_scale": 12.46725845336914, "train/clean_loss_weight": 2.4934518337249756, "train/combined_loss": 0.6063178777694702, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17582938388625594, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5052648782730103 }, { "epoch": 1.758293838862559, "step": 742, "train/alpha": 0.2, "train/clean_ce_loss": 0.083984375, "train/clean_loss_scale": 9.7245512008667, "train/clean_loss_weight": 1.9449102878570557, "train/combined_loss": 0.9800524115562439, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17582938388625594, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.816710352897644 }, { "epoch": 1.758293838862559, "step": 742, "train/alpha": 0.2, "train/clean_ce_loss": 0.0024871826171875, "train/clean_loss_scale": 99.9422607421875, "train/clean_loss_weight": 19.988452911376953, "train/combined_loss": 0.29828956723213196, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17582938388625594, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2485746443271637 }, { "epoch": 1.7606635071090047, "step": 743, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 9.615168571472168, "train/clean_loss_weight": 1.9230337142944336, "train/combined_loss": 0.8169137239456177, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.1760663507109005, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.680761456489563 }, { "epoch": 1.7606635071090047, "step": 743, "train/alpha": 0.2, "train/clean_ce_loss": 0.045654296875, "train/clean_loss_scale": 11.30810832977295, "train/clean_loss_weight": 2.2616217136383057, "train/combined_loss": 0.6195164918899536, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.1760663507109005, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5162637233734131 }, { "epoch": 1.7606635071090047, "step": 743, "train/alpha": 0.2, "train/clean_ce_loss": 0.04638671875, "train/clean_loss_scale": 24.575698852539062, "train/clean_loss_weight": 4.915139675140381, "train/combined_loss": 1.3679832220077515, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.1760663507109005, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1399860382080078 }, { "epoch": 1.7606635071090047, "step": 743, "train/alpha": 0.2, "train/clean_ce_loss": 0.095703125, "train/clean_loss_scale": 9.40244197845459, "train/clean_loss_weight": 1.880488395690918, "train/combined_loss": 1.079811692237854, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.1760663507109005, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8998430371284485 }, { "epoch": 1.7630331753554502, "step": 744, "train/alpha": 0.2, "train/clean_ce_loss": 0.169921875, "train/clean_loss_scale": 10.372777938842773, "train/clean_loss_weight": 2.0745556354522705, "train/combined_loss": 2.115074396133423, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.176303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.7625619173049927 }, { "epoch": 1.7630331753554502, "step": 744, "train/alpha": 0.2, "train/clean_ce_loss": 0.09033203125, "train/clean_loss_scale": 9.309122085571289, "train/clean_loss_weight": 1.8618243932724, "train/combined_loss": 1.00909423828125, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.176303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.840911865234375 }, { "epoch": 1.7630331753554502, "step": 744, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 10.746909141540527, "train/clean_loss_weight": 2.1493818759918213, "train/combined_loss": 1.0327107906341553, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.176303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8605923652648926 }, { "epoch": 1.7630331753554502, "step": 744, "train/alpha": 0.2, "train/clean_ce_loss": 0.1630859375, "train/clean_loss_scale": 9.362556457519531, "train/clean_loss_weight": 1.8725112676620483, "train/combined_loss": 1.8322815895080566, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.176303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.526901364326477 }, { "epoch": 1.7654028436018958, "grad_norm": 23.907318115234375, "learning_rate": 2e-06, "loss": 3.6982, "step": 745 }, { "epoch": 1.7654028436018958, "step": 745, "train/alpha": 0.2, "train/clean_ce_loss": 0.0108642578125, "train/clean_loss_scale": 29.619321823120117, "train/clean_loss_weight": 5.923864364624023, "train/combined_loss": 0.3861503303050995, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17654028436018956, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.32179194688796997 }, { "epoch": 1.7654028436018958, "step": 745, "train/alpha": 0.2, "train/clean_ce_loss": 0.00518798828125, "train/clean_loss_scale": 83.54341125488281, "train/clean_loss_weight": 16.708683013916016, "train/combined_loss": 0.5201066732406616, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17654028436018956, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4334222376346588 }, { "epoch": 1.7654028436018958, "step": 745, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 9.100631713867188, "train/clean_loss_weight": 1.8201264142990112, "train/combined_loss": 0.9118406176567078, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17654028436018956, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7598671913146973 }, { "epoch": 1.7654028436018958, "step": 745, "train/alpha": 0.2, "train/clean_ce_loss": 0.0205078125, "train/clean_loss_scale": 22.376834869384766, "train/clean_loss_weight": 4.475367069244385, "train/combined_loss": 0.5506799221038818, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17654028436018956, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45889991521835327 }, { "epoch": 1.7677725118483414, "step": 746, "train/alpha": 0.2, "train/clean_ce_loss": 0.0634765625, "train/clean_loss_scale": 10.407776832580566, "train/clean_loss_weight": 2.0815553665161133, "train/combined_loss": 0.7927798628807068, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17677725118483412, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6606498956680298 }, { "epoch": 1.7677725118483414, "step": 746, "train/alpha": 0.2, "train/clean_ce_loss": 0.1962890625, "train/clean_loss_scale": 7.162996292114258, "train/clean_loss_weight": 1.4325993061065674, "train/combined_loss": 1.6872212886810303, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17677725118483412, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.406017780303955 }, { "epoch": 1.7677725118483414, "step": 746, "train/alpha": 0.2, "train/clean_ce_loss": 0.060791015625, "train/clean_loss_scale": 16.28874397277832, "train/clean_loss_weight": 3.25774884223938, "train/combined_loss": 1.188251256942749, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17677725118483412, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9902093410491943 }, { "epoch": 1.7677725118483414, "step": 746, "train/alpha": 0.2, "train/clean_ce_loss": 0.015380859375, "train/clean_loss_scale": 25.625791549682617, "train/clean_loss_weight": 5.125158309936523, "train/combined_loss": 0.47297605872154236, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17677725118483412, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.39414671063423157 }, { "epoch": 1.7701421800947867, "step": 747, "train/alpha": 0.2, "train/clean_ce_loss": 0.03564453125, "train/clean_loss_scale": 5.639471530914307, "train/clean_loss_weight": 1.1278942823410034, "train/combined_loss": 0.24121958017349243, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17701421800947867, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.20101632177829742 }, { "epoch": 1.7701421800947867, "step": 747, "train/alpha": 0.2, "train/clean_ce_loss": 0.04541015625, "train/clean_loss_scale": 11.407031059265137, "train/clean_loss_weight": 2.2814061641693115, "train/combined_loss": 0.6215940713882446, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17701421800947867, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5179950594902039 }, { "epoch": 1.7701421800947867, "step": 747, "train/alpha": 0.2, "train/clean_ce_loss": 0.109375, "train/clean_loss_scale": 12.200658798217773, "train/clean_loss_weight": 2.440131902694702, "train/combined_loss": 1.6013364791870117, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17701421800947867, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3344470262527466 }, { "epoch": 1.7701421800947867, "step": 747, "train/alpha": 0.2, "train/clean_ce_loss": 0.059326171875, "train/clean_loss_scale": 10.748412132263184, "train/clean_loss_weight": 2.1496825218200684, "train/combined_loss": 0.7651945948600769, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17701421800947867, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6376621723175049 }, { "epoch": 1.772511848341232, "step": 748, "train/alpha": 0.2, "train/clean_ce_loss": 0.07763671875, "train/clean_loss_scale": 9.975476264953613, "train/clean_loss_weight": 1.9950952529907227, "train/combined_loss": 0.9293558597564697, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17725118483412322, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7744632363319397 }, { "epoch": 1.772511848341232, "step": 748, "train/alpha": 0.2, "train/clean_ce_loss": 0.036865234375, "train/clean_loss_scale": 12.747941017150879, "train/clean_loss_weight": 2.549588203430176, "train/combined_loss": 0.5639470219612122, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17725118483412322, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46995583176612854 }, { "epoch": 1.772511848341232, "step": 748, "train/alpha": 0.2, "train/clean_ce_loss": 0.08544921875, "train/clean_loss_scale": 11.758190155029297, "train/clean_loss_weight": 2.351638078689575, "train/combined_loss": 1.2056738138198853, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17725118483412322, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0047281980514526 }, { "epoch": 1.772511848341232, "step": 748, "train/alpha": 0.2, "train/clean_ce_loss": 0.0140380859375, "train/clean_loss_scale": 46.429527282714844, "train/clean_loss_weight": 9.285905838012695, "train/combined_loss": 0.7821379899978638, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17725118483412322, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6517816781997681 }, { "epoch": 1.7748815165876777, "step": 749, "train/alpha": 0.2, "train/clean_ce_loss": 0.06787109375, "train/clean_loss_scale": 10.05976676940918, "train/clean_loss_weight": 2.011953353881836, "train/combined_loss": 0.8193208575248718, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17748815165876777, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.682767391204834 }, { "epoch": 1.7748815165876777, "step": 749, "train/alpha": 0.2, "train/clean_ce_loss": 0.0001506805419921875, "train/clean_loss_scale": 1605.499755859375, "train/clean_loss_weight": 321.0999450683594, "train/combined_loss": 0.2903010845184326, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17748815165876777, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.24191756546497345 }, { "epoch": 1.7748815165876777, "step": 749, "train/alpha": 0.2, "train/clean_ce_loss": 0.1630859375, "train/clean_loss_scale": 8.143596649169922, "train/clean_loss_weight": 1.6287193298339844, "train/combined_loss": 1.5937272310256958, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17748815165876777, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3281060457229614 }, { "epoch": 1.7748815165876777, "step": 749, "train/alpha": 0.2, "train/clean_ce_loss": 0.00421142578125, "train/clean_loss_scale": 66.50393676757812, "train/clean_loss_weight": 13.300787925720215, "train/combined_loss": 0.33609166741371155, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17748815165876777, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.28007638454437256 }, { "epoch": 1.7772511848341233, "grad_norm": 22.168869018554688, "learning_rate": 2e-06, "loss": 3.5744, "step": 750 }, { "epoch": 1.7772511848341233, "step": 750, "train/alpha": 0.2, "train/clean_ce_loss": 0.1171875, "train/clean_loss_scale": 12.003232955932617, "train/clean_loss_weight": 2.400646686553955, "train/combined_loss": 1.6879546642303467, "train/pos_loss/pos0": 0.3101155161857605, "train/pos_loss/pos1": 1.00975501537323, "train/pos_loss/pos2": 1.7280128002166748, "train/pos_loss/pos3": 2.334993600845337, "train/pos_loss/pos4": 2.990821599960327, "train/pos_loss/pos5": 3.648620367050171, "train/pos_loss/pos6": 3.631781816482544, "train/pos_loss/weight0": 2.198580026626587, "train/pos_loss/weight1": 1.7046440839767456, "train/pos_loss/weight2": 1.237235188484192, "train/pos_loss/weight3": 0.8563472032546997, "train/pos_loss/weight4": 0.5258328914642334, "train/pos_loss/weight5": 0.29405125975608826, "train/pos_loss/weight6": 0.18330931663513184, "train/pos_weight_progress": 0.17772511848341233, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4066288471221924 }, { "epoch": 1.7772511848341233, "step": 750, "train/alpha": 0.2, "train/clean_ce_loss": 0.07373046875, "train/clean_loss_scale": 8.062726974487305, "train/clean_loss_weight": 1.612545371055603, "train/combined_loss": 0.7133623957633972, "train/pos_loss/pos0": 0.282347708940506, "train/pos_loss/pos1": 0.6269937753677368, "train/pos_loss/pos2": 0.962624728679657, "train/pos_loss/pos3": 1.2868198156356812, "train/pos_loss/pos4": 1.743717908859253, "train/pos_loss/pos5": 2.0561435222625732, "train/pos_loss/pos6": 2.343994617462158, "train/pos_loss/weight0": 3.175534248352051, "train/pos_loss/weight1": 2.0843560695648193, "train/pos_loss/weight2": 1.1273159980773926, "train/pos_loss/weight3": 0.45116379857063293, "train/pos_loss/weight4": 0.12696605920791626, "train/pos_loss/weight5": 0.027992667630314827, "train/pos_loss/weight6": 0.006670780945569277, "train/pos_weight_progress": 0.17772511848341233, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5944686532020569 }, { "epoch": 1.7772511848341233, "step": 750, "train/alpha": 0.2, "train/clean_ce_loss": 0.078125, "train/clean_loss_scale": 10.373784065246582, "train/clean_loss_weight": 2.0747568607330322, "train/combined_loss": 0.9725422263145447, "train/pos_loss/pos0": 0.26845964789390564, "train/pos_loss/pos1": 0.6524335145950317, "train/pos_loss/pos2": 1.1154122352600098, "train/pos_loss/pos3": 1.5091034173965454, "train/pos_loss/pos4": 1.8675029277801514, "train/pos_loss/pos5": 2.374330997467041, "train/pos_loss/pos6": 2.7141597270965576, "train/pos_loss/weight0": 2.570967197418213, "train/pos_loss/weight1": 1.8754470348358154, "train/pos_loss/weight2": 1.2199265956878662, "train/pos_loss/weight3": 0.7027919292449951, "train/pos_loss/weight4": 0.3697932958602905, "train/pos_loss/weight5": 0.1736530065536499, "train/pos_loss/weight6": 0.08742110431194305, "train/pos_weight_progress": 0.17772511848341233, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.810451865196228 }, { "epoch": 1.7772511848341233, "step": 750, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 9.151741981506348, "train/clean_loss_weight": 1.8303483724594116, "train/combined_loss": 0.7292794585227966, "train/pos_loss/pos0": 0.2112995684146881, "train/pos_loss/pos1": 0.4605580270290375, "train/pos_loss/pos2": 0.8430320024490356, "train/pos_loss/pos3": 1.2825065851211548, "train/pos_loss/pos4": 1.7146031856536865, "train/pos_loss/pos5": 2.056279182434082, "train/pos_loss/pos6": 2.446394443511963, "train/pos_loss/weight0": 2.707554817199707, "train/pos_loss/weight1": 1.9358699321746826, "train/pos_loss/weight2": 1.2003628015518188, "train/pos_loss/weight3": 0.6643717288970947, "train/pos_loss/weight4": 0.30843833088874817, "train/pos_loss/weight5": 0.12740550935268402, "train/pos_loss/weight6": 0.055997226387262344, "train/pos_weight_progress": 0.17772511848341233, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.607732892036438 }, { "epoch": 1.7796208530805688, "step": 751, "train/alpha": 0.2, "train/clean_ce_loss": 0.150390625, "train/clean_loss_scale": 9.370332717895508, "train/clean_loss_weight": 1.8740665912628174, "train/combined_loss": 1.6910521984100342, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17796208530805688, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.409210205078125 }, { "epoch": 1.7796208530805688, "step": 751, "train/alpha": 0.2, "train/clean_ce_loss": 0.050537109375, "train/clean_loss_scale": 10.590456008911133, "train/clean_loss_weight": 2.118091344833374, "train/combined_loss": 0.6422532200813293, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17796208530805688, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5352110266685486 }, { "epoch": 1.7796208530805688, "step": 751, "train/alpha": 0.2, "train/clean_ce_loss": 0.0576171875, "train/clean_loss_scale": 11.400721549987793, "train/clean_loss_weight": 2.280144453048706, "train/combined_loss": 0.7882530093193054, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17796208530805688, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6568775177001953 }, { "epoch": 1.7796208530805688, "step": 751, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 11.409363746643066, "train/clean_loss_weight": 2.2818727493286133, "train/combined_loss": 1.0829981565475464, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17796208530805688, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9024984836578369 }, { "epoch": 1.7819905213270142, "step": 752, "train/alpha": 0.2, "train/clean_ce_loss": 0.004608154296875, "train/clean_loss_scale": 51.36664581298828, "train/clean_loss_weight": 10.273329734802246, "train/combined_loss": 0.2840465009212494, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17819905213270143, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.23670542240142822 }, { "epoch": 1.7819905213270142, "step": 752, "train/alpha": 0.2, "train/clean_ce_loss": 0.0240478515625, "train/clean_loss_scale": 17.159141540527344, "train/clean_loss_weight": 3.431828260421753, "train/combined_loss": 0.4951685667037964, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17819905213270143, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4126404821872711 }, { "epoch": 1.7819905213270142, "step": 752, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 14.0805082321167, "train/clean_loss_weight": 2.8161017894744873, "train/combined_loss": 1.3530488014221191, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17819905213270143, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1275407075881958 }, { "epoch": 1.7819905213270142, "step": 752, "train/alpha": 0.2, "train/clean_ce_loss": 0.07177734375, "train/clean_loss_scale": 8.13305377960205, "train/clean_loss_weight": 1.6266107559204102, "train/combined_loss": 0.7005228400230408, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17819905213270143, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5837690234184265 }, { "epoch": 1.7843601895734598, "step": 753, "train/alpha": 0.2, "train/clean_ce_loss": 0.1357421875, "train/clean_loss_scale": 8.970057487487793, "train/clean_loss_weight": 1.7940114736557007, "train/combined_loss": 1.4611382484436035, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17843601895734598, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2176152467727661 }, { "epoch": 1.7843601895734598, "step": 753, "train/alpha": 0.2, "train/clean_ce_loss": 0.0673828125, "train/clean_loss_scale": 10.219160079956055, "train/clean_loss_weight": 2.0438320636749268, "train/combined_loss": 0.8263149261474609, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17843601895734598, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6885957717895508 }, { "epoch": 1.7843601895734598, "step": 753, "train/alpha": 0.2, "train/clean_ce_loss": 0.09619140625, "train/clean_loss_scale": 10.071765899658203, "train/clean_loss_weight": 2.0143532752990723, "train/combined_loss": 1.1625807285308838, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17843601895734598, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9688172936439514 }, { "epoch": 1.7843601895734598, "step": 753, "train/alpha": 0.2, "train/clean_ce_loss": 0.169921875, "train/clean_loss_scale": 6.985513687133789, "train/clean_loss_weight": 1.3971027135849, "train/combined_loss": 1.4243898391723633, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17843601895734598, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.186991572380066 }, { "epoch": 1.7867298578199051, "step": 754, "train/alpha": 0.2, "train/clean_ce_loss": 0.1435546875, "train/clean_loss_scale": 8.31447696685791, "train/clean_loss_weight": 1.6628954410552979, "train/combined_loss": 1.4322986602783203, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17867298578199053, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1935821771621704 }, { "epoch": 1.7867298578199051, "step": 754, "train/alpha": 0.2, "train/clean_ce_loss": 0.0947265625, "train/clean_loss_scale": 7.703364849090576, "train/clean_loss_weight": 1.540673017501831, "train/combined_loss": 0.8756558895111084, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17867298578199053, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7297132611274719 }, { "epoch": 1.7867298578199051, "step": 754, "train/alpha": 0.2, "train/clean_ce_loss": 0.01025390625, "train/clean_loss_scale": 18.228601455688477, "train/clean_loss_weight": 3.6457202434539795, "train/combined_loss": 0.22429724037647247, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17867298578199053, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1869143694639206 }, { "epoch": 1.7867298578199051, "step": 754, "train/alpha": 0.2, "train/clean_ce_loss": 0.01287841796875, "train/clean_loss_scale": 29.320438385009766, "train/clean_loss_weight": 5.8640875816345215, "train/combined_loss": 0.45312100648880005, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17867298578199053, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.37760084867477417 }, { "epoch": 1.7890995260663507, "grad_norm": 22.516218185424805, "learning_rate": 2e-06, "loss": 3.465, "step": 755 }, { "epoch": 1.7890995260663507, "step": 755, "train/alpha": 0.2, "train/clean_ce_loss": 0.1005859375, "train/clean_loss_scale": 10.109026908874512, "train/clean_loss_weight": 2.02180552482605, "train/combined_loss": 1.2201911211013794, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17890995260663506, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0168259143829346 }, { "epoch": 1.7890995260663507, "step": 755, "train/alpha": 0.2, "train/clean_ce_loss": 0.14453125, "train/clean_loss_scale": 8.660694122314453, "train/clean_loss_weight": 1.7321388721466064, "train/combined_loss": 1.5020891427993774, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17890995260663506, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2517409324645996 }, { "epoch": 1.7890995260663507, "step": 755, "train/alpha": 0.2, "train/clean_ce_loss": 0.1103515625, "train/clean_loss_scale": 8.359908103942871, "train/clean_loss_weight": 1.671981692314148, "train/combined_loss": 1.107034683227539, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17890995260663506, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9225289225578308 }, { "epoch": 1.7890995260663507, "step": 755, "train/alpha": 0.2, "train/clean_ce_loss": 0.00011348724365234375, "train/clean_loss_scale": 1167.022216796875, "train/clean_loss_weight": 233.40444946289062, "train/combined_loss": 0.15893055498600006, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17890995260663506, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.13244213163852692 }, { "epoch": 1.7914691943127963, "step": 756, "train/alpha": 0.2, "train/clean_ce_loss": 0.058837890625, "train/clean_loss_scale": 18.522180557250977, "train/clean_loss_weight": 3.7044360637664795, "train/combined_loss": 1.3077672719955444, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.1791469194312796, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.089806079864502 }, { "epoch": 1.7914691943127963, "step": 756, "train/alpha": 0.2, "train/clean_ce_loss": 0.08544921875, "train/clean_loss_scale": 6.949416160583496, "train/clean_loss_weight": 1.389883279800415, "train/combined_loss": 0.7125866413116455, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.1791469194312796, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.593822181224823 }, { "epoch": 1.7914691943127963, "step": 756, "train/alpha": 0.2, "train/clean_ce_loss": 0.0947265625, "train/clean_loss_scale": 9.062198638916016, "train/clean_loss_weight": 1.8124397993087769, "train/combined_loss": 1.030117154121399, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.1791469194312796, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8584309220314026 }, { "epoch": 1.7914691943127963, "step": 756, "train/alpha": 0.2, "train/clean_ce_loss": 0.03759765625, "train/clean_loss_scale": 13.969005584716797, "train/clean_loss_weight": 2.7938010692596436, "train/combined_loss": 0.6302422285079956, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.1791469194312796, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5252018570899963 }, { "epoch": 1.7938388625592419, "step": 757, "train/alpha": 0.2, "train/clean_ce_loss": 0.00982666015625, "train/clean_loss_scale": 25.938621520996094, "train/clean_loss_weight": 5.187724590301514, "train/combined_loss": 0.3058680295944214, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17938388625592416, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2548900246620178 }, { "epoch": 1.7938388625592419, "step": 757, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 10.152107238769531, "train/clean_loss_weight": 2.030421495437622, "train/combined_loss": 0.963657021522522, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17938388625592416, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8030475378036499 }, { "epoch": 1.7938388625592419, "step": 757, "train/alpha": 0.2, "train/clean_ce_loss": 0.039306640625, "train/clean_loss_scale": 6.915338516235352, "train/clean_loss_weight": 1.3830677270889282, "train/combined_loss": 0.32618248462677, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17938388625592416, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.27181872725486755 }, { "epoch": 1.7938388625592419, "step": 757, "train/alpha": 0.2, "train/clean_ce_loss": 0.087890625, "train/clean_loss_scale": 8.159934043884277, "train/clean_loss_weight": 1.6319868564605713, "train/combined_loss": 0.860617995262146, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17938388625592416, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7171816825866699 }, { "epoch": 1.7962085308056872, "step": 758, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 13.831084251403809, "train/clean_loss_weight": 2.766216993331909, "train/combined_loss": 1.5397887229919434, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17962085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.283157229423523 }, { "epoch": 1.7962085308056872, "step": 758, "train/alpha": 0.2, "train/clean_ce_loss": 0.044921875, "train/clean_loss_scale": 21.418487548828125, "train/clean_loss_weight": 4.283697605133057, "train/combined_loss": 1.154590368270874, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17962085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9621586203575134 }, { "epoch": 1.7962085308056872, "step": 758, "train/alpha": 0.2, "train/clean_ce_loss": 0.051025390625, "train/clean_loss_scale": 10.692276954650879, "train/clean_loss_weight": 2.138455390930176, "train/combined_loss": 0.6546931266784668, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17962085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5455775856971741 }, { "epoch": 1.7962085308056872, "step": 758, "train/alpha": 0.2, "train/clean_ce_loss": 0.09619140625, "train/clean_loss_scale": 10.161633491516113, "train/clean_loss_weight": 2.0323266983032227, "train/combined_loss": 1.1729542016983032, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17962085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9774618148803711 }, { "epoch": 1.7985781990521326, "step": 759, "train/alpha": 0.2, "train/clean_ce_loss": 0.0673828125, "train/clean_loss_scale": 8.248051643371582, "train/clean_loss_weight": 1.6496104001998901, "train/combined_loss": 0.6669322848320007, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17985781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5557768940925598 }, { "epoch": 1.7985781990521326, "step": 759, "train/alpha": 0.2, "train/clean_ce_loss": 0.042724609375, "train/clean_loss_scale": 16.323591232299805, "train/clean_loss_weight": 3.2647182941436768, "train/combined_loss": 0.8369028568267822, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17985781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6974190473556519 }, { "epoch": 1.7985781990521326, "step": 759, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 5.902859687805176, "train/clean_loss_weight": 1.1805719137191772, "train/combined_loss": 0.5153473019599915, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17985781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.42945608496665955 }, { "epoch": 1.7985781990521326, "step": 759, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 8.030159950256348, "train/clean_loss_weight": 1.6060320138931274, "train/combined_loss": 0.6399034261703491, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.17985781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5332528352737427 }, { "epoch": 1.8009478672985781, "grad_norm": 26.317840576171875, "learning_rate": 2e-06, "loss": 3.6159, "step": 760 }, { "epoch": 1.8009478672985781, "step": 760, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 15.301535606384277, "train/clean_loss_weight": 3.060307264328003, "train/combined_loss": 1.3000328540802002, "train/pos_loss/pos0": 0.26289889216423035, "train/pos_loss/pos1": 0.9120593070983887, "train/pos_loss/pos2": 1.4711018800735474, "train/pos_loss/pos3": 2.086806297302246, "train/pos_loss/pos4": 2.6406915187835693, "train/pos_loss/pos5": 3.0460333824157715, "train/pos_loss/pos6": 3.4023032188415527, "train/pos_loss/weight0": 2.521200656890869, "train/pos_loss/weight1": 1.8450655937194824, "train/pos_loss/weight2": 1.2398234605789185, "train/pos_loss/weight3": 0.7349952459335327, "train/pos_loss/weight4": 0.38598424196243286, "train/pos_loss/weight5": 0.18007011711597443, "train/pos_loss/weight6": 0.09286075085401535, "train/pos_weight_progress": 0.18009478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0833606719970703 }, { "epoch": 1.8009478672985781, "step": 760, "train/alpha": 0.2, "train/clean_ce_loss": 0.125, "train/clean_loss_scale": 6.903542995452881, "train/clean_loss_weight": 1.3807085752487183, "train/combined_loss": 1.0355314016342163, "train/pos_loss/pos0": 0.37442007660865784, "train/pos_loss/pos1": 0.8729755282402039, "train/pos_loss/pos2": 1.4117588996887207, "train/pos_loss/pos3": 1.9281702041625977, "train/pos_loss/pos4": 2.387935161590576, "train/pos_loss/pos5": 2.9723095893859863, "train/pos_loss/pos6": 3.2342491149902344, "train/pos_loss/weight0": 3.0439250469207764, "train/pos_loss/weight1": 2.09834885597229, "train/pos_loss/weight2": 1.1902940273284912, "train/pos_loss/weight3": 0.48064345121383667, "train/pos_loss/weight4": 0.1445770561695099, "train/pos_loss/weight5": 0.034708719700574875, "train/pos_loss/weight6": 0.007502640597522259, "train/pos_weight_progress": 0.18009478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8629428744316101 }, { "epoch": 1.8009478672985781, "step": 760, "train/alpha": 0.2, "train/clean_ce_loss": 0.019287109375, "train/clean_loss_scale": 11.025907516479492, "train/clean_loss_weight": 2.20518159866333, "train/combined_loss": 0.2551894783973694, "train/pos_loss/pos0": 0.04691670462489128, "train/pos_loss/pos1": 0.07718682289123535, "train/pos_loss/pos2": 0.35035717487335205, "train/pos_loss/pos3": 0.7855766415596008, "train/pos_loss/pos4": 1.3623281717300415, "train/pos_loss/pos5": 1.7653281688690186, "train/pos_loss/pos6": 2.3481204509735107, "train/pos_loss/weight0": 3.049025058746338, "train/pos_loss/weight1": 2.025914430618286, "train/pos_loss/weight2": 1.1538043022155762, "train/pos_loss/weight3": 0.5207110643386841, "train/pos_loss/weight4": 0.18329384922981262, "train/pos_loss/weight5": 0.05259332060813904, "train/pos_loss/weight6": 0.014658357948064804, "train/pos_weight_progress": 0.18009478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2126578837633133 }, { "epoch": 1.8009478672985781, "step": 760, "train/alpha": 0.2, "train/clean_ce_loss": 0.048095703125, "train/clean_loss_scale": 26.105491638183594, "train/clean_loss_weight": 5.22109842300415, "train/combined_loss": 1.5066742897033691, "train/pos_loss/pos0": 0.1650363653898239, "train/pos_loss/pos1": 0.7629648447036743, "train/pos_loss/pos2": 1.2201206684112549, "train/pos_loss/pos3": 2.0845236778259277, "train/pos_loss/pos4": 3.1785149574279785, "train/pos_loss/pos5": 3.4978489875793457, "train/pos_loss/pos6": 3.959848403930664, "train/pos_loss/weight0": 2.135488986968994, "train/pos_loss/weight1": 1.627476453781128, "train/pos_loss/weight2": 1.2310538291931152, "train/pos_loss/weight3": 0.8822826147079468, "train/pos_loss/weight4": 0.5636661052703857, "train/pos_loss/weight5": 0.33804622292518616, "train/pos_loss/weight6": 0.22198550403118134, "train/pos_weight_progress": 0.18009478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2555619478225708 }, { "epoch": 1.8033175355450237, "step": 761, "train/alpha": 0.2, "train/clean_ce_loss": 0.051513671875, "train/clean_loss_scale": 8.872572898864746, "train/clean_loss_weight": 1.7745145559310913, "train/combined_loss": 0.5484705567359924, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18033175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45705878734588623 }, { "epoch": 1.8033175355450237, "step": 761, "train/alpha": 0.2, "train/clean_ce_loss": 0.054931640625, "train/clean_loss_scale": 5.60024881362915, "train/clean_loss_weight": 1.1200498342514038, "train/combined_loss": 0.3691570460796356, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18033175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.30763086676597595 }, { "epoch": 1.8033175355450237, "step": 761, "train/alpha": 0.2, "train/clean_ce_loss": 0.01611328125, "train/clean_loss_scale": 16.271385192871094, "train/clean_loss_weight": 3.254276990890503, "train/combined_loss": 0.3146224617958069, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18033175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.26218539476394653 }, { "epoch": 1.8033175355450237, "step": 761, "train/alpha": 0.2, "train/clean_ce_loss": 0.056396484375, "train/clean_loss_scale": 9.34250259399414, "train/clean_loss_weight": 1.8685005903244019, "train/combined_loss": 0.6322611570358276, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18033175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5268843173980713 }, { "epoch": 1.8056872037914693, "step": 762, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 6.9332275390625, "train/clean_loss_weight": 1.3866455554962158, "train/combined_loss": 0.6053032875061035, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18056872037914692, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5044193863868713 }, { "epoch": 1.8056872037914693, "step": 762, "train/alpha": 0.2, "train/clean_ce_loss": 0.138671875, "train/clean_loss_scale": 8.0997896194458, "train/clean_loss_weight": 1.6199579238891602, "train/combined_loss": 1.3478556871414185, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18056872037914692, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1232130527496338 }, { "epoch": 1.8056872037914693, "step": 762, "train/alpha": 0.2, "train/clean_ce_loss": 0.1123046875, "train/clean_loss_scale": 9.740076065063477, "train/clean_loss_weight": 1.9480152130126953, "train/combined_loss": 1.3126274347305298, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18056872037914692, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0938562154769897 }, { "epoch": 1.8056872037914693, "step": 762, "train/alpha": 0.2, "train/clean_ce_loss": 0.103515625, "train/clean_loss_scale": 9.668328285217285, "train/clean_loss_weight": 1.9336656332015991, "train/combined_loss": 1.2009875774383545, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18056872037914692, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0008230209350586 }, { "epoch": 1.8080568720379147, "step": 763, "train/alpha": 0.2, "train/clean_ce_loss": 0.08447265625, "train/clean_loss_scale": 8.033882141113281, "train/clean_loss_weight": 1.606776475906372, "train/combined_loss": 0.8143720030784607, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18080568720379148, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6786433458328247 }, { "epoch": 1.8080568720379147, "step": 763, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 10.658836364746094, "train/clean_loss_weight": 2.1317672729492188, "train/combined_loss": 0.8493759632110596, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18080568720379148, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7078133225440979 }, { "epoch": 1.8080568720379147, "step": 763, "train/alpha": 0.2, "train/clean_ce_loss": 0.142578125, "train/clean_loss_scale": 8.049959182739258, "train/clean_loss_weight": 1.6099919080734253, "train/combined_loss": 1.3772977590560913, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18080568720379148, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1477481126785278 }, { "epoch": 1.8080568720379147, "step": 763, "train/alpha": 0.2, "train/clean_ce_loss": 0.01251220703125, "train/clean_loss_scale": 20.430824279785156, "train/clean_loss_weight": 4.086164951324463, "train/combined_loss": 0.30676162242889404, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18080568720379148, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.25563469529151917 }, { "epoch": 1.8104265402843602, "step": 764, "train/alpha": 0.2, "train/clean_ce_loss": 0.1240234375, "train/clean_loss_scale": 9.585023880004883, "train/clean_loss_weight": 1.9170048236846924, "train/combined_loss": 1.4265210628509521, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18104265402843603, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1887675523757935 }, { "epoch": 1.8104265402843602, "step": 764, "train/alpha": 0.2, "train/clean_ce_loss": 0.023681640625, "train/clean_loss_scale": 12.611726760864258, "train/clean_loss_weight": 2.5223453044891357, "train/combined_loss": 0.3583996593952179, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18104265402843603, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.298666387796402 }, { "epoch": 1.8104265402843602, "step": 764, "train/alpha": 0.2, "train/clean_ce_loss": 0.09130859375, "train/clean_loss_scale": 10.639900207519531, "train/clean_loss_weight": 2.1279799938201904, "train/combined_loss": 1.165817141532898, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18104265402843603, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9715142846107483 }, { "epoch": 1.8104265402843602, "step": 764, "train/alpha": 0.2, "train/clean_ce_loss": 0.051025390625, "train/clean_loss_scale": 18.456811904907227, "train/clean_loss_weight": 3.6913623809814453, "train/combined_loss": 1.1301192045211792, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18104265402843603, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9417660236358643 }, { "epoch": 1.8127962085308056, "grad_norm": 24.242279052734375, "learning_rate": 2e-06, "loss": 3.5573, "step": 765 }, { "epoch": 1.8127962085308056, "step": 765, "train/alpha": 0.2, "train/clean_ce_loss": 0.11865234375, "train/clean_loss_scale": 6.525968551635742, "train/clean_loss_weight": 1.3051937818527222, "train/combined_loss": 0.9291857481002808, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18127962085308058, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.774321436882019 }, { "epoch": 1.8127962085308056, "step": 765, "train/alpha": 0.2, "train/clean_ce_loss": 0.056396484375, "train/clean_loss_scale": 8.713912010192871, "train/clean_loss_weight": 1.742782473564148, "train/combined_loss": 0.5897207856178284, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18127962085308058, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4914339780807495 }, { "epoch": 1.8127962085308056, "step": 765, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 7.878095626831055, "train/clean_loss_weight": 1.575619101524353, "train/combined_loss": 0.8770536184310913, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18127962085308058, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7308779954910278 }, { "epoch": 1.8127962085308056, "step": 765, "train/alpha": 0.2, "train/clean_ce_loss": 0.01519775390625, "train/clean_loss_scale": 13.345939636230469, "train/clean_loss_weight": 2.6691880226135254, "train/combined_loss": 0.24339395761489868, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18127962085308058, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2028283029794693 }, { "epoch": 1.8151658767772512, "step": 766, "train/alpha": 0.2, "train/clean_ce_loss": 0.0947265625, "train/clean_loss_scale": 7.293837547302246, "train/clean_loss_weight": 1.4587675333023071, "train/combined_loss": 0.8291041851043701, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.1815165876777251, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6909201741218567 }, { "epoch": 1.8151658767772512, "step": 766, "train/alpha": 0.2, "train/clean_ce_loss": 0.03564453125, "train/clean_loss_scale": 10.670645713806152, "train/clean_loss_weight": 2.134129285812378, "train/combined_loss": 0.4564202129840851, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.1815165876777251, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.38035017251968384 }, { "epoch": 1.8151658767772512, "step": 766, "train/alpha": 0.2, "train/clean_ce_loss": 0.08935546875, "train/clean_loss_scale": 8.649650573730469, "train/clean_loss_weight": 1.7299301624298096, "train/combined_loss": 0.9274722337722778, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.1815165876777251, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7728935480117798 }, { "epoch": 1.8151658767772512, "step": 766, "train/alpha": 0.2, "train/clean_ce_loss": 0.1416015625, "train/clean_loss_scale": 8.615351676940918, "train/clean_loss_weight": 1.7230703830718994, "train/combined_loss": 1.463936686515808, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.1815165876777251, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2199472188949585 }, { "epoch": 1.8175355450236967, "step": 767, "train/alpha": 0.2, "train/clean_ce_loss": 0.0123291015625, "train/clean_loss_scale": 10.640700340270996, "train/clean_loss_weight": 2.1281402111053467, "train/combined_loss": 0.1574283242225647, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18175355450236966, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.13119027018547058 }, { "epoch": 1.8175355450236967, "step": 767, "train/alpha": 0.2, "train/clean_ce_loss": 0.1455078125, "train/clean_loss_scale": 7.806034564971924, "train/clean_loss_weight": 1.5612069368362427, "train/combined_loss": 1.363006830215454, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18175355450236966, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1358389854431152 }, { "epoch": 1.8175355450236967, "step": 767, "train/alpha": 0.2, "train/clean_ce_loss": 0.068359375, "train/clean_loss_scale": 12.59619140625, "train/clean_loss_weight": 2.519238233566284, "train/combined_loss": 1.0332813262939453, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18175355450236966, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8610677719116211 }, { "epoch": 1.8175355450236967, "step": 767, "train/alpha": 0.2, "train/clean_ce_loss": 0.1083984375, "train/clean_loss_scale": 7.094463348388672, "train/clean_loss_weight": 1.418892741203308, "train/combined_loss": 0.9228345155715942, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18175355450236966, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7690287232398987 }, { "epoch": 1.8199052132701423, "step": 768, "train/alpha": 0.2, "train/clean_ce_loss": 0.072265625, "train/clean_loss_scale": 7.1743059158325195, "train/clean_loss_weight": 1.434861183166504, "train/combined_loss": 0.6221468448638916, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.1819905213270142, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5184556841850281 }, { "epoch": 1.8199052132701423, "step": 768, "train/alpha": 0.2, "train/clean_ce_loss": 0.0281982421875, "train/clean_loss_scale": 7.190850257873535, "train/clean_loss_weight": 1.438170075416565, "train/combined_loss": 0.2433232069015503, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.1819905213270142, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.20276933908462524 }, { "epoch": 1.8199052132701423, "step": 768, "train/alpha": 0.2, "train/clean_ce_loss": 0.0274658203125, "train/clean_loss_scale": 3.8742425441741943, "train/clean_loss_weight": 0.7748485207557678, "train/combined_loss": 0.1276911050081253, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.1819905213270142, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.10640925168991089 }, { "epoch": 1.8199052132701423, "step": 768, "train/alpha": 0.2, "train/clean_ce_loss": 0.1142578125, "train/clean_loss_scale": 6.133357524871826, "train/clean_loss_weight": 1.226671576499939, "train/combined_loss": 0.8409408330917358, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.1819905213270142, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7007840275764465 }, { "epoch": 1.8222748815165877, "step": 769, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 8.322943687438965, "train/clean_loss_weight": 1.6645888090133667, "train/combined_loss": 0.7997828722000122, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18222748815165876, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6664857268333435 }, { "epoch": 1.8222748815165877, "step": 769, "train/alpha": 0.2, "train/clean_ce_loss": 0.03857421875, "train/clean_loss_scale": 11.041715621948242, "train/clean_loss_weight": 2.208343267440796, "train/combined_loss": 0.5111106634140015, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18222748815165876, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4259255528450012 }, { "epoch": 1.8222748815165877, "step": 769, "train/alpha": 0.2, "train/clean_ce_loss": 0.0693359375, "train/clean_loss_scale": 7.671617031097412, "train/clean_loss_weight": 1.5343234539031982, "train/combined_loss": 0.6383025050163269, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18222748815165876, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5319187641143799 }, { "epoch": 1.8222748815165877, "step": 769, "train/alpha": 0.2, "train/clean_ce_loss": 0.0283203125, "train/clean_loss_scale": 6.912214756011963, "train/clean_loss_weight": 1.3824429512023926, "train/combined_loss": 0.23490729928016663, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18222748815165876, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.19575607776641846 }, { "epoch": 1.824644549763033, "grad_norm": 25.633405685424805, "learning_rate": 2e-06, "loss": 3.5097, "step": 770 }, { "epoch": 1.824644549763033, "step": 770, "train/alpha": 0.2, "train/clean_ce_loss": 0.03369140625, "train/clean_loss_scale": 9.21568775177002, "train/clean_loss_weight": 1.8431376218795776, "train/combined_loss": 0.3725873827934265, "train/pos_loss/pos0": 0.06597349047660828, "train/pos_loss/pos1": 0.18479950726032257, "train/pos_loss/pos2": 0.4553636312484741, "train/pos_loss/pos3": 0.9184518456459045, "train/pos_loss/pos4": 1.3911255598068237, "train/pos_loss/pos5": 1.9868345260620117, "train/pos_loss/pos6": 2.4621071815490723, "train/pos_loss/weight0": 2.8532886505126953, "train/pos_loss/weight1": 2.0083935260772705, "train/pos_loss/weight2": 1.1979069709777832, "train/pos_loss/weight3": 0.6247327327728271, "train/pos_loss/weight4": 0.23475652933120728, "train/pos_loss/weight5": 0.06198116019368172, "train/pos_loss/weight6": 0.018940212205052376, "train/pos_weight_progress": 0.18246445497630331, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3104894757270813 }, { "epoch": 1.824644549763033, "step": 770, "train/alpha": 0.2, "train/clean_ce_loss": 0.1474609375, "train/clean_loss_scale": 7.9066267013549805, "train/clean_loss_weight": 1.5813254117965698, "train/combined_loss": 1.3991023302078247, "train/pos_loss/pos0": 0.48793473839759827, "train/pos_loss/pos1": 0.858620285987854, "train/pos_loss/pos2": 1.2232285737991333, "train/pos_loss/pos3": 2.0300374031066895, "train/pos_loss/pos4": 2.2400739192962646, "train/pos_loss/pos5": 2.777944803237915, "train/pos_loss/pos6": 3.240138292312622, "train/pos_loss/weight0": 2.2671751976013184, "train/pos_loss/weight1": 1.725689172744751, "train/pos_loss/weight2": 1.245736002922058, "train/pos_loss/weight3": 0.8433372378349304, "train/pos_loss/weight4": 0.49813395738601685, "train/pos_loss/weight5": 0.2623063623905182, "train/pos_loss/weight6": 0.15762189030647278, "train/pos_weight_progress": 0.18246445497630331, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1659185886383057 }, { "epoch": 1.824644549763033, "step": 770, "train/alpha": 0.2, "train/clean_ce_loss": 0.08447265625, "train/clean_loss_scale": 8.140384674072266, "train/clean_loss_weight": 1.6280769109725952, "train/combined_loss": 0.8251679539680481, "train/pos_loss/pos0": 0.29229438304901123, "train/pos_loss/pos1": 0.6480474472045898, "train/pos_loss/pos2": 1.0719066858291626, "train/pos_loss/pos3": 1.4705575704574585, "train/pos_loss/pos4": 2.0048904418945312, "train/pos_loss/pos5": 2.3461709022521973, "train/pos_loss/pos6": 2.622206211090088, "train/pos_loss/weight0": 3.0273678302764893, "train/pos_loss/weight1": 2.0113332271575928, "train/pos_loss/weight2": 1.1141177415847778, "train/pos_loss/weight3": 0.5630979537963867, "train/pos_loss/weight4": 0.19916892051696777, "train/pos_loss/weight5": 0.06476060301065445, "train/pos_loss/weight6": 0.02015366218984127, "train/pos_weight_progress": 0.18246445497630331, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6876399517059326 }, { "epoch": 1.824644549763033, "step": 770, "train/alpha": 0.2, "train/clean_ce_loss": 0.09765625, "train/clean_loss_scale": 8.940174102783203, "train/clean_loss_weight": 1.7880347967147827, "train/combined_loss": 1.047676682472229, "train/pos_loss/pos0": 0.3406086266040802, "train/pos_loss/pos1": 0.6779751181602478, "train/pos_loss/pos2": 1.1971815824508667, "train/pos_loss/pos3": 1.7292674779891968, "train/pos_loss/pos4": 2.2347350120544434, "train/pos_loss/pos5": 2.7434206008911133, "train/pos_loss/pos6": 2.999180555343628, "train/pos_loss/weight0": 2.6752402782440186, "train/pos_loss/weight1": 1.9098806381225586, "train/pos_loss/weight2": 1.2208669185638428, "train/pos_loss/weight3": 0.6686380505561829, "train/pos_loss/weight4": 0.3270794749259949, "train/pos_loss/weight5": 0.13861367106437683, "train/pos_loss/weight6": 0.05968110263347626, "train/pos_weight_progress": 0.18246445497630331, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8730639219284058 }, { "epoch": 1.8270142180094786, "step": 771, "train/alpha": 0.2, "train/clean_ce_loss": 0.08837890625, "train/clean_loss_scale": 9.450639724731445, "train/clean_loss_weight": 1.8901280164718628, "train/combined_loss": 1.0022846460342407, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18270142180094787, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8352372050285339 }, { "epoch": 1.8270142180094786, "step": 771, "train/alpha": 0.2, "train/clean_ce_loss": 0.01092529296875, "train/clean_loss_scale": 21.32949447631836, "train/clean_loss_weight": 4.265899181365967, "train/combined_loss": 0.2796371579170227, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18270142180094787, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.23303097486495972 }, { "epoch": 1.8270142180094786, "step": 771, "train/alpha": 0.2, "train/clean_ce_loss": 0.06005859375, "train/clean_loss_scale": 9.78384017944336, "train/clean_loss_weight": 1.9567680358886719, "train/combined_loss": 0.7051244378089905, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18270142180094787, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5876036882400513 }, { "epoch": 1.8270142180094786, "step": 771, "train/alpha": 0.2, "train/clean_ce_loss": 0.0693359375, "train/clean_loss_scale": 6.9714274406433105, "train/clean_loss_weight": 1.3942855596542358, "train/combined_loss": 0.5800445675849915, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18270142180094787, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.48337045311927795 }, { "epoch": 1.8293838862559242, "step": 772, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 8.745950698852539, "train/clean_loss_weight": 1.7491902112960815, "train/combined_loss": 0.8558049201965332, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18293838862559242, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7131707668304443 }, { "epoch": 1.8293838862559242, "step": 772, "train/alpha": 0.2, "train/clean_ce_loss": 0.083984375, "train/clean_loss_scale": 7.0395188331604, "train/clean_loss_weight": 1.407903790473938, "train/combined_loss": 0.7094514966011047, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18293838862559242, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5912095904350281 }, { "epoch": 1.8293838862559242, "step": 772, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 5.910393714904785, "train/clean_loss_weight": 1.1820787191390991, "train/combined_loss": 0.47098448872566223, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18293838862559242, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3924870789051056 }, { "epoch": 1.8293838862559242, "step": 772, "train/alpha": 0.2, "train/clean_ce_loss": 0.060546875, "train/clean_loss_scale": 8.305237770080566, "train/clean_loss_weight": 1.6610475778579712, "train/combined_loss": 0.6034274101257324, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18293838862559242, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5028561949729919 }, { "epoch": 1.8317535545023698, "step": 773, "train/alpha": 0.2, "train/clean_ce_loss": 0.1279296875, "train/clean_loss_scale": 6.770905494689941, "train/clean_loss_weight": 1.354181170463562, "train/combined_loss": 1.0394397974014282, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18317535545023697, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8661998510360718 }, { "epoch": 1.8317535545023698, "step": 773, "train/alpha": 0.2, "train/clean_ce_loss": 0.014404296875, "train/clean_loss_scale": 14.430481910705566, "train/clean_loss_weight": 2.886096477508545, "train/combined_loss": 0.24943313002586365, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18317535545023697, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.20786094665527344 }, { "epoch": 1.8317535545023698, "step": 773, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 5.953530311584473, "train/clean_loss_weight": 1.1907061338424683, "train/combined_loss": 0.6244229674339294, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18317535545023697, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5203524827957153 }, { "epoch": 1.8317535545023698, "step": 773, "train/alpha": 0.2, "train/clean_ce_loss": 0.1162109375, "train/clean_loss_scale": 7.0000481605529785, "train/clean_loss_weight": 1.4000096321105957, "train/combined_loss": 0.9761785864830017, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18317535545023697, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8134821653366089 }, { "epoch": 1.8341232227488151, "step": 774, "train/alpha": 0.2, "train/clean_ce_loss": 0.1591796875, "train/clean_loss_scale": 6.1318359375, "train/clean_loss_weight": 1.2263672351837158, "train/combined_loss": 1.1712764501571655, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18341232227488152, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9760637283325195 }, { "epoch": 1.8341232227488151, "step": 774, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 7.24237060546875, "train/clean_loss_weight": 1.4484741687774658, "train/combined_loss": 0.5601521134376526, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18341232227488152, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.466793417930603 }, { "epoch": 1.8341232227488151, "step": 774, "train/alpha": 0.2, "train/clean_ce_loss": 0.005767822265625, "train/clean_loss_scale": 19.812618255615234, "train/clean_loss_weight": 3.9625236988067627, "train/combined_loss": 0.13713079690933228, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18341232227488152, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.11427566409111023 }, { "epoch": 1.8341232227488151, "step": 774, "train/alpha": 0.2, "train/clean_ce_loss": 0.10205078125, "train/clean_loss_scale": 7.794223785400391, "train/clean_loss_weight": 1.558844804763794, "train/combined_loss": 0.9544879794120789, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18341232227488152, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7954066395759583 }, { "epoch": 1.8364928909952607, "grad_norm": 16.445026397705078, "learning_rate": 2e-06, "loss": 3.4511, "step": 775 }, { "epoch": 1.8364928909952607, "step": 775, "train/alpha": 0.2, "train/clean_ce_loss": 0.09912109375, "train/clean_loss_scale": 7.408888816833496, "train/clean_loss_weight": 1.4817777872085571, "train/combined_loss": 0.8812525868415833, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18364928909952608, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7343771457672119 }, { "epoch": 1.8364928909952607, "step": 775, "train/alpha": 0.2, "train/clean_ce_loss": 0.060302734375, "train/clean_loss_scale": 8.591292381286621, "train/clean_loss_weight": 1.7182585000991821, "train/combined_loss": 0.6216941475868225, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18364928909952608, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5180784463882446 }, { "epoch": 1.8364928909952607, "step": 775, "train/alpha": 0.2, "train/clean_ce_loss": 0.04150390625, "train/clean_loss_scale": 10.225052833557129, "train/clean_loss_weight": 2.045010566711426, "train/combined_loss": 0.5092555284500122, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18364928909952608, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4243796169757843 }, { "epoch": 1.8364928909952607, "step": 775, "train/alpha": 0.2, "train/clean_ce_loss": 0.10595703125, "train/clean_loss_scale": 6.586421489715576, "train/clean_loss_weight": 1.317284345626831, "train/combined_loss": 0.8374531865119934, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18364928909952608, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6978776454925537 }, { "epoch": 1.838862559241706, "step": 776, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 8.57486343383789, "train/clean_loss_weight": 1.714972734451294, "train/combined_loss": 0.9546235203742981, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18388625592417063, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7955195903778076 }, { "epoch": 1.838862559241706, "step": 776, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 7.473421573638916, "train/clean_loss_weight": 1.4946843385696411, "train/combined_loss": 0.7312859892845154, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18388625592417063, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.609404981136322 }, { "epoch": 1.838862559241706, "step": 776, "train/alpha": 0.2, "train/clean_ce_loss": 0.072265625, "train/clean_loss_scale": 10.320998191833496, "train/clean_loss_weight": 2.064199686050415, "train/combined_loss": 0.8950240612030029, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18388625592417063, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7458533644676208 }, { "epoch": 1.838862559241706, "step": 776, "train/alpha": 0.2, "train/clean_ce_loss": 0.07861328125, "train/clean_loss_scale": 6.722822189331055, "train/clean_loss_weight": 1.344564437866211, "train/combined_loss": 0.6342037320137024, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18388625592417063, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5285031199455261 }, { "epoch": 1.8412322274881516, "step": 777, "train/alpha": 0.2, "train/clean_ce_loss": 0.076171875, "train/clean_loss_scale": 7.399785041809082, "train/clean_loss_weight": 1.4799569845199585, "train/combined_loss": 0.6763865947723389, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18412322274881518, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5636554956436157 }, { "epoch": 1.8412322274881516, "step": 777, "train/alpha": 0.2, "train/clean_ce_loss": 0.01031494140625, "train/clean_loss_scale": 11.9197416305542, "train/clean_loss_weight": 2.38394832611084, "train/combined_loss": 0.14754171669483185, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18412322274881518, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1229514330625534 }, { "epoch": 1.8412322274881516, "step": 777, "train/alpha": 0.2, "train/clean_ce_loss": 0.0908203125, "train/clean_loss_scale": 6.4514241218566895, "train/clean_loss_weight": 1.2902848720550537, "train/combined_loss": 0.7031043767929077, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18412322274881518, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5859203338623047 }, { "epoch": 1.8412322274881516, "step": 777, "train/alpha": 0.2, "train/clean_ce_loss": 0.010498046875, "train/clean_loss_scale": 24.7208194732666, "train/clean_loss_weight": 4.944163799285889, "train/combined_loss": 0.3114243745803833, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18412322274881518, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2595203220844269 }, { "epoch": 1.8436018957345972, "step": 778, "train/alpha": 0.2, "train/clean_ce_loss": 0.11669921875, "train/clean_loss_scale": 8.305688858032227, "train/clean_loss_weight": 1.6611378192901611, "train/combined_loss": 1.1631208658218384, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.1843601895734597, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9692674279212952 }, { "epoch": 1.8436018957345972, "step": 778, "train/alpha": 0.2, "train/clean_ce_loss": 0.01513671875, "train/clean_loss_scale": 22.279481887817383, "train/clean_loss_weight": 4.455896377563477, "train/combined_loss": 0.404685914516449, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.1843601895734597, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.33723825216293335 }, { "epoch": 1.8436018957345972, "step": 778, "train/alpha": 0.2, "train/clean_ce_loss": 0.02685546875, "train/clean_loss_scale": 6.9541826248168945, "train/clean_loss_weight": 1.3908365964889526, "train/combined_loss": 0.22410939633846283, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.1843601895734597, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.18675783276557922 }, { "epoch": 1.8436018957345972, "step": 778, "train/alpha": 0.2, "train/clean_ce_loss": 0.06884765625, "train/clean_loss_scale": 15.861608505249023, "train/clean_loss_weight": 3.1723217964172363, "train/combined_loss": 1.310441493988037, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.1843601895734597, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0920345783233643 }, { "epoch": 1.8459715639810428, "step": 779, "train/alpha": 0.2, "train/clean_ce_loss": 0.12255859375, "train/clean_loss_scale": 8.294111251831055, "train/clean_loss_weight": 1.6588222980499268, "train/combined_loss": 1.2198176383972168, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18459715639810426, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0165146589279175 }, { "epoch": 1.8459715639810428, "step": 779, "train/alpha": 0.2, "train/clean_ce_loss": 0.022216796875, "train/clean_loss_scale": 7.382552146911621, "train/clean_loss_weight": 1.4765104055404663, "train/combined_loss": 0.1968199908733368, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18459715639810426, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.16401666402816772 }, { "epoch": 1.8459715639810428, "step": 779, "train/alpha": 0.2, "train/clean_ce_loss": 0.0966796875, "train/clean_loss_scale": 6.476907730102539, "train/clean_loss_weight": 1.2953815460205078, "train/combined_loss": 0.7514225244522095, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18459715639810426, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.626185417175293 }, { "epoch": 1.8459715639810428, "step": 779, "train/alpha": 0.2, "train/clean_ce_loss": 0.09619140625, "train/clean_loss_scale": 8.350768089294434, "train/clean_loss_weight": 1.6701536178588867, "train/combined_loss": 0.9639265537261963, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.18459715639810426, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8032721281051636 }, { "epoch": 1.8483412322274881, "grad_norm": 19.113401412963867, "learning_rate": 2e-06, "loss": 3.3781, "step": 780 }, { "epoch": 1.8483412322274881, "step": 780, "train/alpha": 0.2, "train/clean_ce_loss": 0.07470703125, "train/clean_loss_scale": 11.01995849609375, "train/clean_loss_weight": 2.203991651535034, "train/combined_loss": 0.9879220128059387, "train/pos_loss/pos0": 0.29617828130722046, "train/pos_loss/pos1": 0.7801150679588318, "train/pos_loss/pos2": 1.2427433729171753, "train/pos_loss/pos3": 1.8500295877456665, "train/pos_loss/pos4": 2.2764675617218018, "train/pos_loss/pos5": 2.786942481994629, "train/pos_loss/pos6": 3.302396774291992, "train/pos_loss/weight0": 2.920415163040161, "train/pos_loss/weight1": 1.9925739765167236, "train/pos_loss/weight2": 1.1806726455688477, "train/pos_loss/weight3": 0.5827637910842896, "train/pos_loss/weight4": 0.22600719332695007, "train/pos_loss/weight5": 0.07323820888996124, "train/pos_loss/weight6": 0.024329006671905518, "train/pos_weight_progress": 0.1848341232227488, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8232683539390564 }, { "epoch": 1.8483412322274881, "step": 780, "train/alpha": 0.2, "train/clean_ce_loss": 0.1259765625, "train/clean_loss_scale": 6.3511962890625, "train/clean_loss_weight": 1.270239233970642, "train/combined_loss": 0.9601222276687622, "train/pos_loss/pos0": 0.3687690794467926, "train/pos_loss/pos1": 0.7366665005683899, "train/pos_loss/pos2": 1.2095309495925903, "train/pos_loss/pos3": 1.7345229387283325, "train/pos_loss/pos4": 2.1971049308776855, "train/pos_loss/pos5": 2.5752806663513184, "train/pos_loss/pos6": 2.9521658420562744, "train/pos_loss/weight0": 2.9747846126556396, "train/pos_loss/weight1": 2.048902988433838, "train/pos_loss/weight2": 1.1484837532043457, "train/pos_loss/weight3": 0.5408902168273926, "train/pos_loss/weight4": 0.2052168846130371, "train/pos_loss/weight5": 0.061576731503009796, "train/pos_loss/weight6": 0.02014530636370182, "train/pos_weight_progress": 0.1848341232227488, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8001018762588501 }, { "epoch": 1.8483412322274881, "step": 780, "train/alpha": 0.2, "train/clean_ce_loss": 0.0306396484375, "train/clean_loss_scale": 10.862895965576172, "train/clean_loss_weight": 2.172579288482666, "train/combined_loss": 0.399402379989624, "train/pos_loss/pos0": 0.029043519869446754, "train/pos_loss/pos1": 0.2118932604789734, "train/pos_loss/pos2": 0.49588921666145325, "train/pos_loss/pos3": 1.0010610818862915, "train/pos_loss/pos4": 1.6170001029968262, "train/pos_loss/pos5": 2.2008533477783203, "train/pos_loss/pos6": 2.7029387950897217, "train/pos_loss/weight0": 2.92478084564209, "train/pos_loss/weight1": 1.965890884399414, "train/pos_loss/weight2": 1.1639279127120972, "train/pos_loss/weight3": 0.5878203511238098, "train/pos_loss/weight4": 0.23826494812965393, "train/pos_loss/weight5": 0.08709975332021713, "train/pos_loss/weight6": 0.0322154238820076, "train/pos_weight_progress": 0.1848341232227488, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.33283531665802 }, { "epoch": 1.8483412322274881, "step": 780, "train/alpha": 0.2, "train/clean_ce_loss": 0.00830078125, "train/clean_loss_scale": 38.392547607421875, "train/clean_loss_weight": 7.678509712219238, "train/combined_loss": 0.3824257552623749, "train/pos_loss/pos0": 0.00815864372998476, "train/pos_loss/pos1": 0.025746168568730354, "train/pos_loss/pos2": 0.18919938802719116, "train/pos_loss/pos3": 0.6685137748718262, "train/pos_loss/pos4": 1.287229061126709, "train/pos_loss/pos5": 2.04008150100708, "train/pos_loss/pos6": 2.6284031867980957, "train/pos_loss/weight0": 2.3318300247192383, "train/pos_loss/weight1": 1.7710506916046143, "train/pos_loss/weight2": 1.2602417469024658, "train/pos_loss/weight3": 0.8299533128738403, "train/pos_loss/weight4": 0.4607487618923187, "train/pos_loss/weight5": 0.2241210639476776, "train/pos_loss/weight6": 0.12205466628074646, "train/pos_weight_progress": 0.1848341232227488, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.31868812441825867 }, { "epoch": 1.8507109004739335, "step": 781, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 12.065214157104492, "train/clean_loss_weight": 2.4130427837371826, "train/combined_loss": 0.954377293586731, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18507109004739336, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7953144311904907 }, { "epoch": 1.8507109004739335, "step": 781, "train/alpha": 0.2, "train/clean_ce_loss": 0.02294921875, "train/clean_loss_scale": 17.692401885986328, "train/clean_loss_weight": 3.538480520248413, "train/combined_loss": 0.48723217844963074, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18507109004739336, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.40602681040763855 }, { "epoch": 1.8507109004739335, "step": 781, "train/alpha": 0.2, "train/clean_ce_loss": 0.1328125, "train/clean_loss_scale": 11.78857707977295, "train/clean_loss_weight": 2.357715368270874, "train/combined_loss": 1.8788044452667236, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18507109004739336, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.565670371055603 }, { "epoch": 1.8507109004739335, "step": 781, "train/alpha": 0.2, "train/clean_ce_loss": 0.1015625, "train/clean_loss_scale": 11.58552360534668, "train/clean_loss_weight": 2.3171048164367676, "train/combined_loss": 1.4119856357574463, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18507109004739336, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1766546964645386 }, { "epoch": 1.853080568720379, "step": 782, "train/alpha": 0.2, "train/clean_ce_loss": 0.08935546875, "train/clean_loss_scale": 9.696999549865723, "train/clean_loss_weight": 1.9393999576568604, "train/combined_loss": 1.0397759675979614, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.1853080568720379, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8664799332618713 }, { "epoch": 1.853080568720379, "step": 782, "train/alpha": 0.2, "train/clean_ce_loss": 0.0830078125, "train/clean_loss_scale": 9.74047565460205, "train/clean_loss_weight": 1.9480952024459839, "train/combined_loss": 0.9702426791191101, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.1853080568720379, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8085355758666992 }, { "epoch": 1.853080568720379, "step": 782, "train/alpha": 0.2, "train/clean_ce_loss": 0.07666015625, "train/clean_loss_scale": 14.811306953430176, "train/clean_loss_weight": 2.962261438369751, "train/combined_loss": 1.3625245094299316, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.1853080568720379, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1354371309280396 }, { "epoch": 1.853080568720379, "step": 782, "train/alpha": 0.2, "train/clean_ce_loss": 0.09130859375, "train/clean_loss_scale": 9.104846000671387, "train/clean_loss_weight": 1.8209692239761353, "train/combined_loss": 0.9976208209991455, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.1853080568720379, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8313506841659546 }, { "epoch": 1.8554502369668247, "step": 783, "train/alpha": 0.2, "train/clean_ce_loss": 0.032470703125, "train/clean_loss_scale": 14.211000442504883, "train/clean_loss_weight": 2.8422000408172607, "train/combined_loss": 0.5537294149398804, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18554502369668247, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46144118905067444 }, { "epoch": 1.8554502369668247, "step": 783, "train/alpha": 0.2, "train/clean_ce_loss": 0.05224609375, "train/clean_loss_scale": 9.318707466125488, "train/clean_loss_weight": 1.8637415170669556, "train/combined_loss": 0.5842392444610596, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18554502369668247, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4868660569190979 }, { "epoch": 1.8554502369668247, "step": 783, "train/alpha": 0.2, "train/clean_ce_loss": 0.0014495849609375, "train/clean_loss_scale": 232.03042602539062, "train/clean_loss_weight": 46.40608596801758, "train/combined_loss": 0.40361738204956055, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18554502369668247, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3363478183746338 }, { "epoch": 1.8554502369668247, "step": 783, "train/alpha": 0.2, "train/clean_ce_loss": 0.08984375, "train/clean_loss_scale": 9.278141975402832, "train/clean_loss_weight": 1.8556283712387085, "train/combined_loss": 1.0002996921539307, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18554502369668247, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8335830569267273 }, { "epoch": 1.8578199052132702, "step": 784, "train/alpha": 0.2, "train/clean_ce_loss": 0.0595703125, "train/clean_loss_scale": 10.417551040649414, "train/clean_loss_weight": 2.083510160446167, "train/combined_loss": 0.7446921467781067, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18578199052132702, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.620576798915863 }, { "epoch": 1.8578199052132702, "step": 784, "train/alpha": 0.2, "train/clean_ce_loss": 0.09130859375, "train/clean_loss_scale": 9.183748245239258, "train/clean_loss_weight": 1.8367496728897095, "train/combined_loss": 1.0062662363052368, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18578199052132702, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8385551571846008 }, { "epoch": 1.8578199052132702, "step": 784, "train/alpha": 0.2, "train/clean_ce_loss": 0.0250244140625, "train/clean_loss_scale": 12.563776969909668, "train/clean_loss_weight": 2.5127553939819336, "train/combined_loss": 0.3772813677787781, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18578199052132702, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.31440114974975586 }, { "epoch": 1.8578199052132702, "step": 784, "train/alpha": 0.2, "train/clean_ce_loss": 0.057373046875, "train/clean_loss_scale": 11.744673728942871, "train/clean_loss_weight": 2.3489348888397217, "train/combined_loss": 0.8085932731628418, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18578199052132702, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6738277077674866 }, { "epoch": 1.8601895734597158, "grad_norm": 18.36040687561035, "learning_rate": 2e-06, "loss": 3.6008, "step": 785 }, { "epoch": 1.8601895734597158, "step": 785, "train/alpha": 0.2, "train/clean_ce_loss": 0.0595703125, "train/clean_loss_scale": 11.183993339538574, "train/clean_loss_weight": 2.2367987632751465, "train/combined_loss": 0.7994807362556458, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18601895734597157, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6662339568138123 }, { "epoch": 1.8601895734597158, "step": 785, "train/alpha": 0.2, "train/clean_ce_loss": 0.04833984375, "train/clean_loss_scale": 23.134815216064453, "train/clean_loss_weight": 4.626963138580322, "train/combined_loss": 1.3420000076293945, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18601895734597157, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.118333339691162 }, { "epoch": 1.8601895734597158, "step": 785, "train/alpha": 0.2, "train/clean_ce_loss": 0.1640625, "train/clean_loss_scale": 8.14902400970459, "train/clean_loss_weight": 1.6298048496246338, "train/combined_loss": 1.6043391227722168, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18601895734597157, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3369492292404175 }, { "epoch": 1.8601895734597158, "step": 785, "train/alpha": 0.2, "train/clean_ce_loss": 0.126953125, "train/clean_loss_scale": 8.688178062438965, "train/clean_loss_weight": 1.737635612487793, "train/combined_loss": 1.323589563369751, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18601895734597157, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1029913425445557 }, { "epoch": 1.8625592417061612, "step": 786, "train/alpha": 0.2, "train/clean_ce_loss": 0.06005859375, "train/clean_loss_scale": 9.151435852050781, "train/clean_loss_weight": 1.830287218093872, "train/combined_loss": 0.6595468521118164, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18625592417061612, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5496223568916321 }, { "epoch": 1.8625592417061612, "step": 786, "train/alpha": 0.2, "train/clean_ce_loss": 0.06201171875, "train/clean_loss_scale": 15.52968692779541, "train/clean_loss_weight": 3.1059374809265137, "train/combined_loss": 1.1556271314620972, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18625592417061612, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9630225896835327 }, { "epoch": 1.8625592417061612, "step": 786, "train/alpha": 0.2, "train/clean_ce_loss": 0.005279541015625, "train/clean_loss_scale": 88.30245971679688, "train/clean_loss_weight": 17.660491943359375, "train/combined_loss": 0.5594357848167419, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18625592417061612, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4661964774131775 }, { "epoch": 1.8625592417061612, "step": 786, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 17.891244888305664, "train/clean_loss_weight": 3.578248977661133, "train/combined_loss": 1.3837759494781494, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18625592417061612, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1531466245651245 }, { "epoch": 1.8649289099526065, "step": 787, "train/alpha": 0.2, "train/clean_ce_loss": 0.06103515625, "train/clean_loss_scale": 10.032354354858398, "train/clean_loss_weight": 2.0064709186553955, "train/combined_loss": 0.7347915768623352, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18649289099526067, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6123263239860535 }, { "epoch": 1.8649289099526065, "step": 787, "train/alpha": 0.2, "train/clean_ce_loss": 0.18359375, "train/clean_loss_scale": 7.808937072753906, "train/clean_loss_weight": 1.561787486076355, "train/combined_loss": 1.7204065322875977, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18649289099526067, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4336720705032349 }, { "epoch": 1.8649289099526065, "step": 787, "train/alpha": 0.2, "train/clean_ce_loss": 0.039794921875, "train/clean_loss_scale": 12.285318374633789, "train/clean_loss_weight": 2.457063674926758, "train/combined_loss": 0.5866719484329224, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18649289099526067, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4888933002948761 }, { "epoch": 1.8649289099526065, "step": 787, "train/alpha": 0.2, "train/clean_ce_loss": 0.09521484375, "train/clean_loss_scale": 10.330117225646973, "train/clean_loss_weight": 2.066023588180542, "train/combined_loss": 1.180296540260315, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18649289099526067, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.983580470085144 }, { "epoch": 1.867298578199052, "step": 788, "train/alpha": 0.2, "train/clean_ce_loss": 0.07666015625, "train/clean_loss_scale": 19.460758209228516, "train/clean_loss_weight": 3.8921515941619873, "train/combined_loss": 1.7902377843856812, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18672985781990523, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.491864800453186 }, { "epoch": 1.867298578199052, "step": 788, "train/alpha": 0.2, "train/clean_ce_loss": 0.0595703125, "train/clean_loss_scale": 10.243817329406738, "train/clean_loss_weight": 2.0487635135650635, "train/combined_loss": 0.7322728633880615, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18672985781990523, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6102274060249329 }, { "epoch": 1.867298578199052, "step": 788, "train/alpha": 0.2, "train/clean_ce_loss": 0.0732421875, "train/clean_loss_scale": 16.97069549560547, "train/clean_loss_weight": 3.394139051437378, "train/combined_loss": 1.4915649890899658, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18672985781990523, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2429708242416382 }, { "epoch": 1.867298578199052, "step": 788, "train/alpha": 0.2, "train/clean_ce_loss": 0.07666015625, "train/clean_loss_scale": 12.940946578979492, "train/clean_loss_weight": 2.5881893634796143, "train/combined_loss": 1.190466046333313, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18672985781990523, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9920549988746643 }, { "epoch": 1.8696682464454977, "step": 789, "train/alpha": 0.2, "train/clean_ce_loss": 0.054931640625, "train/clean_loss_scale": 10.196575164794922, "train/clean_loss_weight": 2.0393149852752686, "train/combined_loss": 0.6721375584602356, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18696682464454975, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5601146221160889 }, { "epoch": 1.8696682464454977, "step": 789, "train/alpha": 0.2, "train/clean_ce_loss": 0.05078125, "train/clean_loss_scale": 10.967601776123047, "train/clean_loss_weight": 2.1935203075408936, "train/combined_loss": 0.6683382391929626, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18696682464454975, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5569485425949097 }, { "epoch": 1.8696682464454977, "step": 789, "train/alpha": 0.2, "train/clean_ce_loss": 0.087890625, "train/clean_loss_scale": 9.497382164001465, "train/clean_loss_weight": 1.899476408958435, "train/combined_loss": 1.0016770362854004, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18696682464454975, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.834730863571167 }, { "epoch": 1.8696682464454977, "step": 789, "train/alpha": 0.2, "train/clean_ce_loss": 0.0029144287109375, "train/clean_loss_scale": 95.77234649658203, "train/clean_loss_weight": 19.154470443725586, "train/combined_loss": 0.33494600653648376, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.18696682464454975, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.27912166714668274 }, { "epoch": 1.8720379146919433, "grad_norm": 23.931201934814453, "learning_rate": 2e-06, "loss": 3.5842, "step": 790 }, { "epoch": 1.8720379146919433, "step": 790, "train/alpha": 0.2, "train/clean_ce_loss": 0.06787109375, "train/clean_loss_scale": 11.718339920043945, "train/clean_loss_weight": 2.343667984008789, "train/combined_loss": 0.9544038772583008, "train/pos_loss/pos0": 0.2648780047893524, "train/pos_loss/pos1": 0.49125128984451294, "train/pos_loss/pos2": 0.7897652387619019, "train/pos_loss/pos3": 1.176623821258545, "train/pos_loss/pos4": 1.7229546308517456, "train/pos_loss/pos5": 2.140434980392456, "train/pos_loss/pos6": 2.488168716430664, "train/pos_loss/weight0": 2.1496188640594482, "train/pos_loss/weight1": 1.6495699882507324, "train/pos_loss/weight2": 1.2203069925308228, "train/pos_loss/weight3": 0.8865348100662231, "train/pos_loss/weight4": 0.5594004988670349, "train/pos_loss/weight5": 0.32395434379577637, "train/pos_loss/weight6": 0.2106146365404129, "train/pos_weight_progress": 0.1872037914691943, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7953365445137024 }, { "epoch": 1.8720379146919433, "step": 790, "train/alpha": 0.2, "train/clean_ce_loss": 0.051513671875, "train/clean_loss_scale": 11.870285987854004, "train/clean_loss_weight": 2.3740572929382324, "train/combined_loss": 0.7337784171104431, "train/pos_loss/pos0": 0.22318929433822632, "train/pos_loss/pos1": 0.43406492471694946, "train/pos_loss/pos2": 0.8009170293807983, "train/pos_loss/pos3": 1.1809940338134766, "train/pos_loss/pos4": 1.512802243232727, "train/pos_loss/pos5": 1.8551453351974487, "train/pos_loss/pos6": 2.316227436065674, "train/pos_loss/weight0": 2.5749025344848633, "train/pos_loss/weight1": 1.857798457145691, "train/pos_loss/weight2": 1.219381332397461, "train/pos_loss/weight3": 0.7225191593170166, "train/pos_loss/weight4": 0.3745660185813904, "train/pos_loss/weight5": 0.16891300678253174, "train/pos_loss/weight6": 0.08191987872123718, "train/pos_weight_progress": 0.1872037914691943, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6114820241928101 }, { "epoch": 1.8720379146919433, "step": 790, "train/alpha": 0.2, "train/clean_ce_loss": 0.1748046875, "train/clean_loss_scale": 7.556791305541992, "train/clean_loss_weight": 1.5113582611083984, "train/combined_loss": 1.5851550102233887, "train/pos_loss/pos0": 0.5989698171615601, "train/pos_loss/pos1": 1.0075383186340332, "train/pos_loss/pos2": 1.6574219465255737, "train/pos_loss/pos3": 2.1175622940063477, "train/pos_loss/pos4": 2.7809295654296875, "train/pos_loss/pos5": 3.142056941986084, "train/pos_loss/pos6": 3.442566156387329, "train/pos_loss/weight0": 2.4259581565856934, "train/pos_loss/weight1": 1.7968170642852783, "train/pos_loss/weight2": 1.2332926988601685, "train/pos_loss/weight3": 0.775661826133728, "train/pos_loss/weight4": 0.42377158999443054, "train/pos_loss/weight5": 0.21979805827140808, "train/pos_loss/weight6": 0.12470059096813202, "train/pos_weight_progress": 0.1872037914691943, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3209625482559204 }, { "epoch": 1.8720379146919433, "step": 790, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 9.756150245666504, "train/clean_loss_weight": 1.9512300491333008, "train/combined_loss": 0.8288916349411011, "train/pos_loss/pos0": 0.24667441844940186, "train/pos_loss/pos1": 0.6179643869400024, "train/pos_loss/pos2": 1.2345876693725586, "train/pos_loss/pos3": 1.9028875827789307, "train/pos_loss/pos4": 2.7067949771881104, "train/pos_loss/pos5": 3.2126147747039795, "train/pos_loss/pos6": 3.5510764122009277, "train/pos_loss/weight0": 3.2450296878814697, "train/pos_loss/weight1": 2.0135488510131836, "train/pos_loss/weight2": 1.074381709098816, "train/pos_loss/weight3": 0.4606567621231079, "train/pos_loss/weight4": 0.15595048666000366, "train/pos_loss/weight5": 0.039844293147325516, "train/pos_loss/weight6": 0.010588209144771099, "train/pos_weight_progress": 0.1872037914691943, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6907430291175842 }, { "epoch": 1.8744075829383886, "step": 791, "train/alpha": 0.2, "train/clean_ce_loss": 0.11572265625, "train/clean_loss_scale": 6.367475986480713, "train/clean_loss_weight": 1.2734951972961426, "train/combined_loss": 0.8842334747314453, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18744075829383886, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7368612289428711 }, { "epoch": 1.8744075829383886, "step": 791, "train/alpha": 0.2, "train/clean_ce_loss": 0.062255859375, "train/clean_loss_scale": 7.753774642944336, "train/clean_loss_weight": 1.5507549047470093, "train/combined_loss": 0.5792614817619324, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18744075829383886, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.482717901468277 }, { "epoch": 1.8744075829383886, "step": 791, "train/alpha": 0.2, "train/clean_ce_loss": 0.027099609375, "train/clean_loss_scale": 2.934230327606201, "train/clean_loss_weight": 0.5868460536003113, "train/combined_loss": 0.09541979432106018, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18744075829383886, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.07951649278402328 }, { "epoch": 1.8744075829383886, "step": 791, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 6.896304130554199, "train/clean_loss_weight": 1.3792608976364136, "train/combined_loss": 0.6546101570129395, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18744075829383886, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5455084443092346 }, { "epoch": 1.876777251184834, "step": 792, "train/alpha": 0.2, "train/clean_ce_loss": 0.057373046875, "train/clean_loss_scale": 14.720315933227539, "train/clean_loss_weight": 2.944063186645508, "train/combined_loss": 1.0134592056274414, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.1876777251184834, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8445493578910828 }, { "epoch": 1.876777251184834, "step": 792, "train/alpha": 0.2, "train/clean_ce_loss": 0.0830078125, "train/clean_loss_scale": 8.382514953613281, "train/clean_loss_weight": 1.67650306224823, "train/combined_loss": 0.8349770307540894, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.1876777251184834, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6958141922950745 }, { "epoch": 1.876777251184834, "step": 792, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 8.763725280761719, "train/clean_loss_weight": 1.7527450323104858, "train/combined_loss": 0.8524092435836792, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.1876777251184834, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7103410363197327 }, { "epoch": 1.876777251184834, "step": 792, "train/alpha": 0.2, "train/clean_ce_loss": 0.00628662109375, "train/clean_loss_scale": 20.15358543395996, "train/clean_loss_weight": 4.030717372894287, "train/combined_loss": 0.15203754603862762, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.1876777251184834, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.12669795751571655 }, { "epoch": 1.8791469194312795, "step": 793, "train/alpha": 0.2, "train/clean_ce_loss": 0.0048828125, "train/clean_loss_scale": 19.673864364624023, "train/clean_loss_weight": 3.9347729682922363, "train/combined_loss": 0.11527655273675919, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18791469194312796, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.09606379270553589 }, { "epoch": 1.8791469194312795, "step": 793, "train/alpha": 0.2, "train/clean_ce_loss": 0.0830078125, "train/clean_loss_scale": 8.354291915893555, "train/clean_loss_weight": 1.670858383178711, "train/combined_loss": 0.8321657776832581, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18791469194312796, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6934714913368225 }, { "epoch": 1.8791469194312795, "step": 793, "train/alpha": 0.2, "train/clean_ce_loss": 0.029296875, "train/clean_loss_scale": 13.801623344421387, "train/clean_loss_weight": 2.760324716567993, "train/combined_loss": 0.48521333932876587, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18791469194312796, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.40434443950653076 }, { "epoch": 1.8791469194312795, "step": 793, "train/alpha": 0.2, "train/clean_ce_loss": 0.00101470947265625, "train/clean_loss_scale": 85.95886993408203, "train/clean_loss_weight": 17.191774368286133, "train/combined_loss": 0.10466793179512024, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18791469194312796, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.08722327649593353 }, { "epoch": 1.8815165876777251, "step": 794, "train/alpha": 0.2, "train/clean_ce_loss": 0.1103515625, "train/clean_loss_scale": 6.741244792938232, "train/clean_loss_weight": 1.3482489585876465, "train/combined_loss": 0.8926882743835449, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.1881516587677725, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7439069151878357 }, { "epoch": 1.8815165876777251, "step": 794, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 10.05647087097168, "train/clean_loss_weight": 2.01129412651062, "train/combined_loss": 0.960471510887146, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.1881516587677725, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8003929257392883 }, { "epoch": 1.8815165876777251, "step": 794, "train/alpha": 0.2, "train/clean_ce_loss": 0.0634765625, "train/clean_loss_scale": 5.298216342926025, "train/clean_loss_weight": 1.059643268585205, "train/combined_loss": 0.40357506275177, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.1881516587677725, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.33631256222724915 }, { "epoch": 1.8815165876777251, "step": 794, "train/alpha": 0.2, "train/clean_ce_loss": 0.08203125, "train/clean_loss_scale": 5.686304092407227, "train/clean_loss_weight": 1.1372607946395874, "train/combined_loss": 0.5597455501556396, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.1881516587677725, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4664546251296997 }, { "epoch": 1.8838862559241707, "grad_norm": 20.841005325317383, "learning_rate": 2e-06, "loss": 3.2948, "step": 795 }, { "epoch": 1.8838862559241707, "step": 795, "train/alpha": 0.2, "train/clean_ce_loss": 0.1591796875, "train/clean_loss_scale": 6.5570502281188965, "train/clean_loss_weight": 1.3114100694656372, "train/combined_loss": 1.2524991035461426, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18838862559241706, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.043749213218689 }, { "epoch": 1.8838862559241707, "step": 795, "train/alpha": 0.2, "train/clean_ce_loss": 0.055419921875, "train/clean_loss_scale": 7.919750213623047, "train/clean_loss_weight": 1.5839500427246094, "train/combined_loss": 0.5266943573951721, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18838862559241706, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43891194462776184 }, { "epoch": 1.8838862559241707, "step": 795, "train/alpha": 0.2, "train/clean_ce_loss": 0.058349609375, "train/clean_loss_scale": 13.943159103393555, "train/clean_loss_weight": 2.7886319160461426, "train/combined_loss": 0.9762934446334839, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18838862559241706, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8135778903961182 }, { "epoch": 1.8838862559241707, "step": 795, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 6.480156898498535, "train/clean_loss_weight": 1.2960313558578491, "train/combined_loss": 0.6151086091995239, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18838862559241706, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5125905275344849 }, { "epoch": 1.8862559241706163, "step": 796, "train/alpha": 0.2, "train/clean_ce_loss": 0.123046875, "train/clean_loss_scale": 6.544642925262451, "train/clean_loss_weight": 1.3089286088943481, "train/combined_loss": 0.966357409954071, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18862559241706162, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8052978515625 }, { "epoch": 1.8862559241706163, "step": 796, "train/alpha": 0.2, "train/clean_ce_loss": 0.0167236328125, "train/clean_loss_scale": 15.065752029418945, "train/clean_loss_weight": 3.013150453567505, "train/combined_loss": 0.3023449182510376, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18862559241706162, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2519541084766388 }, { "epoch": 1.8862559241706163, "step": 796, "train/alpha": 0.2, "train/clean_ce_loss": 0.0279541015625, "train/clean_loss_scale": 4.574182987213135, "train/clean_loss_weight": 0.914836585521698, "train/combined_loss": 0.15344060957431793, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18862559241706162, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1278671771287918 }, { "epoch": 1.8862559241706163, "step": 796, "train/alpha": 0.2, "train/clean_ce_loss": 0.13671875, "train/clean_loss_scale": 7.034907817840576, "train/clean_loss_weight": 1.4069815874099731, "train/combined_loss": 1.1541645526885986, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18862559241706162, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9618037939071655 }, { "epoch": 1.8886255924170616, "step": 797, "train/alpha": 0.2, "train/clean_ce_loss": 0.04736328125, "train/clean_loss_scale": 16.87990951538086, "train/clean_loss_weight": 3.3759820461273193, "train/combined_loss": 0.9593854546546936, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18886255924170617, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7994878888130188 }, { "epoch": 1.8886255924170616, "step": 797, "train/alpha": 0.2, "train/clean_ce_loss": 0.07763671875, "train/clean_loss_scale": 6.874391078948975, "train/clean_loss_weight": 1.3748782873153687, "train/combined_loss": 0.6404461860656738, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18886255924170617, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5337051749229431 }, { "epoch": 1.8886255924170616, "step": 797, "train/alpha": 0.2, "train/clean_ce_loss": 0.044921875, "train/clean_loss_scale": 5.35996150970459, "train/clean_loss_weight": 1.07199227809906, "train/combined_loss": 0.28893542289733887, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18886255924170617, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.24077951908111572 }, { "epoch": 1.8886255924170616, "step": 797, "train/alpha": 0.2, "train/clean_ce_loss": 0.1611328125, "train/clean_loss_scale": 6.450446128845215, "train/clean_loss_weight": 1.2900892496109009, "train/combined_loss": 1.2472542524337769, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18886255924170617, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0393785238265991 }, { "epoch": 1.890995260663507, "step": 798, "train/alpha": 0.2, "train/clean_ce_loss": 0.08642578125, "train/clean_loss_scale": 5.916788578033447, "train/clean_loss_weight": 1.1833577156066895, "train/combined_loss": 0.6136357188224792, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18909952606635072, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5113630890846252 }, { "epoch": 1.890995260663507, "step": 798, "train/alpha": 0.2, "train/clean_ce_loss": 0.008056640625, "train/clean_loss_scale": 20.476560592651367, "train/clean_loss_weight": 4.095312118530273, "train/combined_loss": 0.19796675443649292, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18909952606635072, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.16497229039669037 }, { "epoch": 1.890995260663507, "step": 798, "train/alpha": 0.2, "train/clean_ce_loss": 0.0189208984375, "train/clean_loss_scale": 5.0853705406188965, "train/clean_loss_weight": 1.0170741081237793, "train/combined_loss": 0.1154637336730957, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18909952606635072, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.09621977806091309 }, { "epoch": 1.890995260663507, "step": 798, "train/alpha": 0.2, "train/clean_ce_loss": 0.08935546875, "train/clean_loss_scale": 8.975197792053223, "train/clean_loss_weight": 1.7950395345687866, "train/combined_loss": 0.9623795747756958, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18909952606635072, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8019829988479614 }, { "epoch": 1.8933649289099526, "step": 799, "train/alpha": 0.2, "train/clean_ce_loss": 0.0169677734375, "train/clean_loss_scale": 8.862432479858398, "train/clean_loss_weight": 1.7724865674972534, "train/combined_loss": 0.18045088648796082, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18933649289099527, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.15037573873996735 }, { "epoch": 1.8933649289099526, "step": 799, "train/alpha": 0.2, "train/clean_ce_loss": 0.04736328125, "train/clean_loss_scale": 9.686226844787598, "train/clean_loss_weight": 1.9372453689575195, "train/combined_loss": 0.5505257844924927, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18933649289099527, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4587714672088623 }, { "epoch": 1.8933649289099526, "step": 799, "train/alpha": 0.2, "train/clean_ce_loss": 0.09912109375, "train/clean_loss_scale": 8.305269241333008, "train/clean_loss_weight": 1.6610538959503174, "train/combined_loss": 0.9878728985786438, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18933649289099527, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8232274055480957 }, { "epoch": 1.8933649289099526, "step": 799, "train/alpha": 0.2, "train/clean_ce_loss": 0.0458984375, "train/clean_loss_scale": 6.950183391571045, "train/clean_loss_weight": 1.390036702156067, "train/combined_loss": 0.3828030824661255, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.18933649289099527, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.31900256872177124 }, { "epoch": 1.8957345971563981, "grad_norm": 12.978764533996582, "learning_rate": 2e-06, "loss": 3.355, "step": 800 }, { "epoch": 1.8957345971563981, "step": 800, "train/alpha": 0.2, "train/clean_ce_loss": 0.083984375, "train/clean_loss_scale": 6.318140983581543, "train/clean_loss_weight": 1.2636282444000244, "train/combined_loss": 0.6367501020431519, "train/pos_loss/pos0": 0.22311678528785706, "train/pos_loss/pos1": 0.4742526710033417, "train/pos_loss/pos2": 0.8091452717781067, "train/pos_loss/pos3": 1.2354857921600342, "train/pos_loss/pos4": 1.6089123487472534, "train/pos_loss/pos5": 2.13348388671875, "train/pos_loss/pos6": 2.413908004760742, "train/pos_loss/weight0": 2.9724228382110596, "train/pos_loss/weight1": 2.010962724685669, "train/pos_loss/weight2": 1.2138694524765015, "train/pos_loss/weight3": 0.5535037517547607, "train/pos_loss/weight4": 0.19342119991779327, "train/pos_loss/weight5": 0.04434119910001755, "train/pos_loss/weight6": 0.01147888507694006, "train/pos_weight_progress": 0.1895734597156398, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5306251049041748 }, { "epoch": 1.8957345971563981, "step": 800, "train/alpha": 0.2, "train/clean_ce_loss": 0.0966796875, "train/clean_loss_scale": 9.509405136108398, "train/clean_loss_weight": 1.9018810987472534, "train/combined_loss": 1.1032395362854004, "train/pos_loss/pos0": 0.3275865316390991, "train/pos_loss/pos1": 0.7239781022071838, "train/pos_loss/pos2": 1.2944566011428833, "train/pos_loss/pos3": 1.6468513011932373, "train/pos_loss/pos4": 2.250220537185669, "train/pos_loss/pos5": 2.617032766342163, "train/pos_loss/pos6": 3.0265862941741943, "train/pos_loss/weight0": 2.586336612701416, "train/pos_loss/weight1": 1.8645564317703247, "train/pos_loss/weight2": 1.2332744598388672, "train/pos_loss/weight3": 0.7175836563110352, "train/pos_loss/weight4": 0.35837647318840027, "train/pos_loss/weight5": 0.16328993439674377, "train/pos_loss/weight6": 0.07658259570598602, "train/pos_weight_progress": 0.1895734597156398, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9193663001060486 }, { "epoch": 1.8957345971563981, "step": 800, "train/alpha": 0.2, "train/clean_ce_loss": 0.019287109375, "train/clean_loss_scale": 16.222412109375, "train/clean_loss_weight": 3.2444825172424316, "train/combined_loss": 0.37546011805534363, "train/pos_loss/pos0": 0.035602398216724396, "train/pos_loss/pos1": 0.13596384227275848, "train/pos_loss/pos2": 0.44925716519355774, "train/pos_loss/pos3": 0.9188476204872131, "train/pos_loss/pos4": 1.506373643875122, "train/pos_loss/pos5": 1.8779124021530151, "train/pos_loss/pos6": 2.356239080429077, "train/pos_loss/weight0": 2.80666184425354, "train/pos_loss/weight1": 1.9429394006729126, "train/pos_loss/weight2": 1.2049129009246826, "train/pos_loss/weight3": 0.6208423972129822, "train/pos_loss/weight4": 0.27847498655319214, "train/pos_loss/weight5": 0.10314157605171204, "train/pos_loss/weight6": 0.0430268794298172, "train/pos_weight_progress": 0.1895734597156398, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3128834366798401 }, { "epoch": 1.8957345971563981, "step": 800, "train/alpha": 0.2, "train/clean_ce_loss": 0.01141357421875, "train/clean_loss_scale": 31.12289810180664, "train/clean_loss_weight": 6.224579811096191, "train/combined_loss": 0.42626821994781494, "train/pos_loss/pos0": 0.007041992619633675, "train/pos_loss/pos1": 0.032566990703344345, "train/pos_loss/pos2": 0.15844939649105072, "train/pos_loss/pos3": 0.5569291114807129, "train/pos_loss/pos4": 1.0749603509902954, "train/pos_loss/pos5": 1.9646098613739014, "train/pos_loss/pos6": 2.6763322353363037, "train/pos_loss/weight0": 2.1578590869903564, "train/pos_loss/weight1": 1.6689956188201904, "train/pos_loss/weight2": 1.2463692426681519, "train/pos_loss/weight3": 0.8696655035018921, "train/pos_loss/weight4": 0.5421794056892395, "train/pos_loss/weight5": 0.31503021717071533, "train/pos_loss/weight6": 0.19990107417106628, "train/pos_weight_progress": 0.1895734597156398, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.35522350668907166 }, { "epoch": 1.8981042654028437, "step": 801, "train/alpha": 0.2, "train/clean_ce_loss": 0.0556640625, "train/clean_loss_scale": 11.088854789733887, "train/clean_loss_weight": 2.217771053314209, "train/combined_loss": 0.7407008409500122, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.18981042654028435, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6172506809234619 }, { "epoch": 1.8981042654028437, "step": 801, "train/alpha": 0.2, "train/clean_ce_loss": 0.057373046875, "train/clean_loss_scale": 10.941823959350586, "train/clean_loss_weight": 2.1883647441864014, "train/combined_loss": 0.7533189058303833, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.18981042654028435, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6277657747268677 }, { "epoch": 1.8981042654028437, "step": 801, "train/alpha": 0.2, "train/clean_ce_loss": 0.08935546875, "train/clean_loss_scale": 14.932893753051758, "train/clean_loss_weight": 2.9865787029266357, "train/combined_loss": 1.6012028455734253, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.18981042654028435, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3343356847763062 }, { "epoch": 1.8981042654028437, "step": 801, "train/alpha": 0.2, "train/clean_ce_loss": 0.0037078857421875, "train/clean_loss_scale": 110.63652801513672, "train/clean_loss_weight": 22.12730598449707, "train/combined_loss": 0.49227312207221985, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.18981042654028435, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.41022759675979614 }, { "epoch": 1.900473933649289, "step": 802, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 12.246466636657715, "train/clean_loss_weight": 2.449293375015259, "train/combined_loss": 0.9543632864952087, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.1900473933649289, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7953027486801147 }, { "epoch": 1.900473933649289, "step": 802, "train/alpha": 0.2, "train/clean_ce_loss": 0.026611328125, "train/clean_loss_scale": 19.80056381225586, "train/clean_loss_weight": 3.9601128101348877, "train/combined_loss": 0.6323031783103943, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.1900473933649289, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5269193053245544 }, { "epoch": 1.900473933649289, "step": 802, "train/alpha": 0.2, "train/clean_ce_loss": 0.033203125, "train/clean_loss_scale": 14.459794044494629, "train/clean_loss_weight": 2.8919589519500732, "train/combined_loss": 0.5761324167251587, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.1900473933649289, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4801103472709656 }, { "epoch": 1.900473933649289, "step": 802, "train/alpha": 0.2, "train/clean_ce_loss": 0.0634765625, "train/clean_loss_scale": 9.028092384338379, "train/clean_loss_weight": 1.8056185245513916, "train/combined_loss": 0.6876866817474365, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.1900473933649289, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5730722546577454 }, { "epoch": 1.9028436018957346, "step": 803, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 9.433540344238281, "train/clean_loss_weight": 1.88670814037323, "train/combined_loss": 0.989416241645813, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.19028436018957345, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8245135545730591 }, { "epoch": 1.9028436018957346, "step": 803, "train/alpha": 0.2, "train/clean_ce_loss": 0.022216796875, "train/clean_loss_scale": 20.452350616455078, "train/clean_loss_weight": 4.090470314025879, "train/combined_loss": 0.54526287317276, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.19028436018957345, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4543857276439667 }, { "epoch": 1.9028436018957346, "step": 803, "train/alpha": 0.2, "train/clean_ce_loss": 0.1953125, "train/clean_loss_scale": 7.0797014236450195, "train/clean_loss_weight": 1.415940284729004, "train/combined_loss": 1.6593050956726074, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.19028436018957345, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3827542066574097 }, { "epoch": 1.9028436018957346, "step": 803, "train/alpha": 0.2, "train/clean_ce_loss": 0.130859375, "train/clean_loss_scale": 8.603376388549805, "train/clean_loss_weight": 1.7206753492355347, "train/combined_loss": 1.350998878479004, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.19028436018957345, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.125832438468933 }, { "epoch": 1.90521327014218, "step": 804, "train/alpha": 0.2, "train/clean_ce_loss": 0.1845703125, "train/clean_loss_scale": 9.109479904174805, "train/clean_loss_weight": 1.821895956993103, "train/combined_loss": 2.0176074504852295, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.190521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.6813396215438843 }, { "epoch": 1.90521327014218, "step": 804, "train/alpha": 0.2, "train/clean_ce_loss": 0.1328125, "train/clean_loss_scale": 9.575491905212402, "train/clean_loss_weight": 1.9150984287261963, "train/combined_loss": 1.5260939598083496, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.190521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.271744966506958 }, { "epoch": 1.90521327014218, "step": 804, "train/alpha": 0.2, "train/clean_ce_loss": 0.0301513671875, "train/clean_loss_scale": 18.815523147583008, "train/clean_loss_weight": 3.7631046772003174, "train/combined_loss": 0.6807764768600464, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.190521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5673137307167053 }, { "epoch": 1.90521327014218, "step": 804, "train/alpha": 0.2, "train/clean_ce_loss": 0.09228515625, "train/clean_loss_scale": 9.130781173706055, "train/clean_loss_weight": 1.8261562585830688, "train/combined_loss": 1.0111626386642456, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.190521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8426355719566345 }, { "epoch": 1.9075829383886256, "grad_norm": 18.306896209716797, "learning_rate": 2e-06, "loss": 3.3007, "step": 805 }, { "epoch": 1.9075829383886256, "step": 805, "train/alpha": 0.2, "train/clean_ce_loss": 0.052001953125, "train/clean_loss_scale": 17.60110092163086, "train/clean_loss_weight": 3.5202202796936035, "train/combined_loss": 1.0983500480651855, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.19075829383886256, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9152916669845581 }, { "epoch": 1.9075829383886256, "step": 805, "train/alpha": 0.2, "train/clean_ce_loss": 0.0517578125, "train/clean_loss_scale": 13.415982246398926, "train/clean_loss_weight": 2.683196544647217, "train/combined_loss": 0.8332582712173462, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.19075829383886256, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6943818926811218 }, { "epoch": 1.9075829383886256, "step": 805, "train/alpha": 0.2, "train/clean_ce_loss": 0.046630859375, "train/clean_loss_scale": 12.31103801727295, "train/clean_loss_weight": 2.462207555770874, "train/combined_loss": 0.6888891458511353, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.19075829383886256, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5740742683410645 }, { "epoch": 1.9075829383886256, "step": 805, "train/alpha": 0.2, "train/clean_ce_loss": 0.048095703125, "train/clean_loss_scale": 12.842391967773438, "train/clean_loss_weight": 2.5684783458709717, "train/combined_loss": 0.7411966323852539, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.19075829383886256, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6176638603210449 }, { "epoch": 1.9099526066350712, "step": 806, "train/alpha": 0.2, "train/clean_ce_loss": 0.1513671875, "train/clean_loss_scale": 9.733942985534668, "train/clean_loss_weight": 1.9467886686325073, "train/combined_loss": 1.7680795192718506, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.1909952606635071, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4733996391296387 }, { "epoch": 1.9099526066350712, "step": 806, "train/alpha": 0.2, "train/clean_ce_loss": 0.040283203125, "train/clean_loss_scale": 12.817025184631348, "train/clean_loss_weight": 2.5634050369262695, "train/combined_loss": 0.6195729970932007, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.1909952606635071, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5163108110427856 }, { "epoch": 1.9099526066350712, "step": 806, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 9.335112571716309, "train/clean_loss_weight": 1.8670225143432617, "train/combined_loss": 0.9353345632553101, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.1909952606635071, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.779445469379425 }, { "epoch": 1.9099526066350712, "step": 806, "train/alpha": 0.2, "train/clean_ce_loss": 0.0196533203125, "train/clean_loss_scale": 19.27229118347168, "train/clean_loss_weight": 3.8544583320617676, "train/combined_loss": 0.4545174241065979, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.1909952606635071, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3787645101547241 }, { "epoch": 1.9123222748815167, "step": 807, "train/alpha": 0.2, "train/clean_ce_loss": 0.056884765625, "train/clean_loss_scale": 10.915453910827637, "train/clean_loss_weight": 2.183090925216675, "train/combined_loss": 0.7451076507568359, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.19123222748815166, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6209230422973633 }, { "epoch": 1.9123222748815167, "step": 807, "train/alpha": 0.2, "train/clean_ce_loss": 0.0002593994140625, "train/clean_loss_scale": 1284.6087646484375, "train/clean_loss_weight": 256.9217529296875, "train/combined_loss": 0.3998721241950989, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.19123222748815166, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3332267701625824 }, { "epoch": 1.9123222748815167, "step": 807, "train/alpha": 0.2, "train/clean_ce_loss": 0.09326171875, "train/clean_loss_scale": 9.945049285888672, "train/clean_loss_weight": 1.9890098571777344, "train/combined_loss": 1.1129908561706543, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.19123222748815166, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9274923801422119 }, { "epoch": 1.9123222748815167, "step": 807, "train/alpha": 0.2, "train/clean_ce_loss": 0.1611328125, "train/clean_loss_scale": 9.705927848815918, "train/clean_loss_weight": 1.9411855936050415, "train/combined_loss": 1.8767321109771729, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.19123222748815166, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5639433860778809 }, { "epoch": 1.914691943127962, "step": 808, "train/alpha": 0.2, "train/clean_ce_loss": 0.1201171875, "train/clean_loss_scale": 11.863654136657715, "train/clean_loss_weight": 2.3727309703826904, "train/combined_loss": 1.7100346088409424, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.19146919431279621, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4250288009643555 }, { "epoch": 1.914691943127962, "step": 808, "train/alpha": 0.2, "train/clean_ce_loss": 0.03271484375, "train/clean_loss_scale": 17.366802215576172, "train/clean_loss_weight": 3.473360538482666, "train/combined_loss": 0.6817827224731445, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.19146919431279621, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5681522488594055 }, { "epoch": 1.914691943127962, "step": 808, "train/alpha": 0.2, "train/clean_ce_loss": 0.000728607177734375, "train/clean_loss_scale": 498.75543212890625, "train/clean_loss_weight": 99.75109100341797, "train/combined_loss": 0.43607616424560547, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.19146919431279621, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3633967936038971 }, { "epoch": 1.914691943127962, "step": 808, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 8.95323657989502, "train/clean_loss_weight": 1.7906473875045776, "train/combined_loss": 0.7029690146446228, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.19146919431279621, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5858075022697449 }, { "epoch": 1.9170616113744074, "step": 809, "train/alpha": 0.2, "train/clean_ce_loss": 0.031494140625, "train/clean_loss_scale": 13.724931716918945, "train/clean_loss_weight": 2.7449862957000732, "train/combined_loss": 0.5187059044837952, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.19170616113744077, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43225494027137756 }, { "epoch": 1.9170616113744074, "step": 809, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 9.37973403930664, "train/clean_loss_weight": 1.8759468793869019, "train/combined_loss": 1.044228196144104, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.19170616113744077, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8701901435852051 }, { "epoch": 1.9170616113744074, "step": 809, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 11.599960327148438, "train/clean_loss_weight": 2.3199920654296875, "train/combined_loss": 1.2914018630981445, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.19170616113744077, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.076168179512024 }, { "epoch": 1.9170616113744074, "step": 809, "train/alpha": 0.2, "train/clean_ce_loss": 0.0283203125, "train/clean_loss_scale": 19.88599395751953, "train/clean_loss_weight": 3.977198839187622, "train/combined_loss": 0.6758130788803101, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.19170616113744077, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5631775856018066 }, { "epoch": 1.919431279620853, "grad_norm": 17.04330062866211, "learning_rate": 2e-06, "loss": 3.3665, "step": 810 }, { "epoch": 1.919431279620853, "step": 810, "train/alpha": 0.2, "train/clean_ce_loss": 0.04150390625, "train/clean_loss_scale": 13.360518455505371, "train/clean_loss_weight": 2.6721036434173584, "train/combined_loss": 0.665416419506073, "train/pos_loss/pos0": 0.06645835936069489, "train/pos_loss/pos1": 0.22492998838424683, "train/pos_loss/pos2": 0.5313528180122375, "train/pos_loss/pos3": 0.9388808012008667, "train/pos_loss/pos4": 1.3551619052886963, "train/pos_loss/pos5": 1.7231017351150513, "train/pos_loss/pos6": 2.0308241844177246, "train/pos_loss/weight0": 2.089056968688965, "train/pos_loss/weight1": 1.6270474195480347, "train/pos_loss/weight2": 1.2179701328277588, "train/pos_loss/weight3": 0.8773391246795654, "train/pos_loss/weight4": 0.5815668106079102, "train/pos_loss/weight5": 0.3703864514827728, "train/pos_loss/weight6": 0.2366325855255127, "train/pos_weight_progress": 0.19194312796208532, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.554513692855835 }, { "epoch": 1.919431279620853, "step": 810, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 14.54581356048584, "train/clean_loss_weight": 2.909162759780884, "train/combined_loss": 1.133550763130188, "train/pos_loss/pos0": 0.17744840681552887, "train/pos_loss/pos1": 0.4666176438331604, "train/pos_loss/pos2": 0.8038369417190552, "train/pos_loss/pos3": 1.8234786987304688, "train/pos_loss/pos4": 2.340876817703247, "train/pos_loss/pos5": 2.81196928024292, "train/pos_loss/pos6": 3.3597469329833984, "train/pos_loss/weight0": 2.142967939376831, "train/pos_loss/weight1": 1.6717818975448608, "train/pos_loss/weight2": 1.247348666191101, "train/pos_loss/weight3": 0.8573578000068665, "train/pos_loss/weight4": 0.5549241900444031, "train/pos_loss/weight5": 0.3237431049346924, "train/pos_loss/weight6": 0.20187638700008392, "train/pos_weight_progress": 0.19194312796208532, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9446256160736084 }, { "epoch": 1.919431279620853, "step": 810, "train/alpha": 0.2, "train/clean_ce_loss": 0.10791015625, "train/clean_loss_scale": 7.144479751586914, "train/clean_loss_weight": 1.4288959503173828, "train/combined_loss": 0.9251543283462524, "train/pos_loss/pos0": 0.29241305589675903, "train/pos_loss/pos1": 0.6168901920318604, "train/pos_loss/pos2": 1.1436574459075928, "train/pos_loss/pos3": 1.6063023805618286, "train/pos_loss/pos4": 2.153907537460327, "train/pos_loss/pos5": 2.562147855758667, "train/pos_loss/pos6": 2.9623165130615234, "train/pos_loss/weight0": 2.7687764167785645, "train/pos_loss/weight1": 1.9356021881103516, "train/pos_loss/weight2": 1.2279539108276367, "train/pos_loss/weight3": 0.6697288751602173, "train/pos_loss/weight4": 0.2839372754096985, "train/pos_loss/weight5": 0.08430549502372742, "train/pos_loss/weight6": 0.02969573251903057, "train/pos_weight_progress": 0.19194312796208532, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7709619402885437 }, { "epoch": 1.919431279620853, "step": 810, "train/alpha": 0.2, "train/clean_ce_loss": 0.0966796875, "train/clean_loss_scale": 7.558125019073486, "train/clean_loss_weight": 1.511625051498413, "train/combined_loss": 0.8768606185913086, "train/pos_loss/pos0": 0.3608783185482025, "train/pos_loss/pos1": 0.6078334450721741, "train/pos_loss/pos2": 1.0782606601715088, "train/pos_loss/pos3": 1.3799331188201904, "train/pos_loss/pos4": 1.8164442777633667, "train/pos_loss/pos5": 2.2271246910095215, "train/pos_loss/pos6": 2.576843023300171, "train/pos_loss/weight0": 2.8510842323303223, "train/pos_loss/weight1": 1.9460333585739136, "train/pos_loss/weight2": 1.1996804475784302, "train/pos_loss/weight3": 0.6234641075134277, "train/pos_loss/weight4": 0.25705036520957947, "train/pos_loss/weight5": 0.08911092579364777, "train/pos_loss/weight6": 0.03357645869255066, "train/pos_weight_progress": 0.19194312796208532, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7307171821594238 }, { "epoch": 1.9218009478672986, "step": 811, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 8.718367576599121, "train/clean_loss_weight": 1.74367356300354, "train/combined_loss": 0.6794196367263794, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.19218009478672987, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5661830306053162 }, { "epoch": 1.9218009478672986, "step": 811, "train/alpha": 0.2, "train/clean_ce_loss": 0.05517578125, "train/clean_loss_scale": 20.617321014404297, "train/clean_loss_weight": 4.123464107513428, "train/combined_loss": 1.365092158317566, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.19218009478672987, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1375768184661865 }, { "epoch": 1.9218009478672986, "step": 811, "train/alpha": 0.2, "train/clean_ce_loss": 0.10693359375, "train/clean_loss_scale": 7.340390205383301, "train/clean_loss_weight": 1.4680780172348022, "train/combined_loss": 0.9419211149215698, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.19218009478672987, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7849342823028564 }, { "epoch": 1.9218009478672986, "step": 811, "train/alpha": 0.2, "train/clean_ce_loss": 0.056396484375, "train/clean_loss_scale": 9.523368835449219, "train/clean_loss_weight": 1.9046738147735596, "train/combined_loss": 0.6445014476776123, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.19218009478672987, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5370845198631287 }, { "epoch": 1.9241706161137442, "step": 812, "train/alpha": 0.2, "train/clean_ce_loss": 0.0267333984375, "train/clean_loss_scale": 19.107370376586914, "train/clean_loss_weight": 3.821474075317383, "train/combined_loss": 0.6129659414291382, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.1924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5108049511909485 }, { "epoch": 1.9241706161137442, "step": 812, "train/alpha": 0.2, "train/clean_ce_loss": 0.0108642578125, "train/clean_loss_scale": 16.350826263427734, "train/clean_loss_weight": 3.270165205001831, "train/combined_loss": 0.21316750347614288, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.1924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.17763958871364594 }, { "epoch": 1.9241706161137442, "step": 812, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 8.090761184692383, "train/clean_loss_weight": 1.6181522607803345, "train/combined_loss": 0.7063614130020142, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.1924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5886344909667969 }, { "epoch": 1.9241706161137442, "step": 812, "train/alpha": 0.2, "train/clean_ce_loss": 0.1064453125, "train/clean_loss_scale": 7.809844017028809, "train/clean_loss_weight": 1.5619688034057617, "train/combined_loss": 0.9975855350494385, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.1924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8313212990760803 }, { "epoch": 1.9265402843601895, "step": 813, "train/alpha": 0.2, "train/clean_ce_loss": 0.045166015625, "train/clean_loss_scale": 15.999618530273438, "train/clean_loss_weight": 3.1999237537384033, "train/combined_loss": 0.8671668171882629, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.19265402843601895, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7226390242576599 }, { "epoch": 1.9265402843601895, "step": 813, "train/alpha": 0.2, "train/clean_ce_loss": 0.0830078125, "train/clean_loss_scale": 8.82579231262207, "train/clean_loss_weight": 1.7651585340499878, "train/combined_loss": 0.8791316747665405, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.19265402843601895, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.732609748840332 }, { "epoch": 1.9265402843601895, "step": 813, "train/alpha": 0.2, "train/clean_ce_loss": 0.000484466552734375, "train/clean_loss_scale": 313.16436767578125, "train/clean_loss_weight": 62.63287353515625, "train/combined_loss": 0.18206119537353516, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.19265402843601895, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1517176628112793 }, { "epoch": 1.9265402843601895, "step": 813, "train/alpha": 0.2, "train/clean_ce_loss": 0.11279296875, "train/clean_loss_scale": 6.991610050201416, "train/clean_loss_weight": 1.3983219861984253, "train/combined_loss": 0.9463253021240234, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.19265402843601895, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7886044383049011 }, { "epoch": 1.9289099526066351, "step": 814, "train/alpha": 0.2, "train/clean_ce_loss": 0.005584716796875, "train/clean_loss_scale": 35.278282165527344, "train/clean_loss_weight": 7.055656433105469, "train/combined_loss": 0.2364230453968048, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.1928909952606635, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.19701920449733734 }, { "epoch": 1.9289099526066351, "step": 814, "train/alpha": 0.2, "train/clean_ce_loss": 0.0732421875, "train/clean_loss_scale": 15.092268943786621, "train/clean_loss_weight": 3.01845383644104, "train/combined_loss": 1.3264689445495605, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.1928909952606635, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1053907871246338 }, { "epoch": 1.9289099526066351, "step": 814, "train/alpha": 0.2, "train/clean_ce_loss": 0.020263671875, "train/clean_loss_scale": 19.560016632080078, "train/clean_loss_weight": 3.9120032787323, "train/combined_loss": 0.4756292998790741, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.1928909952606635, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3963577449321747 }, { "epoch": 1.9289099526066351, "step": 814, "train/alpha": 0.2, "train/clean_ce_loss": 0.1533203125, "train/clean_loss_scale": 7.002843379974365, "train/clean_loss_weight": 1.4005687236785889, "train/combined_loss": 1.2884137630462646, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.1928909952606635, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0736781358718872 }, { "epoch": 1.9312796208530805, "grad_norm": 14.579506874084473, "learning_rate": 2e-06, "loss": 3.3574, "step": 815 }, { "epoch": 1.9312796208530805, "step": 815, "train/alpha": 0.2, "train/clean_ce_loss": 0.058837890625, "train/clean_loss_scale": 9.555696487426758, "train/clean_loss_weight": 1.9111393690109253, "train/combined_loss": 0.6746844053268433, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.19312796208530805, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.562237024307251 }, { "epoch": 1.9312796208530805, "step": 815, "train/alpha": 0.2, "train/clean_ce_loss": 0.1474609375, "train/clean_loss_scale": 7.989620208740234, "train/clean_loss_weight": 1.5979241132736206, "train/combined_loss": 1.4137881994247437, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.19312796208530805, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.178156852722168 }, { "epoch": 1.9312796208530805, "step": 815, "train/alpha": 0.2, "train/clean_ce_loss": 0.11767578125, "train/clean_loss_scale": 11.69290542602539, "train/clean_loss_weight": 2.338581085205078, "train/combined_loss": 1.6511662006378174, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.19312796208530805, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.375971794128418 }, { "epoch": 1.9312796208530805, "step": 815, "train/alpha": 0.2, "train/clean_ce_loss": 0.0908203125, "train/clean_loss_scale": 8.175925254821777, "train/clean_loss_weight": 1.6351851224899292, "train/combined_loss": 0.8910480737686157, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.19312796208530805, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7425400614738464 }, { "epoch": 1.933649289099526, "step": 816, "train/alpha": 0.2, "train/clean_ce_loss": 0.12353515625, "train/clean_loss_scale": 8.14561653137207, "train/clean_loss_weight": 1.629123330116272, "train/combined_loss": 1.207524061203003, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.1933649289099526, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0062700510025024 }, { "epoch": 1.933649289099526, "step": 816, "train/alpha": 0.2, "train/clean_ce_loss": 0.11474609375, "train/clean_loss_scale": 8.40626335144043, "train/clean_loss_weight": 1.6812527179718018, "train/combined_loss": 1.1575031280517578, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.1933649289099526, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9645859003067017 }, { "epoch": 1.933649289099526, "step": 816, "train/alpha": 0.2, "train/clean_ce_loss": 0.16015625, "train/clean_loss_scale": 8.623008728027344, "train/clean_loss_weight": 1.7246017456054688, "train/combined_loss": 1.6572345495224, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.1933649289099526, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3810287714004517 }, { "epoch": 1.933649289099526, "step": 816, "train/alpha": 0.2, "train/clean_ce_loss": 0.044677734375, "train/clean_loss_scale": 9.998034477233887, "train/clean_loss_weight": 1.999606966972351, "train/combined_loss": 0.5360274314880371, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.1933649289099526, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.44668951630592346 }, { "epoch": 1.9360189573459716, "step": 817, "train/alpha": 0.2, "train/clean_ce_loss": 0.12353515625, "train/clean_loss_scale": 9.360093116760254, "train/clean_loss_weight": 1.8720186948776245, "train/combined_loss": 1.3875606060028076, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.19360189573459716, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1563005447387695 }, { "epoch": 1.9360189573459716, "step": 817, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 12.740724563598633, "train/clean_loss_weight": 2.548145055770874, "train/combined_loss": 1.134720802307129, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.19360189573459716, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9456006288528442 }, { "epoch": 1.9360189573459716, "step": 817, "train/alpha": 0.2, "train/clean_ce_loss": 0.08203125, "train/clean_loss_scale": 9.192103385925293, "train/clean_loss_weight": 1.8384207487106323, "train/combined_loss": 0.9048476219177246, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.19360189573459716, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7540397047996521 }, { "epoch": 1.9360189573459716, "step": 817, "train/alpha": 0.2, "train/clean_ce_loss": 0.03662109375, "train/clean_loss_scale": 6.024269104003906, "train/clean_loss_weight": 1.204853892326355, "train/combined_loss": 0.26473838090896606, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.19360189573459716, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.22061532735824585 }, { "epoch": 1.9383886255924172, "step": 818, "train/alpha": 0.2, "train/clean_ce_loss": 0.058837890625, "train/clean_loss_scale": 9.690719604492188, "train/clean_loss_weight": 1.9381439685821533, "train/combined_loss": 0.6842178106307983, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.1938388625592417, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5701814889907837 }, { "epoch": 1.9383886255924172, "step": 818, "train/alpha": 0.2, "train/clean_ce_loss": 0.06982421875, "train/clean_loss_scale": 8.569571495056152, "train/clean_loss_weight": 1.7139142751693726, "train/combined_loss": 0.7180363535881042, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.1938388625592417, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5983636379241943 }, { "epoch": 1.9383886255924172, "step": 818, "train/alpha": 0.2, "train/clean_ce_loss": 0.1044921875, "train/clean_loss_scale": 6.492254734039307, "train/clean_loss_weight": 1.2984509468078613, "train/combined_loss": 0.8140679001808167, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.1938388625592417, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6783899068832397 }, { "epoch": 1.9383886255924172, "step": 818, "train/alpha": 0.2, "train/clean_ce_loss": 0.09228515625, "train/clean_loss_scale": 5.035907745361328, "train/clean_loss_weight": 1.0071815252304077, "train/combined_loss": 0.5576874613761902, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.1938388625592417, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46473953127861023 }, { "epoch": 1.9407582938388626, "step": 819, "train/alpha": 0.2, "train/clean_ce_loss": 0.028564453125, "train/clean_loss_scale": 7.580234527587891, "train/clean_loss_weight": 1.5160468816757202, "train/combined_loss": 0.2598302960395813, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.19407582938388626, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.21652525663375854 }, { "epoch": 1.9407582938388626, "step": 819, "train/alpha": 0.2, "train/clean_ce_loss": 0.0201416015625, "train/clean_loss_scale": 12.573783874511719, "train/clean_loss_weight": 2.514756917953491, "train/combined_loss": 0.3039073646068573, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.19407582938388626, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2532561421394348 }, { "epoch": 1.9407582938388626, "step": 819, "train/alpha": 0.2, "train/clean_ce_loss": 0.0203857421875, "train/clean_loss_scale": 12.738421440124512, "train/clean_loss_weight": 2.54768443107605, "train/combined_loss": 0.3116186261177063, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.19407582938388626, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.25968217849731445 }, { "epoch": 1.9407582938388626, "step": 819, "train/alpha": 0.2, "train/clean_ce_loss": 0.08203125, "train/clean_loss_scale": 12.214346885681152, "train/clean_loss_weight": 2.4428694248199463, "train/combined_loss": 1.2023497819900513, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.19407582938388626, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0019581317901611 }, { "epoch": 1.943127962085308, "grad_norm": 14.699162483215332, "learning_rate": 2e-06, "loss": 3.4416, "step": 820 }, { "epoch": 1.943127962085308, "step": 820, "train/alpha": 0.2, "train/clean_ce_loss": 0.046142578125, "train/clean_loss_scale": 12.844759941101074, "train/clean_loss_weight": 2.5689520835876465, "train/combined_loss": 0.7112284302711487, "train/pos_loss/pos0": 0.25297045707702637, "train/pos_loss/pos1": 0.4852030873298645, "train/pos_loss/pos2": 0.7848260402679443, "train/pos_loss/pos3": 1.0665638446807861, "train/pos_loss/pos4": 1.461159110069275, "train/pos_loss/pos5": 1.8883304595947266, "train/pos_loss/pos6": 2.212275505065918, "train/pos_loss/weight0": 2.700227737426758, "train/pos_loss/weight1": 1.885547399520874, "train/pos_loss/weight2": 1.2101819515228271, "train/pos_loss/weight3": 0.6705144047737122, "train/pos_loss/weight4": 0.32779133319854736, "train/pos_loss/weight5": 0.14208984375, "train/pos_loss/weight6": 0.06364721059799194, "train/pos_weight_progress": 0.1943127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5926903486251831 }, { "epoch": 1.943127962085308, "step": 820, "train/alpha": 0.2, "train/clean_ce_loss": 0.06298828125, "train/clean_loss_scale": 22.57378578186035, "train/clean_loss_weight": 4.51475715637207, "train/combined_loss": 1.7062606811523438, "train/pos_loss/pos0": 0.47458624839782715, "train/pos_loss/pos1": 1.0009952783584595, "train/pos_loss/pos2": 1.7951773405075073, "train/pos_loss/pos3": 2.51499342918396, "train/pos_loss/pos4": 3.3726558685302734, "train/pos_loss/pos5": 3.8107166290283203, "train/pos_loss/pos6": 4.1493611335754395, "train/pos_loss/weight0": 2.3915953636169434, "train/pos_loss/weight1": 1.8248834609985352, "train/pos_loss/weight2": 1.2315430641174316, "train/pos_loss/weight3": 0.7584959268569946, "train/pos_loss/weight4": 0.4361259639263153, "train/pos_loss/weight5": 0.2319926768541336, "train/pos_loss/weight6": 0.12536343932151794, "train/pos_weight_progress": 0.1943127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4218839406967163 }, { "epoch": 1.943127962085308, "step": 820, "train/alpha": 0.2, "train/clean_ce_loss": 0.09326171875, "train/clean_loss_scale": 9.565752983093262, "train/clean_loss_weight": 1.913150668144226, "train/combined_loss": 1.070542335510254, "train/pos_loss/pos0": 0.2797723710536957, "train/pos_loss/pos1": 0.8744039535522461, "train/pos_loss/pos2": 1.8545868396759033, "train/pos_loss/pos3": 2.821162700653076, "train/pos_loss/pos4": 3.3216826915740967, "train/pos_loss/pos5": 3.5524919033050537, "train/pos_loss/pos6": 4.242243766784668, "train/pos_loss/weight0": 3.353069305419922, "train/pos_loss/weight1": 2.0399396419525146, "train/pos_loss/weight2": 1.0932393074035645, "train/pos_loss/weight3": 0.39605364203453064, "train/pos_loss/weight4": 0.10139387845993042, "train/pos_loss/weight5": 0.013808494433760643, "train/pos_loss/weight6": 0.002495828317478299, "train/pos_weight_progress": 0.1943127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8921185731887817 }, { "epoch": 1.943127962085308, "step": 820, "train/alpha": 0.2, "train/clean_ce_loss": 0.0250244140625, "train/clean_loss_scale": 8.715343475341797, "train/clean_loss_weight": 1.7430686950683594, "train/combined_loss": 0.2617156207561493, "train/pos_loss/pos0": 0.02872385084629059, "train/pos_loss/pos1": 0.14535364508628845, "train/pos_loss/pos2": 0.41849860548973083, "train/pos_loss/pos3": 1.0120326280593872, "train/pos_loss/pos4": 1.5340611934661865, "train/pos_loss/pos5": 2.0716607570648193, "train/pos_loss/pos6": 2.6206321716308594, "train/pos_loss/weight0": 3.167816162109375, "train/pos_loss/weight1": 2.1296045780181885, "train/pos_loss/weight2": 1.1417547464370728, "train/pos_loss/weight3": 0.43094009160995483, "train/pos_loss/weight4": 0.1081688329577446, "train/pos_loss/weight5": 0.017938826233148575, "train/pos_loss/weight6": 0.0037764995358884335, "train/pos_weight_progress": 0.1943127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.21809636056423187 }, { "epoch": 1.9454976303317535, "step": 821, "train/alpha": 0.2, "train/clean_ce_loss": 0.10888671875, "train/clean_loss_scale": 8.746420860290527, "train/clean_loss_weight": 1.7492841482162476, "train/combined_loss": 1.1428428888320923, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.19454976303317537, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9523690938949585 }, { "epoch": 1.9454976303317535, "step": 821, "train/alpha": 0.2, "train/clean_ce_loss": 0.056640625, "train/clean_loss_scale": 11.057055473327637, "train/clean_loss_weight": 2.211411237716675, "train/combined_loss": 0.7515342235565186, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.19454976303317537, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6262785196304321 }, { "epoch": 1.9454976303317535, "step": 821, "train/alpha": 0.2, "train/clean_ce_loss": 0.0302734375, "train/clean_loss_scale": 13.834500312805176, "train/clean_loss_weight": 2.766900062561035, "train/combined_loss": 0.5025814771652222, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.19454976303317537, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4188178777694702 }, { "epoch": 1.9454976303317535, "step": 821, "train/alpha": 0.2, "train/clean_ce_loss": 0.10009765625, "train/clean_loss_scale": 8.787020683288574, "train/clean_loss_weight": 1.7574042081832886, "train/combined_loss": 1.0554722547531128, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.19454976303317537, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8795601725578308 }, { "epoch": 1.947867298578199, "step": 822, "train/alpha": 0.2, "train/clean_ce_loss": 0.05712890625, "train/clean_loss_scale": 12.443974494934082, "train/clean_loss_weight": 2.488795042037964, "train/combined_loss": 0.8530927896499634, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.19478672985781992, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7109106779098511 }, { "epoch": 1.947867298578199, "step": 822, "train/alpha": 0.2, "train/clean_ce_loss": 0.09716796875, "train/clean_loss_scale": 8.402597427368164, "train/clean_loss_weight": 1.680519461631775, "train/combined_loss": 0.9797559976577759, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.19478672985781992, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8164633512496948 }, { "epoch": 1.947867298578199, "step": 822, "train/alpha": 0.2, "train/clean_ce_loss": 0.0869140625, "train/clean_loss_scale": 13.499622344970703, "train/clean_loss_weight": 2.6999244689941406, "train/combined_loss": 1.407968521118164, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.19478672985781992, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1733070611953735 }, { "epoch": 1.947867298578199, "step": 822, "train/alpha": 0.2, "train/clean_ce_loss": 0.0673828125, "train/clean_loss_scale": 9.84819221496582, "train/clean_loss_weight": 1.969638466835022, "train/combined_loss": 0.7963186502456665, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.19478672985781992, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.663598895072937 }, { "epoch": 1.9502369668246446, "step": 823, "train/alpha": 0.2, "train/clean_ce_loss": 0.020263671875, "train/clean_loss_scale": 15.664970397949219, "train/clean_loss_weight": 3.1329941749572754, "train/combined_loss": 0.3809157609939575, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.19502369668246444, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3174298107624054 }, { "epoch": 1.9502369668246446, "step": 823, "train/alpha": 0.2, "train/clean_ce_loss": 0.01092529296875, "train/clean_loss_scale": 33.51787567138672, "train/clean_loss_weight": 6.703575134277344, "train/combined_loss": 0.4394311308860779, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.19502369668246444, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.36619260907173157 }, { "epoch": 1.9502369668246446, "step": 823, "train/alpha": 0.2, "train/clean_ce_loss": 0.015625, "train/clean_loss_scale": 36.934818267822266, "train/clean_loss_weight": 7.386963844299316, "train/combined_loss": 0.6925278306007385, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.19502369668246444, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5771065354347229 }, { "epoch": 1.9502369668246446, "step": 823, "train/alpha": 0.2, "train/clean_ce_loss": 0.0908203125, "train/clean_loss_scale": 10.990021705627441, "train/clean_loss_weight": 2.1980044841766357, "train/combined_loss": 1.1977406740188599, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.19502369668246444, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.998117208480835 }, { "epoch": 1.95260663507109, "step": 824, "train/alpha": 0.2, "train/clean_ce_loss": 0.07568359375, "train/clean_loss_scale": 9.360517501831055, "train/clean_loss_weight": 1.8721035718917847, "train/combined_loss": 0.8501251339912415, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.195260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7084376215934753 }, { "epoch": 1.95260663507109, "step": 824, "train/alpha": 0.2, "train/clean_ce_loss": 0.01318359375, "train/clean_loss_scale": 26.552413940429688, "train/clean_loss_weight": 5.310482978820801, "train/combined_loss": 0.4200674891471863, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.195260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3500562310218811 }, { "epoch": 1.95260663507109, "step": 824, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 17.72508430480957, "train/clean_loss_weight": 3.5450170040130615, "train/combined_loss": 1.3916960954666138, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.195260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.159746766090393 }, { "epoch": 1.95260663507109, "step": 824, "train/alpha": 0.2, "train/clean_ce_loss": 0.1025390625, "train/clean_loss_scale": 12.892833709716797, "train/clean_loss_weight": 2.578566789627075, "train/combined_loss": 1.5864229202270508, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.195260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.322019100189209 }, { "epoch": 1.9549763033175356, "grad_norm": 12.785694122314453, "learning_rate": 2e-06, "loss": 3.3355, "step": 825 }, { "epoch": 1.9549763033175356, "step": 825, "train/alpha": 0.2, "train/clean_ce_loss": 0.005218505859375, "train/clean_loss_scale": 68.59637451171875, "train/clean_loss_weight": 13.71927547454834, "train/combined_loss": 0.4295647144317627, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.19549763033175355, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.35797059535980225 }, { "epoch": 1.9549763033175356, "step": 825, "train/alpha": 0.2, "train/clean_ce_loss": 0.0263671875, "train/clean_loss_scale": 13.041572570800781, "train/clean_loss_weight": 2.6083145141601562, "train/combined_loss": 0.41264352202415466, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.19549763033175355, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3438695967197418 }, { "epoch": 1.9549763033175356, "step": 825, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 8.904865264892578, "train/clean_loss_weight": 1.7809730768203735, "train/combined_loss": 0.709606409072876, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.19549763033175355, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5913386940956116 }, { "epoch": 1.9549763033175356, "step": 825, "train/alpha": 0.2, "train/clean_ce_loss": 0.044677734375, "train/clean_loss_scale": 11.110852241516113, "train/clean_loss_weight": 2.22217059135437, "train/combined_loss": 0.5956892371177673, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.19549763033175355, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.496407687664032 }, { "epoch": 1.957345971563981, "step": 826, "train/alpha": 0.2, "train/clean_ce_loss": 0.043701171875, "train/clean_loss_scale": 12.408448219299316, "train/clean_loss_weight": 2.481689691543579, "train/combined_loss": 0.6507164835929871, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.1957345971563981, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5422637462615967 }, { "epoch": 1.957345971563981, "step": 826, "train/alpha": 0.2, "train/clean_ce_loss": 0.146484375, "train/clean_loss_scale": 9.154556274414062, "train/clean_loss_weight": 1.8309112787246704, "train/combined_loss": 1.6091994047164917, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.1957345971563981, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3409994840621948 }, { "epoch": 1.957345971563981, "step": 826, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 17.095680236816406, "train/clean_loss_weight": 3.4191360473632812, "train/combined_loss": 1.4925329685211182, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.1957345971563981, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2437775135040283 }, { "epoch": 1.957345971563981, "step": 826, "train/alpha": 0.2, "train/clean_ce_loss": 0.062255859375, "train/clean_loss_scale": 10.228952407836914, "train/clean_loss_weight": 2.045790433883667, "train/combined_loss": 0.7641746401786804, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.1957345971563981, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6368122100830078 }, { "epoch": 1.9597156398104265, "step": 827, "train/alpha": 0.2, "train/clean_ce_loss": 0.06884765625, "train/clean_loss_scale": 10.876410484313965, "train/clean_loss_weight": 2.1752822399139404, "train/combined_loss": 0.898578405380249, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.19597156398104265, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7488153576850891 }, { "epoch": 1.9597156398104265, "step": 827, "train/alpha": 0.2, "train/clean_ce_loss": 0.1435546875, "train/clean_loss_scale": 9.453240394592285, "train/clean_loss_weight": 1.8906481266021729, "train/combined_loss": 1.62846839427948, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.19597156398104265, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.357056975364685 }, { "epoch": 1.9597156398104265, "step": 827, "train/alpha": 0.2, "train/clean_ce_loss": 0.00142669677734375, "train/clean_loss_scale": 144.87916564941406, "train/clean_loss_weight": 28.975833892822266, "train/combined_loss": 0.2480383664369583, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.19597156398104265, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.20669864118099213 }, { "epoch": 1.9597156398104265, "step": 827, "train/alpha": 0.2, "train/clean_ce_loss": 0.0498046875, "train/clean_loss_scale": 7.157419204711914, "train/clean_loss_weight": 1.4314838647842407, "train/combined_loss": 0.42776763439178467, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.19597156398104265, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.35647302865982056 }, { "epoch": 1.962085308056872, "step": 828, "train/alpha": 0.2, "train/clean_ce_loss": 0.005889892578125, "train/clean_loss_scale": 53.45491027832031, "train/clean_loss_weight": 10.6909818649292, "train/combined_loss": 0.3778124153614044, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.1962085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.31484368443489075 }, { "epoch": 1.962085308056872, "step": 828, "train/alpha": 0.2, "train/clean_ce_loss": 0.09228515625, "train/clean_loss_scale": 10.458608627319336, "train/clean_loss_weight": 2.091721773147583, "train/combined_loss": 1.158209204673767, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.1962085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9651743173599243 }, { "epoch": 1.962085308056872, "step": 828, "train/alpha": 0.2, "train/clean_ce_loss": 0.08642578125, "train/clean_loss_scale": 7.054239273071289, "train/clean_loss_weight": 1.4108479022979736, "train/combined_loss": 0.7316017746925354, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.1962085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6096681356430054 }, { "epoch": 1.962085308056872, "step": 828, "train/alpha": 0.2, "train/clean_ce_loss": 0.002593994140625, "train/clean_loss_scale": 126.86357116699219, "train/clean_loss_weight": 25.37271499633789, "train/combined_loss": 0.39490002393722534, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.1962085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3290833532810211 }, { "epoch": 1.9644549763033177, "step": 829, "train/alpha": 0.2, "train/clean_ce_loss": 0.07177734375, "train/clean_loss_scale": 11.641973495483398, "train/clean_loss_weight": 2.328394651412964, "train/combined_loss": 1.002755880355835, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.19644549763033176, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.835629940032959 }, { "epoch": 1.9644549763033177, "step": 829, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 10.211679458618164, "train/clean_loss_weight": 2.0423359870910645, "train/combined_loss": 0.9752951860427856, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.19644549763033176, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.812745988368988 }, { "epoch": 1.9644549763033177, "step": 829, "train/alpha": 0.2, "train/clean_ce_loss": 0.03662109375, "train/clean_loss_scale": 16.305280685424805, "train/clean_loss_weight": 3.2610561847686768, "train/combined_loss": 0.7165406346321106, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.19644549763033176, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5971171855926514 }, { "epoch": 1.9644549763033177, "step": 829, "train/alpha": 0.2, "train/clean_ce_loss": 0.0205078125, "train/clean_loss_scale": 24.886186599731445, "train/clean_loss_weight": 4.977237224578857, "train/combined_loss": 0.6124334931373596, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.19644549763033176, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5103612542152405 }, { "epoch": 1.966824644549763, "grad_norm": 15.591795921325684, "learning_rate": 2e-06, "loss": 3.1143, "step": 830 }, { "epoch": 1.966824644549763, "step": 830, "train/alpha": 0.2, "train/clean_ce_loss": 0.11962890625, "train/clean_loss_scale": 10.432629585266113, "train/clean_loss_weight": 2.0865259170532227, "train/combined_loss": 1.4976527690887451, "train/pos_loss/pos0": 0.48403292894363403, "train/pos_loss/pos1": 0.9900717735290527, "train/pos_loss/pos2": 1.5004066228866577, "train/pos_loss/pos3": 1.8954017162322998, "train/pos_loss/pos4": 2.3470821380615234, "train/pos_loss/pos5": 3.0000486373901367, "train/pos_loss/pos6": 3.394278049468994, "train/pos_loss/weight0": 2.265389919281006, "train/pos_loss/weight1": 1.7355642318725586, "train/pos_loss/weight2": 1.2442169189453125, "train/pos_loss/weight3": 0.8476042747497559, "train/pos_loss/weight4": 0.49403277039527893, "train/pos_loss/weight5": 0.26211950182914734, "train/pos_loss/weight6": 0.1510726511478424, "train/pos_weight_progress": 0.1966824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2480440139770508 }, { "epoch": 1.966824644549763, "step": 830, "train/alpha": 0.2, "train/clean_ce_loss": 0.1513671875, "train/clean_loss_scale": 7.331516265869141, "train/clean_loss_weight": 1.4663032293319702, "train/combined_loss": 1.3317011594772339, "train/pos_loss/pos0": 0.5494440197944641, "train/pos_loss/pos1": 1.1225286722183228, "train/pos_loss/pos2": 1.6709972620010376, "train/pos_loss/pos3": 2.2550745010375977, "train/pos_loss/pos4": 2.5700485706329346, "train/pos_loss/pos5": 3.0336461067199707, "train/pos_loss/pos6": 3.4573214054107666, "train/pos_loss/weight0": 3.045147180557251, "train/pos_loss/weight1": 2.022584915161133, "train/pos_loss/weight2": 1.1271698474884033, "train/pos_loss/weight3": 0.5265575051307678, "train/pos_loss/weight4": 0.20193065702915192, "train/pos_loss/weight5": 0.058337222784757614, "train/pos_loss/weight6": 0.01827298477292061, "train/pos_weight_progress": 0.1966824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1097509860992432 }, { "epoch": 1.966824644549763, "step": 830, "train/alpha": 0.2, "train/clean_ce_loss": 0.06103515625, "train/clean_loss_scale": 7.388823986053467, "train/clean_loss_weight": 1.4777648448944092, "train/combined_loss": 0.5411736369132996, "train/pos_loss/pos0": 0.20753906667232513, "train/pos_loss/pos1": 0.459551066160202, "train/pos_loss/pos2": 0.6946340799331665, "train/pos_loss/pos3": 0.979401707649231, "train/pos_loss/pos4": 1.4097265005111694, "train/pos_loss/pos5": 1.852040410041809, "train/pos_loss/pos6": 2.2389652729034424, "train/pos_loss/weight0": 3.184197425842285, "train/pos_loss/weight1": 2.0375924110412598, "train/pos_loss/weight2": 1.0907015800476074, "train/pos_loss/weight3": 0.4601789712905884, "train/pos_loss/weight4": 0.1660289466381073, "train/pos_loss/weight5": 0.04703868553042412, "train/pos_loss/weight6": 0.014261598698794842, "train/pos_weight_progress": 0.1966824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45097804069519043 }, { "epoch": 1.966824644549763, "step": 830, "train/alpha": 0.2, "train/clean_ce_loss": 0.08447265625, "train/clean_loss_scale": 12.983505249023438, "train/clean_loss_weight": 2.596701145172119, "train/combined_loss": 1.3161014318466187, "train/pos_loss/pos0": 0.36823517084121704, "train/pos_loss/pos1": 0.7054857015609741, "train/pos_loss/pos2": 1.2509171962738037, "train/pos_loss/pos3": 2.2115964889526367, "train/pos_loss/pos4": 2.9145421981811523, "train/pos_loss/pos5": 2.7726147174835205, "train/pos_loss/pos6": 3.287804365158081, "train/pos_loss/weight0": 2.4168701171875, "train/pos_loss/weight1": 1.8010244369506836, "train/pos_loss/weight2": 1.2348082065582275, "train/pos_loss/weight3": 0.7675730586051941, "train/pos_loss/weight4": 0.4326035976409912, "train/pos_loss/weight5": 0.2243320494890213, "train/pos_loss/weight6": 0.12278838455677032, "train/pos_weight_progress": 0.1966824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0967512130737305 }, { "epoch": 1.9691943127962084, "step": 831, "train/alpha": 0.2, "train/clean_ce_loss": 0.173828125, "train/clean_loss_scale": 5.990509986877441, "train/clean_loss_weight": 1.1981019973754883, "train/combined_loss": 1.2495830059051514, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.19691943127962086, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0413191318511963 }, { "epoch": 1.9691943127962084, "step": 831, "train/alpha": 0.2, "train/clean_ce_loss": 0.146484375, "train/clean_loss_scale": 8.065320014953613, "train/clean_loss_weight": 1.6130640506744385, "train/combined_loss": 1.4177320003509521, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.19691943127962086, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1814433336257935 }, { "epoch": 1.9691943127962084, "step": 831, "train/alpha": 0.2, "train/clean_ce_loss": 0.0157470703125, "train/clean_loss_scale": 11.833022117614746, "train/clean_loss_weight": 2.3666045665740967, "train/combined_loss": 0.2236025184392929, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.19691943127962086, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.18633542954921722 }, { "epoch": 1.9691943127962084, "step": 831, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 8.627753257751465, "train/clean_loss_weight": 1.725550651550293, "train/combined_loss": 0.7380773425102234, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.19691943127962086, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6150644421577454 }, { "epoch": 1.971563981042654, "step": 832, "train/alpha": 0.2, "train/clean_ce_loss": 0.09765625, "train/clean_loss_scale": 6.616847038269043, "train/clean_loss_weight": 1.3233693838119507, "train/combined_loss": 0.7754117250442505, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.1971563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6461764574050903 }, { "epoch": 1.971563981042654, "step": 832, "train/alpha": 0.2, "train/clean_ce_loss": 0.10400390625, "train/clean_loss_scale": 6.000021934509277, "train/clean_loss_weight": 1.2000044584274292, "train/combined_loss": 0.7488308548927307, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.1971563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6240257024765015 }, { "epoch": 1.971563981042654, "step": 832, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 8.121275901794434, "train/clean_loss_weight": 1.6242551803588867, "train/combined_loss": 0.6899911761283875, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.1971563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5749926567077637 }, { "epoch": 1.971563981042654, "step": 832, "train/alpha": 0.2, "train/clean_ce_loss": 0.045166015625, "train/clean_loss_scale": 12.153883934020996, "train/clean_loss_weight": 2.430776834487915, "train/combined_loss": 0.6587309837341309, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.1971563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.548942506313324 }, { "epoch": 1.9739336492890995, "step": 833, "train/alpha": 0.2, "train/clean_ce_loss": 0.10400390625, "train/clean_loss_scale": 5.077585220336914, "train/clean_loss_weight": 1.0155171155929565, "train/combined_loss": 0.6337064504623413, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.19739336492890996, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5280886888504028 }, { "epoch": 1.9739336492890995, "step": 833, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 5.988344192504883, "train/clean_loss_weight": 1.1976689100265503, "train/combined_loss": 0.5122841000556946, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.19739336492890996, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4269034266471863 }, { "epoch": 1.9739336492890995, "step": 833, "train/alpha": 0.2, "train/clean_ce_loss": 0.1376953125, "train/clean_loss_scale": 7.983491897583008, "train/clean_loss_weight": 1.5966984033584595, "train/combined_loss": 1.3191473484039307, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.19739336492890996, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0992894172668457 }, { "epoch": 1.9739336492890995, "step": 833, "train/alpha": 0.2, "train/clean_ce_loss": 0.08251953125, "train/clean_loss_scale": 8.300783157348633, "train/clean_loss_weight": 1.6601566076278687, "train/combined_loss": 0.821972131729126, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.19739336492890996, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6849767565727234 }, { "epoch": 1.9763033175355451, "step": 834, "train/alpha": 0.2, "train/clean_ce_loss": 0.10400390625, "train/clean_loss_scale": 7.4806108474731445, "train/clean_loss_weight": 1.4961222410202026, "train/combined_loss": 0.9336153268814087, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.1976303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.778012752532959 }, { "epoch": 1.9763033175355451, "step": 834, "train/alpha": 0.2, "train/clean_ce_loss": 0.03662109375, "train/clean_loss_scale": 8.572061538696289, "train/clean_loss_weight": 1.7144123315811157, "train/combined_loss": 0.3767019212245941, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.1976303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.31391826272010803 }, { "epoch": 1.9763033175355451, "step": 834, "train/alpha": 0.2, "train/clean_ce_loss": 0.016845703125, "train/clean_loss_scale": 6.037637233734131, "train/clean_loss_weight": 1.2075275182724, "train/combined_loss": 0.12204989045858383, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.1976303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.10170824080705643 }, { "epoch": 1.9763033175355451, "step": 834, "train/alpha": 0.2, "train/clean_ce_loss": 0.1064453125, "train/clean_loss_scale": 6.124016761779785, "train/clean_loss_weight": 1.2248033285140991, "train/combined_loss": 0.7822474241256714, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.1976303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6518728733062744 }, { "epoch": 1.9786729857819905, "grad_norm": 19.268199920654297, "learning_rate": 2e-06, "loss": 3.2651, "step": 835 }, { "epoch": 1.9786729857819905, "step": 835, "train/alpha": 0.2, "train/clean_ce_loss": 0.003509521484375, "train/clean_loss_scale": 29.49025535583496, "train/clean_loss_weight": 5.8980512619018555, "train/combined_loss": 0.12419602274894714, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.19786729857819904, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.10349668562412262 }, { "epoch": 1.9786729857819905, "step": 835, "train/alpha": 0.2, "train/clean_ce_loss": 0.05322265625, "train/clean_loss_scale": 5.635696887969971, "train/clean_loss_weight": 1.1271394491195679, "train/combined_loss": 0.3599361181259155, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.19786729857819904, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.29994675517082214 }, { "epoch": 1.9786729857819905, "step": 835, "train/alpha": 0.2, "train/clean_ce_loss": 0.140625, "train/clean_loss_scale": 7.766322612762451, "train/clean_loss_weight": 1.5532644987106323, "train/combined_loss": 1.3105669021606445, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.19786729857819904, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0921391248703003 }, { "epoch": 1.9786729857819905, "step": 835, "train/alpha": 0.2, "train/clean_ce_loss": 0.038330078125, "train/clean_loss_scale": 7.482827663421631, "train/clean_loss_weight": 1.496565580368042, "train/combined_loss": 0.3441808521747589, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.19786729857819904, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.28681737184524536 }, { "epoch": 1.981042654028436, "step": 836, "train/alpha": 0.2, "train/clean_ce_loss": 0.09130859375, "train/clean_loss_scale": 6.679498672485352, "train/clean_loss_weight": 1.3358997106552124, "train/combined_loss": 0.7318747639656067, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.1981042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.609895646572113 }, { "epoch": 1.981042654028436, "step": 836, "train/alpha": 0.2, "train/clean_ce_loss": 0.061767578125, "train/clean_loss_scale": 8.592804908752441, "train/clean_loss_weight": 1.718561053276062, "train/combined_loss": 0.6369081139564514, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.1981042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5307567715644836 }, { "epoch": 1.981042654028436, "step": 836, "train/alpha": 0.2, "train/clean_ce_loss": 0.1015625, "train/clean_loss_scale": 6.340706825256348, "train/clean_loss_weight": 1.2681413888931274, "train/combined_loss": 0.7727736830711365, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.1981042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6439780592918396 }, { "epoch": 1.981042654028436, "step": 836, "train/alpha": 0.2, "train/clean_ce_loss": 0.1357421875, "train/clean_loss_scale": 6.757848739624023, "train/clean_loss_weight": 1.3515697717666626, "train/combined_loss": 1.10079026222229, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.1981042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9173251986503601 }, { "epoch": 1.9834123222748814, "step": 837, "train/alpha": 0.2, "train/clean_ce_loss": 0.0072021484375, "train/clean_loss_scale": 25.983308792114258, "train/clean_loss_weight": 5.196661949157715, "train/combined_loss": 0.22456277906894684, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.19834123222748815, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.18713565170764923 }, { "epoch": 1.9834123222748814, "step": 837, "train/alpha": 0.2, "train/clean_ce_loss": 0.0732421875, "train/clean_loss_scale": 11.4607515335083, "train/clean_loss_weight": 2.2921502590179443, "train/combined_loss": 1.007292628288269, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.19834123222748815, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8394105434417725 }, { "epoch": 1.9834123222748814, "step": 837, "train/alpha": 0.2, "train/clean_ce_loss": 0.11376953125, "train/clean_loss_scale": 6.834741592407227, "train/clean_loss_weight": 1.3669483661651611, "train/combined_loss": 0.9331023693084717, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.19834123222748815, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7775853276252747 }, { "epoch": 1.9834123222748814, "step": 837, "train/alpha": 0.2, "train/clean_ce_loss": 0.10595703125, "train/clean_loss_scale": 7.4850263595581055, "train/clean_loss_weight": 1.4970053434371948, "train/combined_loss": 0.9517093896865845, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.19834123222748815, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7930911779403687 }, { "epoch": 1.985781990521327, "step": 838, "train/alpha": 0.2, "train/clean_ce_loss": 0.00787353515625, "train/clean_loss_scale": 16.92961311340332, "train/clean_loss_weight": 3.38592267036438, "train/combined_loss": 0.15995508432388306, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.1985781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1332959085702896 }, { "epoch": 1.985781990521327, "step": 838, "train/alpha": 0.2, "train/clean_ce_loss": 0.0026397705078125, "train/clean_loss_scale": 66.25389099121094, "train/clean_loss_weight": 13.250778198242188, "train/combined_loss": 0.20987407863140106, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.1985781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.17489506304264069 }, { "epoch": 1.985781990521327, "step": 838, "train/alpha": 0.2, "train/clean_ce_loss": 0.01025390625, "train/clean_loss_scale": 25.73355484008789, "train/clean_loss_weight": 5.1467108726501465, "train/combined_loss": 0.3166433572769165, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.1985781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2638694643974304 }, { "epoch": 1.985781990521327, "step": 838, "train/alpha": 0.2, "train/clean_ce_loss": 0.01263427734375, "train/clean_loss_scale": 8.89462947845459, "train/clean_loss_weight": 1.778925895690918, "train/combined_loss": 0.13485264778137207, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.1985781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.11237721145153046 }, { "epoch": 1.9881516587677726, "step": 839, "train/alpha": 0.2, "train/clean_ce_loss": 0.12060546875, "train/clean_loss_scale": 7.871638298034668, "train/clean_loss_weight": 1.5743277072906494, "train/combined_loss": 1.1392351388931274, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.19881516587677725, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9493626356124878 }, { "epoch": 1.9881516587677726, "step": 839, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 10.37373161315918, "train/clean_loss_weight": 2.0747463703155518, "train/combined_loss": 0.9846940636634827, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.19881516587677725, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.820578396320343 }, { "epoch": 1.9881516587677726, "step": 839, "train/alpha": 0.2, "train/clean_ce_loss": 0.1640625, "train/clean_loss_scale": 6.705102920532227, "train/clean_loss_weight": 1.3410205841064453, "train/combined_loss": 1.3200671672821045, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.19881516587677725, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1000559329986572 }, { "epoch": 1.9881516587677726, "step": 839, "train/alpha": 0.2, "train/clean_ce_loss": 0.0281982421875, "train/clean_loss_scale": 6.077354907989502, "train/clean_loss_weight": 1.2154710292816162, "train/combined_loss": 0.2056448757648468, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.19881516587677725, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.171370729804039 }, { "epoch": 1.9905213270142181, "grad_norm": 13.729656219482422, "learning_rate": 2e-06, "loss": 3.1409, "step": 840 }, { "epoch": 1.9905213270142181, "step": 840, "train/alpha": 0.2, "train/clean_ce_loss": 0.00982666015625, "train/clean_loss_scale": 8.126893997192383, "train/clean_loss_weight": 1.6253788471221924, "train/combined_loss": 0.09583227336406708, "train/pos_loss/pos0": 0.049302369356155396, "train/pos_loss/pos1": 0.0017608224879950285, "train/pos_loss/pos2": 0.04944775626063347, "train/pos_loss/pos3": 0.2680111527442932, "train/pos_loss/pos4": 0.6760112047195435, "train/pos_loss/pos5": 1.1566120386123657, "train/pos_loss/pos6": 1.7831268310546875, "train/pos_loss/weight0": 3.038649797439575, "train/pos_loss/weight1": 2.0266315937042236, "train/pos_loss/weight2": 1.1468578577041626, "train/pos_loss/weight3": 0.5421136021614075, "train/pos_loss/weight4": 0.18484435975551605, "train/pos_loss/weight5": 0.04650437459349632, "train/pos_loss/weight6": 0.014398494735360146, "train/pos_weight_progress": 0.1990521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.07986022531986237 }, { "epoch": 1.9905213270142181, "step": 840, "train/alpha": 0.2, "train/clean_ce_loss": 0.06103515625, "train/clean_loss_scale": 15.109533309936523, "train/clean_loss_weight": 3.021906614303589, "train/combined_loss": 1.106655240058899, "train/pos_loss/pos0": 0.2415958046913147, "train/pos_loss/pos1": 0.44094860553741455, "train/pos_loss/pos2": 0.8066177368164062, "train/pos_loss/pos3": 1.2473727464675903, "train/pos_loss/pos4": 1.5442317724227905, "train/pos_loss/pos5": 2.018824338912964, "train/pos_loss/pos6": 2.318896770477295, "train/pos_loss/weight0": 1.7597503662109375, "train/pos_loss/weight1": 1.424072504043579, "train/pos_loss/weight2": 1.1460977792739868, "train/pos_loss/weight3": 0.9176610708236694, "train/pos_loss/weight4": 0.7191630601882935, "train/pos_loss/weight5": 0.5711871981620789, "train/pos_loss/weight6": 0.46206796169281006, "train/pos_weight_progress": 0.1990521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9222127199172974 }, { "epoch": 1.9905213270142181, "step": 840, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 7.976034164428711, "train/clean_loss_weight": 1.5952068567276, "train/combined_loss": 0.887956976890564, "train/pos_loss/pos0": 0.22442032396793365, "train/pos_loss/pos1": 0.5529502630233765, "train/pos_loss/pos2": 0.8929682970046997, "train/pos_loss/pos3": 1.3645553588867188, "train/pos_loss/pos4": 1.8264439105987549, "train/pos_loss/pos5": 2.257108211517334, "train/pos_loss/pos6": 2.7068889141082764, "train/pos_loss/weight0": 2.480677366256714, "train/pos_loss/weight1": 1.8129942417144775, "train/pos_loss/weight2": 1.233838438987732, "train/pos_loss/weight3": 0.7607908248901367, "train/pos_loss/weight4": 0.40366601943969727, "train/pos_loss/weight5": 0.19854557514190674, "train/pos_loss/weight6": 0.10948705673217773, "train/pos_weight_progress": 0.1990521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7399641275405884 }, { "epoch": 1.9905213270142181, "step": 840, "train/alpha": 0.2, "train/clean_ce_loss": 0.031982421875, "train/clean_loss_scale": 10.977051734924316, "train/clean_loss_weight": 2.1954104900360107, "train/combined_loss": 0.4212872385978699, "train/pos_loss/pos0": 0.03555532917380333, "train/pos_loss/pos1": 0.17191646993160248, "train/pos_loss/pos2": 0.42076435685157776, "train/pos_loss/pos3": 0.8817583918571472, "train/pos_loss/pos4": 1.277858853340149, "train/pos_loss/pos5": 1.814570665359497, "train/pos_loss/pos6": 2.2233829498291016, "train/pos_loss/weight0": 2.582655906677246, "train/pos_loss/weight1": 1.8932298421859741, "train/pos_loss/weight2": 1.2324490547180176, "train/pos_loss/weight3": 0.7111613750457764, "train/pos_loss/weight4": 0.35061073303222656, "train/pos_loss/weight5": 0.1570650339126587, "train/pos_loss/weight6": 0.07282857596874237, "train/pos_weight_progress": 0.1990521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3510726988315582 }, { "epoch": 1.9928909952606635, "step": 841, "train/alpha": 0.2, "train/clean_ce_loss": 0.0771484375, "train/clean_loss_scale": 9.455633163452148, "train/clean_loss_weight": 1.8911266326904297, "train/combined_loss": 0.8753847479820251, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.19928909952606635, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7294872999191284 }, { "epoch": 1.9928909952606635, "step": 841, "train/alpha": 0.2, "train/clean_ce_loss": 0.00567626953125, "train/clean_loss_scale": 46.60673522949219, "train/clean_loss_weight": 9.3213472366333, "train/combined_loss": 0.31746286153793335, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.19928909952606635, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.26455238461494446 }, { "epoch": 1.9928909952606635, "step": 841, "train/alpha": 0.2, "train/clean_ce_loss": 0.08203125, "train/clean_loss_scale": 9.725419998168945, "train/clean_loss_weight": 1.9450839757919312, "train/combined_loss": 0.9573460817337036, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.19928909952606635, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7977883815765381 }, { "epoch": 1.9928909952606635, "step": 841, "train/alpha": 0.2, "train/clean_ce_loss": 0.10546875, "train/clean_loss_scale": 14.70887279510498, "train/clean_loss_weight": 2.941774606704712, "train/combined_loss": 1.8615916967391968, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.19928909952606635, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5513263940811157 }, { "epoch": 1.9952606635071088, "step": 842, "train/alpha": 0.2, "train/clean_ce_loss": 0.1455078125, "train/clean_loss_scale": 7.79463529586792, "train/clean_loss_weight": 1.558927059173584, "train/combined_loss": 1.3610163927078247, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.1995260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1341803073883057 }, { "epoch": 1.9952606635071088, "step": 842, "train/alpha": 0.2, "train/clean_ce_loss": 0.119140625, "train/clean_loss_scale": 10.135022163391113, "train/clean_loss_weight": 2.0270044803619385, "train/combined_loss": 1.4489914178848267, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.1995260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2074928283691406 }, { "epoch": 1.9952606635071088, "step": 842, "train/alpha": 0.2, "train/clean_ce_loss": 0.11376953125, "train/clean_loss_scale": 11.579586029052734, "train/clean_loss_weight": 2.3159172534942627, "train/combined_loss": 1.5808848142623901, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.1995260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.31740403175354 }, { "epoch": 1.9952606635071088, "step": 842, "train/alpha": 0.2, "train/clean_ce_loss": 0.076171875, "train/clean_loss_scale": 7.485926151275635, "train/clean_loss_weight": 1.497185230255127, "train/combined_loss": 0.6842604279518127, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.1995260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5702170133590698 }, { "epoch": 1.9976303317535544, "step": 843, "train/alpha": 0.2, "train/clean_ce_loss": 0.076171875, "train/clean_loss_scale": 10.406044006347656, "train/clean_loss_weight": 2.0812089443206787, "train/combined_loss": 0.9511774778366089, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.19976303317535546, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7926478981971741 }, { "epoch": 1.9976303317535544, "step": 843, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 9.48777961730957, "train/clean_loss_weight": 1.897555947303772, "train/combined_loss": 0.8283276557922363, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.19976303317535546, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6902730464935303 }, { "epoch": 1.9976303317535544, "step": 843, "train/alpha": 0.2, "train/clean_ce_loss": 0.123046875, "train/clean_loss_scale": 9.72691822052002, "train/clean_loss_weight": 1.9453836679458618, "train/combined_loss": 1.4362401962280273, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.19976303317535546, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1968668699264526 }, { "epoch": 1.9976303317535544, "step": 843, "train/alpha": 0.2, "train/clean_ce_loss": 0.0732421875, "train/clean_loss_scale": 7.78489351272583, "train/clean_loss_weight": 1.556978702545166, "train/combined_loss": 0.6842191219329834, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.19976303317535546, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5701826214790344 }, { "epoch": 2.0, "step": 844, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 12.528775215148926, "train/clean_loss_weight": 2.5057551860809326, "train/combined_loss": 1.2259602546691895, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.2, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.021633505821228 }, { "epoch": 2.0, "step": 844, "train/alpha": 0.2, "train/clean_ce_loss": 0.10986328125, "train/clean_loss_scale": 5.883951187133789, "train/clean_loss_weight": 1.1767902374267578, "train/combined_loss": 0.775716245174408, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.2, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6464301943778992 }, { "epoch": 2.0, "step": 844, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 15.12250804901123, "train/clean_loss_weight": 3.0245015621185303, "train/combined_loss": 1.1696314811706543, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.2, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9746928811073303 }, { "epoch": 2.0, "step": 844, "train/alpha": 0.2, "train/clean_ce_loss": 0.087890625, "train/clean_loss_scale": 7.757712364196777, "train/clean_loss_weight": 1.5515425205230713, "train/combined_loss": 0.818196177482605, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.2, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6818301677703857 }, { "epoch": 2.0023696682464456, "grad_norm": 25.793636322021484, "learning_rate": 2e-06, "loss": 3.2667, "step": 845 }, { "epoch": 2.0023696682464456, "step": 845, "train/alpha": 0.2, "train/clean_ce_loss": 0.006500244140625, "train/clean_loss_scale": 50.97453689575195, "train/clean_loss_weight": 10.194907188415527, "train/combined_loss": 0.39761632680892944, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.20023696682464456, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.33134692907333374 }, { "epoch": 2.0023696682464456, "step": 845, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 9.655694961547852, "train/clean_loss_weight": 1.9311389923095703, "train/combined_loss": 0.8203569054603577, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.20023696682464456, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6836307644844055 }, { "epoch": 2.0023696682464456, "step": 845, "train/alpha": 0.2, "train/clean_ce_loss": 0.02490234375, "train/clean_loss_scale": 22.22011375427246, "train/clean_loss_weight": 4.4440226554870605, "train/combined_loss": 0.6639994978904724, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.20023696682464456, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5533329248428345 }, { "epoch": 2.0023696682464456, "step": 845, "train/alpha": 0.2, "train/clean_ce_loss": 0.091796875, "train/clean_loss_scale": 15.930656433105469, "train/clean_loss_weight": 3.186131238937378, "train/combined_loss": 1.7548613548278809, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.20023696682464456, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4623844623565674 }, { "epoch": 2.004739336492891, "step": 846, "train/alpha": 0.2, "train/clean_ce_loss": 0.05908203125, "train/clean_loss_scale": 11.448787689208984, "train/clean_loss_weight": 2.289757490158081, "train/combined_loss": 0.8117011785507202, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.2004739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6764176487922668 }, { "epoch": 2.004739336492891, "step": 846, "train/alpha": 0.2, "train/clean_ce_loss": 0.047119140625, "train/clean_loss_scale": 14.173606872558594, "train/clean_loss_weight": 2.834721326828003, "train/combined_loss": 0.8014178276062012, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.2004739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6678481698036194 }, { "epoch": 2.004739336492891, "step": 846, "train/alpha": 0.2, "train/clean_ce_loss": 0.061767578125, "train/clean_loss_scale": 16.425769805908203, "train/clean_loss_weight": 3.285154104232788, "train/combined_loss": 1.2174960374832153, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.2004739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0145800113677979 }, { "epoch": 2.004739336492891, "step": 846, "train/alpha": 0.2, "train/clean_ce_loss": 0.000823974609375, "train/clean_loss_scale": 327.6506042480469, "train/clean_loss_weight": 65.53012084960938, "train/combined_loss": 0.3239709436893463, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.2004739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.26997578144073486 }, { "epoch": 2.0071090047393363, "step": 847, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 14.8186616897583, "train/clean_loss_weight": 2.9637324810028076, "train/combined_loss": 1.423980712890625, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.20071090047393364, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1866506338119507 }, { "epoch": 2.0071090047393363, "step": 847, "train/alpha": 0.2, "train/clean_ce_loss": 0.11474609375, "train/clean_loss_scale": 11.082064628601074, "train/clean_loss_weight": 2.2164130210876465, "train/combined_loss": 1.5259482860565186, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.20071090047393364, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2716236114501953 }, { "epoch": 2.0071090047393363, "step": 847, "train/alpha": 0.2, "train/clean_ce_loss": 0.03125, "train/clean_loss_scale": 5.324004650115967, "train/clean_loss_weight": 1.0648009777069092, "train/combined_loss": 0.19965016841888428, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.20071090047393364, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.16637514531612396 }, { "epoch": 2.0071090047393363, "step": 847, "train/alpha": 0.2, "train/clean_ce_loss": 0.05517578125, "train/clean_loss_scale": 12.347542762756348, "train/clean_loss_weight": 2.469508647918701, "train/combined_loss": 0.8175423741340637, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.20071090047393364, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6812853217124939 }, { "epoch": 2.009478672985782, "step": 848, "train/alpha": 0.2, "train/clean_ce_loss": 0.08837890625, "train/clean_loss_scale": 9.265369415283203, "train/clean_loss_weight": 1.8530739545822144, "train/combined_loss": 0.9826358556747437, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.2009478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8188632130622864 }, { "epoch": 2.009478672985782, "step": 848, "train/alpha": 0.2, "train/clean_ce_loss": 0.058349609375, "train/clean_loss_scale": 5.235748291015625, "train/clean_loss_weight": 1.047149658203125, "train/combined_loss": 0.36660465598106384, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.2009478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.30550387501716614 }, { "epoch": 2.009478672985782, "step": 848, "train/alpha": 0.2, "train/clean_ce_loss": 0.15625, "train/clean_loss_scale": 7.908743381500244, "train/clean_loss_weight": 1.5817487239837646, "train/combined_loss": 1.4828894138336182, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.2009478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.235741138458252 }, { "epoch": 2.009478672985782, "step": 848, "train/alpha": 0.2, "train/clean_ce_loss": 0.07373046875, "train/clean_loss_scale": 9.061039924621582, "train/clean_loss_weight": 1.8122080564498901, "train/combined_loss": 0.8016896843910217, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.2009478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6680747270584106 }, { "epoch": 2.0118483412322274, "step": 849, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 10.526735305786133, "train/clean_loss_weight": 2.105347156524658, "train/combined_loss": 0.8265131711959839, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.20118483412322274, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6887609958648682 }, { "epoch": 2.0118483412322274, "step": 849, "train/alpha": 0.2, "train/clean_ce_loss": 0.05810546875, "train/clean_loss_scale": 12.100906372070312, "train/clean_loss_weight": 2.4201812744140625, "train/combined_loss": 0.8437545895576477, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.20118483412322274, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7031288146972656 }, { "epoch": 2.0118483412322274, "step": 849, "train/alpha": 0.2, "train/clean_ce_loss": 0.0027008056640625, "train/clean_loss_scale": 110.22294616699219, "train/clean_loss_weight": 22.04458999633789, "train/combined_loss": 0.3572289049625397, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.20118483412322274, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.297690749168396 }, { "epoch": 2.0118483412322274, "step": 849, "train/alpha": 0.2, "train/clean_ce_loss": 0.0634765625, "train/clean_loss_scale": 18.690324783325195, "train/clean_loss_weight": 3.738065004348755, "train/combined_loss": 1.4236770868301392, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.20118483412322274, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1863975524902344 }, { "epoch": 2.014218009478673, "grad_norm": 16.04393768310547, "learning_rate": 2e-06, "loss": 2.8097, "step": 850 }, { "epoch": 2.014218009478673, "step": 850, "train/alpha": 0.2, "train/clean_ce_loss": 0.12060546875, "train/clean_loss_scale": 10.927624702453613, "train/clean_loss_weight": 2.1855249404907227, "train/combined_loss": 1.5815175771713257, "train/pos_loss/pos0": 0.45480749011039734, "train/pos_loss/pos1": 0.9994423389434814, "train/pos_loss/pos2": 1.5425353050231934, "train/pos_loss/pos3": 2.132540225982666, "train/pos_loss/pos4": 2.4623920917510986, "train/pos_loss/pos5": 2.9853789806365967, "train/pos_loss/pos6": 3.5265212059020996, "train/pos_loss/weight0": 2.205150842666626, "train/pos_loss/weight1": 1.7043875455856323, "train/pos_loss/weight2": 1.2506495714187622, "train/pos_loss/weight3": 0.8532713651657104, "train/pos_loss/weight4": 0.5133689045906067, "train/pos_loss/weight5": 0.2939463257789612, "train/pos_loss/weight6": 0.17922547459602356, "train/pos_weight_progress": 0.2014218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3179312944412231 }, { "epoch": 2.014218009478673, "step": 850, "train/alpha": 0.2, "train/clean_ce_loss": 0.11962890625, "train/clean_loss_scale": 7.87326717376709, "train/clean_loss_weight": 1.5746535062789917, "train/combined_loss": 1.1302443742752075, "train/pos_loss/pos0": 0.5238520503044128, "train/pos_loss/pos1": 0.904167115688324, "train/pos_loss/pos2": 1.452710509300232, "train/pos_loss/pos3": 1.8939740657806396, "train/pos_loss/pos4": 2.354067802429199, "train/pos_loss/pos5": 2.6480090618133545, "train/pos_loss/pos6": 3.079355478286743, "train/pos_loss/weight0": 3.126342296600342, "train/pos_loss/weight1": 2.035810947418213, "train/pos_loss/weight2": 1.1167116165161133, "train/pos_loss/weight3": 0.4945768713951111, "train/pos_loss/weight4": 0.1656998246908188, "train/pos_loss/weight5": 0.04755036160349846, "train/pos_loss/weight6": 0.013308320194482803, "train/pos_weight_progress": 0.2014218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9418703317642212 }, { "epoch": 2.014218009478673, "step": 850, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 7.6648101806640625, "train/clean_loss_weight": 1.5329620838165283, "train/combined_loss": 0.6062984466552734, "train/pos_loss/pos0": 0.21406345069408417, "train/pos_loss/pos1": 0.4418451189994812, "train/pos_loss/pos2": 0.7566263675689697, "train/pos_loss/pos3": 1.1400651931762695, "train/pos_loss/pos4": 1.7140365839004517, "train/pos_loss/pos5": 2.1266331672668457, "train/pos_loss/pos6": 2.6371331214904785, "train/pos_loss/weight0": 3.054917812347412, "train/pos_loss/weight1": 2.0047168731689453, "train/pos_loss/weight2": 1.1307778358459473, "train/pos_loss/weight3": 0.5157127380371094, "train/pos_loss/weight4": 0.20136991143226624, "train/pos_loss/weight5": 0.06831222027540207, "train/pos_loss/weight6": 0.02419322356581688, "train/pos_weight_progress": 0.2014218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5052487254142761 }, { "epoch": 2.014218009478673, "step": 850, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 7.831754207611084, "train/clean_loss_weight": 1.5663508176803589, "train/combined_loss": 0.6103261709213257, "train/pos_loss/pos0": 0.1563817858695984, "train/pos_loss/pos1": 0.3624451160430908, "train/pos_loss/pos2": 0.6396621465682983, "train/pos_loss/pos3": 0.9993249773979187, "train/pos_loss/pos4": 1.3247257471084595, "train/pos_loss/pos5": 1.8702902793884277, "train/pos_loss/pos6": 2.3059184551239014, "train/pos_loss/weight0": 2.574462652206421, "train/pos_loss/weight1": 1.861654281616211, "train/pos_loss/weight2": 1.2249066829681396, "train/pos_loss/weight3": 0.7312053442001343, "train/pos_loss/weight4": 0.3692390024662018, "train/pos_loss/weight5": 0.15983846783638, "train/pos_loss/weight6": 0.07869330793619156, "train/pos_weight_progress": 0.2014218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5086051225662231 }, { "epoch": 2.0165876777251186, "step": 851, "train/alpha": 0.2, "train/clean_ce_loss": 0.06982421875, "train/clean_loss_scale": 9.136544227600098, "train/clean_loss_weight": 1.8273088932037354, "train/combined_loss": 0.7655424475669861, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.20165876777251185, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6379520297050476 }, { "epoch": 2.0165876777251186, "step": 851, "train/alpha": 0.2, "train/clean_ce_loss": 0.0380859375, "train/clean_loss_scale": 9.320707321166992, "train/clean_loss_weight": 1.8641414642333984, "train/combined_loss": 0.4259854555130005, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.20165876777251185, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.35498788952827454 }, { "epoch": 2.0165876777251186, "step": 851, "train/alpha": 0.2, "train/clean_ce_loss": 0.009033203125, "train/clean_loss_scale": 34.681396484375, "train/clean_loss_weight": 6.936279296875, "train/combined_loss": 0.3759409189224243, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.20165876777251185, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.31328409910202026 }, { "epoch": 2.0165876777251186, "step": 851, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 10.086854934692383, "train/clean_loss_weight": 2.0173709392547607, "train/combined_loss": 0.8569886684417725, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.20165876777251185, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.714157223701477 }, { "epoch": 2.018957345971564, "step": 852, "train/alpha": 0.2, "train/clean_ce_loss": 0.06787109375, "train/clean_loss_scale": 13.809203147888184, "train/clean_loss_weight": 2.761840581893921, "train/combined_loss": 1.12469482421875, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.2018957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9372457265853882 }, { "epoch": 2.018957345971564, "step": 852, "train/alpha": 0.2, "train/clean_ce_loss": 0.00323486328125, "train/clean_loss_scale": 70.1795654296875, "train/clean_loss_weight": 14.035913467407227, "train/combined_loss": 0.2724255621433258, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.2018957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.22702129185199738 }, { "epoch": 2.018957345971564, "step": 852, "train/alpha": 0.2, "train/clean_ce_loss": 0.0546875, "train/clean_loss_scale": 10.735105514526367, "train/clean_loss_weight": 2.1470210552215576, "train/combined_loss": 0.7044912576675415, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.2018957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5870760679244995 }, { "epoch": 2.018957345971564, "step": 852, "train/alpha": 0.2, "train/clean_ce_loss": 0.04541015625, "train/clean_loss_scale": 11.368424415588379, "train/clean_loss_weight": 2.2736849784851074, "train/combined_loss": 0.6194902658462524, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.2018957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5162419080734253 }, { "epoch": 2.0213270142180093, "step": 853, "train/alpha": 0.2, "train/clean_ce_loss": 0.103515625, "train/clean_loss_scale": 9.819862365722656, "train/clean_loss_weight": 1.9639724493026733, "train/combined_loss": 1.219810962677002, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.20213270142180095, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0165091753005981 }, { "epoch": 2.0213270142180093, "step": 853, "train/alpha": 0.2, "train/clean_ce_loss": 0.01220703125, "train/clean_loss_scale": 37.721839904785156, "train/clean_loss_weight": 7.544368267059326, "train/combined_loss": 0.5525659918785095, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.20213270142180095, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46047165989875793 }, { "epoch": 2.0213270142180093, "step": 853, "train/alpha": 0.2, "train/clean_ce_loss": 0.0888671875, "train/clean_loss_scale": 14.460868835449219, "train/clean_loss_weight": 2.8921737670898438, "train/combined_loss": 1.5421161651611328, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.20213270142180095, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2850967645645142 }, { "epoch": 2.0213270142180093, "step": 853, "train/alpha": 0.2, "train/clean_ce_loss": 0.06396484375, "train/clean_loss_scale": 17.381832122802734, "train/clean_loss_weight": 3.4763665199279785, "train/combined_loss": 1.3341914415359497, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.20213270142180095, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1118261814117432 }, { "epoch": 2.023696682464455, "step": 854, "train/alpha": 0.2, "train/clean_ce_loss": 0.08251953125, "train/clean_loss_scale": 8.881831169128418, "train/clean_loss_weight": 1.7763662338256836, "train/combined_loss": 0.8795095086097717, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.2023696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7329245805740356 }, { "epoch": 2.023696682464455, "step": 854, "train/alpha": 0.2, "train/clean_ce_loss": 0.07763671875, "train/clean_loss_scale": 11.696806907653809, "train/clean_loss_weight": 2.3393614292144775, "train/combined_loss": 1.0897220373153687, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.2023696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9081017374992371 }, { "epoch": 2.023696682464455, "step": 854, "train/alpha": 0.2, "train/clean_ce_loss": 0.004302978515625, "train/clean_loss_scale": 54.27504348754883, "train/clean_loss_weight": 10.855009078979492, "train/combined_loss": 0.28025323152542114, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.2023696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.23354434967041016 }, { "epoch": 2.023696682464455, "step": 854, "train/alpha": 0.2, "train/clean_ce_loss": 0.049072265625, "train/clean_loss_scale": 13.097411155700684, "train/clean_loss_weight": 2.6194822788238525, "train/combined_loss": 0.7712635397911072, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.2023696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6427196264266968 }, { "epoch": 2.0260663507109005, "grad_norm": 17.169384002685547, "learning_rate": 2e-06, "loss": 3.1077, "step": 855 }, { "epoch": 2.0260663507109005, "step": 855, "train/alpha": 0.2, "train/clean_ce_loss": 0.029052734375, "train/clean_loss_scale": 32.16990280151367, "train/clean_loss_weight": 6.433980464935303, "train/combined_loss": 1.1215484142303467, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.20260663507109006, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.934623658657074 }, { "epoch": 2.0260663507109005, "step": 855, "train/alpha": 0.2, "train/clean_ce_loss": 0.05126953125, "train/clean_loss_scale": 8.279683113098145, "train/clean_loss_weight": 1.655936598777771, "train/combined_loss": 0.5093945264816284, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.20260663507109006, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4244954586029053 }, { "epoch": 2.0260663507109005, "step": 855, "train/alpha": 0.2, "train/clean_ce_loss": 0.06396484375, "train/clean_loss_scale": 10.68972110748291, "train/clean_loss_weight": 2.137944221496582, "train/combined_loss": 0.8205196261405945, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.20260663507109006, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6837663650512695 }, { "epoch": 2.0260663507109005, "step": 855, "train/alpha": 0.2, "train/clean_ce_loss": 0.010009765625, "train/clean_loss_scale": 34.37933349609375, "train/clean_loss_weight": 6.875866889953613, "train/combined_loss": 0.4129548668861389, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.20260663507109006, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3441290557384491 }, { "epoch": 2.028436018957346, "step": 856, "train/alpha": 0.2, "train/clean_ce_loss": 0.01507568359375, "train/clean_loss_scale": 26.98008918762207, "train/clean_loss_weight": 5.396018028259277, "train/combined_loss": 0.48809194564819336, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.2028436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.40674328804016113 }, { "epoch": 2.028436018957346, "step": 856, "train/alpha": 0.2, "train/clean_ce_loss": 0.08447265625, "train/clean_loss_scale": 8.603392601013184, "train/clean_loss_weight": 1.7206785678863525, "train/combined_loss": 0.8721016645431519, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.2028436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7267513871192932 }, { "epoch": 2.028436018957346, "step": 856, "train/alpha": 0.2, "train/clean_ce_loss": 0.169921875, "train/clean_loss_scale": 8.065632820129395, "train/clean_loss_weight": 1.6131266355514526, "train/combined_loss": 1.6446328163146973, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.2028436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3705273866653442 }, { "epoch": 2.028436018957346, "step": 856, "train/alpha": 0.2, "train/clean_ce_loss": 0.06787109375, "train/clean_loss_scale": 10.906020164489746, "train/clean_loss_weight": 2.181204080581665, "train/combined_loss": 0.8882442116737366, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.2028436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7402034997940063 }, { "epoch": 2.0308056872037916, "step": 857, "train/alpha": 0.2, "train/clean_ce_loss": 0.0751953125, "train/clean_loss_scale": 13.175429344177246, "train/clean_loss_weight": 2.6350858211517334, "train/combined_loss": 1.1888766288757324, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.20308056872037913, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9907305240631104 }, { "epoch": 2.0308056872037916, "step": 857, "train/alpha": 0.2, "train/clean_ce_loss": 0.03369140625, "train/clean_loss_scale": 11.334508895874023, "train/clean_loss_weight": 2.266901731491089, "train/combined_loss": 0.45825064182281494, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.20308056872037913, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.38187554478645325 }, { "epoch": 2.0308056872037916, "step": 857, "train/alpha": 0.2, "train/clean_ce_loss": 0.048095703125, "train/clean_loss_scale": 5.263304233551025, "train/clean_loss_weight": 1.0526608228683472, "train/combined_loss": 0.3037707805633545, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.20308056872037913, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2531423270702362 }, { "epoch": 2.0308056872037916, "step": 857, "train/alpha": 0.2, "train/clean_ce_loss": 0.04248046875, "train/clean_loss_scale": 24.386354446411133, "train/clean_loss_weight": 4.8772711753845215, "train/combined_loss": 1.243132472038269, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.20308056872037913, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0359437465667725 }, { "epoch": 2.0331753554502368, "step": 858, "train/alpha": 0.2, "train/clean_ce_loss": 0.06005859375, "train/clean_loss_scale": 10.035500526428223, "train/clean_loss_weight": 2.0071001052856445, "train/combined_loss": 0.7232616543769836, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.2033175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6027180552482605 }, { "epoch": 2.0331753554502368, "step": 858, "train/alpha": 0.2, "train/clean_ce_loss": 0.00897216796875, "train/clean_loss_scale": 39.920310974121094, "train/clean_loss_weight": 7.984062194824219, "train/combined_loss": 0.42980608344078064, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.2033175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3581717312335968 }, { "epoch": 2.0331753554502368, "step": 858, "train/alpha": 0.2, "train/clean_ce_loss": 0.083984375, "train/clean_loss_scale": 6.672447204589844, "train/clean_loss_weight": 1.3344894647598267, "train/combined_loss": 0.6724575757980347, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.2033175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.560381293296814 }, { "epoch": 2.0331753554502368, "step": 858, "train/alpha": 0.2, "train/clean_ce_loss": 0.0732421875, "train/clean_loss_scale": 13.163259506225586, "train/clean_loss_weight": 2.6326520442962646, "train/combined_loss": 1.1569271087646484, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.2033175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9641059041023254 }, { "epoch": 2.0355450236966823, "step": 859, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 8.389906883239746, "train/clean_loss_weight": 1.6779813766479492, "train/combined_loss": 0.7324782609939575, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.20355450236966824, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.610398530960083 }, { "epoch": 2.0355450236966823, "step": 859, "train/alpha": 0.2, "train/clean_ce_loss": 0.10595703125, "train/clean_loss_scale": 3.4398226737976074, "train/clean_loss_weight": 0.6879645586013794, "train/combined_loss": 0.43736809492111206, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.20355450236966824, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3644734025001526 }, { "epoch": 2.0355450236966823, "step": 859, "train/alpha": 0.2, "train/clean_ce_loss": 0.0238037109375, "train/clean_loss_scale": 19.738162994384766, "train/clean_loss_weight": 3.9476325511932373, "train/combined_loss": 0.5638098120689392, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.20355450236966824, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46984151005744934 }, { "epoch": 2.0355450236966823, "step": 859, "train/alpha": 0.2, "train/clean_ce_loss": 0.0634765625, "train/clean_loss_scale": 9.647600173950195, "train/clean_loss_weight": 1.9295200109481812, "train/combined_loss": 0.7348757982254028, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.20355450236966824, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6123964786529541 }, { "epoch": 2.037914691943128, "grad_norm": 26.587966918945312, "learning_rate": 2e-06, "loss": 3.0273, "step": 860 }, { "epoch": 2.037914691943128, "step": 860, "train/alpha": 0.2, "train/clean_ce_loss": 0.091796875, "train/clean_loss_scale": 8.358236312866211, "train/clean_loss_weight": 1.671647310256958, "train/combined_loss": 0.9207119345664978, "train/pos_loss/pos0": 0.2654677927494049, "train/pos_loss/pos1": 0.5513731837272644, "train/pos_loss/pos2": 0.8790159225463867, "train/pos_loss/pos3": 1.2961533069610596, "train/pos_loss/pos4": 1.6860973834991455, "train/pos_loss/pos5": 2.063732385635376, "train/pos_loss/pos6": 2.506420612335205, "train/pos_loss/weight0": 2.338233232498169, "train/pos_loss/weight1": 1.779751181602478, "train/pos_loss/weight2": 1.230385661125183, "train/pos_loss/weight3": 0.7939164638519287, "train/pos_loss/weight4": 0.46274036169052124, "train/pos_loss/weight5": 0.24990516901016235, "train/pos_loss/weight6": 0.14506813883781433, "train/pos_weight_progress": 0.2037914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.767259955406189 }, { "epoch": 2.037914691943128, "step": 860, "train/alpha": 0.2, "train/clean_ce_loss": 0.125, "train/clean_loss_scale": 8.507393836975098, "train/clean_loss_weight": 1.7014788389205933, "train/combined_loss": 1.2761090993881226, "train/pos_loss/pos0": 0.42945897579193115, "train/pos_loss/pos1": 0.9149853587150574, "train/pos_loss/pos2": 1.389356017112732, "train/pos_loss/pos3": 1.8334176540374756, "train/pos_loss/pos4": 2.3148584365844727, "train/pos_loss/pos5": 2.9060001373291016, "train/pos_loss/pos6": 3.301027774810791, "train/pos_loss/weight0": 2.584526777267456, "train/pos_loss/weight1": 1.8501631021499634, "train/pos_loss/weight2": 1.2304359674453735, "train/pos_loss/weight3": 0.7124696373939514, "train/pos_loss/weight4": 0.36610427498817444, "train/pos_loss/weight5": 0.1712685078382492, "train/pos_loss/weight6": 0.08503192663192749, "train/pos_weight_progress": 0.2037914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0634242296218872 }, { "epoch": 2.037914691943128, "step": 860, "train/alpha": 0.2, "train/clean_ce_loss": 0.01556396484375, "train/clean_loss_scale": 14.915850639343262, "train/clean_loss_weight": 2.9831702709198, "train/combined_loss": 0.2785797417163849, "train/pos_loss/pos0": 0.02105426974594593, "train/pos_loss/pos1": 0.10490889102220535, "train/pos_loss/pos2": 0.31954750418663025, "train/pos_loss/pos3": 0.8484868407249451, "train/pos_loss/pos4": 1.4587295055389404, "train/pos_loss/pos5": 2.0174005031585693, "train/pos_loss/pos6": 2.3111584186553955, "train/pos_loss/weight0": 2.9673640727996826, "train/pos_loss/weight1": 2.0178980827331543, "train/pos_loss/weight2": 1.1504828929901123, "train/pos_loss/weight3": 0.5523597002029419, "train/pos_loss/weight4": 0.21775157749652863, "train/pos_loss/weight5": 0.07028169184923172, "train/pos_loss/weight6": 0.023861847817897797, "train/pos_weight_progress": 0.2037914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.23214977979660034 }, { "epoch": 2.037914691943128, "step": 860, "train/alpha": 0.2, "train/clean_ce_loss": 0.033447265625, "train/clean_loss_scale": 33.889251708984375, "train/clean_loss_weight": 6.77785062789917, "train/combined_loss": 1.3602033853530884, "train/pos_loss/pos0": 0.14117076992988586, "train/pos_loss/pos1": 0.5303125381469727, "train/pos_loss/pos2": 1.3466211557388306, "train/pos_loss/pos3": 1.9858813285827637, "train/pos_loss/pos4": 2.5780301094055176, "train/pos_loss/pos5": 3.7171308994293213, "train/pos_loss/pos6": 3.957179069519043, "train/pos_loss/weight0": 2.164116859436035, "train/pos_loss/weight1": 1.6634531021118164, "train/pos_loss/weight2": 1.2352858781814575, "train/pos_loss/weight3": 0.8798518180847168, "train/pos_loss/weight4": 0.5488437414169312, "train/pos_loss/weight5": 0.3138701915740967, "train/pos_loss/weight6": 0.1945783495903015, "train/pos_weight_progress": 0.2037914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1335028409957886 }, { "epoch": 2.0402843601895735, "step": 861, "train/alpha": 0.2, "train/clean_ce_loss": 0.01385498046875, "train/clean_loss_scale": 12.49174690246582, "train/clean_loss_weight": 2.49834942817688, "train/combined_loss": 0.20768748223781586, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20402843601895734, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1730729043483734 }, { "epoch": 2.0402843601895735, "step": 861, "train/alpha": 0.2, "train/clean_ce_loss": 0.0732421875, "train/clean_loss_scale": 6.6303391456604, "train/clean_loss_weight": 1.3260678052902222, "train/combined_loss": 0.5827446579933167, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20402843601895734, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.48562052845954895 }, { "epoch": 2.0402843601895735, "step": 861, "train/alpha": 0.2, "train/clean_ce_loss": 0.041259765625, "train/clean_loss_scale": 8.800410270690918, "train/clean_loss_weight": 1.7600821256637573, "train/combined_loss": 0.4357234537601471, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20402843601895734, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.36310288310050964 }, { "epoch": 2.0402843601895735, "step": 861, "train/alpha": 0.2, "train/clean_ce_loss": 0.037841796875, "train/clean_loss_scale": 7.938337326049805, "train/clean_loss_weight": 1.587667465209961, "train/combined_loss": 0.3604811429977417, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20402843601895734, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3004009425640106 }, { "epoch": 2.042654028436019, "step": 862, "train/alpha": 0.2, "train/clean_ce_loss": 0.07373046875, "train/clean_loss_scale": 7.652334690093994, "train/clean_loss_weight": 1.530466914176941, "train/combined_loss": 0.6770522594451904, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.2042654028436019, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5642102360725403 }, { "epoch": 2.042654028436019, "step": 862, "train/alpha": 0.2, "train/clean_ce_loss": 0.048583984375, "train/clean_loss_scale": 7.349729537963867, "train/clean_loss_weight": 1.4699459075927734, "train/combined_loss": 0.42849498987197876, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.2042654028436019, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3570791482925415 }, { "epoch": 2.042654028436019, "step": 862, "train/alpha": 0.2, "train/clean_ce_loss": 0.018310546875, "train/clean_loss_scale": 10.767806053161621, "train/clean_loss_weight": 2.1535613536834717, "train/combined_loss": 0.23659729957580566, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.2042654028436019, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1971644163131714 }, { "epoch": 2.042654028436019, "step": 862, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 8.255393981933594, "train/clean_loss_weight": 1.6510788202285767, "train/combined_loss": 0.6578517556190491, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.2042654028436019, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5482097864151001 }, { "epoch": 2.0450236966824646, "step": 863, "train/alpha": 0.2, "train/clean_ce_loss": 0.0234375, "train/clean_loss_scale": 10.456831932067871, "train/clean_loss_weight": 2.0913665294647217, "train/combined_loss": 0.2940983772277832, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20450236966824645, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.24508199095726013 }, { "epoch": 2.0450236966824646, "step": 863, "train/alpha": 0.2, "train/clean_ce_loss": 0.072265625, "train/clean_loss_scale": 13.770394325256348, "train/clean_loss_weight": 2.7540788650512695, "train/combined_loss": 1.1941514015197754, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20450236966824645, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9951261281967163 }, { "epoch": 2.0450236966824646, "step": 863, "train/alpha": 0.2, "train/clean_ce_loss": 0.05029296875, "train/clean_loss_scale": 15.062612533569336, "train/clean_loss_weight": 3.0125224590301514, "train/combined_loss": 0.9090521931648254, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20450236966824645, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7575435042381287 }, { "epoch": 2.0450236966824646, "step": 863, "train/alpha": 0.2, "train/clean_ce_loss": 0.001556396484375, "train/clean_loss_scale": 105.26505279541016, "train/clean_loss_weight": 21.053010940551758, "train/combined_loss": 0.1966009885072708, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20450236966824645, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.16383415460586548 }, { "epoch": 2.0473933649289098, "step": 864, "train/alpha": 0.2, "train/clean_ce_loss": 0.055908203125, "train/clean_loss_scale": 12.573137283325195, "train/clean_loss_weight": 2.514627456665039, "train/combined_loss": 0.8435298204421997, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.204739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7029415369033813 }, { "epoch": 2.0473933649289098, "step": 864, "train/alpha": 0.2, "train/clean_ce_loss": 0.171875, "train/clean_loss_scale": 6.22474479675293, "train/clean_loss_weight": 1.2449489831924438, "train/combined_loss": 1.283853530883789, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.204739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0698779821395874 }, { "epoch": 2.0473933649289098, "step": 864, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 9.05872917175293, "train/clean_loss_weight": 1.8117458820343018, "train/combined_loss": 0.716559648513794, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.204739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5971330404281616 }, { "epoch": 2.0473933649289098, "step": 864, "train/alpha": 0.2, "train/clean_ce_loss": 0.1552734375, "train/clean_loss_scale": 7.243781089782715, "train/clean_loss_weight": 1.448756217956543, "train/combined_loss": 1.3497202396392822, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.204739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1247668266296387 }, { "epoch": 2.0497630331753554, "grad_norm": 21.91170310974121, "learning_rate": 2e-06, "loss": 3.21, "step": 865 }, { "epoch": 2.0497630331753554, "step": 865, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 9.74731731414795, "train/clean_loss_weight": 1.9494634866714478, "train/combined_loss": 0.7538940906524658, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20497630331753555, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6282450556755066 }, { "epoch": 2.0497630331753554, "step": 865, "train/alpha": 0.2, "train/clean_ce_loss": 0.07470703125, "train/clean_loss_scale": 9.223082542419434, "train/clean_loss_weight": 1.8446165323257446, "train/combined_loss": 0.8268349170684814, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20497630331753555, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6890290975570679 }, { "epoch": 2.0497630331753554, "step": 865, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 8.384138107299805, "train/clean_loss_weight": 1.6768276691436768, "train/combined_loss": 0.8007506728172302, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20497630331753555, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6672922372817993 }, { "epoch": 2.0497630331753554, "step": 865, "train/alpha": 0.2, "train/clean_ce_loss": 0.1455078125, "train/clean_loss_scale": 6.25221586227417, "train/clean_loss_weight": 1.2504432201385498, "train/combined_loss": 1.0916955471038818, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20497630331753555, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9097462296485901 }, { "epoch": 2.052132701421801, "step": 866, "train/alpha": 0.2, "train/clean_ce_loss": 0.04541015625, "train/clean_loss_scale": 7.945285797119141, "train/clean_loss_weight": 1.589057207107544, "train/combined_loss": 0.4329559803009033, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.2052132701421801, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.36079666018486023 }, { "epoch": 2.052132701421801, "step": 866, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 6.767563343048096, "train/clean_loss_weight": 1.3535126447677612, "train/combined_loss": 0.5789439082145691, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.2052132701421801, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.48245325684547424 }, { "epoch": 2.052132701421801, "step": 866, "train/alpha": 0.2, "train/clean_ce_loss": 0.11279296875, "train/clean_loss_scale": 7.098994255065918, "train/clean_loss_weight": 1.4197988510131836, "train/combined_loss": 0.9608599543571472, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.2052132701421801, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8007166385650635 }, { "epoch": 2.052132701421801, "step": 866, "train/alpha": 0.2, "train/clean_ce_loss": 0.11328125, "train/clean_loss_scale": 9.232773780822754, "train/clean_loss_weight": 1.8465547561645508, "train/combined_loss": 1.2550801038742065, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.2052132701421801, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0459001064300537 }, { "epoch": 2.0545023696682465, "step": 867, "train/alpha": 0.2, "train/clean_ce_loss": 0.0247802734375, "train/clean_loss_scale": 26.7277889251709, "train/clean_loss_weight": 5.345557689666748, "train/combined_loss": 0.7947863340377808, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20545023696682466, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.662321925163269 }, { "epoch": 2.0545023696682465, "step": 867, "train/alpha": 0.2, "train/clean_ce_loss": 0.06689453125, "train/clean_loss_scale": 10.423358917236328, "train/clean_loss_weight": 2.08467173576355, "train/combined_loss": 0.8367187976837158, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20545023696682466, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6972656846046448 }, { "epoch": 2.0545023696682465, "step": 867, "train/alpha": 0.2, "train/clean_ce_loss": 0.0093994140625, "train/clean_loss_scale": 19.87254524230957, "train/clean_loss_weight": 3.9745090007781982, "train/combined_loss": 0.22414834797382355, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20545023696682466, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.18679028749465942 }, { "epoch": 2.0545023696682465, "step": 867, "train/alpha": 0.2, "train/clean_ce_loss": 0.0693359375, "train/clean_loss_scale": 6.6876115798950195, "train/clean_loss_weight": 1.3375223875045776, "train/combined_loss": 0.5564302206039429, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20545023696682466, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4636918306350708 }, { "epoch": 2.056872037914692, "step": 868, "train/alpha": 0.2, "train/clean_ce_loss": 0.1064453125, "train/clean_loss_scale": 9.00987434387207, "train/clean_loss_weight": 1.801974892616272, "train/combined_loss": 1.150870680809021, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20568720379146918, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.959058940410614 }, { "epoch": 2.056872037914692, "step": 868, "train/alpha": 0.2, "train/clean_ce_loss": 0.049560546875, "train/clean_loss_scale": 8.009573936462402, "train/clean_loss_weight": 1.6019147634506226, "train/combined_loss": 0.47635066509246826, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20568720379146918, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3969588875770569 }, { "epoch": 2.056872037914692, "step": 868, "train/alpha": 0.2, "train/clean_ce_loss": 0.111328125, "train/clean_loss_scale": 8.400415420532227, "train/clean_loss_weight": 1.680083155632019, "train/combined_loss": 1.122243046760559, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20568720379146918, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9352025389671326 }, { "epoch": 2.056872037914692, "step": 868, "train/alpha": 0.2, "train/clean_ce_loss": 0.047607421875, "train/clean_loss_scale": 8.639062881469727, "train/clean_loss_weight": 1.727812647819519, "train/combined_loss": 0.4935402274131775, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20568720379146918, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4112835228443146 }, { "epoch": 2.0592417061611377, "step": 869, "train/alpha": 0.2, "train/clean_ce_loss": 0.11865234375, "train/clean_loss_scale": 5.81320858001709, "train/clean_loss_weight": 1.1626417636871338, "train/combined_loss": 0.8277010321617126, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20592417061611373, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6897508502006531 }, { "epoch": 2.0592417061611377, "step": 869, "train/alpha": 0.2, "train/clean_ce_loss": 0.0966796875, "train/clean_loss_scale": 6.332570552825928, "train/clean_loss_weight": 1.2665141820907593, "train/combined_loss": 0.7346771359443665, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20592417061611373, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6122309565544128 }, { "epoch": 2.0592417061611377, "step": 869, "train/alpha": 0.2, "train/clean_ce_loss": 0.04833984375, "train/clean_loss_scale": 6.926030158996582, "train/clean_loss_weight": 1.3852061033248901, "train/combined_loss": 0.4017638564109802, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20592417061611373, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3348032236099243 }, { "epoch": 2.0592417061611377, "step": 869, "train/alpha": 0.2, "train/clean_ce_loss": 0.056640625, "train/clean_loss_scale": 15.054413795471191, "train/clean_loss_weight": 3.01088285446167, "train/combined_loss": 1.023229718208313, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20592417061611373, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8526914119720459 }, { "epoch": 2.061611374407583, "grad_norm": 24.3290958404541, "learning_rate": 2e-06, "loss": 3.3667, "step": 870 }, { "epoch": 2.061611374407583, "step": 870, "train/alpha": 0.2, "train/clean_ce_loss": 0.00927734375, "train/clean_loss_scale": 22.398107528686523, "train/clean_loss_weight": 4.479621410369873, "train/combined_loss": 0.24935391545295715, "train/pos_loss/pos0": 0.013794353231787682, "train/pos_loss/pos1": 0.06905260682106018, "train/pos_loss/pos2": 0.32755666971206665, "train/pos_loss/pos3": 0.7376052737236023, "train/pos_loss/pos4": 1.308149814605713, "train/pos_loss/pos5": 1.7153853178024292, "train/pos_loss/pos6": 2.0973851680755615, "train/pos_loss/weight0": 2.8282456398010254, "train/pos_loss/weight1": 2.0227622985839844, "train/pos_loss/weight2": 1.2431336641311646, "train/pos_loss/weight3": 0.6126923561096191, "train/pos_loss/weight4": 0.22436970472335815, "train/pos_loss/weight5": 0.05480225384235382, "train/pos_loss/weight6": 0.013994302600622177, "train/pos_weight_progress": 0.20616113744075829, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2077949345111847 }, { "epoch": 2.061611374407583, "step": 870, "train/alpha": 0.2, "train/clean_ce_loss": 0.134765625, "train/clean_loss_scale": 8.05979061126709, "train/clean_loss_weight": 1.6119581460952759, "train/combined_loss": 1.3034192323684692, "train/pos_loss/pos0": 0.40855923295021057, "train/pos_loss/pos1": 0.7359291315078735, "train/pos_loss/pos2": 1.2124462127685547, "train/pos_loss/pos3": 1.558098554611206, "train/pos_loss/pos4": 2.0425100326538086, "train/pos_loss/pos5": 2.484076976776123, "train/pos_loss/pos6": 2.570246696472168, "train/pos_loss/weight0": 2.0832433700561523, "train/pos_loss/weight1": 1.6193403005599976, "train/pos_loss/weight2": 1.2413740158081055, "train/pos_loss/weight3": 0.8763642311096191, "train/pos_loss/weight4": 0.5891318917274475, "train/pos_loss/weight5": 0.35263851284980774, "train/pos_loss/weight6": 0.23790788650512695, "train/pos_weight_progress": 0.20616113744075829, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.086182713508606 }, { "epoch": 2.061611374407583, "step": 870, "train/alpha": 0.2, "train/clean_ce_loss": 0.03076171875, "train/clean_loss_scale": 15.184113502502441, "train/clean_loss_weight": 3.03682279586792, "train/combined_loss": 0.5605072975158691, "train/pos_loss/pos0": 0.15297432243824005, "train/pos_loss/pos1": 0.2733071446418762, "train/pos_loss/pos2": 0.6027801036834717, "train/pos_loss/pos3": 1.1687806844711304, "train/pos_loss/pos4": 1.7114458084106445, "train/pos_loss/pos5": 2.2267045974731445, "train/pos_loss/pos6": 2.4931325912475586, "train/pos_loss/weight0": 2.808088541030884, "train/pos_loss/weight1": 1.9267972707748413, "train/pos_loss/weight2": 1.1872366666793823, "train/pos_loss/weight3": 0.6262676119804382, "train/pos_loss/weight4": 0.28937920928001404, "train/pos_loss/weight5": 0.11550553888082504, "train/pos_loss/weight6": 0.04672550782561302, "train/pos_weight_progress": 0.20616113744075829, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4670894145965576 }, { "epoch": 2.061611374407583, "step": 870, "train/alpha": 0.2, "train/clean_ce_loss": 0.0400390625, "train/clean_loss_scale": 27.478382110595703, "train/clean_loss_weight": 5.495676517486572, "train/combined_loss": 1.320250391960144, "train/pos_loss/pos0": 0.2428009808063507, "train/pos_loss/pos1": 0.650912880897522, "train/pos_loss/pos2": 1.4844704866409302, "train/pos_loss/pos3": 2.086416482925415, "train/pos_loss/pos4": 2.828122138977051, "train/pos_loss/pos5": 3.238217353820801, "train/pos_loss/pos6": 3.669121026992798, "train/pos_loss/weight0": 2.3951945304870605, "train/pos_loss/weight1": 1.7628228664398193, "train/pos_loss/weight2": 1.2422666549682617, "train/pos_loss/weight3": 0.8002280592918396, "train/pos_loss/weight4": 0.44830334186553955, "train/pos_loss/weight5": 0.2269597053527832, "train/pos_loss/weight6": 0.12422451376914978, "train/pos_weight_progress": 0.20616113744075829, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1002086400985718 }, { "epoch": 2.0639810426540284, "step": 871, "train/alpha": 0.2, "train/clean_ce_loss": 0.0128173828125, "train/clean_loss_scale": 5.670738220214844, "train/clean_loss_weight": 1.1341476440429688, "train/combined_loss": 0.08722082525491714, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.20639810426540284, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.07268401980400085 }, { "epoch": 2.0639810426540284, "step": 871, "train/alpha": 0.2, "train/clean_ce_loss": 0.013671875, "train/clean_loss_scale": 18.77593994140625, "train/clean_loss_weight": 3.75518798828125, "train/combined_loss": 0.3080427646636963, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.20639810426540284, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2567023038864136 }, { "epoch": 2.0639810426540284, "step": 871, "train/alpha": 0.2, "train/clean_ce_loss": 0.041015625, "train/clean_loss_scale": 9.052577018737793, "train/clean_loss_weight": 1.8105154037475586, "train/combined_loss": 0.44555652141571045, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.20639810426540284, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.37129709124565125 }, { "epoch": 2.0639810426540284, "step": 871, "train/alpha": 0.2, "train/clean_ce_loss": 0.08642578125, "train/clean_loss_scale": 8.652652740478516, "train/clean_loss_weight": 1.7305306196212769, "train/combined_loss": 0.8973747491836548, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.20639810426540284, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7478122711181641 }, { "epoch": 2.066350710900474, "step": 872, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 6.8722710609436035, "train/clean_loss_weight": 1.3744542598724365, "train/combined_loss": 0.6603823304176331, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.2066350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5503185987472534 }, { "epoch": 2.066350710900474, "step": 872, "train/alpha": 0.2, "train/clean_ce_loss": 0.01043701171875, "train/clean_loss_scale": 22.005828857421875, "train/clean_loss_weight": 4.401165962219238, "train/combined_loss": 0.275610089302063, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.2066350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2296750843524933 }, { "epoch": 2.066350710900474, "step": 872, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 6.470658302307129, "train/clean_loss_weight": 1.2941316366195679, "train/combined_loss": 0.5497531890869141, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.2066350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4581276476383209 }, { "epoch": 2.066350710900474, "step": 872, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 9.948161125183105, "train/clean_loss_weight": 1.989632248878479, "train/combined_loss": 0.8452051281929016, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.2066350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7043375968933105 }, { "epoch": 2.0687203791469195, "step": 873, "train/alpha": 0.2, "train/clean_ce_loss": 0.04150390625, "train/clean_loss_scale": 4.259390354156494, "train/clean_loss_weight": 0.8518781065940857, "train/combined_loss": 0.2121376097202301, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.20687203791469194, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.17678134143352509 }, { "epoch": 2.0687203791469195, "step": 873, "train/alpha": 0.2, "train/clean_ce_loss": 0.134765625, "train/clean_loss_scale": 7.621928691864014, "train/clean_loss_weight": 1.5243858098983765, "train/combined_loss": 1.2326087951660156, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.20687203791469194, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0271739959716797 }, { "epoch": 2.0687203791469195, "step": 873, "train/alpha": 0.2, "train/clean_ce_loss": 0.0191650390625, "train/clean_loss_scale": 18.736652374267578, "train/clean_loss_weight": 3.7473304271698, "train/combined_loss": 0.43090638518333435, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.20687203791469194, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.359088659286499 }, { "epoch": 2.0687203791469195, "step": 873, "train/alpha": 0.2, "train/clean_ce_loss": 0.004852294921875, "train/clean_loss_scale": 30.067689895629883, "train/clean_loss_weight": 6.013537883758545, "train/combined_loss": 0.17507675290107727, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.20687203791469194, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1458972990512848 }, { "epoch": 2.071090047393365, "step": 874, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 7.932722091674805, "train/clean_loss_weight": 1.586544394493103, "train/combined_loss": 0.6925638318061829, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.2071090047393365, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5771365165710449 }, { "epoch": 2.071090047393365, "step": 874, "train/alpha": 0.2, "train/clean_ce_loss": 0.0001678466796875, "train/clean_loss_scale": 285.6698913574219, "train/clean_loss_weight": 57.13397979736328, "train/combined_loss": 0.057538487017154694, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.2071090047393365, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.04794874042272568 }, { "epoch": 2.071090047393365, "step": 874, "train/alpha": 0.2, "train/clean_ce_loss": 0.0106201171875, "train/clean_loss_scale": 19.016098022460938, "train/clean_loss_weight": 3.8032195568084717, "train/combined_loss": 0.24234382808208466, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.2071090047393365, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.20195318758487701 }, { "epoch": 2.071090047393365, "step": 874, "train/alpha": 0.2, "train/clean_ce_loss": 0.046630859375, "train/clean_loss_scale": 7.690342903137207, "train/clean_loss_weight": 1.5380686521530151, "train/combined_loss": 0.43032875657081604, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.2071090047393365, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.35860729217529297 }, { "epoch": 2.0734597156398102, "grad_norm": 13.618721008300781, "learning_rate": 2e-06, "loss": 3.2222, "step": 875 }, { "epoch": 2.0734597156398102, "step": 875, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 7.220898151397705, "train/clean_loss_weight": 1.444179654121399, "train/combined_loss": 0.6854212284088135, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.20734597156398105, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5711843371391296 }, { "epoch": 2.0734597156398102, "step": 875, "train/alpha": 0.2, "train/clean_ce_loss": 0.091796875, "train/clean_loss_scale": 8.399553298950195, "train/clean_loss_weight": 1.679910659790039, "train/combined_loss": 0.9252632856369019, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.20734597156398105, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7710527181625366 }, { "epoch": 2.0734597156398102, "step": 875, "train/alpha": 0.2, "train/clean_ce_loss": 0.006134033203125, "train/clean_loss_scale": 21.617328643798828, "train/clean_loss_weight": 4.323465824127197, "train/combined_loss": 0.15912169218063354, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.20734597156398105, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.13260141015052795 }, { "epoch": 2.0734597156398102, "step": 875, "train/alpha": 0.2, "train/clean_ce_loss": 0.10791015625, "train/clean_loss_scale": 7.558919429779053, "train/clean_loss_weight": 1.5117839574813843, "train/combined_loss": 0.9788210391998291, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.20734597156398105, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8156841993331909 }, { "epoch": 2.075829383886256, "step": 876, "train/alpha": 0.2, "train/clean_ce_loss": 0.06298828125, "train/clean_loss_scale": 6.8492512702941895, "train/clean_loss_weight": 1.3698502779006958, "train/combined_loss": 0.517707109451294, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.2075829383886256, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43142256140708923 }, { "epoch": 2.075829383886256, "step": 876, "train/alpha": 0.2, "train/clean_ce_loss": 0.0625, "train/clean_loss_scale": 8.61181926727295, "train/clean_loss_weight": 1.722363829612732, "train/combined_loss": 0.6458864212036133, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.2075829383886256, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5382387042045593 }, { "epoch": 2.075829383886256, "step": 876, "train/alpha": 0.2, "train/clean_ce_loss": 0.01300048828125, "train/clean_loss_scale": 10.039007186889648, "train/clean_loss_weight": 2.0078015327453613, "train/combined_loss": 0.15661439299583435, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.2075829383886256, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.13051199913024902 }, { "epoch": 2.075829383886256, "step": 876, "train/alpha": 0.2, "train/clean_ce_loss": 0.08203125, "train/clean_loss_scale": 10.461822509765625, "train/clean_loss_weight": 2.092364549636841, "train/combined_loss": 1.0298357009887695, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.2075829383886256, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8581963777542114 }, { "epoch": 2.0781990521327014, "step": 877, "train/alpha": 0.2, "train/clean_ce_loss": 0.0177001953125, "train/clean_loss_scale": 19.409242630004883, "train/clean_loss_weight": 3.8818485736846924, "train/combined_loss": 0.4122568368911743, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.20781990521327015, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.34354737401008606 }, { "epoch": 2.0781990521327014, "step": 877, "train/alpha": 0.2, "train/clean_ce_loss": 0.04638671875, "train/clean_loss_scale": 8.64504337310791, "train/clean_loss_weight": 1.729008674621582, "train/combined_loss": 0.48121821880340576, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.20781990521327015, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.40101519227027893 }, { "epoch": 2.0781990521327014, "step": 877, "train/alpha": 0.2, "train/clean_ce_loss": 0.1142578125, "train/clean_loss_scale": 6.276449680328369, "train/clean_loss_weight": 1.255289912223816, "train/combined_loss": 0.8605600595474243, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.20781990521327015, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7171334028244019 }, { "epoch": 2.0781990521327014, "step": 877, "train/alpha": 0.2, "train/clean_ce_loss": 0.00098419189453125, "train/clean_loss_scale": 172.32351684570312, "train/clean_loss_weight": 34.46470260620117, "train/combined_loss": 0.2035192996263504, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.20781990521327015, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.16959941387176514 }, { "epoch": 2.080568720379147, "step": 878, "train/alpha": 0.2, "train/clean_ce_loss": 0.0034027099609375, "train/clean_loss_scale": 64.30126190185547, "train/clean_loss_weight": 12.860252380371094, "train/combined_loss": 0.262558251619339, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.2080568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.21879853308200836 }, { "epoch": 2.080568720379147, "step": 878, "train/alpha": 0.2, "train/clean_ce_loss": 0.08447265625, "train/clean_loss_scale": 7.417124271392822, "train/clean_loss_weight": 1.4834249019622803, "train/combined_loss": 0.7518529891967773, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.2080568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.626544177532196 }, { "epoch": 2.080568720379147, "step": 878, "train/alpha": 0.2, "train/clean_ce_loss": 0.11767578125, "train/clean_loss_scale": 6.026581764221191, "train/clean_loss_weight": 1.205316424369812, "train/combined_loss": 0.8510192632675171, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.2080568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7091827392578125 }, { "epoch": 2.080568720379147, "step": 878, "train/alpha": 0.2, "train/clean_ce_loss": 0.0220947265625, "train/clean_loss_scale": 9.524652481079102, "train/clean_loss_weight": 1.9049304723739624, "train/combined_loss": 0.25253352522850037, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.2080568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2104445993900299 }, { "epoch": 2.0829383886255926, "step": 879, "train/alpha": 0.2, "train/clean_ce_loss": 0.031494140625, "train/clean_loss_scale": 8.468181610107422, "train/clean_loss_weight": 1.6936362981796265, "train/combined_loss": 0.32003772258758545, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.20829383886255926, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.26669809222221375 }, { "epoch": 2.0829383886255926, "step": 879, "train/alpha": 0.2, "train/clean_ce_loss": 0.11083984375, "train/clean_loss_scale": 1.9467984437942505, "train/clean_loss_weight": 0.3893596827983856, "train/combined_loss": 0.2589394152164459, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.20829383886255926, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.21578283607959747 }, { "epoch": 2.0829383886255926, "step": 879, "train/alpha": 0.2, "train/clean_ce_loss": 0.103515625, "train/clean_loss_scale": 8.560403823852539, "train/clean_loss_weight": 1.7120808362960815, "train/combined_loss": 1.063362717628479, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.20829383886255926, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8861355781555176 }, { "epoch": 2.0829383886255926, "step": 879, "train/alpha": 0.2, "train/clean_ce_loss": 0.09228515625, "train/clean_loss_scale": 7.936243057250977, "train/clean_loss_weight": 1.587248682975769, "train/combined_loss": 0.8788769245147705, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.20829383886255926, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7323974370956421 }, { "epoch": 2.085308056872038, "grad_norm": 24.667821884155273, "learning_rate": 2e-06, "loss": 3.263, "step": 880 }, { "epoch": 2.085308056872038, "step": 880, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 8.394200325012207, "train/clean_loss_weight": 1.6788400411605835, "train/combined_loss": 0.7131791114807129, "train/pos_loss/pos0": 0.2532119154930115, "train/pos_loss/pos1": 0.5194299221038818, "train/pos_loss/pos2": 0.9713302850723267, "train/pos_loss/pos3": 1.3012669086456299, "train/pos_loss/pos4": 1.6671974658966064, "train/pos_loss/pos5": 2.1491775512695312, "train/pos_loss/pos6": 2.424100637435913, "train/pos_loss/weight0": 2.9462666511535645, "train/pos_loss/weight1": 2.0357983112335205, "train/pos_loss/weight2": 1.1972724199295044, "train/pos_loss/weight3": 0.5643261671066284, "train/pos_loss/weight4": 0.19472549855709076, "train/pos_loss/weight5": 0.04781825467944145, "train/pos_loss/weight6": 0.013792714104056358, "train/pos_weight_progress": 0.20853080568720378, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5943159461021423 }, { "epoch": 2.085308056872038, "step": 880, "train/alpha": 0.2, "train/clean_ce_loss": 0.0106201171875, "train/clean_loss_scale": 30.11929702758789, "train/clean_loss_weight": 6.02385950088501, "train/combined_loss": 0.3838445544242859, "train/pos_loss/pos0": 0.0294047798961401, "train/pos_loss/pos1": 0.08811189234256744, "train/pos_loss/pos2": 0.24949926137924194, "train/pos_loss/pos3": 0.713935911655426, "train/pos_loss/pos4": 1.3656190633773804, "train/pos_loss/pos5": 2.008936882019043, "train/pos_loss/pos6": 2.584815263748169, "train/pos_loss/weight0": 2.5175940990448, "train/pos_loss/weight1": 1.825974941253662, "train/pos_loss/weight2": 1.2455339431762695, "train/pos_loss/weight3": 0.7291443943977356, "train/pos_loss/weight4": 0.3917449712753296, "train/pos_loss/weight5": 0.19197718799114227, "train/pos_loss/weight6": 0.09803011268377304, "train/pos_weight_progress": 0.20853080568720378, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3198704719543457 }, { "epoch": 2.085308056872038, "step": 880, "train/alpha": 0.2, "train/clean_ce_loss": 0.09765625, "train/clean_loss_scale": 15.020011901855469, "train/clean_loss_weight": 3.004002332687378, "train/combined_loss": 1.7601577043533325, "train/pos_loss/pos0": 0.4696624279022217, "train/pos_loss/pos1": 1.02007257938385, "train/pos_loss/pos2": 1.594113826751709, "train/pos_loss/pos3": 2.341298818588257, "train/pos_loss/pos4": 2.8817012310028076, "train/pos_loss/pos5": 3.309100866317749, "train/pos_loss/pos6": 3.7202651500701904, "train/pos_loss/weight0": 2.121622323989868, "train/pos_loss/weight1": 1.6535022258758545, "train/pos_loss/weight2": 1.2161762714385986, "train/pos_loss/weight3": 0.8797174692153931, "train/pos_loss/weight4": 0.5645070672035217, "train/pos_loss/weight5": 0.3377090394496918, "train/pos_loss/weight6": 0.2267654538154602, "train/pos_weight_progress": 0.20853080568720378, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4667980670928955 }, { "epoch": 2.085308056872038, "step": 880, "train/alpha": 0.2, "train/clean_ce_loss": 0.1025390625, "train/clean_loss_scale": 5.099879264831543, "train/clean_loss_weight": 1.0199759006500244, "train/combined_loss": 0.627524197101593, "train/pos_loss/pos0": 0.24780575931072235, "train/pos_loss/pos1": 0.48814627528190613, "train/pos_loss/pos2": 0.8718434572219849, "train/pos_loss/pos3": 1.249204158782959, "train/pos_loss/pos4": 1.7617015838623047, "train/pos_loss/pos5": 2.116651773452759, "train/pos_loss/pos6": 2.395433187484741, "train/pos_loss/weight0": 3.206536293029785, "train/pos_loss/weight1": 2.0398714542388916, "train/pos_loss/weight2": 1.118107557296753, "train/pos_loss/weight3": 0.46363991498947144, "train/pos_loss/weight4": 0.1362585872411728, "train/pos_loss/weight5": 0.029195450246334076, "train/pos_loss/weight6": 0.006391096394509077, "train/pos_weight_progress": 0.20853080568720378, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5229368209838867 }, { "epoch": 2.0876777251184833, "step": 881, "train/alpha": 0.2, "train/clean_ce_loss": 0.04248046875, "train/clean_loss_scale": 8.994730949401855, "train/clean_loss_weight": 1.7989462614059448, "train/combined_loss": 0.45852047204971313, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.20876777251184833, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.382100373506546 }, { "epoch": 2.0876777251184833, "step": 881, "train/alpha": 0.2, "train/clean_ce_loss": 0.0673828125, "train/clean_loss_scale": 12.846837997436523, "train/clean_loss_weight": 2.5693676471710205, "train/combined_loss": 1.0387872457504272, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.20876777251184833, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8656560778617859 }, { "epoch": 2.0876777251184833, "step": 881, "train/alpha": 0.2, "train/clean_ce_loss": 0.0615234375, "train/clean_loss_scale": 7.5631937980651855, "train/clean_loss_weight": 1.512638807296753, "train/combined_loss": 0.5583764314651489, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.20876777251184833, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4653136730194092 }, { "epoch": 2.0876777251184833, "step": 881, "train/alpha": 0.2, "train/clean_ce_loss": 0.04052734375, "train/clean_loss_scale": 11.240612983703613, "train/clean_loss_weight": 2.2481226921081543, "train/combined_loss": 0.5466626286506653, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.20876777251184833, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45555219054222107 }, { "epoch": 2.090047393364929, "step": 882, "train/alpha": 0.2, "train/clean_ce_loss": 0.037353515625, "train/clean_loss_scale": 8.533467292785645, "train/clean_loss_weight": 1.7066935300827026, "train/combined_loss": 0.38250601291656494, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.20900473933649288, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.31875500082969666 }, { "epoch": 2.090047393364929, "step": 882, "train/alpha": 0.2, "train/clean_ce_loss": 0.03564453125, "train/clean_loss_scale": 11.23320198059082, "train/clean_loss_weight": 2.24664044380188, "train/combined_loss": 0.48048266768455505, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.20900473933649288, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4004022181034088 }, { "epoch": 2.090047393364929, "step": 882, "train/alpha": 0.2, "train/clean_ce_loss": 0.0849609375, "train/clean_loss_scale": 9.053699493408203, "train/clean_loss_weight": 1.8107398748397827, "train/combined_loss": 0.923052966594696, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.20900473933649288, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7692108154296875 }, { "epoch": 2.090047393364929, "step": 882, "train/alpha": 0.2, "train/clean_ce_loss": 0.054443359375, "train/clean_loss_scale": 18.526260375976562, "train/clean_loss_weight": 3.705252170562744, "train/combined_loss": 1.2103581428527832, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.20900473933649288, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0086318254470825 }, { "epoch": 2.0924170616113744, "step": 883, "train/alpha": 0.2, "train/clean_ce_loss": 0.04150390625, "train/clean_loss_scale": 10.856791496276855, "train/clean_loss_weight": 2.171358346939087, "train/combined_loss": 0.5407190918922424, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.20924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4505992531776428 }, { "epoch": 2.0924170616113744, "step": 883, "train/alpha": 0.2, "train/clean_ce_loss": 0.07568359375, "train/clean_loss_scale": 11.259891510009766, "train/clean_loss_weight": 2.2519783973693848, "train/combined_loss": 1.0226268768310547, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.20924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8521890640258789 }, { "epoch": 2.0924170616113744, "step": 883, "train/alpha": 0.2, "train/clean_ce_loss": 0.06884765625, "train/clean_loss_scale": 8.753891944885254, "train/clean_loss_weight": 1.7507784366607666, "train/combined_loss": 0.7232219576835632, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.20924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6026849746704102 }, { "epoch": 2.0924170616113744, "step": 883, "train/alpha": 0.2, "train/clean_ce_loss": 0.07470703125, "train/clean_loss_scale": 14.281725883483887, "train/clean_loss_weight": 2.8563451766967773, "train/combined_loss": 1.2803343534469604, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.20924170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0669453144073486 }, { "epoch": 2.09478672985782, "step": 884, "train/alpha": 0.2, "train/clean_ce_loss": 0.09521484375, "train/clean_loss_scale": 9.235432624816895, "train/clean_loss_weight": 1.8470865488052368, "train/combined_loss": 1.055220365524292, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.209478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8793502449989319 }, { "epoch": 2.09478672985782, "step": 884, "train/alpha": 0.2, "train/clean_ce_loss": 0.059326171875, "train/clean_loss_scale": 8.691910743713379, "train/clean_loss_weight": 1.7383822202682495, "train/combined_loss": 0.6187893152236938, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.209478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5156577825546265 }, { "epoch": 2.09478672985782, "step": 884, "train/alpha": 0.2, "train/clean_ce_loss": 0.0751953125, "train/clean_loss_scale": 13.395613670349121, "train/clean_loss_weight": 2.6791226863861084, "train/combined_loss": 1.2087448835372925, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.209478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0072873830795288 }, { "epoch": 2.09478672985782, "step": 884, "train/alpha": 0.2, "train/clean_ce_loss": 0.09716796875, "train/clean_loss_scale": 10.396722793579102, "train/clean_loss_weight": 2.0793445110321045, "train/combined_loss": 1.2122740745544434, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.209478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0102283954620361 }, { "epoch": 2.0971563981042656, "grad_norm": 19.81438446044922, "learning_rate": 2e-06, "loss": 3.1767, "step": 885 }, { "epoch": 2.0971563981042656, "step": 885, "train/alpha": 0.2, "train/clean_ce_loss": 0.03466796875, "train/clean_loss_scale": 7.320372581481934, "train/clean_loss_weight": 1.4640744924545288, "train/combined_loss": 0.30453893542289734, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.20971563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2537824511528015 }, { "epoch": 2.0971563981042656, "step": 885, "train/alpha": 0.2, "train/clean_ce_loss": 0.06298828125, "train/clean_loss_scale": 10.62210750579834, "train/clean_loss_weight": 2.1244215965270996, "train/combined_loss": 0.8028819561004639, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.20971563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6690682768821716 }, { "epoch": 2.0971563981042656, "step": 885, "train/alpha": 0.2, "train/clean_ce_loss": 0.02490234375, "train/clean_loss_scale": 9.537735939025879, "train/clean_loss_weight": 1.9075472354888916, "train/combined_loss": 0.28501439094543457, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.20971563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.23751197755336761 }, { "epoch": 2.0971563981042656, "step": 885, "train/alpha": 0.2, "train/clean_ce_loss": 0.07666015625, "train/clean_loss_scale": 10.992502212524414, "train/clean_loss_weight": 2.198500394821167, "train/combined_loss": 1.0112242698669434, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.20971563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8426869511604309 }, { "epoch": 2.0995260663507107, "step": 886, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 13.744938850402832, "train/clean_loss_weight": 2.748987913131714, "train/combined_loss": 1.3369101285934448, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.2099526066350711, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1140917539596558 }, { "epoch": 2.0995260663507107, "step": 886, "train/alpha": 0.2, "train/clean_ce_loss": 0.0302734375, "train/clean_loss_scale": 12.869142532348633, "train/clean_loss_weight": 2.5738284587860107, "train/combined_loss": 0.4675118327140808, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.2099526066350711, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3895931839942932 }, { "epoch": 2.0995260663507107, "step": 886, "train/alpha": 0.2, "train/clean_ce_loss": 0.0869140625, "train/clean_loss_scale": 5.686753273010254, "train/clean_loss_weight": 1.1373506784439087, "train/combined_loss": 0.5931105613708496, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.2099526066350711, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4942588210105896 }, { "epoch": 2.0995260663507107, "step": 886, "train/alpha": 0.2, "train/clean_ce_loss": 0.0390625, "train/clean_loss_scale": 10.169742584228516, "train/clean_loss_weight": 2.0339486598968506, "train/combined_loss": 0.47670668363571167, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.2099526066350711, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3972555696964264 }, { "epoch": 2.1018957345971563, "step": 887, "train/alpha": 0.2, "train/clean_ce_loss": 0.076171875, "train/clean_loss_scale": 8.859457015991211, "train/clean_loss_weight": 1.771891474723816, "train/combined_loss": 0.8098097443580627, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.21018957345971565, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6748414635658264 }, { "epoch": 2.1018957345971563, "step": 887, "train/alpha": 0.2, "train/clean_ce_loss": 0.0294189453125, "train/clean_loss_scale": 10.707815170288086, "train/clean_loss_weight": 2.1415631771087646, "train/combined_loss": 0.3780151605606079, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.21018957345971565, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3150126338005066 }, { "epoch": 2.1018957345971563, "step": 887, "train/alpha": 0.2, "train/clean_ce_loss": 0.0693359375, "train/clean_loss_scale": 10.849553108215332, "train/clean_loss_weight": 2.1699106693267822, "train/combined_loss": 0.9027167558670044, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.21018957345971565, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7522639632225037 }, { "epoch": 2.1018957345971563, "step": 887, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 10.28420639038086, "train/clean_loss_weight": 2.0568413734436035, "train/combined_loss": 0.8134968280792236, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.21018957345971565, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.677914023399353 }, { "epoch": 2.104265402843602, "step": 888, "train/alpha": 0.2, "train/clean_ce_loss": 0.10693359375, "train/clean_loss_scale": 10.758683204650879, "train/clean_loss_weight": 2.1517367362976074, "train/combined_loss": 1.3805575370788574, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.2104265402843602, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.150464653968811 }, { "epoch": 2.104265402843602, "step": 888, "train/alpha": 0.2, "train/clean_ce_loss": 0.0027923583984375, "train/clean_loss_scale": 61.30677795410156, "train/clean_loss_weight": 12.26135540008545, "train/combined_loss": 0.2054286003112793, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.2104265402843602, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.17119050025939941 }, { "epoch": 2.104265402843602, "step": 888, "train/alpha": 0.2, "train/clean_ce_loss": 0.0439453125, "train/clean_loss_scale": 11.872896194458008, "train/clean_loss_weight": 2.3745791912078857, "train/combined_loss": 0.626109778881073, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.2104265402843602, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5217581391334534 }, { "epoch": 2.104265402843602, "step": 888, "train/alpha": 0.2, "train/clean_ce_loss": 0.076171875, "train/clean_loss_scale": 8.429880142211914, "train/clean_loss_weight": 1.6859760284423828, "train/combined_loss": 0.7705436944961548, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.2104265402843602, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6421197652816772 }, { "epoch": 2.1066350710900474, "step": 889, "train/alpha": 0.2, "train/clean_ce_loss": 0.09716796875, "train/clean_loss_scale": 10.32003116607666, "train/clean_loss_weight": 2.0640063285827637, "train/combined_loss": 1.2033318281173706, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.21066350710900475, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0027765035629272 }, { "epoch": 2.1066350710900474, "step": 889, "train/alpha": 0.2, "train/clean_ce_loss": 0.1005859375, "train/clean_loss_scale": 10.041234970092773, "train/clean_loss_weight": 2.008247137069702, "train/combined_loss": 1.2120084762573242, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.21066350710900475, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0100070238113403 }, { "epoch": 2.1066350710900474, "step": 889, "train/alpha": 0.2, "train/clean_ce_loss": 0.0693359375, "train/clean_loss_scale": 9.659224510192871, "train/clean_loss_weight": 1.93184494972229, "train/combined_loss": 0.8036776781082153, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.21066350710900475, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6697313785552979 }, { "epoch": 2.1066350710900474, "step": 889, "train/alpha": 0.2, "train/clean_ce_loss": 0.083984375, "train/clean_loss_scale": 9.734898567199707, "train/clean_loss_weight": 1.9469797611236572, "train/combined_loss": 0.9810952544212341, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.21066350710900475, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8175793886184692 }, { "epoch": 2.109004739336493, "grad_norm": 25.164810180664062, "learning_rate": 2e-06, "loss": 3.1958, "step": 890 }, { "epoch": 2.109004739336493, "step": 890, "train/alpha": 0.2, "train/clean_ce_loss": 0.091796875, "train/clean_loss_scale": 9.333137512207031, "train/clean_loss_weight": 1.86662757396698, "train/combined_loss": 1.0281034708023071, "train/pos_loss/pos0": 0.3007192313671112, "train/pos_loss/pos1": 0.5707884430885315, "train/pos_loss/pos2": 1.2992600202560425, "train/pos_loss/pos3": 1.5686427354812622, "train/pos_loss/pos4": 2.134265899658203, "train/pos_loss/pos5": 2.6924171447753906, "train/pos_loss/pos6": 3.3247790336608887, "train/pos_loss/weight0": 2.609515905380249, "train/pos_loss/weight1": 1.8645163774490356, "train/pos_loss/weight2": 1.2269285917282104, "train/pos_loss/weight3": 0.6999553442001343, "train/pos_loss/weight4": 0.3574221134185791, "train/pos_loss/weight5": 0.16349683701992035, "train/pos_loss/weight6": 0.07816509157419205, "train/pos_weight_progress": 0.2109004739336493, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.856752872467041 }, { "epoch": 2.109004739336493, "step": 890, "train/alpha": 0.2, "train/clean_ce_loss": 0.0556640625, "train/clean_loss_scale": 9.59603214263916, "train/clean_loss_weight": 1.9192065000534058, "train/combined_loss": 0.6409849524497986, "train/pos_loss/pos0": 0.19040757417678833, "train/pos_loss/pos1": 0.4307132661342621, "train/pos_loss/pos2": 0.8238632678985596, "train/pos_loss/pos3": 1.2189724445343018, "train/pos_loss/pos4": 1.6126539707183838, "train/pos_loss/pos5": 1.9295554161071777, "train/pos_loss/pos6": 2.2926406860351562, "train/pos_loss/weight0": 2.8850290775299072, "train/pos_loss/weight1": 1.9315072298049927, "train/pos_loss/weight2": 1.235573649406433, "train/pos_loss/weight3": 0.5893064737319946, "train/pos_loss/weight4": 0.24772411584854126, "train/pos_loss/weight5": 0.08143267035484314, "train/pos_loss/weight6": 0.029426738619804382, "train/pos_weight_progress": 0.2109004739336493, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5341541171073914 }, { "epoch": 2.109004739336493, "step": 890, "train/alpha": 0.2, "train/clean_ce_loss": 0.03173828125, "train/clean_loss_scale": 25.32602310180664, "train/clean_loss_weight": 5.065204620361328, "train/combined_loss": 0.9645653367042542, "train/pos_loss/pos0": 0.1309526115655899, "train/pos_loss/pos1": 0.4211352467536926, "train/pos_loss/pos2": 1.040987253189087, "train/pos_loss/pos3": 1.5455796718597412, "train/pos_loss/pos4": 2.320481777191162, "train/pos_loss/pos5": 3.0358388423919678, "train/pos_loss/pos6": 3.4556190967559814, "train/pos_loss/weight0": 2.420020580291748, "train/pos_loss/weight1": 1.8021206855773926, "train/pos_loss/weight2": 1.2335741519927979, "train/pos_loss/weight3": 0.7781153321266174, "train/pos_loss/weight4": 0.42531824111938477, "train/pos_loss/weight5": 0.2216247320175171, "train/pos_loss/weight6": 0.11922622472047806, "train/pos_weight_progress": 0.2109004739336493, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8038044571876526 }, { "epoch": 2.109004739336493, "step": 890, "train/alpha": 0.2, "train/clean_ce_loss": 0.0181884765625, "train/clean_loss_scale": 8.911050796508789, "train/clean_loss_weight": 1.7822102308273315, "train/combined_loss": 0.1944941282272339, "train/pos_loss/pos0": 0.01026767585426569, "train/pos_loss/pos1": 0.044556695967912674, "train/pos_loss/pos2": 0.17315244674682617, "train/pos_loss/pos3": 0.5541670918464661, "train/pos_loss/pos4": 0.9247039556503296, "train/pos_loss/pos5": 1.3680137395858765, "train/pos_loss/pos6": 1.7761353254318237, "train/pos_loss/weight0": 2.723891019821167, "train/pos_loss/weight1": 1.9399397373199463, "train/pos_loss/weight2": 1.2982728481292725, "train/pos_loss/weight3": 0.6341251134872437, "train/pos_loss/weight4": 0.2773151099681854, "train/pos_loss/weight5": 0.09204402565956116, "train/pos_loss/weight6": 0.03441207855939865, "train/pos_weight_progress": 0.2109004739336493, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.16207844018936157 }, { "epoch": 2.1113744075829386, "step": 891, "train/alpha": 0.2, "train/clean_ce_loss": 0.061767578125, "train/clean_loss_scale": 20.200929641723633, "train/clean_loss_weight": 4.040185928344727, "train/combined_loss": 1.4973149299621582, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21113744075829383, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2477624416351318 }, { "epoch": 2.1113744075829386, "step": 891, "train/alpha": 0.2, "train/clean_ce_loss": 0.006317138671875, "train/clean_loss_scale": 73.60466003417969, "train/clean_loss_weight": 14.720932006835938, "train/combined_loss": 0.5579650402069092, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21113744075829383, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4649708569049835 }, { "epoch": 2.1113744075829386, "step": 891, "train/alpha": 0.2, "train/clean_ce_loss": 0.034912109375, "train/clean_loss_scale": 17.573766708374023, "train/clean_loss_weight": 3.5147533416748047, "train/combined_loss": 0.7362446784973145, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21113744075829383, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6135372519493103 }, { "epoch": 2.1113744075829386, "step": 891, "train/alpha": 0.2, "train/clean_ce_loss": 0.0400390625, "train/clean_loss_scale": 15.496528625488281, "train/clean_loss_weight": 3.0993058681488037, "train/combined_loss": 0.7445597648620605, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21113744075829383, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6204664707183838 }, { "epoch": 2.1137440758293837, "step": 892, "train/alpha": 0.2, "train/clean_ce_loss": 0.0034332275390625, "train/clean_loss_scale": 170.06260681152344, "train/clean_loss_weight": 34.01252365112305, "train/combined_loss": 0.7006363272666931, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21137440758293838, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5838636159896851 }, { "epoch": 2.1137440758293837, "step": 892, "train/alpha": 0.2, "train/clean_ce_loss": 0.0556640625, "train/clean_loss_scale": 19.90289878845215, "train/clean_loss_weight": 3.9805798530578613, "train/combined_loss": 1.3294514417648315, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21137440758293838, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.107876181602478 }, { "epoch": 2.1137440758293837, "step": 892, "train/alpha": 0.2, "train/clean_ce_loss": 0.000629425048828125, "train/clean_loss_scale": 555.3265380859375, "train/clean_loss_weight": 111.0653076171875, "train/combined_loss": 0.4194437265396118, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21137440758293838, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3495364487171173 }, { "epoch": 2.1137440758293837, "step": 892, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 9.015985488891602, "train/clean_loss_weight": 1.8031971454620361, "train/combined_loss": 0.802986204624176, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21137440758293838, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6691551804542542 }, { "epoch": 2.1161137440758293, "step": 893, "train/alpha": 0.2, "train/clean_ce_loss": 0.11376953125, "train/clean_loss_scale": 9.04310417175293, "train/clean_loss_weight": 1.808620810508728, "train/combined_loss": 1.2345956563949585, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21161137440758293, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0288296937942505 }, { "epoch": 2.1161137440758293, "step": 893, "train/alpha": 0.2, "train/clean_ce_loss": 0.0283203125, "train/clean_loss_scale": 26.892667770385742, "train/clean_loss_weight": 5.378533840179443, "train/combined_loss": 0.9139305353164673, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21161137440758293, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7616087794303894 }, { "epoch": 2.1161137440758293, "step": 893, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 16.23210906982422, "train/clean_loss_weight": 3.2464218139648438, "train/combined_loss": 1.626381278038025, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21161137440758293, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3553177118301392 }, { "epoch": 2.1161137440758293, "step": 893, "train/alpha": 0.2, "train/clean_ce_loss": 0.04638671875, "train/clean_loss_scale": 25.453052520751953, "train/clean_loss_weight": 5.090610504150391, "train/combined_loss": 1.4168202877044678, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21161137440758293, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1806836128234863 }, { "epoch": 2.118483412322275, "step": 894, "train/alpha": 0.2, "train/clean_ce_loss": 0.0732421875, "train/clean_loss_scale": 16.544584274291992, "train/clean_loss_weight": 3.3089168071746826, "train/combined_loss": 1.4541139602661133, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21184834123222748, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2117615938186646 }, { "epoch": 2.118483412322275, "step": 894, "train/alpha": 0.2, "train/clean_ce_loss": 0.06787109375, "train/clean_loss_scale": 11.033590316772461, "train/clean_loss_weight": 2.2067182064056396, "train/combined_loss": 0.8986341953277588, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21184834123222748, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7488618493080139 }, { "epoch": 2.118483412322275, "step": 894, "train/alpha": 0.2, "train/clean_ce_loss": 0.060302734375, "train/clean_loss_scale": 16.839706420898438, "train/clean_loss_weight": 3.367941379547119, "train/combined_loss": 1.218576431274414, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21184834123222748, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.015480399131775 }, { "epoch": 2.118483412322275, "step": 894, "train/alpha": 0.2, "train/clean_ce_loss": 0.04736328125, "train/clean_loss_scale": 17.590482711791992, "train/clean_loss_weight": 3.518096685409546, "train/combined_loss": 0.9997715353965759, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21184834123222748, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8331429362297058 }, { "epoch": 2.1208530805687205, "grad_norm": 19.902790069580078, "learning_rate": 2e-06, "loss": 3.5673, "step": 895 }, { "epoch": 2.1208530805687205, "step": 895, "train/alpha": 0.2, "train/clean_ce_loss": 0.0849609375, "train/clean_loss_scale": 12.38044548034668, "train/clean_loss_weight": 2.4760892391204834, "train/combined_loss": 1.2622251510620117, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21208530805687204, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0518542528152466 }, { "epoch": 2.1208530805687205, "step": 895, "train/alpha": 0.2, "train/clean_ce_loss": 0.020263671875, "train/clean_loss_scale": 33.15629959106445, "train/clean_loss_weight": 6.631259918212891, "train/combined_loss": 0.806242048740387, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21208530805687204, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6718683838844299 }, { "epoch": 2.1208530805687205, "step": 895, "train/alpha": 0.2, "train/clean_ce_loss": 0.0179443359375, "train/clean_loss_scale": 64.56250762939453, "train/clean_loss_weight": 12.912501335144043, "train/combined_loss": 1.39023756980896, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21208530805687204, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1585313081741333 }, { "epoch": 2.1208530805687205, "step": 895, "train/alpha": 0.2, "train/clean_ce_loss": 0.004791259765625, "train/clean_loss_scale": 123.79747009277344, "train/clean_loss_weight": 24.75949478149414, "train/combined_loss": 0.711775004863739, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21208530805687204, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5931458473205566 }, { "epoch": 2.123222748815166, "step": 896, "train/alpha": 0.2, "train/clean_ce_loss": 0.12158203125, "train/clean_loss_scale": 11.071759223937988, "train/clean_loss_weight": 2.2143518924713135, "train/combined_loss": 1.6153523921966553, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.2123222748815166, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3461270332336426 }, { "epoch": 2.123222748815166, "step": 896, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 10.964422225952148, "train/clean_loss_weight": 2.1928844451904297, "train/combined_loss": 0.9379720687866211, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.2123222748815166, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7816433906555176 }, { "epoch": 2.123222748815166, "step": 896, "train/alpha": 0.2, "train/clean_ce_loss": 0.000743865966796875, "train/clean_loss_scale": 553.439453125, "train/clean_loss_weight": 110.6878890991211, "train/combined_loss": 0.4940217435359955, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.2123222748815166, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4116847813129425 }, { "epoch": 2.123222748815166, "step": 896, "train/alpha": 0.2, "train/clean_ce_loss": 0.1259765625, "train/clean_loss_scale": 10.5029296875, "train/clean_loss_weight": 2.1005859375, "train/combined_loss": 1.587747573852539, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.2123222748815166, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3231229782104492 }, { "epoch": 2.125592417061611, "step": 897, "train/alpha": 0.2, "train/clean_ce_loss": 0.0283203125, "train/clean_loss_scale": 36.456356048583984, "train/clean_loss_weight": 7.291271209716797, "train/combined_loss": 1.238946557044983, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21255924170616114, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0324554443359375 }, { "epoch": 2.125592417061611, "step": 897, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 15.145867347717285, "train/clean_loss_weight": 3.0291736125946045, "train/combined_loss": 1.2956815958023071, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21255924170616114, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0797346830368042 }, { "epoch": 2.125592417061611, "step": 897, "train/alpha": 0.2, "train/clean_ce_loss": 0.1064453125, "train/clean_loss_scale": 10.945348739624023, "train/clean_loss_weight": 2.1890697479248047, "train/combined_loss": 1.398097276687622, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21255924170616114, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1650810241699219 }, { "epoch": 2.125592417061611, "step": 897, "train/alpha": 0.2, "train/clean_ce_loss": 0.1005859375, "train/clean_loss_scale": 13.482828140258789, "train/clean_loss_weight": 2.696565628051758, "train/combined_loss": 1.6274194717407227, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21255924170616114, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3561829328536987 }, { "epoch": 2.1279620853080567, "step": 898, "train/alpha": 0.2, "train/clean_ce_loss": 0.1083984375, "train/clean_loss_scale": 12.979525566101074, "train/clean_loss_weight": 2.595905065536499, "train/combined_loss": 1.688352346420288, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.2127962085308057, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4069602489471436 }, { "epoch": 2.1279620853080567, "step": 898, "train/alpha": 0.2, "train/clean_ce_loss": 0.08056640625, "train/clean_loss_scale": 12.314842224121094, "train/clean_loss_weight": 2.462968587875366, "train/combined_loss": 1.1905951499938965, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.2127962085308057, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9921625852584839 }, { "epoch": 2.1279620853080567, "step": 898, "train/alpha": 0.2, "train/clean_ce_loss": 0.046630859375, "train/clean_loss_scale": 17.83476448059082, "train/clean_loss_weight": 3.56695294380188, "train/combined_loss": 0.997980535030365, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.2127962085308057, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.83165043592453 }, { "epoch": 2.1279620853080567, "step": 898, "train/alpha": 0.2, "train/clean_ce_loss": 0.055908203125, "train/clean_loss_scale": 11.033567428588867, "train/clean_loss_weight": 2.2067134380340576, "train/combined_loss": 0.7402403354644775, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.2127962085308057, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.616866946220398 }, { "epoch": 2.1303317535545023, "step": 899, "train/alpha": 0.2, "train/clean_ce_loss": 0.00011110305786132812, "train/clean_loss_scale": 2988.89208984375, "train/clean_loss_weight": 597.7784423828125, "train/combined_loss": 0.3984900712966919, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21303317535545024, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3320750594139099 }, { "epoch": 2.1303317535545023, "step": 899, "train/alpha": 0.2, "train/clean_ce_loss": 0.0296630859375, "train/clean_loss_scale": 20.228418350219727, "train/clean_loss_weight": 4.045683860778809, "train/combined_loss": 0.7200447916984558, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21303317535545024, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6000373363494873 }, { "epoch": 2.1303317535545023, "step": 899, "train/alpha": 0.2, "train/clean_ce_loss": 0.036376953125, "train/clean_loss_scale": 16.452205657958984, "train/clean_loss_weight": 3.2904412746429443, "train/combined_loss": 0.718177318572998, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21303317535545024, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5984811186790466 }, { "epoch": 2.1303317535545023, "step": 899, "train/alpha": 0.2, "train/clean_ce_loss": 0.04736328125, "train/clean_loss_scale": 16.152572631835938, "train/clean_loss_weight": 3.2305145263671875, "train/combined_loss": 0.9180465936660767, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.21303317535545024, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7650388479232788 }, { "epoch": 2.132701421800948, "grad_norm": 47.80033493041992, "learning_rate": 2e-06, "loss": 3.654, "step": 900 }, { "epoch": 2.132701421800948, "step": 900, "train/alpha": 0.2, "train/clean_ce_loss": 0.0908203125, "train/clean_loss_scale": 8.675736427307129, "train/clean_loss_weight": 1.7351473569869995, "train/combined_loss": 0.9455196857452393, "train/pos_loss/pos0": 0.08656778186559677, "train/pos_loss/pos1": 0.26817843317985535, "train/pos_loss/pos2": 0.6621039509773254, "train/pos_loss/pos3": 1.1570236682891846, "train/pos_loss/pos4": 1.6949431896209717, "train/pos_loss/pos5": 2.0995283126831055, "train/pos_loss/pos6": 2.6672353744506836, "train/pos_loss/weight0": 1.8752179145812988, "train/pos_loss/weight1": 1.5077117681503296, "train/pos_loss/weight2": 1.1804053783416748, "train/pos_loss/weight3": 0.9261980652809143, "train/pos_loss/weight4": 0.6793449521064758, "train/pos_loss/weight5": 0.47943925857543945, "train/pos_loss/weight6": 0.3516824543476105, "train/pos_weight_progress": 0.2132701421800948, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7879330515861511 }, { "epoch": 2.132701421800948, "step": 900, "train/alpha": 0.2, "train/clean_ce_loss": 0.08984375, "train/clean_loss_scale": 8.992992401123047, "train/clean_loss_weight": 1.7985985279083252, "train/combined_loss": 0.969556987285614, "train/pos_loss/pos0": 0.25563275814056396, "train/pos_loss/pos1": 0.5733183026313782, "train/pos_loss/pos2": 0.9894970655441284, "train/pos_loss/pos3": 1.3590631484985352, "train/pos_loss/pos4": 1.88768470287323, "train/pos_loss/pos5": 2.2834978103637695, "train/pos_loss/pos6": 2.5930190086364746, "train/pos_loss/weight0": 2.3582370281219482, "train/pos_loss/weight1": 1.7722595930099487, "train/pos_loss/weight2": 1.2625898122787476, "train/pos_loss/weight3": 0.8032269477844238, "train/pos_loss/weight4": 0.4492773413658142, "train/pos_loss/weight5": 0.22802799940109253, "train/pos_loss/weight6": 0.12638144195079803, "train/pos_weight_progress": 0.2132701421800948, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8079641461372375 }, { "epoch": 2.132701421800948, "step": 900, "train/alpha": 0.2, "train/clean_ce_loss": 0.0361328125, "train/clean_loss_scale": 18.05869483947754, "train/clean_loss_weight": 3.611738920211792, "train/combined_loss": 0.7830137014389038, "train/pos_loss/pos0": 0.21673017740249634, "train/pos_loss/pos1": 0.35716792941093445, "train/pos_loss/pos2": 0.6990859508514404, "train/pos_loss/pos3": 1.4266879558563232, "train/pos_loss/pos4": 1.9983153343200684, "train/pos_loss/pos5": 2.9896247386932373, "train/pos_loss/pos6": 3.966179609298706, "train/pos_loss/weight0": 2.603067636489868, "train/pos_loss/weight1": 1.8857970237731934, "train/pos_loss/weight2": 1.2141740322113037, "train/pos_loss/weight3": 0.709046483039856, "train/pos_loss/weight4": 0.3544449806213379, "train/pos_loss/weight5": 0.15832190215587616, "train/pos_loss/weight6": 0.07514803111553192, "train/pos_weight_progress": 0.2132701421800948, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6525114178657532 }, { "epoch": 2.132701421800948, "step": 900, "train/alpha": 0.2, "train/clean_ce_loss": 0.021240234375, "train/clean_loss_scale": 14.585131645202637, "train/clean_loss_weight": 2.9170262813568115, "train/combined_loss": 0.3717499375343323, "train/pos_loss/pos0": 0.014142449013888836, "train/pos_loss/pos1": 0.12080851942300797, "train/pos_loss/pos2": 0.29400834441185, "train/pos_loss/pos3": 0.8441765308380127, "train/pos_loss/pos4": 1.4904890060424805, "train/pos_loss/pos5": 2.0015299320220947, "train/pos_loss/pos6": 2.5102944374084473, "train/pos_loss/weight0": 2.617197275161743, "train/pos_loss/weight1": 1.8721027374267578, "train/pos_loss/weight2": 1.2251594066619873, "train/pos_loss/weight3": 0.7552547454833984, "train/pos_loss/weight4": 0.3517100214958191, "train/pos_loss/weight5": 0.12525972723960876, "train/pos_loss/weight6": 0.05331601947546005, "train/pos_weight_progress": 0.2132701421800948, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.309791624546051 }, { "epoch": 2.1350710900473935, "step": 901, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 13.179292678833008, "train/clean_loss_weight": 2.6358585357666016, "train/combined_loss": 1.0347803831100464, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21350710900473935, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8623170256614685 }, { "epoch": 2.1350710900473935, "step": 901, "train/alpha": 0.2, "train/clean_ce_loss": 0.055908203125, "train/clean_loss_scale": 10.453063011169434, "train/clean_loss_weight": 2.0906126499176025, "train/combined_loss": 0.7012943625450134, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21350710900473935, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5844119787216187 }, { "epoch": 2.1350710900473935, "step": 901, "train/alpha": 0.2, "train/clean_ce_loss": 0.0703125, "train/clean_loss_scale": 8.326519966125488, "train/clean_loss_weight": 1.6653040647506714, "train/combined_loss": 0.7025501728057861, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21350710900473935, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5854584574699402 }, { "epoch": 2.1350710900473935, "step": 901, "train/alpha": 0.2, "train/clean_ce_loss": 0.049560546875, "train/clean_loss_scale": 14.169276237487793, "train/clean_loss_weight": 2.833855390548706, "train/combined_loss": 0.8426845073699951, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21350710900473935, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.702237069606781 }, { "epoch": 2.137440758293839, "step": 902, "train/alpha": 0.2, "train/clean_ce_loss": 0.036376953125, "train/clean_loss_scale": 22.42833137512207, "train/clean_loss_weight": 4.485666275024414, "train/combined_loss": 0.9790492057800293, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21374407582938387, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8158743381500244 }, { "epoch": 2.137440758293839, "step": 902, "train/alpha": 0.2, "train/clean_ce_loss": 0.1005859375, "train/clean_loss_scale": 11.350996971130371, "train/clean_loss_weight": 2.2701995372772217, "train/combined_loss": 1.3701008558273315, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21374407582938387, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.141750693321228 }, { "epoch": 2.137440758293839, "step": 902, "train/alpha": 0.2, "train/clean_ce_loss": 0.10498046875, "train/clean_loss_scale": 9.969292640686035, "train/clean_loss_weight": 1.9938585758209229, "train/combined_loss": 1.2558972835540771, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21374407582938387, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0465810298919678 }, { "epoch": 2.137440758293839, "step": 902, "train/alpha": 0.2, "train/clean_ce_loss": 0.007110595703125, "train/clean_loss_scale": 46.8958625793457, "train/clean_loss_weight": 9.379172325134277, "train/combined_loss": 0.40014904737472534, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21374407582938387, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3334575295448303 }, { "epoch": 2.139810426540284, "step": 903, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 9.816587448120117, "train/clean_loss_weight": 1.9633175134658813, "train/combined_loss": 0.8340264558792114, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21398104265402842, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6950220465660095 }, { "epoch": 2.139810426540284, "step": 903, "train/alpha": 0.2, "train/clean_ce_loss": 0.03857421875, "train/clean_loss_scale": 11.713357925415039, "train/clean_loss_weight": 2.3426716327667236, "train/combined_loss": 0.5422003865242004, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21398104265402842, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4518336355686188 }, { "epoch": 2.139810426540284, "step": 903, "train/alpha": 0.2, "train/clean_ce_loss": 0.0615234375, "train/clean_loss_scale": 9.77042007446289, "train/clean_loss_weight": 1.954084038734436, "train/combined_loss": 0.721331775188446, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21398104265402842, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6011098027229309 }, { "epoch": 2.139810426540284, "step": 903, "train/alpha": 0.2, "train/clean_ce_loss": 0.0291748046875, "train/clean_loss_scale": 19.455358505249023, "train/clean_loss_weight": 3.8910717964172363, "train/combined_loss": 0.6811275482177734, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21398104265402842, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5676062703132629 }, { "epoch": 2.1421800947867298, "step": 904, "train/alpha": 0.2, "train/clean_ce_loss": 0.134765625, "train/clean_loss_scale": 8.538941383361816, "train/clean_loss_weight": 1.707788348197937, "train/combined_loss": 1.3809069395065308, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21421800947867298, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.150755763053894 }, { "epoch": 2.1421800947867298, "step": 904, "train/alpha": 0.2, "train/clean_ce_loss": 0.1640625, "train/clean_loss_scale": 8.598799705505371, "train/clean_loss_weight": 1.7197599411010742, "train/combined_loss": 1.6928887367248535, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21421800947867298, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.410740613937378 }, { "epoch": 2.1421800947867298, "step": 904, "train/alpha": 0.2, "train/clean_ce_loss": 0.08447265625, "train/clean_loss_scale": 10.56353759765625, "train/clean_loss_weight": 2.1127076148986816, "train/combined_loss": 1.070796012878418, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21421800947867298, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8923300504684448 }, { "epoch": 2.1421800947867298, "step": 904, "train/alpha": 0.2, "train/clean_ce_loss": 0.0019683837890625, "train/clean_loss_scale": 153.46971130371094, "train/clean_loss_weight": 30.69394302368164, "train/combined_loss": 0.3625047206878662, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21421800947867298, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.30208727717399597 }, { "epoch": 2.1445497630331753, "grad_norm": 19.38420867919922, "learning_rate": 2e-06, "loss": 3.3848, "step": 905 }, { "epoch": 2.1445497630331753, "step": 905, "train/alpha": 0.2, "train/clean_ce_loss": 0.06005859375, "train/clean_loss_scale": 10.077117919921875, "train/clean_loss_weight": 2.015423536300659, "train/combined_loss": 0.7262610197067261, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21445497630331753, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6052175164222717 }, { "epoch": 2.1445497630331753, "step": 905, "train/alpha": 0.2, "train/clean_ce_loss": 0.1240234375, "train/clean_loss_scale": 9.844684600830078, "train/clean_loss_weight": 1.9689369201660156, "train/combined_loss": 1.4651658535003662, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21445497630331753, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2209715843200684 }, { "epoch": 2.1445497630331753, "step": 905, "train/alpha": 0.2, "train/clean_ce_loss": 0.08447265625, "train/clean_loss_scale": 11.323829650878906, "train/clean_loss_weight": 2.264765977859497, "train/combined_loss": 1.1478646993637085, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21445497630331753, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9565539360046387 }, { "epoch": 2.1445497630331753, "step": 905, "train/alpha": 0.2, "train/clean_ce_loss": 0.1787109375, "train/clean_loss_scale": 7.418941497802734, "train/clean_loss_weight": 1.4837883710861206, "train/combined_loss": 1.591015100479126, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21445497630331753, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3258459568023682 }, { "epoch": 2.146919431279621, "step": 906, "train/alpha": 0.2, "train/clean_ce_loss": 0.060791015625, "train/clean_loss_scale": 20.211814880371094, "train/clean_loss_weight": 4.042363166809082, "train/combined_loss": 1.4744360446929932, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21469194312796208, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2286967039108276 }, { "epoch": 2.146919431279621, "step": 906, "train/alpha": 0.2, "train/clean_ce_loss": 0.007293701171875, "train/clean_loss_scale": 29.18984603881836, "train/clean_loss_weight": 5.8379693031311035, "train/combined_loss": 0.25548240542411804, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21469194312796208, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2129020094871521 }, { "epoch": 2.146919431279621, "step": 906, "train/alpha": 0.2, "train/clean_ce_loss": 0.0177001953125, "train/clean_loss_scale": 13.295900344848633, "train/clean_loss_weight": 2.659180164337158, "train/combined_loss": 0.2824080288410187, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21469194312796208, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.23534002900123596 }, { "epoch": 2.146919431279621, "step": 906, "train/alpha": 0.2, "train/clean_ce_loss": 0.154296875, "train/clean_loss_scale": 7.658671855926514, "train/clean_loss_weight": 1.5317343473434448, "train/combined_loss": 1.41805100440979, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21469194312796208, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1817091703414917 }, { "epoch": 2.1492890995260665, "step": 907, "train/alpha": 0.2, "train/clean_ce_loss": 0.049072265625, "train/clean_loss_scale": 12.251233100891113, "train/clean_loss_weight": 2.450246572494507, "train/combined_loss": 0.7214348912239075, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21492890995260663, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.601195752620697 }, { "epoch": 2.1492890995260665, "step": 907, "train/alpha": 0.2, "train/clean_ce_loss": 0.00616455078125, "train/clean_loss_scale": 54.0379638671875, "train/clean_loss_weight": 10.80759334564209, "train/combined_loss": 0.3997437357902527, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21492890995260663, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.33311977982521057 }, { "epoch": 2.1492890995260665, "step": 907, "train/alpha": 0.2, "train/clean_ce_loss": 8.106231689453125e-06, "train/clean_loss_scale": 19090.794921875, "train/clean_loss_weight": 3818.158935546875, "train/combined_loss": 0.1857052743434906, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21492890995260663, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1547544002532959 }, { "epoch": 2.1492890995260665, "step": 907, "train/alpha": 0.2, "train/clean_ce_loss": 0.0106201171875, "train/clean_loss_scale": 39.340721130371094, "train/clean_loss_weight": 7.868144512176514, "train/combined_loss": 0.5013636350631714, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21492890995260663, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4178030490875244 }, { "epoch": 2.1516587677725116, "step": 908, "train/alpha": 0.2, "train/clean_ce_loss": 0.032958984375, "train/clean_loss_scale": 14.90359115600586, "train/clean_loss_weight": 2.9807183742523193, "train/combined_loss": 0.5894486904144287, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21516587677725119, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4912072420120239 }, { "epoch": 2.1516587677725116, "step": 908, "train/alpha": 0.2, "train/clean_ce_loss": 0.004791259765625, "train/clean_loss_scale": 58.02057647705078, "train/clean_loss_weight": 11.60411548614502, "train/combined_loss": 0.33358997106552124, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21516587677725119, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2779916524887085 }, { "epoch": 2.1516587677725116, "step": 908, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 10.432891845703125, "train/clean_loss_weight": 2.086578369140625, "train/combined_loss": 1.1614742279052734, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21516587677725119, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9678952097892761 }, { "epoch": 2.1516587677725116, "step": 908, "train/alpha": 0.2, "train/clean_ce_loss": 0.0673828125, "train/clean_loss_scale": 10.259078025817871, "train/clean_loss_weight": 2.0518157482147217, "train/combined_loss": 0.8295426368713379, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21516587677725119, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6912855505943298 }, { "epoch": 2.154028436018957, "step": 909, "train/alpha": 0.2, "train/clean_ce_loss": 0.091796875, "train/clean_loss_scale": 9.82308292388916, "train/clean_loss_weight": 1.9646166563034058, "train/combined_loss": 1.0820739269256592, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21540284360189574, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9017282724380493 }, { "epoch": 2.154028436018957, "step": 909, "train/alpha": 0.2, "train/clean_ce_loss": 0.12255859375, "train/clean_loss_scale": 10.816119194030762, "train/clean_loss_weight": 2.1632239818573, "train/combined_loss": 1.59073007106781, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21540284360189574, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3256083726882935 }, { "epoch": 2.154028436018957, "step": 909, "train/alpha": 0.2, "train/clean_ce_loss": 0.0223388671875, "train/clean_loss_scale": 18.788429260253906, "train/clean_loss_weight": 3.757685899734497, "train/combined_loss": 0.5036546587944031, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21540284360189574, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.41971221566200256 }, { "epoch": 2.154028436018957, "step": 909, "train/alpha": 0.2, "train/clean_ce_loss": 0.038330078125, "train/clean_loss_scale": 27.125289916992188, "train/clean_loss_weight": 5.425057888031006, "train/combined_loss": 1.247657299041748, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.21540284360189574, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0397144556045532 }, { "epoch": 2.156398104265403, "grad_norm": 20.696609497070312, "learning_rate": 2e-06, "loss": 3.2011, "step": 910 }, { "epoch": 2.156398104265403, "step": 910, "train/alpha": 0.2, "train/clean_ce_loss": 0.056396484375, "train/clean_loss_scale": 11.813496589660645, "train/clean_loss_weight": 2.362699270248413, "train/combined_loss": 0.7994875907897949, "train/pos_loss/pos0": 0.20813995599746704, "train/pos_loss/pos1": 0.4421141743659973, "train/pos_loss/pos2": 0.7588568925857544, "train/pos_loss/pos3": 1.0794172286987305, "train/pos_loss/pos4": 1.385333776473999, "train/pos_loss/pos5": 1.73016357421875, "train/pos_loss/pos6": 2.1200411319732666, "train/pos_loss/weight0": 2.222639799118042, "train/pos_loss/weight1": 1.6946851015090942, "train/pos_loss/weight2": 1.2352253198623657, "train/pos_loss/weight3": 0.870679497718811, "train/pos_loss/weight4": 0.5239367485046387, "train/pos_loss/weight5": 0.2836662530899048, "train/pos_loss/weight6": 0.16916778683662415, "train/pos_weight_progress": 0.2156398104265403, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6662396788597107 }, { "epoch": 2.156398104265403, "step": 910, "train/alpha": 0.2, "train/clean_ce_loss": 0.15625, "train/clean_loss_scale": 7.7241621017456055, "train/clean_loss_weight": 1.544832468032837, "train/combined_loss": 1.4482804536819458, "train/pos_loss/pos0": 0.48945489525794983, "train/pos_loss/pos1": 0.9204696416854858, "train/pos_loss/pos2": 1.3950016498565674, "train/pos_loss/pos3": 1.9452567100524902, "train/pos_loss/pos4": 2.63372540473938, "train/pos_loss/pos5": 2.953780174255371, "train/pos_loss/pos6": 3.395158529281616, "train/pos_loss/weight0": 2.363574266433716, "train/pos_loss/weight1": 1.7716076374053955, "train/pos_loss/weight2": 1.2332355976104736, "train/pos_loss/weight3": 0.7871922850608826, "train/pos_loss/weight4": 0.45626795291900635, "train/pos_loss/weight5": 0.24647074937820435, "train/pos_loss/weight6": 0.14165163040161133, "train/pos_weight_progress": 0.2156398104265403, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2069003582000732 }, { "epoch": 2.156398104265403, "step": 910, "train/alpha": 0.2, "train/clean_ce_loss": 0.08544921875, "train/clean_loss_scale": 7.350528717041016, "train/clean_loss_weight": 1.470105767250061, "train/combined_loss": 0.7537163496017456, "train/pos_loss/pos0": 0.2752310037612915, "train/pos_loss/pos1": 0.6213318705558777, "train/pos_loss/pos2": 1.019142746925354, "train/pos_loss/pos3": 1.452998161315918, "train/pos_loss/pos4": 1.8888585567474365, "train/pos_loss/pos5": 2.208603620529175, "train/pos_loss/pos6": 2.5174715518951416, "train/pos_loss/weight0": 3.1629083156585693, "train/pos_loss/weight1": 2.0376670360565186, "train/pos_loss/weight2": 1.090018391609192, "train/pos_loss/weight3": 0.48319587111473083, "train/pos_loss/weight4": 0.17181138694286346, "train/pos_loss/weight5": 0.04284583777189255, "train/pos_loss/weight6": 0.011553903110325336, "train/pos_weight_progress": 0.2156398104265403, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6280969381332397 }, { "epoch": 2.156398104265403, "step": 910, "train/alpha": 0.2, "train/clean_ce_loss": 0.058837890625, "train/clean_loss_scale": 10.717730522155762, "train/clean_loss_weight": 2.1435461044311523, "train/combined_loss": 0.7567304372787476, "train/pos_loss/pos0": 0.20349720120429993, "train/pos_loss/pos1": 0.4750721752643585, "train/pos_loss/pos2": 0.8765066862106323, "train/pos_loss/pos3": 1.2704194784164429, "train/pos_loss/pos4": 1.612990140914917, "train/pos_loss/pos5": 2.064349889755249, "train/pos_loss/pos6": 2.381164312362671, "train/pos_loss/weight0": 2.6308236122131348, "train/pos_loss/weight1": 1.9044032096862793, "train/pos_loss/weight2": 1.2076199054718018, "train/pos_loss/weight3": 0.6836473345756531, "train/pos_loss/weight4": 0.34919244050979614, "train/pos_loss/weight5": 0.1530306041240692, "train/pos_loss/weight6": 0.07128316164016724, "train/pos_weight_progress": 0.2156398104265403, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6306086778640747 }, { "epoch": 2.1587677725118484, "step": 911, "train/alpha": 0.2, "train/clean_ce_loss": 0.04248046875, "train/clean_loss_scale": 13.081711769104004, "train/clean_loss_weight": 2.616342306137085, "train/combined_loss": 0.6668606996536255, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21587677725118484, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5557172298431396 }, { "epoch": 2.1587677725118484, "step": 911, "train/alpha": 0.2, "train/clean_ce_loss": 0.034423828125, "train/clean_loss_scale": 3.140188694000244, "train/clean_loss_weight": 0.6280377507209778, "train/combined_loss": 0.12971678376197815, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21587677725118484, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.10809731483459473 }, { "epoch": 2.1587677725118484, "step": 911, "train/alpha": 0.2, "train/clean_ce_loss": 0.009521484375, "train/clean_loss_scale": 37.98373031616211, "train/clean_loss_weight": 7.59674596786499, "train/combined_loss": 0.43399378657341003, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21587677725118484, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.36166149377822876 }, { "epoch": 2.1587677725118484, "step": 911, "train/alpha": 0.2, "train/clean_ce_loss": 0.03125, "train/clean_loss_scale": 14.48952865600586, "train/clean_loss_weight": 2.8979058265686035, "train/combined_loss": 0.5433573126792908, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21587677725118484, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4527977705001831 }, { "epoch": 2.161137440758294, "step": 912, "train/alpha": 0.2, "train/clean_ce_loss": 0.024169921875, "train/clean_loss_scale": 24.015594482421875, "train/clean_loss_weight": 4.80311918258667, "train/combined_loss": 0.6965460777282715, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.2161137440758294, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5804550647735596 }, { "epoch": 2.161137440758294, "step": 912, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 9.969202995300293, "train/clean_loss_weight": 1.9938405752182007, "train/combined_loss": 0.946295440196991, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.2161137440758294, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.788579523563385 }, { "epoch": 2.161137440758294, "step": 912, "train/alpha": 0.2, "train/clean_ce_loss": 0.1015625, "train/clean_loss_scale": 6.796653747558594, "train/clean_loss_weight": 1.3593307733535767, "train/combined_loss": 0.8283421993255615, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.2161137440758294, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6902851462364197 }, { "epoch": 2.161137440758294, "step": 912, "train/alpha": 0.2, "train/clean_ce_loss": 0.007110595703125, "train/clean_loss_scale": 28.59174156188965, "train/clean_loss_weight": 5.718348503112793, "train/combined_loss": 0.24396517872810364, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.2161137440758294, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.20330432057380676 }, { "epoch": 2.1635071090047395, "step": 913, "train/alpha": 0.2, "train/clean_ce_loss": 0.032958984375, "train/clean_loss_scale": 13.249238967895508, "train/clean_loss_weight": 2.6498477458953857, "train/combined_loss": 0.5240177512168884, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21635071090047395, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43668144941329956 }, { "epoch": 2.1635071090047395, "step": 913, "train/alpha": 0.2, "train/clean_ce_loss": 0.01092529296875, "train/clean_loss_scale": 20.732955932617188, "train/clean_loss_weight": 4.1465911865234375, "train/combined_loss": 0.27181634306907654, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21635071090047395, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.22651360929012299 }, { "epoch": 2.1635071090047395, "step": 913, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 8.141693115234375, "train/clean_loss_weight": 1.6283386945724487, "train/combined_loss": 0.9063994884490967, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21635071090047395, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.755332887172699 }, { "epoch": 2.1635071090047395, "step": 913, "train/alpha": 0.2, "train/clean_ce_loss": 0.0908203125, "train/clean_loss_scale": 7.956089496612549, "train/clean_loss_weight": 1.5912178754806519, "train/combined_loss": 0.8670894503593445, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21635071090047395, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7225745320320129 }, { "epoch": 2.1658767772511847, "step": 914, "train/alpha": 0.2, "train/clean_ce_loss": 0.1337890625, "train/clean_loss_scale": 8.227941513061523, "train/clean_loss_weight": 1.6455882787704468, "train/combined_loss": 1.3209702968597412, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21658767772511847, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1008086204528809 }, { "epoch": 2.1658767772511847, "step": 914, "train/alpha": 0.2, "train/clean_ce_loss": 0.1044921875, "train/clean_loss_scale": 8.311233520507812, "train/clean_loss_weight": 1.6622467041015625, "train/combined_loss": 1.0421507358551025, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21658767772511847, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8684589266777039 }, { "epoch": 2.1658767772511847, "step": 914, "train/alpha": 0.2, "train/clean_ce_loss": 0.0390625, "train/clean_loss_scale": 18.141630172729492, "train/clean_loss_weight": 3.628326177597046, "train/combined_loss": 0.8503889441490173, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21658767772511847, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.708657443523407 }, { "epoch": 2.1658767772511847, "step": 914, "train/alpha": 0.2, "train/clean_ce_loss": 0.021240234375, "train/clean_loss_scale": 28.256412506103516, "train/clean_loss_weight": 5.651282787322998, "train/combined_loss": 0.7202073931694031, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21658767772511847, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6001728177070618 }, { "epoch": 2.1682464454976302, "grad_norm": 15.33160400390625, "learning_rate": 2e-06, "loss": 3.2272, "step": 915 }, { "epoch": 2.1682464454976302, "step": 915, "train/alpha": 0.2, "train/clean_ce_loss": 0.1240234375, "train/clean_loss_scale": 9.75255298614502, "train/clean_loss_weight": 1.9505106210708618, "train/combined_loss": 1.4514541625976562, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21682464454976302, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2095451354980469 }, { "epoch": 2.1682464454976302, "step": 915, "train/alpha": 0.2, "train/clean_ce_loss": 0.0966796875, "train/clean_loss_scale": 8.770395278930664, "train/clean_loss_weight": 1.7540791034698486, "train/combined_loss": 1.0175029039382935, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21682464454976302, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8479190468788147 }, { "epoch": 2.1682464454976302, "step": 915, "train/alpha": 0.2, "train/clean_ce_loss": 0.008544921875, "train/clean_loss_scale": 25.725345611572266, "train/clean_loss_weight": 5.145069122314453, "train/combined_loss": 0.2637852728366852, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21682464454976302, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.21982106566429138 }, { "epoch": 2.1682464454976302, "step": 915, "train/alpha": 0.2, "train/clean_ce_loss": 0.060302734375, "train/clean_loss_scale": 9.505097389221191, "train/clean_loss_weight": 1.9010194540023804, "train/combined_loss": 0.6878200173377991, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21682464454976302, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5731833577156067 }, { "epoch": 2.170616113744076, "step": 916, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 8.54051685333252, "train/clean_loss_weight": 1.7081034183502197, "train/combined_loss": 0.7306145429611206, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21706161137440758, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6088454723358154 }, { "epoch": 2.170616113744076, "step": 916, "train/alpha": 0.2, "train/clean_ce_loss": 0.07666015625, "train/clean_loss_scale": 8.22978687286377, "train/clean_loss_weight": 1.645957350730896, "train/combined_loss": 0.7570760846138, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21706161137440758, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6308967471122742 }, { "epoch": 2.170616113744076, "step": 916, "train/alpha": 0.2, "train/clean_ce_loss": 0.0673828125, "train/clean_loss_scale": 12.604240417480469, "train/clean_loss_weight": 2.520848035812378, "train/combined_loss": 1.0191709995269775, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21706161137440758, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8493091464042664 }, { "epoch": 2.170616113744076, "step": 916, "train/alpha": 0.2, "train/clean_ce_loss": 0.060791015625, "train/clean_loss_scale": 9.286409378051758, "train/clean_loss_weight": 1.8572819232940674, "train/combined_loss": 0.6774362921714783, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21706161137440758, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5645302534103394 }, { "epoch": 2.1729857819905214, "step": 917, "train/alpha": 0.2, "train/clean_ce_loss": 0.06689453125, "train/clean_loss_scale": 9.978658676147461, "train/clean_loss_weight": 1.9957317113876343, "train/combined_loss": 0.8010212779045105, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21729857819905213, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6675177216529846 }, { "epoch": 2.1729857819905214, "step": 917, "train/alpha": 0.2, "train/clean_ce_loss": 0.006439208984375, "train/clean_loss_scale": 40.24709701538086, "train/clean_loss_weight": 8.049419403076172, "train/combined_loss": 0.3109913766384125, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21729857819905213, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.25915947556495667 }, { "epoch": 2.1729857819905214, "step": 917, "train/alpha": 0.2, "train/clean_ce_loss": 0.0181884765625, "train/clean_loss_scale": 35.66767883300781, "train/clean_loss_weight": 7.133535861968994, "train/combined_loss": 0.7784889340400696, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21729857819905213, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6487407684326172 }, { "epoch": 2.1729857819905214, "step": 917, "train/alpha": 0.2, "train/clean_ce_loss": 0.07666015625, "train/clean_loss_scale": 11.863597869873047, "train/clean_loss_weight": 2.3727195262908936, "train/combined_loss": 1.0913583040237427, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21729857819905213, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9094652533531189 }, { "epoch": 2.175355450236967, "step": 918, "train/alpha": 0.2, "train/clean_ce_loss": 0.0908203125, "train/clean_loss_scale": 8.625597953796387, "train/clean_loss_weight": 1.7251195907592773, "train/combined_loss": 0.9400553703308105, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21753554502369668, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7833794951438904 }, { "epoch": 2.175355450236967, "step": 918, "train/alpha": 0.2, "train/clean_ce_loss": 0.018798828125, "train/clean_loss_scale": 13.658463478088379, "train/clean_loss_weight": 2.7316927909851074, "train/combined_loss": 0.30811572074890137, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21753554502369668, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2567631006240845 }, { "epoch": 2.175355450236967, "step": 918, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 5.826995372772217, "train/clean_loss_weight": 1.1653990745544434, "train/combined_loss": 0.518966794013977, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21753554502369668, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4324723184108734 }, { "epoch": 2.175355450236967, "step": 918, "train/alpha": 0.2, "train/clean_ce_loss": 0.0390625, "train/clean_loss_scale": 7.865306854248047, "train/clean_loss_weight": 1.5730613470077515, "train/combined_loss": 0.3686862587928772, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21753554502369668, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.30723854899406433 }, { "epoch": 2.177725118483412, "step": 919, "train/alpha": 0.2, "train/clean_ce_loss": 0.05078125, "train/clean_loss_scale": 17.73126983642578, "train/clean_loss_weight": 3.5462539196014404, "train/combined_loss": 1.080499291419983, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21777251184834123, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9004160761833191 }, { "epoch": 2.177725118483412, "step": 919, "train/alpha": 0.2, "train/clean_ce_loss": 0.072265625, "train/clean_loss_scale": 9.013970375061035, "train/clean_loss_weight": 1.802794098854065, "train/combined_loss": 0.7816802263259888, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21777251184834123, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6514002084732056 }, { "epoch": 2.177725118483412, "step": 919, "train/alpha": 0.2, "train/clean_ce_loss": 0.10400390625, "train/clean_loss_scale": 8.51148796081543, "train/clean_loss_weight": 1.702297568321228, "train/combined_loss": 1.062273621559143, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21777251184834123, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8852280378341675 }, { "epoch": 2.177725118483412, "step": 919, "train/alpha": 0.2, "train/clean_ce_loss": 0.0498046875, "train/clean_loss_scale": 11.960652351379395, "train/clean_loss_weight": 2.3921306133270264, "train/combined_loss": 0.7148358821868896, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21777251184834123, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5956965684890747 }, { "epoch": 2.1800947867298577, "grad_norm": 18.63193130493164, "learning_rate": 2e-06, "loss": 3.231, "step": 920 }, { "epoch": 2.1800947867298577, "step": 920, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 7.698782444000244, "train/clean_loss_weight": 1.5397565364837646, "train/combined_loss": 0.6540957689285278, "train/pos_loss/pos0": 0.1492014080286026, "train/pos_loss/pos1": 0.3700632154941559, "train/pos_loss/pos2": 0.7176386117935181, "train/pos_loss/pos3": 0.9904311895370483, "train/pos_loss/pos4": 1.3998013734817505, "train/pos_loss/pos5": 1.8315165042877197, "train/pos_loss/pos6": 2.2375717163085938, "train/pos_loss/weight0": 2.4887704849243164, "train/pos_loss/weight1": 1.8296104669570923, "train/pos_loss/weight2": 1.251021385192871, "train/pos_loss/weight3": 0.7348074316978455, "train/pos_loss/weight4": 0.3997729420661926, "train/pos_loss/weight5": 0.19769267737865448, "train/pos_loss/weight6": 0.0983242392539978, "train/pos_weight_progress": 0.21800947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5450798273086548 }, { "epoch": 2.1800947867298577, "step": 920, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 8.729903221130371, "train/clean_loss_weight": 1.7459806203842163, "train/combined_loss": 0.7621614933013916, "train/pos_loss/pos0": 0.19678349792957306, "train/pos_loss/pos1": 0.3831161558628082, "train/pos_loss/pos2": 0.7635220885276794, "train/pos_loss/pos3": 1.1487629413604736, "train/pos_loss/pos4": 1.4919660091400146, "train/pos_loss/pos5": 1.9265105724334717, "train/pos_loss/pos6": 2.2741103172302246, "train/pos_loss/weight0": 2.370326042175293, "train/pos_loss/weight1": 1.7723313570022583, "train/pos_loss/weight2": 1.2410204410552979, "train/pos_loss/weight3": 0.790174126625061, "train/pos_loss/weight4": 0.44621801376342773, "train/pos_loss/weight5": 0.2424294650554657, "train/pos_loss/weight6": 0.13750073313713074, "train/pos_weight_progress": 0.21800947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6351345777511597 }, { "epoch": 2.1800947867298577, "step": 920, "train/alpha": 0.2, "train/clean_ce_loss": 0.1064453125, "train/clean_loss_scale": 11.109748840332031, "train/clean_loss_weight": 2.221949815750122, "train/combined_loss": 1.419096827507019, "train/pos_loss/pos0": 0.31625327467918396, "train/pos_loss/pos1": 0.8514384031295776, "train/pos_loss/pos2": 1.446009874343872, "train/pos_loss/pos3": 2.076885461807251, "train/pos_loss/pos4": 2.6678969860076904, "train/pos_loss/pos5": 3.2045321464538574, "train/pos_loss/pos6": 3.6101839542388916, "train/pos_loss/weight0": 2.3406167030334473, "train/pos_loss/weight1": 1.7480037212371826, "train/pos_loss/weight2": 1.244853138923645, "train/pos_loss/weight3": 0.8006643652915955, "train/pos_loss/weight4": 0.4585846960544586, "train/pos_loss/weight5": 0.25881513953208923, "train/pos_loss/weight6": 0.14846210181713104, "train/pos_weight_progress": 0.21800947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1825807094573975 }, { "epoch": 2.1800947867298577, "step": 920, "train/alpha": 0.2, "train/clean_ce_loss": 0.055419921875, "train/clean_loss_scale": 6.417287826538086, "train/clean_loss_weight": 1.283457636833191, "train/combined_loss": 0.4267747104167938, "train/pos_loss/pos0": 0.14296923577785492, "train/pos_loss/pos1": 0.2976043224334717, "train/pos_loss/pos2": 0.591588020324707, "train/pos_loss/pos3": 0.830366313457489, "train/pos_loss/pos4": 1.1939610242843628, "train/pos_loss/pos5": 1.6738674640655518, "train/pos_loss/pos6": 2.061478853225708, "train/pos_loss/weight0": 2.998790979385376, "train/pos_loss/weight1": 2.0691521167755127, "train/pos_loss/weight2": 1.1664402484893799, "train/pos_loss/weight3": 0.5234009027481079, "train/pos_loss/weight4": 0.18263228237628937, "train/pos_loss/weight5": 0.04668230563402176, "train/pos_loss/weight6": 0.012901343405246735, "train/pos_weight_progress": 0.21800947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3556455969810486 }, { "epoch": 2.1824644549763033, "step": 921, "train/alpha": 0.2, "train/clean_ce_loss": 0.10107421875, "train/clean_loss_scale": 8.02840805053711, "train/clean_loss_weight": 1.6056816577911377, "train/combined_loss": 0.9737580418586731, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.21824644549763034, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8114650249481201 }, { "epoch": 2.1824644549763033, "step": 921, "train/alpha": 0.2, "train/clean_ce_loss": 0.040771484375, "train/clean_loss_scale": 14.59628963470459, "train/clean_loss_weight": 2.919257879257202, "train/combined_loss": 0.7141348719596863, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.21824644549763034, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5951123833656311 }, { "epoch": 2.1824644549763033, "step": 921, "train/alpha": 0.2, "train/clean_ce_loss": 0.06787109375, "train/clean_loss_scale": 18.275510787963867, "train/clean_loss_weight": 3.655102252960205, "train/combined_loss": 1.4884545803070068, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.21824644549763034, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2403788566589355 }, { "epoch": 2.1824644549763033, "step": 921, "train/alpha": 0.2, "train/clean_ce_loss": 0.166015625, "train/clean_loss_scale": 7.693223476409912, "train/clean_loss_weight": 1.5386446714401245, "train/combined_loss": 1.5326343774795532, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.21824644549763034, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2771953344345093 }, { "epoch": 2.184834123222749, "step": 922, "train/alpha": 0.2, "train/clean_ce_loss": 0.0703125, "train/clean_loss_scale": 16.556367874145508, "train/clean_loss_weight": 3.3112735748291016, "train/combined_loss": 1.3969435691833496, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.2184834123222749, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1641196012496948 }, { "epoch": 2.184834123222749, "step": 922, "train/alpha": 0.2, "train/clean_ce_loss": 0.1103515625, "train/clean_loss_scale": 8.701526641845703, "train/clean_loss_weight": 1.7403053045272827, "train/combined_loss": 1.1522724628448486, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.2184834123222749, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9602270722389221 }, { "epoch": 2.184834123222749, "step": 922, "train/alpha": 0.2, "train/clean_ce_loss": 0.055908203125, "train/clean_loss_scale": 9.744392395019531, "train/clean_loss_weight": 1.948878526687622, "train/combined_loss": 0.6537497639656067, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.2184834123222749, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5447914600372314 }, { "epoch": 2.184834123222749, "step": 922, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 16.37710952758789, "train/clean_loss_weight": 3.2754218578338623, "train/combined_loss": 1.295454978942871, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.2184834123222749, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0795458555221558 }, { "epoch": 2.1872037914691944, "step": 923, "train/alpha": 0.2, "train/clean_ce_loss": 0.040283203125, "train/clean_loss_scale": 6.05406379699707, "train/clean_loss_weight": 1.2108128070831299, "train/combined_loss": 0.2926524877548218, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.21872037914691944, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2438770830631256 }, { "epoch": 2.1872037914691944, "step": 923, "train/alpha": 0.2, "train/clean_ce_loss": 0.059326171875, "train/clean_loss_scale": 8.87259292602539, "train/clean_loss_weight": 1.774518609046936, "train/combined_loss": 0.6316523551940918, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.21872037914691944, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5263769626617432 }, { "epoch": 2.1872037914691944, "step": 923, "train/alpha": 0.2, "train/clean_ce_loss": 0.001373291015625, "train/clean_loss_scale": 165.52822875976562, "train/clean_loss_weight": 33.105648040771484, "train/combined_loss": 0.2727821171283722, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.21872037914691944, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2273184359073639 }, { "epoch": 2.1872037914691944, "step": 923, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 10.081231117248535, "train/clean_loss_weight": 2.0162463188171387, "train/combined_loss": 0.8565108776092529, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.21872037914691944, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7137590646743774 }, { "epoch": 2.18957345971564, "step": 924, "train/alpha": 0.2, "train/clean_ce_loss": 0.01251220703125, "train/clean_loss_scale": 84.02379608154297, "train/clean_loss_weight": 16.804759979248047, "train/combined_loss": 1.2615878582000732, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.218957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0513231754302979 }, { "epoch": 2.18957345971564, "step": 924, "train/alpha": 0.2, "train/clean_ce_loss": 0.060302734375, "train/clean_loss_scale": 9.598340034484863, "train/clean_loss_weight": 1.9196680784225464, "train/combined_loss": 0.6945673823356628, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.218957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5788061618804932 }, { "epoch": 2.18957345971564, "step": 924, "train/alpha": 0.2, "train/clean_ce_loss": 0.04296875, "train/clean_loss_scale": 10.96293830871582, "train/clean_loss_weight": 2.1925876140594482, "train/combined_loss": 0.5652765035629272, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.218957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4710637629032135 }, { "epoch": 2.18957345971564, "step": 924, "train/alpha": 0.2, "train/clean_ce_loss": 0.05029296875, "train/clean_loss_scale": 20.421091079711914, "train/clean_loss_weight": 4.084218502044678, "train/combined_loss": 1.2324447631835938, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.218957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.027037262916565 }, { "epoch": 2.191943127962085, "grad_norm": 21.099214553833008, "learning_rate": 2e-06, "loss": 3.3266, "step": 925 }, { "epoch": 2.191943127962085, "step": 925, "train/alpha": 0.2, "train/clean_ce_loss": 0.134765625, "train/clean_loss_scale": 8.481681823730469, "train/clean_loss_weight": 1.6963363885879517, "train/combined_loss": 1.3716468811035156, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.21919431279620852, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1430391073226929 }, { "epoch": 2.191943127962085, "step": 925, "train/alpha": 0.2, "train/clean_ce_loss": 0.06005859375, "train/clean_loss_scale": 10.426766395568848, "train/clean_loss_weight": 2.085353374481201, "train/combined_loss": 0.7514603137969971, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.21919431279620852, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6262169480323792 }, { "epoch": 2.191943127962085, "step": 925, "train/alpha": 0.2, "train/clean_ce_loss": 0.053466796875, "train/clean_loss_scale": 12.706401824951172, "train/clean_loss_weight": 2.541280508041382, "train/combined_loss": 0.8152446746826172, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.21919431279620852, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6793705821037292 }, { "epoch": 2.191943127962085, "step": 925, "train/alpha": 0.2, "train/clean_ce_loss": 1.3947486877441406e-05, "train/clean_loss_scale": 12108.150390625, "train/clean_loss_weight": 2421.630126953125, "train/combined_loss": 0.2026539295911789, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.21919431279620852, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.16887827217578888 }, { "epoch": 2.1943127962085307, "step": 926, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 6.182058334350586, "train/clean_loss_weight": 1.236411690711975, "train/combined_loss": 0.6013017892837524, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.21943127962085307, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5010848045349121 }, { "epoch": 2.1943127962085307, "step": 926, "train/alpha": 0.2, "train/clean_ce_loss": 0.09326171875, "train/clean_loss_scale": 9.423637390136719, "train/clean_loss_weight": 1.8847274780273438, "train/combined_loss": 1.0546375513076782, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.21943127962085307, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8788646459579468 }, { "epoch": 2.1943127962085307, "step": 926, "train/alpha": 0.2, "train/clean_ce_loss": 0.1015625, "train/clean_loss_scale": 8.774919509887695, "train/clean_loss_weight": 1.754983901977539, "train/combined_loss": 1.0694433450698853, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.21943127962085307, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8912028074264526 }, { "epoch": 2.1943127962085307, "step": 926, "train/alpha": 0.2, "train/clean_ce_loss": 0.06787109375, "train/clean_loss_scale": 7.210524082183838, "train/clean_loss_weight": 1.4421048164367676, "train/combined_loss": 0.5872634053230286, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.21943127962085307, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4893861711025238 }, { "epoch": 2.1966824644549763, "step": 927, "train/alpha": 0.2, "train/clean_ce_loss": 0.107421875, "train/clean_loss_scale": 8.651718139648438, "train/clean_loss_weight": 1.7303436994552612, "train/combined_loss": 1.1152604818344116, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.21966824644549762, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9293837547302246 }, { "epoch": 2.1966824644549763, "step": 927, "train/alpha": 0.2, "train/clean_ce_loss": 0.00531005859375, "train/clean_loss_scale": 33.34904479980469, "train/clean_loss_weight": 6.669808864593506, "train/combined_loss": 0.21250246465206146, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.21966824644549762, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.17708538472652435 }, { "epoch": 2.1966824644549763, "step": 927, "train/alpha": 0.2, "train/clean_ce_loss": 0.09326171875, "train/clean_loss_scale": 8.505545616149902, "train/clean_loss_weight": 1.7011091709136963, "train/combined_loss": 0.9518901705741882, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.21966824644549762, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7932417988777161 }, { "epoch": 2.1966824644549763, "step": 927, "train/alpha": 0.2, "train/clean_ce_loss": 0.04345703125, "train/clean_loss_scale": 22.097795486450195, "train/clean_loss_weight": 4.419559001922607, "train/combined_loss": 1.1523654460906982, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.21966824644549762, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9603045582771301 }, { "epoch": 2.199052132701422, "step": 928, "train/alpha": 0.2, "train/clean_ce_loss": 0.009033203125, "train/clean_loss_scale": 35.877525329589844, "train/clean_loss_weight": 7.1755051612854, "train/combined_loss": 0.3889067769050598, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.21990521327014217, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.324088990688324 }, { "epoch": 2.199052132701422, "step": 928, "train/alpha": 0.2, "train/clean_ce_loss": 0.06396484375, "train/clean_loss_scale": 13.144526481628418, "train/clean_loss_weight": 2.6289052963256836, "train/combined_loss": 1.008945107460022, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.21990521327014217, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8407875895500183 }, { "epoch": 2.199052132701422, "step": 928, "train/alpha": 0.2, "train/clean_ce_loss": 0.1328125, "train/clean_loss_scale": 9.121190071105957, "train/clean_loss_weight": 1.8242380619049072, "train/combined_loss": 1.4536895751953125, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.21990521327014217, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.211408019065857 }, { "epoch": 2.199052132701422, "step": 928, "train/alpha": 0.2, "train/clean_ce_loss": 0.0380859375, "train/clean_loss_scale": 17.699466705322266, "train/clean_loss_weight": 3.539893388748169, "train/combined_loss": 0.8089209198951721, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.21990521327014217, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6741007566452026 }, { "epoch": 2.2014218009478674, "step": 929, "train/alpha": 0.2, "train/clean_ce_loss": 0.08447265625, "train/clean_loss_scale": 11.075339317321777, "train/clean_loss_weight": 2.2150678634643555, "train/combined_loss": 1.1226760149002075, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.22014218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.935563325881958 }, { "epoch": 2.2014218009478674, "step": 929, "train/alpha": 0.2, "train/clean_ce_loss": 0.031982421875, "train/clean_loss_scale": 21.889083862304688, "train/clean_loss_weight": 4.377816677093506, "train/combined_loss": 0.8400790691375732, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.22014218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7000659108161926 }, { "epoch": 2.2014218009478674, "step": 929, "train/alpha": 0.2, "train/clean_ce_loss": 0.053955078125, "train/clean_loss_scale": 14.82669734954834, "train/clean_loss_weight": 2.965339422225952, "train/combined_loss": 0.9599707722663879, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.22014218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7999756336212158 }, { "epoch": 2.2014218009478674, "step": 929, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 8.856488227844238, "train/clean_loss_weight": 1.7712976932525635, "train/combined_loss": 0.8458638191223145, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.22014218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7048864960670471 }, { "epoch": 2.2037914691943126, "grad_norm": 15.800542831420898, "learning_rate": 2e-06, "loss": 2.9021, "step": 930 }, { "epoch": 2.2037914691943126, "step": 930, "train/alpha": 0.2, "train/clean_ce_loss": 0.076171875, "train/clean_loss_scale": 11.276426315307617, "train/clean_loss_weight": 2.2552852630615234, "train/combined_loss": 1.0307358503341675, "train/pos_loss/pos0": 0.2588610351085663, "train/pos_loss/pos1": 0.6081480979919434, "train/pos_loss/pos2": 1.0331321954727173, "train/pos_loss/pos3": 1.4322906732559204, "train/pos_loss/pos4": 1.9669300317764282, "train/pos_loss/pos5": 2.3082621097564697, "train/pos_loss/pos6": 2.5996737480163574, "train/pos_loss/weight0": 2.3203976154327393, "train/pos_loss/weight1": 1.749570608139038, "train/pos_loss/weight2": 1.247701644897461, "train/pos_loss/weight3": 0.805484414100647, "train/pos_loss/weight4": 0.47047945857048035, "train/pos_loss/weight5": 0.2574828565120697, "train/pos_loss/weight6": 0.14888329803943634, "train/pos_weight_progress": 0.22037914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8589465022087097 }, { "epoch": 2.2037914691943126, "step": 930, "train/alpha": 0.2, "train/clean_ce_loss": 0.0294189453125, "train/clean_loss_scale": 9.913701057434082, "train/clean_loss_weight": 1.9827402830123901, "train/combined_loss": 0.34998074173927307, "train/pos_loss/pos0": 0.01004498265683651, "train/pos_loss/pos1": 0.07000543922185898, "train/pos_loss/pos2": 0.3620349168777466, "train/pos_loss/pos3": 0.7961897850036621, "train/pos_loss/pos4": 1.3934779167175293, "train/pos_loss/pos5": 1.757028579711914, "train/pos_loss/pos6": 2.264216423034668, "train/pos_loss/weight0": 2.6287739276885986, "train/pos_loss/weight1": 1.8958079814910889, "train/pos_loss/weight2": 1.2313286066055298, "train/pos_loss/weight3": 0.6855273842811584, "train/pos_loss/weight4": 0.3425693213939667, "train/pos_loss/weight5": 0.14973300695419312, "train/pos_loss/weight6": 0.06626032292842865, "train/pos_weight_progress": 0.22037914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2916506230831146 }, { "epoch": 2.2037914691943126, "step": 930, "train/alpha": 0.2, "train/clean_ce_loss": 0.0311279296875, "train/clean_loss_scale": 18.80422019958496, "train/clean_loss_weight": 3.7608439922332764, "train/combined_loss": 0.702403724193573, "train/pos_loss/pos0": 0.06163438409566879, "train/pos_loss/pos1": 0.13400113582611084, "train/pos_loss/pos2": 0.5364124774932861, "train/pos_loss/pos3": 0.9374011754989624, "train/pos_loss/pos4": 1.667508840560913, "train/pos_loss/pos5": 2.137158155441284, "train/pos_loss/pos6": 2.478297710418701, "train/pos_loss/weight0": 2.1001105308532715, "train/pos_loss/weight1": 1.6248550415039062, "train/pos_loss/weight2": 1.2406809329986572, "train/pos_loss/weight3": 0.8894965648651123, "train/pos_loss/weight4": 0.5783764123916626, "train/pos_loss/weight5": 0.34386512637138367, "train/pos_loss/weight6": 0.2226150929927826, "train/pos_weight_progress": 0.22037914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.585336446762085 }, { "epoch": 2.2037914691943126, "step": 930, "train/alpha": 0.2, "train/clean_ce_loss": 0.099609375, "train/clean_loss_scale": 10.033113479614258, "train/clean_loss_weight": 2.006622791290283, "train/combined_loss": 1.1992706060409546, "train/pos_loss/pos0": 0.350582480430603, "train/pos_loss/pos1": 0.7051150798797607, "train/pos_loss/pos2": 1.120725393295288, "train/pos_loss/pos3": 1.7640811204910278, "train/pos_loss/pos4": 2.1732542514801025, "train/pos_loss/pos5": 2.7655200958251953, "train/pos_loss/pos6": 3.259719133377075, "train/pos_loss/weight0": 2.3347256183624268, "train/pos_loss/weight1": 1.7473236322402954, "train/pos_loss/weight2": 1.2694199085235596, "train/pos_loss/weight3": 0.8184077739715576, "train/pos_loss/weight4": 0.47318944334983826, "train/pos_loss/weight5": 0.22623062133789062, "train/pos_loss/weight6": 0.1307026743888855, "train/pos_weight_progress": 0.22037914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9993921518325806 }, { "epoch": 2.206161137440758, "step": 931, "train/alpha": 0.2, "train/clean_ce_loss": 0.07763671875, "train/clean_loss_scale": 5.634805679321289, "train/clean_loss_weight": 1.1269611120224, "train/combined_loss": 0.5249613523483276, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22061611374407583, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4374678134918213 }, { "epoch": 2.206161137440758, "step": 931, "train/alpha": 0.2, "train/clean_ce_loss": 0.038330078125, "train/clean_loss_scale": 8.918845176696777, "train/clean_loss_weight": 1.7837690114974976, "train/combined_loss": 0.4102320373058319, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22061611374407583, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3418600261211395 }, { "epoch": 2.206161137440758, "step": 931, "train/alpha": 0.2, "train/clean_ce_loss": 0.056884765625, "train/clean_loss_scale": 6.0613789558410645, "train/clean_loss_weight": 1.2122758626937866, "train/combined_loss": 0.41376012563705444, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22061611374407583, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.34480011463165283 }, { "epoch": 2.206161137440758, "step": 931, "train/alpha": 0.2, "train/clean_ce_loss": 0.06982421875, "train/clean_loss_scale": 8.234376907348633, "train/clean_loss_weight": 1.6468753814697266, "train/combined_loss": 0.689950704574585, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22061611374407583, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5749589204788208 }, { "epoch": 2.2085308056872037, "step": 932, "train/alpha": 0.2, "train/clean_ce_loss": 0.0517578125, "train/clean_loss_scale": 9.647581100463867, "train/clean_loss_weight": 1.9295161962509155, "train/combined_loss": 0.5992052555084229, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22085308056872038, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.49933770298957825 }, { "epoch": 2.2085308056872037, "step": 932, "train/alpha": 0.2, "train/clean_ce_loss": 0.07470703125, "train/clean_loss_scale": 7.773234844207764, "train/clean_loss_weight": 1.5546469688415527, "train/combined_loss": 0.6968583464622498, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22085308056872038, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5807152986526489 }, { "epoch": 2.2085308056872037, "step": 932, "train/alpha": 0.2, "train/clean_ce_loss": 0.00107574462890625, "train/clean_loss_scale": 122.08643341064453, "train/clean_loss_weight": 24.417287826538086, "train/combined_loss": 0.15760059654712677, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22085308056872038, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1313338279724121 }, { "epoch": 2.2085308056872037, "step": 932, "train/alpha": 0.2, "train/clean_ce_loss": 0.0205078125, "train/clean_loss_scale": 8.332006454467773, "train/clean_loss_weight": 1.6664012670516968, "train/combined_loss": 0.20504547655582428, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22085308056872038, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.17087122797966003 }, { "epoch": 2.2109004739336493, "step": 933, "train/alpha": 0.2, "train/clean_ce_loss": 0.11572265625, "train/clean_loss_scale": 5.931756496429443, "train/clean_loss_weight": 1.1863512992858887, "train/combined_loss": 0.8237263560295105, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22109004739336494, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6864386200904846 }, { "epoch": 2.2109004739336493, "step": 933, "train/alpha": 0.2, "train/clean_ce_loss": 0.08251953125, "train/clean_loss_scale": 9.551615715026855, "train/clean_loss_weight": 1.910323143005371, "train/combined_loss": 0.9458338022232056, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22109004739336494, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7881948351860046 }, { "epoch": 2.2109004739336493, "step": 933, "train/alpha": 0.2, "train/clean_ce_loss": 0.044189453125, "train/clean_loss_scale": 10.407642364501953, "train/clean_loss_weight": 2.081528425216675, "train/combined_loss": 0.5518895983695984, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22109004739336494, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45990800857543945 }, { "epoch": 2.2109004739336493, "step": 933, "train/alpha": 0.2, "train/clean_ce_loss": 0.022705078125, "train/clean_loss_scale": 17.390466690063477, "train/clean_loss_weight": 3.478093385696411, "train/combined_loss": 0.47382229566574097, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22109004739336494, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3948519229888916 }, { "epoch": 2.213270142180095, "step": 934, "train/alpha": 0.2, "train/clean_ce_loss": 0.00909423828125, "train/clean_loss_scale": 15.662150382995605, "train/clean_loss_weight": 3.132430076599121, "train/combined_loss": 0.1709223985671997, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.2213270142180095, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.14243532717227936 }, { "epoch": 2.213270142180095, "step": 934, "train/alpha": 0.2, "train/clean_ce_loss": 0.04638671875, "train/clean_loss_scale": 5.8960280418396, "train/clean_loss_weight": 1.1792056560516357, "train/combined_loss": 0.3281968832015991, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.2213270142180095, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.27349740266799927 }, { "epoch": 2.213270142180095, "step": 934, "train/alpha": 0.2, "train/clean_ce_loss": 0.0089111328125, "train/clean_loss_scale": 23.668710708618164, "train/clean_loss_weight": 4.7337422370910645, "train/combined_loss": 0.2530980408191681, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.2213270142180095, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.21091502904891968 }, { "epoch": 2.213270142180095, "step": 934, "train/alpha": 0.2, "train/clean_ce_loss": 0.01312255859375, "train/clean_loss_scale": 13.257028579711914, "train/clean_loss_weight": 2.6514058113098145, "train/combined_loss": 0.2087593674659729, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.2213270142180095, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.17396613955497742 }, { "epoch": 2.2156398104265405, "grad_norm": 22.7853946685791, "learning_rate": 2e-06, "loss": 2.8086, "step": 935 }, { "epoch": 2.2156398104265405, "step": 935, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 8.347705841064453, "train/clean_loss_weight": 1.6695412397384644, "train/combined_loss": 0.7287938594818115, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22156398104265404, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6073282361030579 }, { "epoch": 2.2156398104265405, "step": 935, "train/alpha": 0.2, "train/clean_ce_loss": 0.12060546875, "train/clean_loss_scale": 6.751094818115234, "train/clean_loss_weight": 1.3502190113067627, "train/combined_loss": 0.9770627021789551, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22156398104265404, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8142189383506775 }, { "epoch": 2.2156398104265405, "step": 935, "train/alpha": 0.2, "train/clean_ce_loss": 0.038818359375, "train/clean_loss_scale": 3.7583441734313965, "train/clean_loss_weight": 0.7516688704490662, "train/combined_loss": 0.17507129907608032, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22156398104265404, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.14589275419712067 }, { "epoch": 2.2156398104265405, "step": 935, "train/alpha": 0.2, "train/clean_ce_loss": 0.0223388671875, "train/clean_loss_scale": 5.83133602142334, "train/clean_loss_weight": 1.1662672758102417, "train/combined_loss": 0.1563185304403305, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22156398104265404, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.13026544451713562 }, { "epoch": 2.2180094786729856, "step": 936, "train/alpha": 0.2, "train/clean_ce_loss": 0.038818359375, "train/clean_loss_scale": 14.133091926574707, "train/clean_loss_weight": 2.8266184329986572, "train/combined_loss": 0.6583481431007385, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22180094786729856, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5486234426498413 }, { "epoch": 2.2180094786729856, "step": 936, "train/alpha": 0.2, "train/clean_ce_loss": 0.049560546875, "train/clean_loss_scale": 22.991792678833008, "train/clean_loss_weight": 4.598358631134033, "train/combined_loss": 1.3673830032348633, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22180094786729856, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1394858360290527 }, { "epoch": 2.2180094786729856, "step": 936, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 8.98650074005127, "train/clean_loss_weight": 1.7973002195358276, "train/combined_loss": 1.0004502534866333, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22180094786729856, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8337085843086243 }, { "epoch": 2.2180094786729856, "step": 936, "train/alpha": 0.2, "train/clean_ce_loss": 0.017822265625, "train/clean_loss_scale": 11.944920539855957, "train/clean_loss_weight": 2.388984203338623, "train/combined_loss": 0.255462646484375, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22180094786729856, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2128855437040329 }, { "epoch": 2.220379146919431, "step": 937, "train/alpha": 0.2, "train/clean_ce_loss": 0.022216796875, "train/clean_loss_scale": 9.089550971984863, "train/clean_loss_weight": 1.8179101943969727, "train/combined_loss": 0.24232885241508484, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22203791469194312, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.20194071531295776 }, { "epoch": 2.220379146919431, "step": 937, "train/alpha": 0.2, "train/clean_ce_loss": 0.0751953125, "train/clean_loss_scale": 11.228337287902832, "train/clean_loss_weight": 2.2456674575805664, "train/combined_loss": 1.0131820440292358, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22203791469194312, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8443183302879333 }, { "epoch": 2.220379146919431, "step": 937, "train/alpha": 0.2, "train/clean_ce_loss": 0.00872802734375, "train/clean_loss_scale": 26.613237380981445, "train/clean_loss_weight": 5.322647571563721, "train/combined_loss": 0.27873727679252625, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22203791469194312, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.23228105902671814 }, { "epoch": 2.220379146919431, "step": 937, "train/alpha": 0.2, "train/clean_ce_loss": 0.04052734375, "train/clean_loss_scale": 17.630008697509766, "train/clean_loss_weight": 3.5260016918182373, "train/combined_loss": 0.8573969602584839, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22203791469194312, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.714497447013855 }, { "epoch": 2.2227488151658767, "step": 938, "train/alpha": 0.2, "train/clean_ce_loss": 0.007080078125, "train/clean_loss_scale": 19.75242805480957, "train/clean_loss_weight": 3.9504857063293457, "train/combined_loss": 0.16781848669052124, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22227488151658767, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1398487389087677 }, { "epoch": 2.2227488151658767, "step": 938, "train/alpha": 0.2, "train/clean_ce_loss": 0.01556396484375, "train/clean_loss_scale": 12.303987503051758, "train/clean_loss_weight": 2.4607975482940674, "train/combined_loss": 0.2297986000776291, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22227488151658767, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.19149883091449738 }, { "epoch": 2.2227488151658767, "step": 938, "train/alpha": 0.2, "train/clean_ce_loss": 0.08447265625, "train/clean_loss_scale": 9.081365585327148, "train/clean_loss_weight": 1.8162730932235718, "train/combined_loss": 0.9205524921417236, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22227488151658767, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7671270966529846 }, { "epoch": 2.2227488151658767, "step": 938, "train/alpha": 0.2, "train/clean_ce_loss": 0.07470703125, "train/clean_loss_scale": 7.293189525604248, "train/clean_loss_weight": 1.4586379528045654, "train/combined_loss": 0.6538230776786804, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22227488151658767, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5448525547981262 }, { "epoch": 2.2251184834123223, "step": 939, "train/alpha": 0.2, "train/clean_ce_loss": 0.000820159912109375, "train/clean_loss_scale": 74.10846710205078, "train/clean_loss_weight": 14.821693420410156, "train/combined_loss": 0.07293695211410522, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22251184834123222, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.06078079342842102 }, { "epoch": 2.2251184834123223, "step": 939, "train/alpha": 0.2, "train/clean_ce_loss": 0.01531982421875, "train/clean_loss_scale": 9.824938774108887, "train/clean_loss_weight": 1.9649877548217773, "train/combined_loss": 0.18061959743499756, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22251184834123222, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1505163311958313 }, { "epoch": 2.2251184834123223, "step": 939, "train/alpha": 0.2, "train/clean_ce_loss": 0.0849609375, "train/clean_loss_scale": 8.223554611206055, "train/clean_loss_weight": 1.644710898399353, "train/combined_loss": 0.838417112827301, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22251184834123222, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6986809372901917 }, { "epoch": 2.2251184834123223, "step": 939, "train/alpha": 0.2, "train/clean_ce_loss": 0.09912109375, "train/clean_loss_scale": 7.919116497039795, "train/clean_loss_weight": 1.583823323249817, "train/combined_loss": 0.9419417977333069, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22251184834123222, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7849515080451965 }, { "epoch": 2.227488151658768, "grad_norm": 18.874656677246094, "learning_rate": 2e-06, "loss": 3.1829, "step": 940 }, { "epoch": 2.227488151658768, "step": 940, "train/alpha": 0.2, "train/clean_ce_loss": 0.040771484375, "train/clean_loss_scale": 15.943852424621582, "train/clean_loss_weight": 3.1887705326080322, "train/combined_loss": 0.7800654172897339, "train/pos_loss/pos0": 0.18741118907928467, "train/pos_loss/pos1": 0.5389790534973145, "train/pos_loss/pos2": 0.9480761289596558, "train/pos_loss/pos3": 1.587636113166809, "train/pos_loss/pos4": 2.3669164180755615, "train/pos_loss/pos5": 2.7853286266326904, "train/pos_loss/pos6": 3.071680784225464, "train/pos_loss/weight0": 2.9027998447418213, "train/pos_loss/weight1": 1.9651529788970947, "train/pos_loss/weight2": 1.1757889986038208, "train/pos_loss/weight3": 0.605470597743988, "train/pos_loss/weight4": 0.24009762704372406, "train/pos_loss/weight5": 0.08223478496074677, "train/pos_loss/weight6": 0.028454944491386414, "train/pos_weight_progress": 0.22274881516587677, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6500545144081116 }, { "epoch": 2.227488151658768, "step": 940, "train/alpha": 0.2, "train/clean_ce_loss": 0.016845703125, "train/clean_loss_scale": 6.521206378936768, "train/clean_loss_weight": 1.3042412996292114, "train/combined_loss": 0.13182516396045685, "train/pos_loss/pos0": 0.02650568075478077, "train/pos_loss/pos1": 0.013905846513807774, "train/pos_loss/pos2": 0.09545186161994934, "train/pos_loss/pos3": 0.25872117280960083, "train/pos_loss/pos4": 0.7648781538009644, "train/pos_loss/pos5": 1.5715792179107666, "train/pos_loss/pos6": 2.3408892154693604, "train/pos_loss/weight0": 2.8031787872314453, "train/pos_loss/weight1": 1.9890689849853516, "train/pos_loss/weight2": 1.2137482166290283, "train/pos_loss/weight3": 0.6330007910728455, "train/pos_loss/weight4": 0.2507704496383667, "train/pos_loss/weight5": 0.08015958964824677, "train/pos_loss/weight6": 0.030072707682847977, "train/pos_weight_progress": 0.22274881516587677, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1098543033003807 }, { "epoch": 2.227488151658768, "step": 940, "train/alpha": 0.2, "train/clean_ce_loss": 0.041015625, "train/clean_loss_scale": 26.876070022583008, "train/clean_loss_weight": 5.375214099884033, "train/combined_loss": 1.3228065967559814, "train/pos_loss/pos0": 0.27612993121147156, "train/pos_loss/pos1": 0.580128014087677, "train/pos_loss/pos2": 1.0820016860961914, "train/pos_loss/pos3": 1.5264561176300049, "train/pos_loss/pos4": 2.0814757347106934, "train/pos_loss/pos5": 2.7055306434631348, "train/pos_loss/pos6": 3.3646907806396484, "train/pos_loss/weight0": 1.9614893198013306, "train/pos_loss/weight1": 1.5479851961135864, "train/pos_loss/weight2": 1.194622278213501, "train/pos_loss/weight3": 0.8708335161209106, "train/pos_loss/weight4": 0.6555173397064209, "train/pos_loss/weight5": 0.4454137980937958, "train/pos_loss/weight6": 0.32413846254348755, "train/pos_weight_progress": 0.22274881516587677, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1023387908935547 }, { "epoch": 2.227488151658768, "step": 940, "train/alpha": 0.2, "train/clean_ce_loss": 0.0322265625, "train/clean_loss_scale": 22.069921493530273, "train/clean_loss_weight": 4.413984298706055, "train/combined_loss": 0.8534852862358093, "train/pos_loss/pos0": 0.16714859008789062, "train/pos_loss/pos1": 0.5972172021865845, "train/pos_loss/pos2": 1.109161615371704, "train/pos_loss/pos3": 1.68181312084198, "train/pos_loss/pos4": 2.158639430999756, "train/pos_loss/pos5": 2.9264981746673584, "train/pos_loss/pos6": 3.171020746231079, "train/pos_loss/weight0": 2.7839245796203613, "train/pos_loss/weight1": 1.9827907085418701, "train/pos_loss/weight2": 1.2281792163848877, "train/pos_loss/weight3": 0.6230947375297546, "train/pos_loss/weight4": 0.25671038031578064, "train/pos_loss/weight5": 0.09155359864234924, "train/pos_loss/weight6": 0.03374744951725006, "train/pos_weight_progress": 0.22274881516587677, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7112377285957336 }, { "epoch": 2.229857819905213, "step": 941, "train/alpha": 0.2, "train/clean_ce_loss": 0.0067138671875, "train/clean_loss_scale": 37.79114532470703, "train/clean_loss_weight": 7.558228969573975, "train/combined_loss": 0.30446967482566833, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22298578199052133, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.25372472405433655 }, { "epoch": 2.229857819905213, "step": 941, "train/alpha": 0.2, "train/clean_ce_loss": 0.038818359375, "train/clean_loss_scale": 7.555026054382324, "train/clean_loss_weight": 1.5110052824020386, "train/combined_loss": 0.3519284725189209, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22298578199052133, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2932737171649933 }, { "epoch": 2.229857819905213, "step": 941, "train/alpha": 0.2, "train/clean_ce_loss": 0.10791015625, "train/clean_loss_scale": 9.851243019104004, "train/clean_loss_weight": 1.9702485799789429, "train/combined_loss": 1.2756590843200684, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22298578199052133, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0630491971969604 }, { "epoch": 2.229857819905213, "step": 941, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 7.487959861755371, "train/clean_loss_weight": 1.4975919723510742, "train/combined_loss": 0.7502585053443909, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22298578199052133, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6252154111862183 }, { "epoch": 2.2322274881516586, "step": 942, "train/alpha": 0.2, "train/clean_ce_loss": 0.1064453125, "train/clean_loss_scale": 8.850958824157715, "train/clean_loss_weight": 1.7701917886734009, "train/combined_loss": 1.130571722984314, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22322274881516588, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9421430826187134 }, { "epoch": 2.2322274881516586, "step": 942, "train/alpha": 0.2, "train/clean_ce_loss": 0.0263671875, "train/clean_loss_scale": 12.548980712890625, "train/clean_loss_weight": 2.509796142578125, "train/combined_loss": 0.39705759286880493, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22322274881516588, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3308813273906708 }, { "epoch": 2.2322274881516586, "step": 942, "train/alpha": 0.2, "train/clean_ce_loss": 0.002166748046875, "train/clean_loss_scale": 81.64412689208984, "train/clean_loss_weight": 16.328824996948242, "train/combined_loss": 0.2122827023267746, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22322274881516588, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.17690224945545197 }, { "epoch": 2.2322274881516586, "step": 942, "train/alpha": 0.2, "train/clean_ce_loss": 0.103515625, "train/clean_loss_scale": 6.310369491577148, "train/clean_loss_weight": 1.2620738744735718, "train/combined_loss": 0.7838662266731262, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22322274881516588, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.653221845626831 }, { "epoch": 2.234597156398104, "step": 943, "train/alpha": 0.2, "train/clean_ce_loss": 0.1162109375, "train/clean_loss_scale": 8.111557960510254, "train/clean_loss_weight": 1.6223115921020508, "train/combined_loss": 1.131182074546814, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22345971563981043, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9426517486572266 }, { "epoch": 2.234597156398104, "step": 943, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 14.729516983032227, "train/clean_loss_weight": 2.9459035396575928, "train/combined_loss": 1.2514336109161377, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22345971563981043, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0428613424301147 }, { "epoch": 2.234597156398104, "step": 943, "train/alpha": 0.2, "train/clean_ce_loss": 0.0004634857177734375, "train/clean_loss_scale": 97.2669677734375, "train/clean_loss_weight": 19.453393936157227, "train/combined_loss": 0.0540982186794281, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22345971563981043, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.04508185014128685 }, { "epoch": 2.234597156398104, "step": 943, "train/alpha": 0.2, "train/clean_ce_loss": 0.006561279296875, "train/clean_loss_scale": 17.444965362548828, "train/clean_loss_weight": 3.4889931678771973, "train/combined_loss": 0.1373535394668579, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22345971563981043, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.11446128785610199 }, { "epoch": 2.2369668246445498, "step": 944, "train/alpha": 0.2, "train/clean_ce_loss": 0.056884765625, "train/clean_loss_scale": 13.576980590820312, "train/clean_loss_weight": 2.7153961658477783, "train/combined_loss": 0.9267880320549011, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22369668246445498, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7723233699798584 }, { "epoch": 2.2369668246445498, "step": 944, "train/alpha": 0.2, "train/clean_ce_loss": 0.13671875, "train/clean_loss_scale": 7.5695624351501465, "train/clean_loss_weight": 1.513912558555603, "train/combined_loss": 1.2418813705444336, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22369668246445498, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0349011421203613 }, { "epoch": 2.2369668246445498, "step": 944, "train/alpha": 0.2, "train/clean_ce_loss": 0.0001392364501953125, "train/clean_loss_scale": 986.3153686523438, "train/clean_loss_weight": 197.26307678222656, "train/combined_loss": 0.16479726135730743, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22369668246445498, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1373310536146164 }, { "epoch": 2.2369668246445498, "step": 944, "train/alpha": 0.2, "train/clean_ce_loss": 0.10302734375, "train/clean_loss_scale": 7.230994701385498, "train/clean_loss_weight": 1.4461989402770996, "train/combined_loss": 0.8939881920814514, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22369668246445498, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7449901700019836 }, { "epoch": 2.2393364928909953, "grad_norm": 20.613685607910156, "learning_rate": 2e-06, "loss": 3.2307, "step": 945 }, { "epoch": 2.2393364928909953, "step": 945, "train/alpha": 0.2, "train/clean_ce_loss": 0.11279296875, "train/clean_loss_scale": 7.978662014007568, "train/clean_loss_weight": 1.5957324504852295, "train/combined_loss": 1.0799243450164795, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22393364928909953, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8999369740486145 }, { "epoch": 2.2393364928909953, "step": 945, "train/alpha": 0.2, "train/clean_ce_loss": 0.061279296875, "train/clean_loss_scale": 6.206877708435059, "train/clean_loss_weight": 1.2413755655288696, "train/combined_loss": 0.45642369985580444, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22393364928909953, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.38035309314727783 }, { "epoch": 2.2393364928909953, "step": 945, "train/alpha": 0.2, "train/clean_ce_loss": 0.039794921875, "train/clean_loss_scale": 12.162323951721191, "train/clean_loss_weight": 2.432464838027954, "train/combined_loss": 0.5807984471321106, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22393364928909953, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.48399871587753296 }, { "epoch": 2.2393364928909953, "step": 945, "train/alpha": 0.2, "train/clean_ce_loss": 0.12255859375, "train/clean_loss_scale": 7.145331859588623, "train/clean_loss_weight": 1.4290664196014404, "train/combined_loss": 1.0508661270141602, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22393364928909953, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.87572181224823 }, { "epoch": 2.241706161137441, "step": 946, "train/alpha": 0.2, "train/clean_ce_loss": 0.1201171875, "train/clean_loss_scale": 7.451510906219482, "train/clean_loss_weight": 1.4903022050857544, "train/combined_loss": 1.0740654468536377, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.2241706161137441, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8950545191764832 }, { "epoch": 2.241706161137441, "step": 946, "train/alpha": 0.2, "train/clean_ce_loss": 0.1279296875, "train/clean_loss_scale": 8.732500076293945, "train/clean_loss_weight": 1.746500015258789, "train/combined_loss": 1.3405752182006836, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.2241706161137441, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1171460151672363 }, { "epoch": 2.241706161137441, "step": 946, "train/alpha": 0.2, "train/clean_ce_loss": 0.0185546875, "train/clean_loss_scale": 18.700363159179688, "train/clean_loss_weight": 3.740072727203369, "train/combined_loss": 0.4163752496242523, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.2241706161137441, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.34697937965393066 }, { "epoch": 2.241706161137441, "step": 946, "train/alpha": 0.2, "train/clean_ce_loss": 0.040283203125, "train/clean_loss_scale": 20.55950927734375, "train/clean_loss_weight": 4.111901760101318, "train/combined_loss": 0.9938434958457947, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.2241706161137441, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8282029032707214 }, { "epoch": 2.244075829383886, "step": 947, "train/alpha": 0.2, "train/clean_ce_loss": 0.053955078125, "train/clean_loss_scale": 7.369138240814209, "train/clean_loss_weight": 1.4738277196884155, "train/combined_loss": 0.4771229326725006, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22440758293838864, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3976024389266968 }, { "epoch": 2.244075829383886, "step": 947, "train/alpha": 0.2, "train/clean_ce_loss": 0.00067901611328125, "train/clean_loss_scale": 217.1141357421875, "train/clean_loss_weight": 43.422828674316406, "train/combined_loss": 0.17690879106521606, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22440758293838864, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.14742399752140045 }, { "epoch": 2.244075829383886, "step": 947, "train/alpha": 0.2, "train/clean_ce_loss": 0.0194091796875, "train/clean_loss_scale": 28.223125457763672, "train/clean_loss_weight": 5.644625186920166, "train/combined_loss": 0.6573452949523926, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22440758293838864, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5477877259254456 }, { "epoch": 2.244075829383886, "step": 947, "train/alpha": 0.2, "train/clean_ce_loss": 0.11083984375, "train/clean_loss_scale": 7.143816947937012, "train/clean_loss_weight": 1.4287633895874023, "train/combined_loss": 0.9501835107803345, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22440758293838864, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7918195724487305 }, { "epoch": 2.2464454976303316, "step": 948, "train/alpha": 0.2, "train/clean_ce_loss": 6.4849853515625e-05, "train/clean_loss_scale": 1783.999755859375, "train/clean_loss_weight": 356.7999572753906, "train/combined_loss": 0.1388305425643921, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22464454976303316, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1156921237707138 }, { "epoch": 2.2464454976303316, "step": 948, "train/alpha": 0.2, "train/clean_ce_loss": 0.09375, "train/clean_loss_scale": 7.233606815338135, "train/clean_loss_weight": 1.4467214345932007, "train/combined_loss": 0.8137807846069336, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22464454976303316, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6781506538391113 }, { "epoch": 2.2464454976303316, "step": 948, "train/alpha": 0.2, "train/clean_ce_loss": 0.012939453125, "train/clean_loss_scale": 19.732131958007812, "train/clean_loss_weight": 3.9464263916015625, "train/combined_loss": 0.30638760328292847, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22464454976303316, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.25532299280166626 }, { "epoch": 2.2464454976303316, "step": 948, "train/alpha": 0.2, "train/clean_ce_loss": 0.04541015625, "train/clean_loss_scale": 18.423965454101562, "train/clean_loss_weight": 3.68479323387146, "train/combined_loss": 1.0039621591567993, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22464454976303316, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.836635172367096 }, { "epoch": 2.248815165876777, "step": 949, "train/alpha": 0.2, "train/clean_ce_loss": 0.142578125, "train/clean_loss_scale": 6.700236797332764, "train/clean_loss_weight": 1.3400473594665527, "train/combined_loss": 1.1463686227798462, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22488151658767772, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9553071856498718 }, { "epoch": 2.248815165876777, "step": 949, "train/alpha": 0.2, "train/clean_ce_loss": 0.1337890625, "train/clean_loss_scale": 7.740890026092529, "train/clean_loss_weight": 1.5481780767440796, "train/combined_loss": 1.2427756786346436, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22488151658767772, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0356464385986328 }, { "epoch": 2.248815165876777, "step": 949, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 7.863406181335449, "train/clean_loss_weight": 1.5726813077926636, "train/combined_loss": 0.6081852912902832, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22488151658767772, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5068210959434509 }, { "epoch": 2.248815165876777, "step": 949, "train/alpha": 0.2, "train/clean_ce_loss": 0.10400390625, "train/clean_loss_scale": 8.735288619995117, "train/clean_loss_weight": 1.7470577955245972, "train/combined_loss": 1.0902049541473389, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22488151658767772, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9085041284561157 }, { "epoch": 2.251184834123223, "grad_norm": 30.66469383239746, "learning_rate": 2e-06, "loss": 3.3436, "step": 950 }, { "epoch": 2.251184834123223, "step": 950, "train/alpha": 0.2, "train/clean_ce_loss": 8.678436279296875e-05, "train/clean_loss_scale": 1080.4107666015625, "train/clean_loss_weight": 216.0821533203125, "train/combined_loss": 0.11251531541347504, "train/pos_loss/pos0": 0.0005831002490594983, "train/pos_loss/pos1": 0.002328654285520315, "train/pos_loss/pos2": 0.0385461300611496, "train/pos_loss/pos3": 0.2524033188819885, "train/pos_loss/pos4": 0.6116406917572021, "train/pos_loss/pos5": 1.4159611463546753, "train/pos_loss/pos6": 2.5452401638031006, "train/pos_loss/weight0": 2.678874969482422, "train/pos_loss/weight1": 1.9781638383865356, "train/pos_loss/weight2": 1.2295016050338745, "train/pos_loss/weight3": 0.6798160076141357, "train/pos_loss/weight4": 0.287140816450119, "train/pos_loss/weight5": 0.10575074702501297, "train/pos_loss/weight6": 0.04075193405151367, "train/pos_weight_progress": 0.22511848341232227, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.09376276284456253 }, { "epoch": 2.251184834123223, "step": 950, "train/alpha": 0.2, "train/clean_ce_loss": 0.0732421875, "train/clean_loss_scale": 13.755362510681152, "train/clean_loss_weight": 2.751072645187378, "train/combined_loss": 1.2089674472808838, "train/pos_loss/pos0": 0.2773677408695221, "train/pos_loss/pos1": 0.6117348074913025, "train/pos_loss/pos2": 0.958612859249115, "train/pos_loss/pos3": 1.3731893301010132, "train/pos_loss/pos4": 1.8649485111236572, "train/pos_loss/pos5": 2.3624789714813232, "train/pos_loss/pos6": 2.6306819915771484, "train/pos_loss/weight0": 1.9319177865982056, "train/pos_loss/weight1": 1.544966220855713, "train/pos_loss/weight2": 1.1918944120407104, "train/pos_loss/weight3": 0.8798872232437134, "train/pos_loss/weight4": 0.6201439499855042, "train/pos_loss/weight5": 0.4500141143798828, "train/pos_loss/weight6": 0.3811761438846588, "train/pos_weight_progress": 0.22511848341232227, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0074728727340698 }, { "epoch": 2.251184834123223, "step": 950, "train/alpha": 0.2, "train/clean_ce_loss": 0.057373046875, "train/clean_loss_scale": 8.704544067382812, "train/clean_loss_weight": 1.7409088611602783, "train/combined_loss": 0.5992874503135681, "train/pos_loss/pos0": 0.09998831897974014, "train/pos_loss/pos1": 0.34599193930625916, "train/pos_loss/pos2": 0.6688232421875, "train/pos_loss/pos3": 1.0357167720794678, "train/pos_loss/pos4": 1.5865259170532227, "train/pos_loss/pos5": 1.9202451705932617, "train/pos_loss/pos6": 2.286850690841675, "train/pos_loss/weight0": 2.613616943359375, "train/pos_loss/weight1": 1.8556360006332397, "train/pos_loss/weight2": 1.2358639240264893, "train/pos_loss/weight3": 0.7125753164291382, "train/pos_loss/weight4": 0.35076963901519775, "train/pos_loss/weight5": 0.15691892802715302, "train/pos_loss/weight6": 0.0746193677186966, "train/pos_weight_progress": 0.22511848341232227, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.49940621852874756 }, { "epoch": 2.251184834123223, "step": 950, "train/alpha": 0.2, "train/clean_ce_loss": 0.005340576171875, "train/clean_loss_scale": 64.47348022460938, "train/clean_loss_weight": 12.894696235656738, "train/combined_loss": 0.4131906032562256, "train/pos_loss/pos0": 0.018929634243249893, "train/pos_loss/pos1": 0.11119908094406128, "train/pos_loss/pos2": 0.2354537546634674, "train/pos_loss/pos3": 0.718599259853363, "train/pos_loss/pos4": 1.317272663116455, "train/pos_loss/pos5": 1.980634331703186, "train/pos_loss/pos6": 2.5159153938293457, "train/pos_loss/weight0": 2.384500503540039, "train/pos_loss/weight1": 1.8097704648971558, "train/pos_loss/weight2": 1.2538666725158691, "train/pos_loss/weight3": 0.7766751646995544, "train/pos_loss/weight4": 0.43630868196487427, "train/pos_loss/weight5": 0.21927949786186218, "train/pos_loss/weight6": 0.11959931254386902, "train/pos_weight_progress": 0.22511848341232227, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3443255126476288 }, { "epoch": 2.2535545023696684, "step": 951, "train/alpha": 0.2, "train/clean_ce_loss": 0.051025390625, "train/clean_loss_scale": 7.510364532470703, "train/clean_loss_weight": 1.5020729303359985, "train/combined_loss": 0.4598631262779236, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22535545023696682, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.38321927189826965 }, { "epoch": 2.2535545023696684, "step": 951, "train/alpha": 0.2, "train/clean_ce_loss": 0.05224609375, "train/clean_loss_scale": 20.819028854370117, "train/clean_loss_weight": 4.163805961608887, "train/combined_loss": 1.3052555322647095, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22535545023696682, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0877128839492798 }, { "epoch": 2.2535545023696684, "step": 951, "train/alpha": 0.2, "train/clean_ce_loss": 0.022705078125, "train/clean_loss_scale": 21.795494079589844, "train/clean_loss_weight": 4.3590989112854, "train/combined_loss": 0.593842089176178, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22535545023696682, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4948683977127075 }, { "epoch": 2.2535545023696684, "step": 951, "train/alpha": 0.2, "train/clean_ce_loss": 0.0634765625, "train/clean_loss_scale": 10.184699058532715, "train/clean_loss_weight": 2.036939859390259, "train/combined_loss": 0.7757875919342041, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22535545023696682, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.646489679813385 }, { "epoch": 2.2559241706161135, "step": 952, "train/alpha": 0.2, "train/clean_ce_loss": 0.146484375, "train/clean_loss_scale": 8.191529273986816, "train/clean_loss_weight": 1.638305902481079, "train/combined_loss": 1.4399172067642212, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22559241706161137, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.199931025505066 }, { "epoch": 2.2559241706161135, "step": 952, "train/alpha": 0.2, "train/clean_ce_loss": 0.030517578125, "train/clean_loss_scale": 16.274024963378906, "train/clean_loss_weight": 3.254805088043213, "train/combined_loss": 0.5959725975990295, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22559241706161137, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4966438412666321 }, { "epoch": 2.2559241706161135, "step": 952, "train/alpha": 0.2, "train/clean_ce_loss": 0.0201416015625, "train/clean_loss_scale": 30.2431697845459, "train/clean_loss_weight": 6.048634052276611, "train/combined_loss": 0.7309750318527222, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22559241706161137, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6091458797454834 }, { "epoch": 2.2559241706161135, "step": 952, "train/alpha": 0.2, "train/clean_ce_loss": 0.115234375, "train/clean_loss_scale": 11.669763565063477, "train/clean_loss_weight": 2.3339526653289795, "train/combined_loss": 1.6137094497680664, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22559241706161137, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3447579145431519 }, { "epoch": 2.258293838862559, "step": 953, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 10.422158241271973, "train/clean_loss_weight": 2.0844316482543945, "train/combined_loss": 1.0015043020248413, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22582938388625592, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8345869183540344 }, { "epoch": 2.258293838862559, "step": 953, "train/alpha": 0.2, "train/clean_ce_loss": 0.01171875, "train/clean_loss_scale": 24.3795166015625, "train/clean_loss_weight": 4.875903606414795, "train/combined_loss": 0.3428369462490082, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22582938388625592, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.28569746017456055 }, { "epoch": 2.258293838862559, "step": 953, "train/alpha": 0.2, "train/clean_ce_loss": 0.1494140625, "train/clean_loss_scale": 7.9722747802734375, "train/clean_loss_weight": 1.5944550037384033, "train/combined_loss": 1.4294040203094482, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22582938388625592, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1911699771881104 }, { "epoch": 2.258293838862559, "step": 953, "train/alpha": 0.2, "train/clean_ce_loss": 0.0135498046875, "train/clean_loss_scale": 37.40034484863281, "train/clean_loss_weight": 7.480069160461426, "train/combined_loss": 0.608120858669281, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22582938388625592, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5067673921585083 }, { "epoch": 2.2606635071090047, "step": 954, "train/alpha": 0.2, "train/clean_ce_loss": 0.04638671875, "train/clean_loss_scale": 20.797260284423828, "train/clean_loss_weight": 4.159451961517334, "train/combined_loss": 1.1576600074768066, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22606635071090048, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9647166728973389 }, { "epoch": 2.2606635071090047, "step": 954, "train/alpha": 0.2, "train/clean_ce_loss": 0.07861328125, "train/clean_loss_scale": 9.90522289276123, "train/clean_loss_weight": 1.9810446500778198, "train/combined_loss": 0.9344184398651123, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22606635071090048, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7786820530891418 }, { "epoch": 2.2606635071090047, "step": 954, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 10.480849266052246, "train/clean_loss_weight": 2.096169948577881, "train/combined_loss": 0.8966038823127747, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22606635071090048, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7471699118614197 }, { "epoch": 2.2606635071090047, "step": 954, "train/alpha": 0.2, "train/clean_ce_loss": 0.1494140625, "train/clean_loss_scale": 9.932759284973145, "train/clean_loss_weight": 1.9865518808364868, "train/combined_loss": 1.7809126377105713, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22606635071090048, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4840939044952393 }, { "epoch": 2.2630331753554502, "grad_norm": 29.02385902404785, "learning_rate": 2e-06, "loss": 3.028, "step": 955 }, { "epoch": 2.2630331753554502, "step": 955, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 9.004806518554688, "train/clean_loss_weight": 1.8009613752365112, "train/combined_loss": 0.6964654922485352, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22630331753554503, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5803878903388977 }, { "epoch": 2.2630331753554502, "step": 955, "train/alpha": 0.2, "train/clean_ce_loss": 0.07666015625, "train/clean_loss_scale": 13.429373741149902, "train/clean_loss_weight": 2.6858747005462646, "train/combined_loss": 1.235397458076477, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22630331753554503, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0294978618621826 }, { "epoch": 2.2630331753554502, "step": 955, "train/alpha": 0.2, "train/clean_ce_loss": 0.06298828125, "train/clean_loss_scale": 20.992841720581055, "train/clean_loss_weight": 4.198568344116211, "train/combined_loss": 1.586763620376587, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22630331753554503, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.322303056716919 }, { "epoch": 2.2630331753554502, "step": 955, "train/alpha": 0.2, "train/clean_ce_loss": 0.076171875, "train/clean_loss_scale": 10.287129402160645, "train/clean_loss_weight": 2.0574259757995605, "train/combined_loss": 0.9403079152107239, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22630331753554503, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7835898995399475 }, { "epoch": 2.265402843601896, "step": 956, "train/alpha": 0.2, "train/clean_ce_loss": 0.10302734375, "train/clean_loss_scale": 12.535741806030273, "train/clean_loss_weight": 2.507148504257202, "train/combined_loss": 1.5498290061950684, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22654028436018958, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2915241718292236 }, { "epoch": 2.265402843601896, "step": 956, "train/alpha": 0.2, "train/clean_ce_loss": 0.10107421875, "train/clean_loss_scale": 10.806662559509277, "train/clean_loss_weight": 2.161332607269287, "train/combined_loss": 1.31072998046875, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22654028436018958, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0922750234603882 }, { "epoch": 2.265402843601896, "step": 956, "train/alpha": 0.2, "train/clean_ce_loss": 0.04296875, "train/clean_loss_scale": 16.739408493041992, "train/clean_loss_weight": 3.34788179397583, "train/combined_loss": 0.8631256818771362, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22654028436018958, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7192714214324951 }, { "epoch": 2.265402843601896, "step": 956, "train/alpha": 0.2, "train/clean_ce_loss": 0.000484466552734375, "train/clean_loss_scale": 576.1254272460938, "train/clean_loss_weight": 115.22509002685547, "train/combined_loss": 0.3349362015724182, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22654028436018958, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2791135013103485 }, { "epoch": 2.2677725118483414, "step": 957, "train/alpha": 0.2, "train/clean_ce_loss": 0.1455078125, "train/clean_loss_scale": 7.955226421356201, "train/clean_loss_weight": 1.5910452604293823, "train/combined_loss": 1.3890571594238281, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22677725118483413, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1575475931167603 }, { "epoch": 2.2677725118483414, "step": 957, "train/alpha": 0.2, "train/clean_ce_loss": 0.028076171875, "train/clean_loss_scale": 26.45905876159668, "train/clean_loss_weight": 5.291811943054199, "train/combined_loss": 0.8914428949356079, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22677725118483413, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7428690791130066 }, { "epoch": 2.2677725118483414, "step": 957, "train/alpha": 0.2, "train/clean_ce_loss": 0.0115966796875, "train/clean_loss_scale": 30.372758865356445, "train/clean_loss_weight": 6.074552059173584, "train/combined_loss": 0.42266780138015747, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22677725118483413, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.35222315788269043 }, { "epoch": 2.2677725118483414, "step": 957, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 9.355910301208496, "train/clean_loss_weight": 1.8711820840835571, "train/combined_loss": 0.729103147983551, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22677725118483413, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6075859665870667 }, { "epoch": 2.270142180094787, "step": 958, "train/alpha": 0.2, "train/clean_ce_loss": 0.018310546875, "train/clean_loss_scale": 45.7214469909668, "train/clean_loss_weight": 9.14428997039795, "train/combined_loss": 1.0046217441558838, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22701421800947869, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.837184727191925 }, { "epoch": 2.270142180094787, "step": 958, "train/alpha": 0.2, "train/clean_ce_loss": 0.078125, "train/clean_loss_scale": 10.120844841003418, "train/clean_loss_weight": 2.0241689682006836, "train/combined_loss": 0.9488292336463928, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22701421800947869, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7906910181045532 }, { "epoch": 2.270142180094787, "step": 958, "train/alpha": 0.2, "train/clean_ce_loss": 0.060302734375, "train/clean_loss_scale": 12.12453556060791, "train/clean_loss_weight": 2.4249072074890137, "train/combined_loss": 0.8773711919784546, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22701421800947869, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7311426401138306 }, { "epoch": 2.270142180094787, "step": 958, "train/alpha": 0.2, "train/clean_ce_loss": 0.08935546875, "train/clean_loss_scale": 10.539011001586914, "train/clean_loss_weight": 2.107802152633667, "train/combined_loss": 1.1300618648529053, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22701421800947869, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9417182803153992 }, { "epoch": 2.272511848341232, "step": 959, "train/alpha": 0.2, "train/clean_ce_loss": 0.034912109375, "train/clean_loss_scale": 33.817230224609375, "train/clean_loss_weight": 6.76344633102417, "train/combined_loss": 1.4167569875717163, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.2272511848341232, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1806308031082153 }, { "epoch": 2.272511848341232, "step": 959, "train/alpha": 0.2, "train/clean_ce_loss": 0.014892578125, "train/clean_loss_scale": 33.34556579589844, "train/clean_loss_weight": 6.6691131591796875, "train/combined_loss": 0.5959217548370361, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.2272511848341232, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4966014325618744 }, { "epoch": 2.272511848341232, "step": 959, "train/alpha": 0.2, "train/clean_ce_loss": 0.05810546875, "train/clean_loss_scale": 10.879541397094727, "train/clean_loss_weight": 2.175908327102661, "train/combined_loss": 0.7585930228233337, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.2272511848341232, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6321608424186707 }, { "epoch": 2.272511848341232, "step": 959, "train/alpha": 0.2, "train/clean_ce_loss": 0.000385284423828125, "train/clean_loss_scale": 699.8117065429688, "train/clean_loss_weight": 139.96234130859375, "train/combined_loss": 0.32355186343193054, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.2272511848341232, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.26962655782699585 }, { "epoch": 2.2748815165876777, "grad_norm": 19.53838539123535, "learning_rate": 2e-06, "loss": 3.2016, "step": 960 }, { "epoch": 2.2748815165876777, "step": 960, "train/alpha": 0.2, "train/clean_ce_loss": 0.10400390625, "train/clean_loss_scale": 10.40697193145752, "train/clean_loss_weight": 2.0813944339752197, "train/combined_loss": 1.2988388538360596, "train/pos_loss/pos0": 0.30873051285743713, "train/pos_loss/pos1": 0.7435789108276367, "train/pos_loss/pos2": 1.1505178213119507, "train/pos_loss/pos3": 1.6381100416183472, "train/pos_loss/pos4": 2.2428572177886963, "train/pos_loss/pos5": 2.6306936740875244, "train/pos_loss/pos6": 3.154294490814209, "train/pos_loss/weight0": 2.1182122230529785, "train/pos_loss/weight1": 1.6648914813995361, "train/pos_loss/weight2": 1.2232017517089844, "train/pos_loss/weight3": 0.8711011409759521, "train/pos_loss/weight4": 0.5600275993347168, "train/pos_loss/weight5": 0.33989471197128296, "train/pos_loss/weight6": 0.22267135977745056, "train/pos_weight_progress": 0.22748815165876776, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0823657512664795 }, { "epoch": 2.2748815165876777, "step": 960, "train/alpha": 0.2, "train/clean_ce_loss": 0.054931640625, "train/clean_loss_scale": 9.240283966064453, "train/clean_loss_weight": 1.8480567932128906, "train/combined_loss": 0.6091007590293884, "train/pos_loss/pos0": 0.13347306847572327, "train/pos_loss/pos1": 0.33837369084358215, "train/pos_loss/pos2": 0.8155345916748047, "train/pos_loss/pos3": 1.2353003025054932, "train/pos_loss/pos4": 1.8942607641220093, "train/pos_loss/pos5": 2.55952525138855, "train/pos_loss/pos6": 2.902401924133301, "train/pos_loss/weight0": 2.8592917919158936, "train/pos_loss/weight1": 1.9906296730041504, "train/pos_loss/weight2": 1.1768652200698853, "train/pos_loss/weight3": 0.5995312333106995, "train/pos_loss/weight4": 0.255565345287323, "train/pos_loss/weight5": 0.08645888417959213, "train/pos_loss/weight6": 0.03165765851736069, "train/pos_weight_progress": 0.22748815165876776, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5075839757919312 }, { "epoch": 2.2748815165876777, "step": 960, "train/alpha": 0.2, "train/clean_ce_loss": 0.1484375, "train/clean_loss_scale": 7.08087158203125, "train/clean_loss_weight": 1.416174292564392, "train/combined_loss": 1.2612802982330322, "train/pos_loss/pos0": 0.40724074840545654, "train/pos_loss/pos1": 0.8240991830825806, "train/pos_loss/pos2": 1.3875726461410522, "train/pos_loss/pos3": 2.0348801612854004, "train/pos_loss/pos4": 2.406308650970459, "train/pos_loss/pos5": 2.8950536251068115, "train/pos_loss/pos6": 3.1678433418273926, "train/pos_loss/weight0": 2.5622940063476562, "train/pos_loss/weight1": 1.8588261604309082, "train/pos_loss/weight2": 1.2503077983856201, "train/pos_loss/weight3": 0.7397327423095703, "train/pos_loss/weight4": 0.36472296714782715, "train/pos_loss/weight5": 0.15594270825386047, "train/pos_loss/weight6": 0.06817382574081421, "train/pos_weight_progress": 0.22748815165876776, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0510668754577637 }, { "epoch": 2.2748815165876777, "step": 960, "train/alpha": 0.2, "train/clean_ce_loss": 0.00848388671875, "train/clean_loss_scale": 14.101029396057129, "train/clean_loss_weight": 2.8202059268951416, "train/combined_loss": 0.1435578465461731, "train/pos_loss/pos0": 0.005769635085016489, "train/pos_loss/pos1": 0.028376461938023567, "train/pos_loss/pos2": 0.15917283296585083, "train/pos_loss/pos3": 0.367997944355011, "train/pos_loss/pos4": 0.8929265141487122, "train/pos_loss/pos5": 1.795508861541748, "train/pos_loss/pos6": 2.36195969581604, "train/pos_loss/weight0": 2.963261127471924, "train/pos_loss/weight1": 1.9938526153564453, "train/pos_loss/weight2": 1.1803218126296997, "train/pos_loss/weight3": 0.5559884309768677, "train/pos_loss/weight4": 0.21233490109443665, "train/pos_loss/weight5": 0.07083046436309814, "train/pos_loss/weight6": 0.023410676047205925, "train/pos_weight_progress": 0.22748815165876776, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.11963153630495071 }, { "epoch": 2.2772511848341233, "step": 961, "train/alpha": 0.2, "train/clean_ce_loss": 0.005767822265625, "train/clean_loss_scale": 75.37857055664062, "train/clean_loss_weight": 15.075714111328125, "train/combined_loss": 0.5217242240905762, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22772511848341231, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4347701966762543 }, { "epoch": 2.2772511848341233, "step": 961, "train/alpha": 0.2, "train/clean_ce_loss": 0.03466796875, "train/clean_loss_scale": 32.123191833496094, "train/clean_loss_weight": 6.424638271331787, "train/combined_loss": 1.3363749980926514, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22772511848341231, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1136457920074463 }, { "epoch": 2.2772511848341233, "step": 961, "train/alpha": 0.2, "train/clean_ce_loss": 0.0111083984375, "train/clean_loss_scale": 101.84717559814453, "train/clean_loss_weight": 20.369436264038086, "train/combined_loss": 1.357630729675293, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22772511848341231, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1313589811325073 }, { "epoch": 2.2772511848341233, "step": 961, "train/alpha": 0.2, "train/clean_ce_loss": 0.05712890625, "train/clean_loss_scale": 12.561208724975586, "train/clean_loss_weight": 2.512241840362549, "train/combined_loss": 0.8611297011375427, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22772511848341231, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7176080942153931 }, { "epoch": 2.279620853080569, "step": 962, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 12.646003723144531, "train/clean_loss_weight": 2.529200792312622, "train/combined_loss": 1.1262847185134888, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22796208530805687, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9385705590248108 }, { "epoch": 2.279620853080569, "step": 962, "train/alpha": 0.2, "train/clean_ce_loss": 0.02978515625, "train/clean_loss_scale": 27.989913940429688, "train/clean_loss_weight": 5.597982883453369, "train/combined_loss": 1.0004208087921143, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22796208530805687, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.833683967590332 }, { "epoch": 2.279620853080569, "step": 962, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 18.73233985900879, "train/clean_loss_weight": 3.7464680671691895, "train/combined_loss": 1.4927332401275635, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22796208530805687, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2439444065093994 }, { "epoch": 2.279620853080569, "step": 962, "train/alpha": 0.2, "train/clean_ce_loss": 0.01556396484375, "train/clean_loss_scale": 28.644792556762695, "train/clean_loss_weight": 5.728958606719971, "train/combined_loss": 0.5349918603897095, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22796208530805687, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.44582653045654297 }, { "epoch": 2.2819905213270144, "step": 963, "train/alpha": 0.2, "train/clean_ce_loss": 0.002838134765625, "train/clean_loss_scale": 122.21990203857422, "train/clean_loss_weight": 24.443981170654297, "train/combined_loss": 0.41625186800956726, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22819905213270142, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3468765616416931 }, { "epoch": 2.2819905213270144, "step": 963, "train/alpha": 0.2, "train/clean_ce_loss": 0.0174560546875, "train/clean_loss_scale": 27.749515533447266, "train/clean_loss_weight": 5.549903392791748, "train/combined_loss": 0.5812764763832092, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22819905213270142, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4843970537185669 }, { "epoch": 2.2819905213270144, "step": 963, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 10.526545524597168, "train/clean_loss_weight": 2.105309247970581, "train/combined_loss": 1.1040537357330322, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22819905213270142, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9200447797775269 }, { "epoch": 2.2819905213270144, "step": 963, "train/alpha": 0.2, "train/clean_ce_loss": 0.0625, "train/clean_loss_scale": 13.197890281677246, "train/clean_loss_weight": 2.639578104019165, "train/combined_loss": 0.9898417592048645, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22819905213270142, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8248681426048279 }, { "epoch": 2.2843601895734595, "step": 964, "train/alpha": 0.2, "train/clean_ce_loss": 0.050048828125, "train/clean_loss_scale": 21.000606536865234, "train/clean_loss_weight": 4.2001214027404785, "train/combined_loss": 1.2612669467926025, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22843601895734597, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0510557889938354 }, { "epoch": 2.2843601895734595, "step": 964, "train/alpha": 0.2, "train/clean_ce_loss": 7.2479248046875e-05, "train/clean_loss_scale": 3781.0341796875, "train/clean_loss_weight": 756.2068481445312, "train/combined_loss": 0.32885581254959106, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22843601895734597, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.27404651045799255 }, { "epoch": 2.2843601895734595, "step": 964, "train/alpha": 0.2, "train/clean_ce_loss": 0.0247802734375, "train/clean_loss_scale": 23.009639739990234, "train/clean_loss_weight": 4.601928234100342, "train/combined_loss": 0.6842222213745117, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22843601895734597, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5701851844787598 }, { "epoch": 2.2843601895734595, "step": 964, "train/alpha": 0.2, "train/clean_ce_loss": 0.009521484375, "train/clean_loss_scale": 105.19165802001953, "train/clean_loss_weight": 21.038331985473633, "train/combined_loss": 1.2018969058990479, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22843601895734597, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0015807151794434 }, { "epoch": 2.286729857819905, "grad_norm": 22.35969352722168, "learning_rate": 2e-06, "loss": 3.2724, "step": 965 }, { "epoch": 2.286729857819905, "step": 965, "train/alpha": 0.2, "train/clean_ce_loss": 0.04541015625, "train/clean_loss_scale": 17.506132125854492, "train/clean_loss_weight": 3.5012264251708984, "train/combined_loss": 0.9539474248886108, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22867298578199052, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7949562072753906 }, { "epoch": 2.286729857819905, "step": 965, "train/alpha": 0.2, "train/clean_ce_loss": 0.05712890625, "train/clean_loss_scale": 10.57478141784668, "train/clean_loss_weight": 2.1149563789367676, "train/combined_loss": 0.7249507904052734, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22867298578199052, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6041256785392761 }, { "epoch": 2.286729857819905, "step": 965, "train/alpha": 0.2, "train/clean_ce_loss": 0.115234375, "train/clean_loss_scale": 11.666370391845703, "train/clean_loss_weight": 2.3332741260528564, "train/combined_loss": 1.6132402420043945, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22867298578199052, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3443669080734253 }, { "epoch": 2.286729857819905, "step": 965, "train/alpha": 0.2, "train/clean_ce_loss": 0.01031494140625, "train/clean_loss_scale": 33.9829216003418, "train/clean_loss_weight": 6.796584606170654, "train/combined_loss": 0.42063820362091064, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22867298578199052, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.35053184628486633 }, { "epoch": 2.2890995260663507, "step": 966, "train/alpha": 0.2, "train/clean_ce_loss": 0.16015625, "train/clean_loss_scale": 7.78200626373291, "train/clean_loss_weight": 1.556401252746582, "train/combined_loss": 1.4956042766571045, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22890995260663508, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2463369369506836 }, { "epoch": 2.2890995260663507, "step": 966, "train/alpha": 0.2, "train/clean_ce_loss": 0.005157470703125, "train/clean_loss_scale": 192.741455078125, "train/clean_loss_weight": 38.54829025268555, "train/combined_loss": 1.1928701400756836, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22890995260663508, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9940584301948547 }, { "epoch": 2.2890995260663507, "step": 966, "train/alpha": 0.2, "train/clean_ce_loss": 0.07373046875, "train/clean_loss_scale": 9.29662799835205, "train/clean_loss_weight": 1.859325647354126, "train/combined_loss": 0.8225337266921997, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22890995260663508, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6854447722434998 }, { "epoch": 2.2890995260663507, "step": 966, "train/alpha": 0.2, "train/clean_ce_loss": 0.091796875, "train/clean_loss_scale": 11.740917205810547, "train/clean_loss_weight": 2.3481833934783936, "train/combined_loss": 1.2933354377746582, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22890995260663508, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0777795314788818 }, { "epoch": 2.2914691943127963, "step": 967, "train/alpha": 0.2, "train/clean_ce_loss": 0.07568359375, "train/clean_loss_scale": 15.665249824523926, "train/clean_loss_weight": 3.133049964904785, "train/combined_loss": 1.4227229356765747, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22914691943127963, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1856024265289307 }, { "epoch": 2.2914691943127963, "step": 967, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 23.316253662109375, "train/clean_loss_weight": 4.663250923156738, "train/combined_loss": 1.8580139875411987, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22914691943127963, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5483449697494507 }, { "epoch": 2.2914691943127963, "step": 967, "train/alpha": 0.2, "train/clean_ce_loss": 0.0087890625, "train/clean_loss_scale": 57.61270523071289, "train/clean_loss_weight": 11.522541046142578, "train/combined_loss": 0.6076340079307556, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22914691943127963, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5063616633415222 }, { "epoch": 2.2914691943127963, "step": 967, "train/alpha": 0.2, "train/clean_ce_loss": 0.003326416015625, "train/clean_loss_scale": 139.30226135253906, "train/clean_loss_weight": 27.86045265197754, "train/combined_loss": 0.5560527443885803, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22914691943127963, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4633772671222687 }, { "epoch": 2.293838862559242, "step": 968, "train/alpha": 0.2, "train/clean_ce_loss": 0.0272216796875, "train/clean_loss_scale": 17.205598831176758, "train/clean_loss_weight": 3.441119909286499, "train/combined_loss": 0.5620383620262146, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22938388625592418, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46836528182029724 }, { "epoch": 2.293838862559242, "step": 968, "train/alpha": 0.2, "train/clean_ce_loss": 0.07470703125, "train/clean_loss_scale": 15.432195663452148, "train/clean_loss_weight": 3.0864391326904297, "train/combined_loss": 1.383472204208374, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22938388625592418, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1528935432434082 }, { "epoch": 2.293838862559242, "step": 968, "train/alpha": 0.2, "train/clean_ce_loss": 0.0458984375, "train/clean_loss_scale": 7.141408920288086, "train/clean_loss_weight": 1.4282817840576172, "train/combined_loss": 0.39333540201187134, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22938388625592418, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.32777950167655945 }, { "epoch": 2.293838862559242, "step": 968, "train/alpha": 0.2, "train/clean_ce_loss": 0.1494140625, "train/clean_loss_scale": 10.647618293762207, "train/clean_loss_weight": 2.129523754119873, "train/combined_loss": 1.909084677696228, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22938388625592418, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.590903878211975 }, { "epoch": 2.2962085308056874, "step": 969, "train/alpha": 0.2, "train/clean_ce_loss": 0.06982421875, "train/clean_loss_scale": 10.25139045715332, "train/clean_loss_weight": 2.0502781867980957, "train/combined_loss": 0.8589544296264648, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22962085308056873, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7157953381538391 }, { "epoch": 2.2962085308056874, "step": 969, "train/alpha": 0.2, "train/clean_ce_loss": 0.11279296875, "train/clean_loss_scale": 12.068561553955078, "train/clean_loss_weight": 2.4137122631073, "train/combined_loss": 1.6334986686706543, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22962085308056873, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3612488508224487 }, { "epoch": 2.2962085308056874, "step": 969, "train/alpha": 0.2, "train/clean_ce_loss": 0.0220947265625, "train/clean_loss_scale": 55.04083251953125, "train/clean_loss_weight": 11.008166313171387, "train/combined_loss": 1.4593346118927002, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22962085308056873, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2161121368408203 }, { "epoch": 2.2962085308056874, "step": 969, "train/alpha": 0.2, "train/clean_ce_loss": 0.1708984375, "train/clean_loss_scale": 9.25090217590332, "train/clean_loss_weight": 1.8501805067062378, "train/combined_loss": 1.8971577882766724, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22962085308056873, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5809648036956787 }, { "epoch": 2.2985781990521326, "grad_norm": 16.382253646850586, "learning_rate": 2e-06, "loss": 3.6335, "step": 970 }, { "epoch": 2.2985781990521326, "step": 970, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 15.984845161437988, "train/clean_loss_weight": 3.1969690322875977, "train/combined_loss": 1.4236502647399902, "train/pos_loss/pos0": 0.3139497935771942, "train/pos_loss/pos1": 0.7955438494682312, "train/pos_loss/pos2": 1.1971445083618164, "train/pos_loss/pos3": 1.7850958108901978, "train/pos_loss/pos4": 2.344510316848755, "train/pos_loss/pos5": 2.6558797359466553, "train/pos_loss/pos6": 3.02331280708313, "train/pos_loss/weight0": 2.012927770614624, "train/pos_loss/weight1": 1.587325930595398, "train/pos_loss/weight2": 1.210402488708496, "train/pos_loss/weight3": 0.8802268505096436, "train/pos_loss/weight4": 0.6114360094070435, "train/pos_loss/weight5": 0.4100261926651001, "train/pos_loss/weight6": 0.2876550257205963, "train/pos_weight_progress": 0.22985781990521326, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1863752603530884 }, { "epoch": 2.2985781990521326, "step": 970, "train/alpha": 0.2, "train/clean_ce_loss": 0.1142578125, "train/clean_loss_scale": 10.169715881347656, "train/clean_loss_weight": 2.0339431762695312, "train/combined_loss": 1.3943634033203125, "train/pos_loss/pos0": 0.519483745098114, "train/pos_loss/pos1": 0.9730931520462036, "train/pos_loss/pos2": 1.6589384078979492, "train/pos_loss/pos3": 2.5725579261779785, "train/pos_loss/pos4": 3.3177614212036133, "train/pos_loss/pos5": 3.4296395778656006, "train/pos_loss/pos6": 3.792933702468872, "train/pos_loss/weight0": 2.852116584777832, "train/pos_loss/weight1": 2.0107243061065674, "train/pos_loss/weight2": 1.1940447092056274, "train/pos_loss/weight3": 0.579225480556488, "train/pos_loss/weight4": 0.2552417516708374, "train/pos_loss/weight5": 0.0789145827293396, "train/pos_loss/weight6": 0.029732588678598404, "train/pos_weight_progress": 0.22985781990521326, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.161969542503357 }, { "epoch": 2.2985781990521326, "step": 970, "train/alpha": 0.2, "train/clean_ce_loss": 0.140625, "train/clean_loss_scale": 5.78367805480957, "train/clean_loss_weight": 1.1567356586456299, "train/combined_loss": 0.9759957194328308, "train/pos_loss/pos0": 0.413409948348999, "train/pos_loss/pos1": 0.8149308562278748, "train/pos_loss/pos2": 1.3241738080978394, "train/pos_loss/pos3": 1.9490000009536743, "train/pos_loss/pos4": 2.373934507369995, "train/pos_loss/pos5": 2.851390838623047, "train/pos_loss/pos6": 3.1099441051483154, "train/pos_loss/weight0": 3.2832212448120117, "train/pos_loss/weight1": 2.0529844760894775, "train/pos_loss/weight2": 1.0655392408370972, "train/pos_loss/weight3": 0.4358205199241638, "train/pos_loss/weight4": 0.1271386742591858, "train/pos_loss/weight5": 0.028453879058361053, "train/pos_loss/weight6": 0.006842222064733505, "train/pos_weight_progress": 0.22985781990521326, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8133297562599182 }, { "epoch": 2.2985781990521326, "step": 970, "train/alpha": 0.2, "train/clean_ce_loss": 0.1015625, "train/clean_loss_scale": 8.72462272644043, "train/clean_loss_weight": 1.744924545288086, "train/combined_loss": 1.063313364982605, "train/pos_loss/pos0": 0.4498524069786072, "train/pos_loss/pos1": 0.8922513723373413, "train/pos_loss/pos2": 1.321824073791504, "train/pos_loss/pos3": 1.7959407567977905, "train/pos_loss/pos4": 2.2096571922302246, "train/pos_loss/pos5": 2.426492929458618, "train/pos_loss/pos6": 2.8666887283325195, "train/pos_loss/weight0": 3.0792787075042725, "train/pos_loss/weight1": 2.0064353942871094, "train/pos_loss/weight2": 1.1262085437774658, "train/pos_loss/weight3": 0.5365553498268127, "train/pos_loss/weight4": 0.18596132099628448, "train/pos_loss/weight5": 0.050717346370220184, "train/pos_loss/weight6": 0.014843259938061237, "train/pos_weight_progress": 0.22985781990521326, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8860945105552673 }, { "epoch": 2.300947867298578, "step": 971, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 8.557239532470703, "train/clean_loss_weight": 1.7114479541778564, "train/combined_loss": 0.6819050312042236, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.2300947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5682541728019714 }, { "epoch": 2.300947867298578, "step": 971, "train/alpha": 0.2, "train/clean_ce_loss": 0.060302734375, "train/clean_loss_scale": 8.231623649597168, "train/clean_loss_weight": 1.6463247537612915, "train/combined_loss": 0.59566730260849, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.2300947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4963894188404083 }, { "epoch": 2.300947867298578, "step": 971, "train/alpha": 0.2, "train/clean_ce_loss": 0.06982421875, "train/clean_loss_scale": 9.81788444519043, "train/clean_loss_weight": 1.9635769128799438, "train/combined_loss": 0.8226313591003418, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.2300947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6855261325836182 }, { "epoch": 2.300947867298578, "step": 971, "train/alpha": 0.2, "train/clean_ce_loss": 0.005218505859375, "train/clean_loss_scale": 53.074954986572266, "train/clean_loss_weight": 10.614991188049316, "train/combined_loss": 0.33236634731292725, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.2300947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2769719660282135 }, { "epoch": 2.3033175355450237, "step": 972, "train/alpha": 0.2, "train/clean_ce_loss": 0.04833984375, "train/clean_loss_scale": 8.543295860290527, "train/clean_loss_weight": 1.7086591720581055, "train/combined_loss": 0.495577871799469, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23033175355450236, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4129815697669983 }, { "epoch": 2.3033175355450237, "step": 972, "train/alpha": 0.2, "train/clean_ce_loss": 0.00799560546875, "train/clean_loss_scale": 30.794790267944336, "train/clean_loss_weight": 6.1589579582214355, "train/combined_loss": 0.2954675853252411, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23033175355450236, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.24622298777103424 }, { "epoch": 2.3033175355450237, "step": 972, "train/alpha": 0.2, "train/clean_ce_loss": 0.061279296875, "train/clean_loss_scale": 13.821029663085938, "train/clean_loss_weight": 2.7642059326171875, "train/combined_loss": 1.0163315534591675, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23033175355450236, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8469429612159729 }, { "epoch": 2.3033175355450237, "step": 972, "train/alpha": 0.2, "train/clean_ce_loss": 0.11328125, "train/clean_loss_scale": 7.682487964630127, "train/clean_loss_weight": 1.5364975929260254, "train/combined_loss": 1.0443382263183594, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23033175355450236, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8702818155288696 }, { "epoch": 2.3056872037914693, "step": 973, "train/alpha": 0.2, "train/clean_ce_loss": 0.040283203125, "train/clean_loss_scale": 15.706023216247559, "train/clean_loss_weight": 3.141204595565796, "train/combined_loss": 0.7592267394065857, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.2305687203791469, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6326889395713806 }, { "epoch": 2.3056872037914693, "step": 973, "train/alpha": 0.2, "train/clean_ce_loss": 0.0771484375, "train/clean_loss_scale": 7.16148567199707, "train/clean_loss_weight": 1.4322971105575562, "train/combined_loss": 0.6629969477653503, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.2305687203791469, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5524974465370178 }, { "epoch": 2.3056872037914693, "step": 973, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 8.930526733398438, "train/clean_loss_weight": 1.7861053943634033, "train/combined_loss": 0.8529350757598877, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.2305687203791469, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7107791900634766 }, { "epoch": 2.3056872037914693, "step": 973, "train/alpha": 0.2, "train/clean_ce_loss": 0.044189453125, "train/clean_loss_scale": 9.332895278930664, "train/clean_loss_weight": 1.8665790557861328, "train/combined_loss": 0.49489864706993103, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.2305687203791469, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4124155342578888 }, { "epoch": 2.308056872037915, "step": 974, "train/alpha": 0.2, "train/clean_ce_loss": 0.1337890625, "train/clean_loss_scale": 6.567002773284912, "train/clean_loss_weight": 1.3134006261825562, "train/combined_loss": 1.054311752319336, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23080568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8785931468009949 }, { "epoch": 2.308056872037915, "step": 974, "train/alpha": 0.2, "train/clean_ce_loss": 0.052978515625, "train/clean_loss_scale": 6.602190971374512, "train/clean_loss_weight": 1.320438265800476, "train/combined_loss": 0.4197291135787964, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23080568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3497742712497711 }, { "epoch": 2.308056872037915, "step": 974, "train/alpha": 0.2, "train/clean_ce_loss": 0.0732421875, "train/clean_loss_scale": 6.997153282165527, "train/clean_loss_weight": 1.3994306325912476, "train/combined_loss": 0.6149841547012329, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23080568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5124868154525757 }, { "epoch": 2.308056872037915, "step": 974, "train/alpha": 0.2, "train/clean_ce_loss": 0.05859375, "train/clean_loss_scale": 8.132743835449219, "train/clean_loss_weight": 1.6265487670898438, "train/combined_loss": 0.5718335509300232, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23080568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.47652795910835266 }, { "epoch": 2.31042654028436, "grad_norm": 20.291345596313477, "learning_rate": 2e-06, "loss": 2.8824, "step": 975 }, { "epoch": 2.31042654028436, "step": 975, "train/alpha": 0.2, "train/clean_ce_loss": 0.0859375, "train/clean_loss_scale": 12.12851333618164, "train/clean_loss_weight": 2.4257028102874756, "train/combined_loss": 1.2507529258728027, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23104265402843602, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0422941446304321 }, { "epoch": 2.31042654028436, "step": 975, "train/alpha": 0.2, "train/clean_ce_loss": 0.10498046875, "train/clean_loss_scale": 8.300626754760742, "train/clean_loss_weight": 1.6601253747940063, "train/combined_loss": 1.0456844568252563, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23104265402843602, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.871403694152832 }, { "epoch": 2.31042654028436, "step": 975, "train/alpha": 0.2, "train/clean_ce_loss": 0.02978515625, "train/clean_loss_scale": 19.559677124023438, "train/clean_loss_weight": 3.911935567855835, "train/combined_loss": 0.6991056203842163, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23104265402843602, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5825880169868469 }, { "epoch": 2.31042654028436, "step": 975, "train/alpha": 0.2, "train/clean_ce_loss": 0.0001983642578125, "train/clean_loss_scale": 384.0419616699219, "train/clean_loss_weight": 76.80839538574219, "train/combined_loss": 0.09141623973846436, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23104265402843602, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.07618019729852676 }, { "epoch": 2.3127962085308056, "step": 976, "train/alpha": 0.2, "train/clean_ce_loss": 0.017333984375, "train/clean_loss_scale": 16.38363265991211, "train/clean_loss_weight": 3.276726484298706, "train/combined_loss": 0.34079235792160034, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23127962085308057, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2839936316013336 }, { "epoch": 2.3127962085308056, "step": 976, "train/alpha": 0.2, "train/clean_ce_loss": 0.00616455078125, "train/clean_loss_scale": 26.902061462402344, "train/clean_loss_weight": 5.380412578582764, "train/combined_loss": 0.19900694489479065, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23127962085308057, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.16583912074565887 }, { "epoch": 2.3127962085308056, "step": 976, "train/alpha": 0.2, "train/clean_ce_loss": 0.00421142578125, "train/clean_loss_scale": 30.891420364379883, "train/clean_loss_weight": 6.178284168243408, "train/combined_loss": 0.1561163067817688, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23127962085308057, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1300969272851944 }, { "epoch": 2.3127962085308056, "step": 976, "train/alpha": 0.2, "train/clean_ce_loss": 0.1650390625, "train/clean_loss_scale": 6.402495384216309, "train/clean_loss_weight": 1.2804991006851196, "train/combined_loss": 1.2679941654205322, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23127962085308057, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.05666184425354 }, { "epoch": 2.315165876777251, "step": 977, "train/alpha": 0.2, "train/clean_ce_loss": 0.083984375, "train/clean_loss_scale": 9.233546257019043, "train/clean_loss_weight": 1.8467092514038086, "train/combined_loss": 0.9305683374404907, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23151658767772512, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7754735946655273 }, { "epoch": 2.315165876777251, "step": 977, "train/alpha": 0.2, "train/clean_ce_loss": 0.00347900390625, "train/clean_loss_scale": 35.42210006713867, "train/clean_loss_weight": 7.084420204162598, "train/combined_loss": 0.14788034558296204, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23151658767772512, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1232336238026619 }, { "epoch": 2.315165876777251, "step": 977, "train/alpha": 0.2, "train/clean_ce_loss": 0.09228515625, "train/clean_loss_scale": 6.450651168823242, "train/clean_loss_weight": 1.2901302576065063, "train/combined_loss": 0.7143592238426208, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23151658767772512, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5952993631362915 }, { "epoch": 2.315165876777251, "step": 977, "train/alpha": 0.2, "train/clean_ce_loss": 0.043701171875, "train/clean_loss_scale": 14.015114784240723, "train/clean_loss_weight": 2.803023099899292, "train/combined_loss": 0.734972357749939, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23151658767772512, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6124769449234009 }, { "epoch": 2.3175355450236967, "step": 978, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 6.679485321044922, "train/clean_loss_weight": 1.3358970880508423, "train/combined_loss": 0.6496843099594116, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23175355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.541403591632843 }, { "epoch": 2.3175355450236967, "step": 978, "train/alpha": 0.2, "train/clean_ce_loss": 0.02978515625, "train/clean_loss_scale": 8.009834289550781, "train/clean_loss_weight": 1.6019668579101562, "train/combined_loss": 0.2862890064716339, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23175355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.23857417702674866 }, { "epoch": 2.3175355450236967, "step": 978, "train/alpha": 0.2, "train/clean_ce_loss": 0.10888671875, "train/clean_loss_scale": 7.814538955688477, "train/clean_loss_weight": 1.5629078149795532, "train/combined_loss": 1.021079421043396, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23175355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8508995175361633 }, { "epoch": 2.3175355450236967, "step": 978, "train/alpha": 0.2, "train/clean_ce_loss": 0.1005859375, "train/clean_loss_scale": 9.26771068572998, "train/clean_loss_weight": 1.8535422086715698, "train/combined_loss": 1.1186416149139404, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23175355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9322013258934021 }, { "epoch": 2.3199052132701423, "step": 979, "train/alpha": 0.2, "train/clean_ce_loss": 0.07470703125, "train/clean_loss_scale": 9.601757049560547, "train/clean_loss_weight": 1.9203513860702515, "train/combined_loss": 0.8607825040817261, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23199052132701423, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7173187732696533 }, { "epoch": 2.3199052132701423, "step": 979, "train/alpha": 0.2, "train/clean_ce_loss": 0.1318359375, "train/clean_loss_scale": 7.773336410522461, "train/clean_loss_weight": 1.554667353630066, "train/combined_loss": 1.2297661304473877, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23199052132701423, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0248050689697266 }, { "epoch": 2.3199052132701423, "step": 979, "train/alpha": 0.2, "train/clean_ce_loss": 0.0732421875, "train/clean_loss_scale": 7.634078025817871, "train/clean_loss_weight": 1.52681565284729, "train/combined_loss": 0.6709638833999634, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23199052132701423, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5591365694999695 }, { "epoch": 2.3199052132701423, "step": 979, "train/alpha": 0.2, "train/clean_ce_loss": 0.0966796875, "train/clean_loss_scale": 5.744094371795654, "train/clean_loss_weight": 1.148818850517273, "train/combined_loss": 0.6664047241210938, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23199052132701423, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5553372502326965 }, { "epoch": 2.322274881516588, "grad_norm": 21.97747039794922, "learning_rate": 2e-06, "loss": 2.9638, "step": 980 }, { "epoch": 2.322274881516588, "step": 980, "train/alpha": 0.2, "train/clean_ce_loss": 0.1357421875, "train/clean_loss_scale": 6.189042568206787, "train/clean_loss_weight": 1.2378085851669312, "train/combined_loss": 1.0081369876861572, "train/pos_loss/pos0": 0.40966665744781494, "train/pos_loss/pos1": 0.7471479177474976, "train/pos_loss/pos2": 1.1302669048309326, "train/pos_loss/pos3": 1.6091387271881104, "train/pos_loss/pos4": 2.1223397254943848, "train/pos_loss/pos5": 2.555490732192993, "train/pos_loss/pos6": 2.9650063514709473, "train/pos_loss/weight0": 2.8573555946350098, "train/pos_loss/weight1": 1.9673349857330322, "train/pos_loss/weight2": 1.156607747077942, "train/pos_loss/weight3": 0.6022844910621643, "train/pos_loss/weight4": 0.2688447833061218, "train/pos_loss/weight5": 0.10432226210832596, "train/pos_loss/weight6": 0.04325004667043686, "train/pos_weight_progress": 0.23222748815165878, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.840114176273346 }, { "epoch": 2.322274881516588, "step": 980, "train/alpha": 0.2, "train/clean_ce_loss": 0.14453125, "train/clean_loss_scale": 8.059981346130371, "train/clean_loss_weight": 1.6119962930679321, "train/combined_loss": 1.3979029655456543, "train/pos_loss/pos0": 0.5499358177185059, "train/pos_loss/pos1": 1.1289812326431274, "train/pos_loss/pos2": 1.6596344709396362, "train/pos_loss/pos3": 2.2973179817199707, "train/pos_loss/pos4": 2.6592016220092773, "train/pos_loss/pos5": 3.3196399211883545, "train/pos_loss/pos6": 3.7204413414001465, "train/pos_loss/weight0": 2.930349826812744, "train/pos_loss/weight1": 1.9580934047698975, "train/pos_loss/weight2": 1.1534466743469238, "train/pos_loss/weight3": 0.5966455936431885, "train/pos_loss/weight4": 0.2483394294977188, "train/pos_loss/weight5": 0.08395743370056152, "train/pos_loss/weight6": 0.02916698530316353, "train/pos_weight_progress": 0.23222748815165878, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.164919137954712 }, { "epoch": 2.322274881516588, "step": 980, "train/alpha": 0.2, "train/clean_ce_loss": 0.028564453125, "train/clean_loss_scale": 9.583639144897461, "train/clean_loss_weight": 1.916727900505066, "train/combined_loss": 0.32850170135498047, "train/pos_loss/pos0": 0.0370820127427578, "train/pos_loss/pos1": 0.21948179602622986, "train/pos_loss/pos2": 0.52713543176651, "train/pos_loss/pos3": 1.0156797170639038, "train/pos_loss/pos4": 1.5460669994354248, "train/pos_loss/pos5": 2.172792911529541, "train/pos_loss/pos6": 2.869033098220825, "train/pos_loss/weight0": 3.2198266983032227, "train/pos_loss/weight1": 2.066497325897217, "train/pos_loss/weight2": 1.0801610946655273, "train/pos_loss/weight3": 0.4505906105041504, "train/pos_loss/weight4": 0.13878926634788513, "train/pos_loss/weight5": 0.035695482045412064, "train/pos_loss/weight6": 0.008439804427325726, "train/pos_weight_progress": 0.23222748815165878, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2737514078617096 }, { "epoch": 2.322274881516588, "step": 980, "train/alpha": 0.2, "train/clean_ce_loss": 0.083984375, "train/clean_loss_scale": 7.992626667022705, "train/clean_loss_weight": 1.5985254049301147, "train/combined_loss": 0.8055069446563721, "train/pos_loss/pos0": 0.2230180948972702, "train/pos_loss/pos1": 0.47504574060440063, "train/pos_loss/pos2": 0.7160115838050842, "train/pos_loss/pos3": 1.0339621305465698, "train/pos_loss/pos4": 1.5868842601776123, "train/pos_loss/pos5": 1.903759241104126, "train/pos_loss/pos6": 2.2964158058166504, "train/pos_loss/weight0": 2.270660877227783, "train/pos_loss/weight1": 1.7436496019363403, "train/pos_loss/weight2": 1.2624491453170776, "train/pos_loss/weight3": 0.8337168097496033, "train/pos_loss/weight4": 0.48193037509918213, "train/pos_loss/weight5": 0.2584267854690552, "train/pos_loss/weight6": 0.14916646480560303, "train/pos_weight_progress": 0.23222748815165878, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6712557673454285 }, { "epoch": 2.324644549763033, "step": 981, "train/alpha": 0.2, "train/clean_ce_loss": 0.00579833984375, "train/clean_loss_scale": 60.459503173828125, "train/clean_loss_weight": 12.091900825500488, "train/combined_loss": 0.42067769169807434, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23246445497630333, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.35056474804878235 }, { "epoch": 2.324644549763033, "step": 981, "train/alpha": 0.2, "train/clean_ce_loss": 0.0380859375, "train/clean_loss_scale": 13.540366172790527, "train/clean_loss_weight": 2.708073377609253, "train/combined_loss": 0.6188370585441589, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23246445497630333, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5156975388526917 }, { "epoch": 2.324644549763033, "step": 981, "train/alpha": 0.2, "train/clean_ce_loss": 0.0004425048828125, "train/clean_loss_scale": 389.5958251953125, "train/clean_loss_weight": 77.9191665649414, "train/combined_loss": 0.2068776786327362, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23246445497630333, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.17239806056022644 }, { "epoch": 2.324644549763033, "step": 981, "train/alpha": 0.2, "train/clean_ce_loss": 0.07861328125, "train/clean_loss_scale": 9.89149284362793, "train/clean_loss_weight": 1.978298544883728, "train/combined_loss": 0.9331232309341431, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23246445497630333, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7776026725769043 }, { "epoch": 2.3270142180094786, "step": 982, "train/alpha": 0.2, "train/clean_ce_loss": 0.0849609375, "train/clean_loss_scale": 7.962604999542236, "train/clean_loss_weight": 1.592521071434021, "train/combined_loss": 0.8118124604225159, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23270142180094786, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6765103936195374 }, { "epoch": 2.3270142180094786, "step": 982, "train/alpha": 0.2, "train/clean_ce_loss": 0.10595703125, "train/clean_loss_scale": 7.071135520935059, "train/clean_loss_weight": 1.4142271280288696, "train/combined_loss": 0.8990838527679443, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23270142180094786, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.749236524105072 }, { "epoch": 2.3270142180094786, "step": 982, "train/alpha": 0.2, "train/clean_ce_loss": 0.06396484375, "train/clean_loss_scale": 12.066839218139648, "train/clean_loss_weight": 2.413367986679077, "train/combined_loss": 0.9262242317199707, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23270142180094786, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.771853506565094 }, { "epoch": 2.3270142180094786, "step": 982, "train/alpha": 0.2, "train/clean_ce_loss": 0.07568359375, "train/clean_loss_scale": 6.264718055725098, "train/clean_loss_weight": 1.2529436349868774, "train/combined_loss": 0.5689636468887329, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23270142180094786, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4741363823413849 }, { "epoch": 2.329383886255924, "step": 983, "train/alpha": 0.2, "train/clean_ce_loss": 0.0057373046875, "train/clean_loss_scale": 34.54100799560547, "train/clean_loss_weight": 6.908201694488525, "train/combined_loss": 0.23780673742294312, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.2329383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.19817228615283966 }, { "epoch": 2.329383886255924, "step": 983, "train/alpha": 0.2, "train/clean_ce_loss": 0.006195068359375, "train/clean_loss_scale": 24.33873176574707, "train/clean_loss_weight": 4.867746353149414, "train/combined_loss": 0.18093612790107727, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.2329383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1507801115512848 }, { "epoch": 2.329383886255924, "step": 983, "train/alpha": 0.2, "train/clean_ce_loss": 0.09521484375, "train/clean_loss_scale": 9.557109832763672, "train/clean_loss_weight": 1.9114220142364502, "train/combined_loss": 1.0919744968414307, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.2329383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9099787473678589 }, { "epoch": 2.329383886255924, "step": 983, "train/alpha": 0.2, "train/clean_ce_loss": 0.0703125, "train/clean_loss_scale": 5.786838054656982, "train/clean_loss_weight": 1.1573675870895386, "train/combined_loss": 0.4882644712924957, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.2329383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4068870544433594 }, { "epoch": 2.3317535545023698, "step": 984, "train/alpha": 0.2, "train/clean_ce_loss": 0.11572265625, "train/clean_loss_scale": 8.787484169006348, "train/clean_loss_weight": 1.7574968338012695, "train/combined_loss": 1.2202932834625244, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23317535545023696, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0169110298156738 }, { "epoch": 2.3317535545023698, "step": 984, "train/alpha": 0.2, "train/clean_ce_loss": 0.08837890625, "train/clean_loss_scale": 8.467696189880371, "train/clean_loss_weight": 1.6935392618179321, "train/combined_loss": 0.898038923740387, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23317535545023696, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7483657598495483 }, { "epoch": 2.3317535545023698, "step": 984, "train/alpha": 0.2, "train/clean_ce_loss": 0.05224609375, "train/clean_loss_scale": 18.147581100463867, "train/clean_loss_weight": 3.629516363143921, "train/combined_loss": 1.137768268585205, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23317535545023696, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9481402635574341 }, { "epoch": 2.3317535545023698, "step": 984, "train/alpha": 0.2, "train/clean_ce_loss": 0.1494140625, "train/clean_loss_scale": 7.687225341796875, "train/clean_loss_weight": 1.537445068359375, "train/combined_loss": 1.3782955408096313, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23317535545023696, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1485795974731445 }, { "epoch": 2.3341232227488153, "grad_norm": 17.227333068847656, "learning_rate": 2e-06, "loss": 3.272, "step": 985 }, { "epoch": 2.3341232227488153, "step": 985, "train/alpha": 0.2, "train/clean_ce_loss": 0.07373046875, "train/clean_loss_scale": 11.68053913116455, "train/clean_loss_weight": 2.3361079692840576, "train/combined_loss": 1.0334539413452148, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.2334123222748815, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8612115979194641 }, { "epoch": 2.3341232227488153, "step": 985, "train/alpha": 0.2, "train/clean_ce_loss": 0.01165771484375, "train/clean_loss_scale": 44.29206848144531, "train/clean_loss_weight": 8.858413696289062, "train/combined_loss": 0.6196131706237793, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.2334123222748815, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5163443088531494 }, { "epoch": 2.3341232227488153, "step": 985, "train/alpha": 0.2, "train/clean_ce_loss": 0.0498046875, "train/clean_loss_scale": 9.40172004699707, "train/clean_loss_weight": 1.880344033241272, "train/combined_loss": 0.5618996620178223, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.2334123222748815, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46824973821640015 }, { "epoch": 2.3341232227488153, "step": 985, "train/alpha": 0.2, "train/clean_ce_loss": 0.14453125, "train/clean_loss_scale": 7.671265602111816, "train/clean_loss_weight": 1.5342531204223633, "train/combined_loss": 1.3304851055145264, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.2334123222748815, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.108737587928772 }, { "epoch": 2.3364928909952605, "step": 986, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 12.350791931152344, "train/clean_loss_weight": 2.470158338546753, "train/combined_loss": 0.9842037558555603, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23364928909952606, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8201698064804077 }, { "epoch": 2.3364928909952605, "step": 986, "train/alpha": 0.2, "train/clean_ce_loss": 0.01385498046875, "train/clean_loss_scale": 8.832270622253418, "train/clean_loss_weight": 1.7664541006088257, "train/combined_loss": 0.14684511721134186, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23364928909952606, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.12237093597650528 }, { "epoch": 2.3364928909952605, "step": 986, "train/alpha": 0.2, "train/clean_ce_loss": 0.00909423828125, "train/clean_loss_scale": 24.7651424407959, "train/clean_loss_weight": 4.953028678894043, "train/combined_loss": 0.2702641487121582, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23364928909952606, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.22522011399269104 }, { "epoch": 2.3364928909952605, "step": 986, "train/alpha": 0.2, "train/clean_ce_loss": 0.0020599365234375, "train/clean_loss_scale": 107.73236846923828, "train/clean_loss_weight": 21.54647445678711, "train/combined_loss": 0.26630622148513794, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23364928909952606, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.22192184627056122 }, { "epoch": 2.338862559241706, "step": 987, "train/alpha": 0.2, "train/clean_ce_loss": 0.027099609375, "train/clean_loss_scale": 11.55286979675293, "train/clean_loss_weight": 2.3105740547180176, "train/combined_loss": 0.3756939172744751, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23388625592417062, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.31307825446128845 }, { "epoch": 2.338862559241706, "step": 987, "train/alpha": 0.2, "train/clean_ce_loss": 0.0211181640625, "train/clean_loss_scale": 23.57704734802246, "train/clean_loss_weight": 4.715409755706787, "train/combined_loss": 0.5974847078323364, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23388625592417062, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4979039430618286 }, { "epoch": 2.338862559241706, "step": 987, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 18.727449417114258, "train/clean_loss_weight": 3.7454898357391357, "train/combined_loss": 1.4703974723815918, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23388625592417062, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.22533118724823 }, { "epoch": 2.338862559241706, "step": 987, "train/alpha": 0.2, "train/clean_ce_loss": 0.0322265625, "train/clean_loss_scale": 11.973161697387695, "train/clean_loss_weight": 2.394632339477539, "train/combined_loss": 0.4630246162414551, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23388625592417062, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3858538568019867 }, { "epoch": 2.3412322274881516, "step": 988, "train/alpha": 0.2, "train/clean_ce_loss": 0.007293701171875, "train/clean_loss_scale": 44.99884796142578, "train/clean_loss_weight": 8.999770164489746, "train/combined_loss": 0.39384979009628296, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23412322274881517, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.32820814847946167 }, { "epoch": 2.3412322274881516, "step": 988, "train/alpha": 0.2, "train/clean_ce_loss": 0.04345703125, "train/clean_loss_scale": 22.069122314453125, "train/clean_loss_weight": 4.413824558258057, "train/combined_loss": 1.1508702039718628, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23412322274881517, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9590585231781006 }, { "epoch": 2.3412322274881516, "step": 988, "train/alpha": 0.2, "train/clean_ce_loss": 0.0869140625, "train/clean_loss_scale": 9.31731128692627, "train/clean_loss_weight": 1.8634623289108276, "train/combined_loss": 0.971766471862793, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23412322274881517, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8098053932189941 }, { "epoch": 2.3412322274881516, "step": 988, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 10.437337875366211, "train/clean_loss_weight": 2.087467670440674, "train/combined_loss": 0.8194941282272339, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23412322274881517, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.68291175365448 }, { "epoch": 2.343601895734597, "step": 989, "train/alpha": 0.2, "train/clean_ce_loss": 0.002532958984375, "train/clean_loss_scale": 136.38621520996094, "train/clean_loss_weight": 27.27724266052246, "train/combined_loss": 0.41455280780792236, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23436018957345972, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3454606831073761 }, { "epoch": 2.343601895734597, "step": 989, "train/alpha": 0.2, "train/clean_ce_loss": 0.09521484375, "train/clean_loss_scale": 12.745124816894531, "train/clean_loss_weight": 2.549025058746338, "train/combined_loss": 1.4562300443649292, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23436018957345972, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2135250568389893 }, { "epoch": 2.343601895734597, "step": 989, "train/alpha": 0.2, "train/clean_ce_loss": 0.0673828125, "train/clean_loss_scale": 8.696011543273926, "train/clean_loss_weight": 1.7392023801803589, "train/combined_loss": 0.7031540274620056, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23436018957345972, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5859616994857788 }, { "epoch": 2.343601895734597, "step": 989, "train/alpha": 0.2, "train/clean_ce_loss": 0.0021209716796875, "train/clean_loss_scale": 130.92477416992188, "train/clean_loss_weight": 26.184955596923828, "train/combined_loss": 0.333225280046463, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23436018957345972, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2776877284049988 }, { "epoch": 2.345971563981043, "grad_norm": 35.73973083496094, "learning_rate": 2e-06, "loss": 3.2171, "step": 990 }, { "epoch": 2.345971563981043, "step": 990, "train/alpha": 0.2, "train/clean_ce_loss": 0.004364013671875, "train/clean_loss_scale": 48.86375045776367, "train/clean_loss_weight": 9.772749900817871, "train/combined_loss": 0.25589048862457275, "train/pos_loss/pos0": 0.00409855367615819, "train/pos_loss/pos1": 0.017142485827207565, "train/pos_loss/pos2": 0.22530141472816467, "train/pos_loss/pos3": 0.43981966376304626, "train/pos_loss/pos4": 0.8666809797286987, "train/pos_loss/pos5": 1.3896756172180176, "train/pos_loss/pos6": 1.913398265838623, "train/pos_loss/weight0": 2.471442222595215, "train/pos_loss/weight1": 1.8330955505371094, "train/pos_loss/weight2": 1.223659634590149, "train/pos_loss/weight3": 0.7480671405792236, "train/pos_loss/weight4": 0.4144466817378998, "train/pos_loss/weight5": 0.20065239071846008, "train/pos_loss/weight6": 0.10863662511110306, "train/pos_weight_progress": 0.23459715639810427, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2132420688867569 }, { "epoch": 2.345971563981043, "step": 990, "train/alpha": 0.2, "train/clean_ce_loss": 0.061767578125, "train/clean_loss_scale": 12.332233428955078, "train/clean_loss_weight": 2.4664466381073, "train/combined_loss": 0.9140786528587341, "train/pos_loss/pos0": 0.14629387855529785, "train/pos_loss/pos1": 0.3629973828792572, "train/pos_loss/pos2": 0.6960057020187378, "train/pos_loss/pos3": 1.0700706243515015, "train/pos_loss/pos4": 1.5005521774291992, "train/pos_loss/pos5": 1.9822514057159424, "train/pos_loss/pos6": 2.445058584213257, "train/pos_loss/weight0": 1.9254276752471924, "train/pos_loss/weight1": 1.5256319046020508, "train/pos_loss/weight2": 1.2126617431640625, "train/pos_loss/weight3": 0.8965080380439758, "train/pos_loss/weight4": 0.6563910245895386, "train/pos_loss/weight5": 0.44451987743377686, "train/pos_loss/weight6": 0.3388596773147583, "train/pos_weight_progress": 0.23459715639810427, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7617322206497192 }, { "epoch": 2.345971563981043, "step": 990, "train/alpha": 0.2, "train/clean_ce_loss": 0.054443359375, "train/clean_loss_scale": 6.038465976715088, "train/clean_loss_weight": 1.2076932191848755, "train/combined_loss": 0.39450523257255554, "train/pos_loss/pos0": 0.03772755712270737, "train/pos_loss/pos1": 0.11109200865030289, "train/pos_loss/pos2": 0.29211950302124023, "train/pos_loss/pos3": 0.7242364287376404, "train/pos_loss/pos4": 1.1096510887145996, "train/pos_loss/pos5": 1.7234861850738525, "train/pos_loss/pos6": 2.1451077461242676, "train/pos_loss/weight0": 2.4239606857299805, "train/pos_loss/weight1": 1.8009443283081055, "train/pos_loss/weight2": 1.2500605583190918, "train/pos_loss/weight3": 0.774097204208374, "train/pos_loss/weight4": 0.4202994406223297, "train/pos_loss/weight5": 0.2154543101787567, "train/pos_loss/weight6": 0.11518315225839615, "train/pos_weight_progress": 0.23459715639810427, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.32875436544418335 }, { "epoch": 2.345971563981043, "step": 990, "train/alpha": 0.2, "train/clean_ce_loss": 0.0130615234375, "train/clean_loss_scale": 57.395084381103516, "train/clean_loss_weight": 11.47901725769043, "train/combined_loss": 0.8996006846427917, "train/pos_loss/pos0": 0.0884527936577797, "train/pos_loss/pos1": 0.2577296793460846, "train/pos_loss/pos2": 0.784389853477478, "train/pos_loss/pos3": 1.2683132886886597, "train/pos_loss/pos4": 1.6971051692962646, "train/pos_loss/pos5": 2.3808133602142334, "train/pos_loss/pos6": 2.8389670848846436, "train/pos_loss/weight0": 2.073012351989746, "train/pos_loss/weight1": 1.6221181154251099, "train/pos_loss/weight2": 1.2214479446411133, "train/pos_loss/weight3": 0.8822270035743713, "train/pos_loss/weight4": 0.5884799957275391, "train/pos_loss/weight5": 0.36874040961265564, "train/pos_loss/weight6": 0.24397431313991547, "train/pos_weight_progress": 0.23459715639810427, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.749667227268219 }, { "epoch": 2.3483412322274884, "step": 991, "train/alpha": 0.2, "train/clean_ce_loss": 0.015380859375, "train/clean_loss_scale": 19.14552879333496, "train/clean_loss_weight": 3.829105854034424, "train/combined_loss": 0.3533696234226227, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23483412322274883, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.29447469115257263 }, { "epoch": 2.3483412322274884, "step": 991, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 9.385436058044434, "train/clean_loss_weight": 1.8770872354507446, "train/combined_loss": 1.044862985610962, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23483412322274883, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8707191348075867 }, { "epoch": 2.3483412322274884, "step": 991, "train/alpha": 0.2, "train/clean_ce_loss": 0.0546875, "train/clean_loss_scale": 8.91222095489502, "train/clean_loss_weight": 1.7824442386627197, "train/combined_loss": 0.5848644971847534, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23483412322274883, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4873870611190796 }, { "epoch": 2.3483412322274884, "step": 991, "train/alpha": 0.2, "train/clean_ce_loss": 0.123046875, "train/clean_loss_scale": 8.714852333068848, "train/clean_loss_weight": 1.7429704666137695, "train/combined_loss": 1.2868024110794067, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23483412322274883, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0723353624343872 }, { "epoch": 2.3507109004739335, "step": 992, "train/alpha": 0.2, "train/clean_ce_loss": 0.1357421875, "train/clean_loss_scale": 8.529450416564941, "train/clean_loss_weight": 1.7058900594711304, "train/combined_loss": 1.3893675804138184, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23507109004739338, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1578062772750854 }, { "epoch": 2.3507109004739335, "step": 992, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 8.714357376098633, "train/clean_loss_weight": 1.7428715229034424, "train/combined_loss": 0.8527135252952576, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23507109004739338, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7105945944786072 }, { "epoch": 2.3507109004739335, "step": 992, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 8.798824310302734, "train/clean_loss_weight": 1.7597649097442627, "train/combined_loss": 0.6856896281242371, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23507109004739338, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5714080333709717 }, { "epoch": 2.3507109004739335, "step": 992, "train/alpha": 0.2, "train/clean_ce_loss": 0.0026397705078125, "train/clean_loss_scale": 57.95730209350586, "train/clean_loss_weight": 11.591460227966309, "train/combined_loss": 0.1835927665233612, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23507109004739338, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.15299397706985474 }, { "epoch": 2.353080568720379, "step": 993, "train/alpha": 0.2, "train/clean_ce_loss": 0.0194091796875, "train/clean_loss_scale": 14.903814315795898, "train/clean_loss_weight": 2.9807629585266113, "train/combined_loss": 0.34712499380111694, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.2353080568720379, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.28927081823349 }, { "epoch": 2.353080568720379, "step": 993, "train/alpha": 0.2, "train/clean_ce_loss": 0.006134033203125, "train/clean_loss_scale": 55.11182403564453, "train/clean_loss_weight": 11.022364616394043, "train/combined_loss": 0.40566930174827576, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.2353080568720379, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3380577564239502 }, { "epoch": 2.353080568720379, "step": 993, "train/alpha": 0.2, "train/clean_ce_loss": 0.08203125, "train/clean_loss_scale": 8.713722229003906, "train/clean_loss_weight": 1.7427444458007812, "train/combined_loss": 0.8577569723129272, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.2353080568720379, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7147974967956543 }, { "epoch": 2.353080568720379, "step": 993, "train/alpha": 0.2, "train/clean_ce_loss": 0.1220703125, "train/clean_loss_scale": 11.012727737426758, "train/clean_loss_weight": 2.202545642852783, "train/combined_loss": 1.6131925582885742, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.2353080568720379, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3443270921707153 }, { "epoch": 2.3554502369668247, "step": 994, "train/alpha": 0.2, "train/clean_ce_loss": 0.11279296875, "train/clean_loss_scale": 10.097335815429688, "train/clean_loss_weight": 2.0194671154022217, "train/combined_loss": 1.3666901588439941, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23554502369668245, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1389085054397583 }, { "epoch": 2.3554502369668247, "step": 994, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 7.709282875061035, "train/clean_loss_weight": 1.5418566465377808, "train/combined_loss": 0.7724339962005615, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23554502369668245, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6436949968338013 }, { "epoch": 2.3554502369668247, "step": 994, "train/alpha": 0.2, "train/clean_ce_loss": 0.01226806640625, "train/clean_loss_scale": 18.03382110595703, "train/clean_loss_weight": 3.606764316558838, "train/combined_loss": 0.2654881179332733, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23554502369668245, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2212401032447815 }, { "epoch": 2.3554502369668247, "step": 994, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 11.43062686920166, "train/clean_loss_weight": 2.286125421524048, "train/combined_loss": 1.0850166082382202, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23554502369668245, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9041804671287537 }, { "epoch": 2.3578199052132702, "grad_norm": 14.28442096710205, "learning_rate": 2e-06, "loss": 3.2939, "step": 995 }, { "epoch": 2.3578199052132702, "step": 995, "train/alpha": 0.2, "train/clean_ce_loss": 0.10595703125, "train/clean_loss_scale": 10.459734916687012, "train/clean_loss_weight": 2.091947078704834, "train/combined_loss": 1.3299388885498047, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.235781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.108282446861267 }, { "epoch": 2.3578199052132702, "step": 995, "train/alpha": 0.2, "train/clean_ce_loss": 0.0966796875, "train/clean_loss_scale": 12.067461967468262, "train/clean_loss_weight": 2.413492441177368, "train/combined_loss": 1.4000141620635986, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.235781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1666784286499023 }, { "epoch": 2.3578199052132702, "step": 995, "train/alpha": 0.2, "train/clean_ce_loss": 0.003509521484375, "train/clean_loss_scale": 59.193092346191406, "train/clean_loss_weight": 11.838618278503418, "train/combined_loss": 0.24928730726242065, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.235781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2077394276857376 }, { "epoch": 2.3578199052132702, "step": 995, "train/alpha": 0.2, "train/clean_ce_loss": 0.0693359375, "train/clean_loss_scale": 11.1083345413208, "train/clean_loss_weight": 2.2216670513153076, "train/combined_loss": 0.9242481589317322, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.235781990521327, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7702068090438843 }, { "epoch": 2.360189573459716, "step": 996, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 14.10726261138916, "train/clean_loss_weight": 2.8214526176452637, "train/combined_loss": 1.3556197881698608, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23601895734597156, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1296831369400024 }, { "epoch": 2.360189573459716, "step": 996, "train/alpha": 0.2, "train/clean_ce_loss": 0.050537109375, "train/clean_loss_scale": 17.710474014282227, "train/clean_loss_weight": 3.5420949459075928, "train/combined_loss": 1.0740433931350708, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23601895734597156, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8950361609458923 }, { "epoch": 2.360189573459716, "step": 996, "train/alpha": 0.2, "train/clean_ce_loss": 0.0223388671875, "train/clean_loss_scale": 15.80756950378418, "train/clean_loss_weight": 3.1615140438079834, "train/combined_loss": 0.42374783754348755, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23601895734597156, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.35312318801879883 }, { "epoch": 2.360189573459716, "step": 996, "train/alpha": 0.2, "train/clean_ce_loss": 0.01116943359375, "train/clean_loss_scale": 52.716583251953125, "train/clean_loss_weight": 10.543316841125488, "train/combined_loss": 0.7065772414207458, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23601895734597156, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.588814377784729 }, { "epoch": 2.362559241706161, "step": 997, "train/alpha": 0.2, "train/clean_ce_loss": 0.023681640625, "train/clean_loss_scale": 8.381269454956055, "train/clean_loss_weight": 1.6762539148330688, "train/combined_loss": 0.23817864060401917, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.2362559241706161, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1984822005033493 }, { "epoch": 2.362559241706161, "step": 997, "train/alpha": 0.2, "train/clean_ce_loss": 0.08935546875, "train/clean_loss_scale": 11.565045356750488, "train/clean_loss_weight": 2.313009023666382, "train/combined_loss": 1.2400801181793213, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.2362559241706161, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.033400058746338 }, { "epoch": 2.362559241706161, "step": 997, "train/alpha": 0.2, "train/clean_ce_loss": 0.12890625, "train/clean_loss_scale": 6.460917949676514, "train/clean_loss_weight": 1.2921836376190186, "train/combined_loss": 0.9994232654571533, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.2362559241706161, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8328527212142944 }, { "epoch": 2.362559241706161, "step": 997, "train/alpha": 0.2, "train/clean_ce_loss": 0.15234375, "train/clean_loss_scale": 7.245150089263916, "train/clean_loss_weight": 1.4490300416946411, "train/combined_loss": 1.324504017829895, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.2362559241706161, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1037533283233643 }, { "epoch": 2.3649289099526065, "step": 998, "train/alpha": 0.2, "train/clean_ce_loss": 0.048095703125, "train/clean_loss_scale": 9.056587219238281, "train/clean_loss_weight": 1.8113174438476562, "train/combined_loss": 0.5226995348930359, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23649289099526066, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4355829358100891 }, { "epoch": 2.3649289099526065, "step": 998, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 9.46546459197998, "train/clean_loss_weight": 1.8930929899215698, "train/combined_loss": 0.7376406788825989, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23649289099526066, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6147005558013916 }, { "epoch": 2.3649289099526065, "step": 998, "train/alpha": 0.2, "train/clean_ce_loss": 0.0238037109375, "train/clean_loss_scale": 3.403485059738159, "train/clean_loss_weight": 0.6806970238685608, "train/combined_loss": 0.09721868485212326, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23649289099526066, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.08101557195186615 }, { "epoch": 2.3649289099526065, "step": 998, "train/alpha": 0.2, "train/clean_ce_loss": 0.0123291015625, "train/clean_loss_scale": 20.871463775634766, "train/clean_loss_weight": 4.174293041229248, "train/combined_loss": 0.3087916672229767, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23649289099526066, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2573263943195343 }, { "epoch": 2.367298578199052, "step": 999, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 7.824667453765869, "train/clean_loss_weight": 1.5649335384368896, "train/combined_loss": 0.6051890850067139, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23672985781990522, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5043242573738098 }, { "epoch": 2.367298578199052, "step": 999, "train/alpha": 0.2, "train/clean_ce_loss": 0.0013427734375, "train/clean_loss_scale": 213.17608642578125, "train/clean_loss_weight": 42.63521957397461, "train/combined_loss": 0.3434966206550598, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23672985781990522, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.286247193813324 }, { "epoch": 2.367298578199052, "step": 999, "train/alpha": 0.2, "train/clean_ce_loss": 0.037353515625, "train/clean_loss_scale": 11.806487083435059, "train/clean_loss_weight": 2.361297369003296, "train/combined_loss": 0.5292165279388428, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23672985781990522, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.44101378321647644 }, { "epoch": 2.367298578199052, "step": 999, "train/alpha": 0.2, "train/clean_ce_loss": 0.042236328125, "train/clean_loss_scale": 3.597137689590454, "train/clean_loss_weight": 0.7194275259971619, "train/combined_loss": 0.182315856218338, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23672985781990522, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.15192988514900208 }, { "epoch": 2.3696682464454977, "grad_norm": 23.398784637451172, "learning_rate": 2e-06, "loss": 3.0029, "step": 1000 }, { "epoch": 2.3696682464454977, "step": 1000, "train/alpha": 0.2, "train/clean_ce_loss": 0.10400390625, "train/clean_loss_scale": 10.279690742492676, "train/clean_loss_weight": 2.055938243865967, "train/combined_loss": 1.2829536199569702, "train/pos_loss/pos0": 0.5217992663383484, "train/pos_loss/pos1": 0.8887786865234375, "train/pos_loss/pos2": 1.3258159160614014, "train/pos_loss/pos3": 1.7596118450164795, "train/pos_loss/pos4": 2.168837785720825, "train/pos_loss/pos5": 2.4815335273742676, "train/pos_loss/pos6": 2.9300332069396973, "train/pos_loss/weight0": 2.4896507263183594, "train/pos_loss/weight1": 1.8165931701660156, "train/pos_loss/weight2": 1.2753533124923706, "train/pos_loss/weight3": 0.7671267986297607, "train/pos_loss/weight4": 0.39043092727661133, "train/pos_loss/weight5": 0.17779825627803802, "train/pos_loss/weight6": 0.08304646611213684, "train/pos_weight_progress": 0.23696682464454977, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0691280364990234 }, { "epoch": 2.3696682464454977, "step": 1000, "train/alpha": 0.2, "train/clean_ce_loss": 0.06787109375, "train/clean_loss_scale": 14.580940246582031, "train/clean_loss_weight": 2.9161880016326904, "train/combined_loss": 1.1875492334365845, "train/pos_loss/pos0": 0.21861103177070618, "train/pos_loss/pos1": 1.0892677307128906, "train/pos_loss/pos2": 1.6751238107681274, "train/pos_loss/pos3": 2.1246843338012695, "train/pos_loss/pos4": 2.717226505279541, "train/pos_loss/pos5": 3.4451112747192383, "train/pos_loss/pos6": 3.6255428791046143, "train/pos_loss/weight0": 2.9825620651245117, "train/pos_loss/weight1": 1.955326795578003, "train/pos_loss/weight2": 1.1460357904434204, "train/pos_loss/weight3": 0.5658745765686035, "train/pos_loss/weight4": 0.23428533971309662, "train/pos_loss/weight5": 0.0838857963681221, "train/pos_loss/weight6": 0.03202884644269943, "train/pos_weight_progress": 0.23696682464454977, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9896243810653687 }, { "epoch": 2.3696682464454977, "step": 1000, "train/alpha": 0.2, "train/clean_ce_loss": 0.0888671875, "train/clean_loss_scale": 11.037872314453125, "train/clean_loss_weight": 2.2075746059417725, "train/combined_loss": 1.1770856380462646, "train/pos_loss/pos0": 0.2935176193714142, "train/pos_loss/pos1": 0.9270815849304199, "train/pos_loss/pos2": 1.6860382556915283, "train/pos_loss/pos3": 2.4618003368377686, "train/pos_loss/pos4": 2.6127398014068604, "train/pos_loss/pos5": 3.2259888648986816, "train/pos_loss/pos6": 3.7778825759887695, "train/pos_loss/weight0": 2.9013009071350098, "train/pos_loss/weight1": 2.10385799407959, "train/pos_loss/weight2": 1.202074408531189, "train/pos_loss/weight3": 0.5437596440315247, "train/pos_loss/weight4": 0.18346887826919556, "train/pos_loss/weight5": 0.05194944515824318, "train/pos_loss/weight6": 0.01358866598457098, "train/pos_weight_progress": 0.23696682464454977, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9809046387672424 }, { "epoch": 2.3696682464454977, "step": 1000, "train/alpha": 0.2, "train/clean_ce_loss": 0.00811767578125, "train/clean_loss_scale": 14.218816757202148, "train/clean_loss_weight": 2.8437633514404297, "train/combined_loss": 0.13850849866867065, "train/pos_loss/pos0": 0.007774239871650934, "train/pos_loss/pos1": 0.04538211598992348, "train/pos_loss/pos2": 0.14763471484184265, "train/pos_loss/pos3": 0.5115926265716553, "train/pos_loss/pos4": 1.113169550895691, "train/pos_loss/pos5": 1.7160804271697998, "train/pos_loss/pos6": 2.040811777114868, "train/pos_loss/weight0": 3.0917961597442627, "train/pos_loss/weight1": 2.009167194366455, "train/pos_loss/weight2": 1.1922743320465088, "train/pos_loss/weight3": 0.5024449825286865, "train/pos_loss/weight4": 0.15507292747497559, "train/pos_loss/weight5": 0.03994528576731682, "train/pos_loss/weight6": 0.009298746474087238, "train/pos_weight_progress": 0.23696682464454977, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.11542374640703201 }, { "epoch": 2.3720379146919433, "step": 1001, "train/alpha": 0.2, "train/clean_ce_loss": 0.050537109375, "train/clean_loss_scale": 13.80256175994873, "train/clean_loss_weight": 2.760512351989746, "train/combined_loss": 0.8370499014854431, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23720379146919432, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6975415945053101 }, { "epoch": 2.3720379146919433, "step": 1001, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 10.518108367919922, "train/clean_loss_weight": 2.1036217212677, "train/combined_loss": 0.8319987654685974, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23720379146919432, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.693332314491272 }, { "epoch": 2.3720379146919433, "step": 1001, "train/alpha": 0.2, "train/clean_ce_loss": 0.06298828125, "train/clean_loss_scale": 17.868009567260742, "train/clean_loss_weight": 3.5736019611358643, "train/combined_loss": 1.3505702018737793, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23720379146919432, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1254751682281494 }, { "epoch": 2.3720379146919433, "step": 1001, "train/alpha": 0.2, "train/clean_ce_loss": 0.07177734375, "train/clean_loss_scale": 9.458711624145508, "train/clean_loss_weight": 1.8917423486709595, "train/combined_loss": 0.814705491065979, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23720379146919432, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6789212226867676 }, { "epoch": 2.374407582938389, "step": 1002, "train/alpha": 0.2, "train/clean_ce_loss": 0.01434326171875, "train/clean_loss_scale": 35.31156539916992, "train/clean_loss_weight": 7.062313079833984, "train/combined_loss": 0.6077796220779419, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23744075829383887, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5064830183982849 }, { "epoch": 2.374407582938389, "step": 1002, "train/alpha": 0.2, "train/clean_ce_loss": 0.0263671875, "train/clean_loss_scale": 37.2098388671875, "train/clean_loss_weight": 7.441967964172363, "train/combined_loss": 1.1773425340652466, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23744075829383887, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9811187982559204 }, { "epoch": 2.374407582938389, "step": 1002, "train/alpha": 0.2, "train/clean_ce_loss": 0.05419921875, "train/clean_loss_scale": 14.060669898986816, "train/clean_loss_weight": 2.812134027481079, "train/combined_loss": 0.9144927859306335, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23744075829383887, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7620773315429688 }, { "epoch": 2.374407582938389, "step": 1002, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 10.239102363586426, "train/clean_loss_weight": 2.047820568084717, "train/combined_loss": 0.9119200706481934, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23744075829383887, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.759933352470398 }, { "epoch": 2.376777251184834, "step": 1003, "train/alpha": 0.2, "train/clean_ce_loss": 0.06201171875, "train/clean_loss_scale": 9.09591007232666, "train/clean_loss_weight": 1.81918203830719, "train/combined_loss": 0.6768636107444763, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23767772511848342, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5640529990196228 }, { "epoch": 2.376777251184834, "step": 1003, "train/alpha": 0.2, "train/clean_ce_loss": 0.060791015625, "train/clean_loss_scale": 13.310354232788086, "train/clean_loss_weight": 2.6620709896087646, "train/combined_loss": 0.9709799885749817, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23767772511848342, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.809149980545044 }, { "epoch": 2.376777251184834, "step": 1003, "train/alpha": 0.2, "train/clean_ce_loss": 0.0625, "train/clean_loss_scale": 10.65575122833252, "train/clean_loss_weight": 2.131150245666504, "train/combined_loss": 0.799181342124939, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23767772511848342, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6659844517707825 }, { "epoch": 2.376777251184834, "step": 1003, "train/alpha": 0.2, "train/clean_ce_loss": 0.0242919921875, "train/clean_loss_scale": 21.425092697143555, "train/clean_loss_weight": 4.285018444061279, "train/combined_loss": 0.6245498061180115, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23767772511848342, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5204581618309021 }, { "epoch": 2.3791469194312795, "step": 1004, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 10.911218643188477, "train/clean_loss_weight": 2.182243824005127, "train/combined_loss": 1.0421066284179688, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23791469194312795, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8684222102165222 }, { "epoch": 2.3791469194312795, "step": 1004, "train/alpha": 0.2, "train/clean_ce_loss": 0.07470703125, "train/clean_loss_scale": 11.062664985656738, "train/clean_loss_weight": 2.2125329971313477, "train/combined_loss": 0.9917506575584412, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23791469194312795, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8264588713645935 }, { "epoch": 2.3791469194312795, "step": 1004, "train/alpha": 0.2, "train/clean_ce_loss": 0.07470703125, "train/clean_loss_scale": 10.823620796203613, "train/clean_loss_weight": 2.164724111557007, "train/combined_loss": 0.9703207015991211, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23791469194312795, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8086006045341492 }, { "epoch": 2.3791469194312795, "step": 1004, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 12.311491966247559, "train/clean_loss_weight": 2.4622983932495117, "train/combined_loss": 1.2335537672042847, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23791469194312795, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0279614925384521 }, { "epoch": 2.381516587677725, "grad_norm": 26.529050827026367, "learning_rate": 2e-06, "loss": 3.4524, "step": 1005 }, { "epoch": 2.381516587677725, "step": 1005, "train/alpha": 0.2, "train/clean_ce_loss": 0.08837890625, "train/clean_loss_scale": 9.139580726623535, "train/clean_loss_weight": 1.827916145324707, "train/combined_loss": 0.96929532289505, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.2381516587677725, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8077461123466492 }, { "epoch": 2.381516587677725, "step": 1005, "train/alpha": 0.2, "train/clean_ce_loss": 0.14453125, "train/clean_loss_scale": 9.775699615478516, "train/clean_loss_weight": 1.9551399946212769, "train/combined_loss": 1.6954729557037354, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.2381516587677725, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4128941297531128 }, { "epoch": 2.381516587677725, "step": 1005, "train/alpha": 0.2, "train/clean_ce_loss": 0.03125, "train/clean_loss_scale": 15.074597358703613, "train/clean_loss_weight": 3.0149195194244385, "train/combined_loss": 0.5652974247932434, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.2381516587677725, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4710811674594879 }, { "epoch": 2.381516587677725, "step": 1005, "train/alpha": 0.2, "train/clean_ce_loss": 0.06298828125, "train/clean_loss_scale": 17.8350772857666, "train/clean_loss_weight": 3.5670154094696045, "train/combined_loss": 1.3480809926986694, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.2381516587677725, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1234008073806763 }, { "epoch": 2.3838862559241707, "step": 1006, "train/alpha": 0.2, "train/clean_ce_loss": 0.06201171875, "train/clean_loss_scale": 25.988187789916992, "train/clean_loss_weight": 5.197637557983398, "train/combined_loss": 1.9338865280151367, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23838862559241705, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.6115721464157104 }, { "epoch": 2.3838862559241707, "step": 1006, "train/alpha": 0.2, "train/clean_ce_loss": 0.052001953125, "train/clean_loss_scale": 13.570924758911133, "train/clean_loss_weight": 2.7141849994659424, "train/combined_loss": 0.846857488155365, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23838862559241705, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7057145833969116 }, { "epoch": 2.3838862559241707, "step": 1006, "train/alpha": 0.2, "train/clean_ce_loss": 0.022705078125, "train/clean_loss_scale": 8.446346282958984, "train/clean_loss_weight": 1.6892693042755127, "train/combined_loss": 0.23012994229793549, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23838862559241705, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.19177494943141937 }, { "epoch": 2.3838862559241707, "step": 1006, "train/alpha": 0.2, "train/clean_ce_loss": 0.12109375, "train/clean_loss_scale": 11.484090805053711, "train/clean_loss_weight": 2.296818256378174, "train/combined_loss": 1.6687818765640259, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23838862559241705, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3906515836715698 }, { "epoch": 2.3862559241706163, "step": 1007, "train/alpha": 0.2, "train/clean_ce_loss": 0.053466796875, "train/clean_loss_scale": 16.78303337097168, "train/clean_loss_weight": 3.3566067218780518, "train/combined_loss": 1.0768020153045654, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.2386255924170616, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8973350524902344 }, { "epoch": 2.3862559241706163, "step": 1007, "train/alpha": 0.2, "train/clean_ce_loss": 0.0625, "train/clean_loss_scale": 1.9797989130020142, "train/clean_loss_weight": 0.3959597945213318, "train/combined_loss": 0.14848491549491882, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.2386255924170616, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.12373743206262589 }, { "epoch": 2.3862559241706163, "step": 1007, "train/alpha": 0.2, "train/clean_ce_loss": 0.05126953125, "train/clean_loss_scale": 12.463571548461914, "train/clean_loss_weight": 2.4927144050598145, "train/combined_loss": 0.7668017745018005, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.2386255924170616, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6390014886856079 }, { "epoch": 2.3862559241706163, "step": 1007, "train/alpha": 0.2, "train/clean_ce_loss": 0.03466796875, "train/clean_loss_scale": 10.677408218383789, "train/clean_loss_weight": 2.135481595993042, "train/combined_loss": 0.444196879863739, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.2386255924170616, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.37016406655311584 }, { "epoch": 2.3886255924170614, "step": 1008, "train/alpha": 0.2, "train/clean_ce_loss": 0.08251953125, "train/clean_loss_scale": 12.979331970214844, "train/clean_loss_weight": 2.5958664417266846, "train/combined_loss": 1.2852580547332764, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23886255924170616, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.071048378944397 }, { "epoch": 2.3886255924170614, "step": 1008, "train/alpha": 0.2, "train/clean_ce_loss": 0.00958251953125, "train/clean_loss_scale": 53.480262756347656, "train/clean_loss_weight": 10.696052551269531, "train/combined_loss": 0.6149708032608032, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23886255924170616, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5124756693840027 }, { "epoch": 2.3886255924170614, "step": 1008, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 8.86345386505127, "train/clean_loss_weight": 1.772690773010254, "train/combined_loss": 0.7894013524055481, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23886255924170616, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6578344702720642 }, { "epoch": 2.3886255924170614, "step": 1008, "train/alpha": 0.2, "train/clean_ce_loss": 0.06396484375, "train/clean_loss_scale": 4.586851596832275, "train/clean_loss_weight": 0.9173703193664551, "train/combined_loss": 0.3520767092704773, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23886255924170616, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2933972477912903 }, { "epoch": 2.390995260663507, "step": 1009, "train/alpha": 0.2, "train/clean_ce_loss": 0.0230712890625, "train/clean_loss_scale": 25.888105392456055, "train/clean_loss_weight": 5.177621364593506, "train/combined_loss": 0.7167263627052307, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.2390995260663507, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5972719788551331 }, { "epoch": 2.390995260663507, "step": 1009, "train/alpha": 0.2, "train/clean_ce_loss": 0.04638671875, "train/clean_loss_scale": 13.295406341552734, "train/clean_loss_weight": 2.659081220626831, "train/combined_loss": 0.7400763034820557, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.2390995260663507, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.616730272769928 }, { "epoch": 2.390995260663507, "step": 1009, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 5.585121154785156, "train/clean_loss_weight": 1.117024302482605, "train/combined_loss": 0.47778964042663574, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.2390995260663507, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3981580436229706 }, { "epoch": 2.390995260663507, "step": 1009, "train/alpha": 0.2, "train/clean_ce_loss": 0.0016937255859375, "train/clean_loss_scale": 335.100830078125, "train/clean_loss_weight": 67.0201644897461, "train/combined_loss": 0.6810826063156128, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.2390995260663507, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.567568838596344 }, { "epoch": 2.3933649289099526, "grad_norm": 33.71124267578125, "learning_rate": 2e-06, "loss": 3.4039, "step": 1010 }, { "epoch": 2.3933649289099526, "step": 1010, "train/alpha": 0.2, "train/clean_ce_loss": 0.0289306640625, "train/clean_loss_scale": 23.276504516601562, "train/clean_loss_weight": 4.655301094055176, "train/combined_loss": 0.8080856800079346, "train/pos_loss/pos0": 0.05614657700061798, "train/pos_loss/pos1": 0.17739933729171753, "train/pos_loss/pos2": 0.58961421251297, "train/pos_loss/pos3": 1.1195993423461914, "train/pos_loss/pos4": 1.6937004327774048, "train/pos_loss/pos5": 2.0763564109802246, "train/pos_loss/pos6": 2.623948335647583, "train/pos_loss/weight0": 2.0182971954345703, "train/pos_loss/weight1": 1.5922513008117676, "train/pos_loss/weight2": 1.1969279050827026, "train/pos_loss/weight3": 0.8916484117507935, "train/pos_loss/weight4": 0.6203201413154602, "train/pos_loss/weight5": 0.4070182144641876, "train/pos_loss/weight6": 0.2735365033149719, "train/pos_weight_progress": 0.23933649289099526, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6734047532081604 }, { "epoch": 2.3933649289099526, "step": 1010, "train/alpha": 0.2, "train/clean_ce_loss": 0.0673828125, "train/clean_loss_scale": 9.406973838806152, "train/clean_loss_weight": 1.8813947439193726, "train/combined_loss": 0.7606419920921326, "train/pos_loss/pos0": 0.23788891732692719, "train/pos_loss/pos1": 0.4027840495109558, "train/pos_loss/pos2": 0.6921169757843018, "train/pos_loss/pos3": 1.072649598121643, "train/pos_loss/pos4": 1.4544997215270996, "train/pos_loss/pos5": 1.88350510597229, "train/pos_loss/pos6": 2.1952929496765137, "train/pos_loss/weight0": 2.320369005203247, "train/pos_loss/weight1": 1.7474621534347534, "train/pos_loss/weight2": 1.2711467742919922, "train/pos_loss/weight3": 0.8256055116653442, "train/pos_loss/weight4": 0.46374866366386414, "train/pos_loss/weight5": 0.23749090731143951, "train/pos_loss/weight6": 0.13417640328407288, "train/pos_weight_progress": 0.23933649289099526, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6338683366775513 }, { "epoch": 2.3933649289099526, "step": 1010, "train/alpha": 0.2, "train/clean_ce_loss": 0.0045166015625, "train/clean_loss_scale": 57.108558654785156, "train/clean_loss_weight": 11.421711921691895, "train/combined_loss": 0.30952391028404236, "train/pos_loss/pos0": 0.02151372656226158, "train/pos_loss/pos1": 0.0817757174372673, "train/pos_loss/pos2": 0.1983443945646286, "train/pos_loss/pos3": 0.530277669429779, "train/pos_loss/pos4": 0.9429436326026917, "train/pos_loss/pos5": 1.3464877605438232, "train/pos_loss/pos6": 1.8211174011230469, "train/pos_loss/weight0": 2.4135501384735107, "train/pos_loss/weight1": 1.783821702003479, "train/pos_loss/weight2": 1.2314118146896362, "train/pos_loss/weight3": 0.777031660079956, "train/pos_loss/weight4": 0.4386870861053467, "train/pos_loss/weight5": 0.22917447984218597, "train/pos_loss/weight6": 0.1263229250907898, "train/pos_weight_progress": 0.23933649289099526, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.25793659687042236 }, { "epoch": 2.3933649289099526, "step": 1010, "train/alpha": 0.2, "train/clean_ce_loss": 0.0859375, "train/clean_loss_scale": 12.857144355773926, "train/clean_loss_weight": 2.5714290142059326, "train/combined_loss": 1.3258930444717407, "train/pos_loss/pos0": 0.3164236843585968, "train/pos_loss/pos1": 0.6930869221687317, "train/pos_loss/pos2": 1.139347791671753, "train/pos_loss/pos3": 1.5567235946655273, "train/pos_loss/pos4": 1.9216855764389038, "train/pos_loss/pos5": 2.2962799072265625, "train/pos_loss/pos6": 2.78310227394104, "train/pos_loss/weight0": 1.8963146209716797, "train/pos_loss/weight1": 1.528760552406311, "train/pos_loss/weight2": 1.1856474876403809, "train/pos_loss/weight3": 0.9148461222648621, "train/pos_loss/weight4": 0.6668109893798828, "train/pos_loss/weight5": 0.46763306856155396, "train/pos_loss/weight6": 0.3399869501590729, "train/pos_weight_progress": 0.23933649289099526, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1049108505249023 }, { "epoch": 2.395734597156398, "step": 1011, "train/alpha": 0.2, "train/clean_ce_loss": 0.07568359375, "train/clean_loss_scale": 15.568666458129883, "train/clean_loss_weight": 3.1137332916259766, "train/combined_loss": 1.4139511585235596, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.2395734597156398, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1782926321029663 }, { "epoch": 2.395734597156398, "step": 1011, "train/alpha": 0.2, "train/clean_ce_loss": 0.0751953125, "train/clean_loss_scale": 11.980908393859863, "train/clean_loss_weight": 2.39618182182312, "train/combined_loss": 1.0810898542404175, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.2395734597156398, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9009081721305847 }, { "epoch": 2.395734597156398, "step": 1011, "train/alpha": 0.2, "train/clean_ce_loss": 0.019775390625, "train/clean_loss_scale": 14.046989440917969, "train/clean_loss_weight": 2.8093979358673096, "train/combined_loss": 0.33334165811538696, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.2395734597156398, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.27778470516204834 }, { "epoch": 2.395734597156398, "step": 1011, "train/alpha": 0.2, "train/clean_ce_loss": 0.060302734375, "train/clean_loss_scale": 5.981203556060791, "train/clean_loss_weight": 1.196240782737732, "train/combined_loss": 0.43281951546669006, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.2395734597156398, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3606829345226288 }, { "epoch": 2.3981042654028437, "step": 1012, "train/alpha": 0.2, "train/clean_ce_loss": 0.049072265625, "train/clean_loss_scale": 10.635913848876953, "train/clean_loss_weight": 2.127182722091675, "train/combined_loss": 0.6263140439987183, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.23981042654028437, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5219283699989319 }, { "epoch": 2.3981042654028437, "step": 1012, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 12.684493064880371, "train/clean_loss_weight": 2.536898612976074, "train/combined_loss": 0.9810662269592285, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.23981042654028437, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8175551891326904 }, { "epoch": 2.3981042654028437, "step": 1012, "train/alpha": 0.2, "train/clean_ce_loss": 0.0986328125, "train/clean_loss_scale": 7.586462020874023, "train/clean_loss_weight": 1.5172923803329468, "train/combined_loss": 0.8979288935661316, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.23981042654028437, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7482740879058838 }, { "epoch": 2.3981042654028437, "step": 1012, "train/alpha": 0.2, "train/clean_ce_loss": 0.03955078125, "train/clean_loss_scale": 8.634674072265625, "train/clean_loss_weight": 1.726934790611267, "train/combined_loss": 0.4098097085952759, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.23981042654028437, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.34150809049606323 }, { "epoch": 2.4004739336492893, "step": 1013, "train/alpha": 0.2, "train/clean_ce_loss": 0.05078125, "train/clean_loss_scale": 17.68960952758789, "train/clean_loss_weight": 3.537921905517578, "train/combined_loss": 1.0779606103897095, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.24004739336492892, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8983005285263062 }, { "epoch": 2.4004739336492893, "step": 1013, "train/alpha": 0.2, "train/clean_ce_loss": 0.00982666015625, "train/clean_loss_scale": 31.1256160736084, "train/clean_loss_weight": 6.225123405456543, "train/combined_loss": 0.3670330047607422, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.24004739336492892, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.30586084723472595 }, { "epoch": 2.4004739336492893, "step": 1013, "train/alpha": 0.2, "train/clean_ce_loss": 0.076171875, "train/clean_loss_scale": 7.243922233581543, "train/clean_loss_weight": 1.4487844705581665, "train/combined_loss": 0.6621397733688354, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.24004739336492892, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5517831444740295 }, { "epoch": 2.4004739336492893, "step": 1013, "train/alpha": 0.2, "train/clean_ce_loss": 0.01519775390625, "train/clean_loss_scale": 15.89030647277832, "train/clean_loss_weight": 3.1780612468719482, "train/combined_loss": 0.2897963523864746, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.24004739336492892, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2414969652891159 }, { "epoch": 2.4028436018957344, "step": 1014, "train/alpha": 0.2, "train/clean_ce_loss": 0.0003147125244140625, "train/clean_loss_scale": 358.5364074707031, "train/clean_loss_weight": 71.70728302001953, "train/combined_loss": 0.1354030817747116, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.24028436018957347, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.11283589899539948 }, { "epoch": 2.4028436018957344, "step": 1014, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 7.250145435333252, "train/clean_loss_weight": 1.4500291347503662, "train/combined_loss": 0.5650016069412231, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.24028436018957347, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4708346426486969 }, { "epoch": 2.4028436018957344, "step": 1014, "train/alpha": 0.2, "train/clean_ce_loss": 0.0245361328125, "train/clean_loss_scale": 13.538227081298828, "train/clean_loss_weight": 2.7076454162597656, "train/combined_loss": 0.3986108899116516, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.24028436018957347, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.33217573165893555 }, { "epoch": 2.4028436018957344, "step": 1014, "train/alpha": 0.2, "train/clean_ce_loss": 0.00188446044921875, "train/clean_loss_scale": 139.1022491455078, "train/clean_loss_weight": 27.820449829101562, "train/combined_loss": 0.3145592212677002, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.24028436018957347, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2621326744556427 }, { "epoch": 2.40521327014218, "grad_norm": 19.709739685058594, "learning_rate": 2e-06, "loss": 3.3118, "step": 1015 }, { "epoch": 2.40521327014218, "step": 1015, "train/alpha": 0.2, "train/clean_ce_loss": 0.0010833740234375, "train/clean_loss_scale": 109.30506134033203, "train/clean_loss_weight": 21.861013412475586, "train/combined_loss": 0.14210191369056702, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.24052132701421802, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.11841826140880585 }, { "epoch": 2.40521327014218, "step": 1015, "train/alpha": 0.2, "train/clean_ce_loss": 0.083984375, "train/clean_loss_scale": 8.19715690612793, "train/clean_loss_weight": 1.639431357383728, "train/combined_loss": 0.8261197209358215, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.24052132701421802, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6884331107139587 }, { "epoch": 2.40521327014218, "step": 1015, "train/alpha": 0.2, "train/clean_ce_loss": 0.07373046875, "train/clean_loss_scale": 8.643851280212402, "train/clean_loss_weight": 1.7287702560424805, "train/combined_loss": 0.7647782564163208, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.24052132701421802, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6373152136802673 }, { "epoch": 2.40521327014218, "step": 1015, "train/alpha": 0.2, "train/clean_ce_loss": 0.1044921875, "train/clean_loss_scale": 10.040555953979492, "train/clean_loss_weight": 2.0081112384796143, "train/combined_loss": 1.2589915990829468, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.24052132701421802, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0491596460342407 }, { "epoch": 2.4075829383886256, "step": 1016, "train/alpha": 0.2, "train/clean_ce_loss": 0.01312255859375, "train/clean_loss_scale": 12.88287353515625, "train/clean_loss_weight": 2.5765748023986816, "train/combined_loss": 0.2028675228357315, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.24075829383886255, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1690562665462494 }, { "epoch": 2.4075829383886256, "step": 1016, "train/alpha": 0.2, "train/clean_ce_loss": 0.07666015625, "train/clean_loss_scale": 9.793829917907715, "train/clean_loss_weight": 1.958765983581543, "train/combined_loss": 0.9009557962417603, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.24075829383886255, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7507964968681335 }, { "epoch": 2.4075829383886256, "step": 1016, "train/alpha": 0.2, "train/clean_ce_loss": 0.1474609375, "train/clean_loss_scale": 8.712727546691895, "train/clean_loss_weight": 1.742545485496521, "train/combined_loss": 1.541744351387024, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.24075829383886255, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2847869396209717 }, { "epoch": 2.4075829383886256, "step": 1016, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 6.519155025482178, "train/clean_loss_weight": 1.3038309812545776, "train/combined_loss": 0.637909471988678, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.24075829383886255, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5315912365913391 }, { "epoch": 2.409952606635071, "step": 1017, "train/alpha": 0.2, "train/clean_ce_loss": 0.000804901123046875, "train/clean_loss_scale": 218.14979553222656, "train/clean_loss_weight": 43.62995910644531, "train/combined_loss": 0.21070681512355804, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.2409952606635071, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.17558901011943817 }, { "epoch": 2.409952606635071, "step": 1017, "train/alpha": 0.2, "train/clean_ce_loss": 0.0159912109375, "train/clean_loss_scale": 19.134586334228516, "train/clean_loss_weight": 3.8269174098968506, "train/combined_loss": 0.36718225479125977, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.2409952606635071, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3059852123260498 }, { "epoch": 2.409952606635071, "step": 1017, "train/alpha": 0.2, "train/clean_ce_loss": 0.1103515625, "train/clean_loss_scale": 9.534196853637695, "train/clean_loss_weight": 1.906839370727539, "train/combined_loss": 1.2625362873077393, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.2409952606635071, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0521135330200195 }, { "epoch": 2.409952606635071, "step": 1017, "train/alpha": 0.2, "train/clean_ce_loss": 0.04443359375, "train/clean_loss_scale": 16.47040367126465, "train/clean_loss_weight": 3.2940807342529297, "train/combined_loss": 0.8782070875167847, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.2409952606635071, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7318392395973206 }, { "epoch": 2.4123222748815167, "step": 1018, "train/alpha": 0.2, "train/clean_ce_loss": 0.04296875, "train/clean_loss_scale": 10.995361328125, "train/clean_loss_weight": 2.1990723609924316, "train/combined_loss": 0.5669482946395874, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.24123222748815165, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4724569320678711 }, { "epoch": 2.4123222748815167, "step": 1018, "train/alpha": 0.2, "train/clean_ce_loss": 0.0576171875, "train/clean_loss_scale": 10.775739669799805, "train/clean_loss_weight": 2.1551480293273926, "train/combined_loss": 0.7450413703918457, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.24123222748815165, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6208677887916565 }, { "epoch": 2.4123222748815167, "step": 1018, "train/alpha": 0.2, "train/clean_ce_loss": 0.126953125, "train/clean_loss_scale": 6.02865743637085, "train/clean_loss_weight": 1.2057315111160278, "train/combined_loss": 0.9184283018112183, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.24123222748815165, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7653568983078003 }, { "epoch": 2.4123222748815167, "step": 1018, "train/alpha": 0.2, "train/clean_ce_loss": 0.15625, "train/clean_loss_scale": 7.082029819488525, "train/clean_loss_weight": 1.4164060354232788, "train/combined_loss": 1.327880620956421, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.24123222748815165, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.106567144393921 }, { "epoch": 2.414691943127962, "step": 1019, "train/alpha": 0.2, "train/clean_ce_loss": 0.057373046875, "train/clean_loss_scale": 8.864282608032227, "train/clean_loss_weight": 1.772856593132019, "train/combined_loss": 0.6102851033210754, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.2414691943127962, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5085709095001221 }, { "epoch": 2.414691943127962, "step": 1019, "train/alpha": 0.2, "train/clean_ce_loss": 0.0830078125, "train/clean_loss_scale": 8.753600120544434, "train/clean_loss_weight": 1.7507200241088867, "train/combined_loss": 0.8719406723976135, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.2414691943127962, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7266172170639038 }, { "epoch": 2.414691943127962, "step": 1019, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 9.993582725524902, "train/clean_loss_weight": 1.9987165927886963, "train/combined_loss": 0.9486096501350403, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.2414691943127962, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7905080318450928 }, { "epoch": 2.414691943127962, "step": 1019, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 10.656606674194336, "train/clean_loss_weight": 2.131321430206299, "train/combined_loss": 1.1176948547363281, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.2414691943127962, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.931412398815155 }, { "epoch": 2.4170616113744074, "grad_norm": 30.482990264892578, "learning_rate": 2e-06, "loss": 3.595, "step": 1020 }, { "epoch": 2.4170616113744074, "step": 1020, "train/alpha": 0.2, "train/clean_ce_loss": 0.119140625, "train/clean_loss_scale": 7.543093681335449, "train/clean_loss_weight": 1.508618712425232, "train/combined_loss": 1.078426718711853, "train/pos_loss/pos0": 0.34247025847435, "train/pos_loss/pos1": 0.6886533498764038, "train/pos_loss/pos2": 1.230053186416626, "train/pos_loss/pos3": 1.8103309869766235, "train/pos_loss/pos4": 2.3820996284484863, "train/pos_loss/pos5": 2.955108165740967, "train/pos_loss/pos6": 3.094762086868286, "train/pos_loss/weight0": 2.6966326236724854, "train/pos_loss/weight1": 1.903664469718933, "train/pos_loss/weight2": 1.2108620405197144, "train/pos_loss/weight3": 0.668318510055542, "train/pos_loss/weight4": 0.3186228275299072, "train/pos_loss/weight5": 0.13831305503845215, "train/pos_loss/weight6": 0.06358632445335388, "train/pos_weight_progress": 0.24170616113744076, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8986889123916626 }, { "epoch": 2.4170616113744074, "step": 1020, "train/alpha": 0.2, "train/clean_ce_loss": 0.10986328125, "train/clean_loss_scale": 6.460304260253906, "train/clean_loss_weight": 1.2920608520507812, "train/combined_loss": 0.8517003059387207, "train/pos_loss/pos0": 0.3419955372810364, "train/pos_loss/pos1": 0.6611588001251221, "train/pos_loss/pos2": 1.0241385698318481, "train/pos_loss/pos3": 1.3453905582427979, "train/pos_loss/pos4": 1.7004505395889282, "train/pos_loss/pos5": 2.147054672241211, "train/pos_loss/pos6": 2.3561275005340576, "train/pos_loss/weight0": 2.8871028423309326, "train/pos_loss/weight1": 1.9730610847473145, "train/pos_loss/weight2": 1.1803163290023804, "train/pos_loss/weight3": 0.6087706685066223, "train/pos_loss/weight4": 0.24197684228420258, "train/pos_loss/weight5": 0.07995233684778214, "train/pos_loss/weight6": 0.02881988324224949, "train/pos_weight_progress": 0.24170616113744076, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.709750235080719 }, { "epoch": 2.4170616113744074, "step": 1020, "train/alpha": 0.2, "train/clean_ce_loss": 0.0849609375, "train/clean_loss_scale": 7.247847557067871, "train/clean_loss_weight": 1.449569582939148, "train/combined_loss": 0.7389407157897949, "train/pos_loss/pos0": 0.2802006006240845, "train/pos_loss/pos1": 0.47220659255981445, "train/pos_loss/pos2": 0.7956511974334717, "train/pos_loss/pos3": 1.0745155811309814, "train/pos_loss/pos4": 1.5198581218719482, "train/pos_loss/pos5": 1.9110848903656006, "train/pos_loss/pos6": 2.25308895111084, "train/pos_loss/weight0": 2.6510767936706543, "train/pos_loss/weight1": 1.883094072341919, "train/pos_loss/weight2": 1.2095918655395508, "train/pos_loss/weight3": 0.6884634494781494, "train/pos_loss/weight4": 0.3386908769607544, "train/pos_loss/weight5": 0.15660503506660461, "train/pos_loss/weight6": 0.07247816026210785, "train/pos_weight_progress": 0.24170616113744076, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6157839298248291 }, { "epoch": 2.4170616113744074, "step": 1020, "train/alpha": 0.2, "train/clean_ce_loss": 0.13671875, "train/clean_loss_scale": 7.782604217529297, "train/clean_loss_weight": 1.5565208196640015, "train/combined_loss": 1.2768335342407227, "train/pos_loss/pos0": 0.45656928420066833, "train/pos_loss/pos1": 0.8162228465080261, "train/pos_loss/pos2": 1.279358148574829, "train/pos_loss/pos3": 1.8493261337280273, "train/pos_loss/pos4": 2.2641334533691406, "train/pos_loss/pos5": 2.6960370540618896, "train/pos_loss/pos6": 3.1946120262145996, "train/pos_loss/weight0": 2.4684977531433105, "train/pos_loss/weight1": 1.8034076690673828, "train/pos_loss/weight2": 1.2280582189559937, "train/pos_loss/weight3": 0.7564550042152405, "train/pos_loss/weight4": 0.4159855246543884, "train/pos_loss/weight5": 0.21292951703071594, "train/pos_loss/weight6": 0.11466608196496964, "train/pos_weight_progress": 0.24170616113744076, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0640279054641724 }, { "epoch": 2.419431279620853, "step": 1021, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 8.506607055664062, "train/clean_loss_weight": 1.7013214826583862, "train/combined_loss": 0.8174318075180054, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.2419431279620853, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6811931729316711 }, { "epoch": 2.419431279620853, "step": 1021, "train/alpha": 0.2, "train/clean_ce_loss": 0.1455078125, "train/clean_loss_scale": 7.5422468185424805, "train/clean_loss_weight": 1.5084494352340698, "train/combined_loss": 1.3169469833374023, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.2419431279620853, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0974558591842651 }, { "epoch": 2.419431279620853, "step": 1021, "train/alpha": 0.2, "train/clean_ce_loss": 0.130859375, "train/clean_loss_scale": 8.48420524597168, "train/clean_loss_weight": 1.6968411207199097, "train/combined_loss": 1.3322854042053223, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.2419431279620853, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1102378368377686 }, { "epoch": 2.419431279620853, "step": 1021, "train/alpha": 0.2, "train/clean_ce_loss": 0.054443359375, "train/clean_loss_scale": 17.277807235717773, "train/clean_loss_weight": 3.455561399459839, "train/combined_loss": 1.1287941932678223, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.2419431279620853, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9406618475914001 }, { "epoch": 2.4218009478672986, "step": 1022, "train/alpha": 0.2, "train/clean_ce_loss": 0.0279541015625, "train/clean_loss_scale": 4.73853874206543, "train/clean_loss_weight": 0.9477077722549438, "train/combined_loss": 0.15895390510559082, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.24218009478672986, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.13246159255504608 }, { "epoch": 2.4218009478672986, "step": 1022, "train/alpha": 0.2, "train/clean_ce_loss": 0.0164794921875, "train/clean_loss_scale": 2.108081817626953, "train/clean_loss_weight": 0.4216163754463196, "train/combined_loss": 0.041688140481710434, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.24218009478672986, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.03474011644721031 }, { "epoch": 2.4218009478672986, "step": 1022, "train/alpha": 0.2, "train/clean_ce_loss": 0.11083984375, "train/clean_loss_scale": 8.325322151184082, "train/clean_loss_weight": 1.6650644540786743, "train/combined_loss": 1.1073329448699951, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.24218009478672986, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9227774143218994 }, { "epoch": 2.4218009478672986, "step": 1022, "train/alpha": 0.2, "train/clean_ce_loss": 0.04296875, "train/clean_loss_scale": 1.3994513750076294, "train/clean_loss_weight": 0.2798902690410614, "train/combined_loss": 0.07215920835733414, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.24218009478672986, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.060132674872875214 }, { "epoch": 2.424170616113744, "step": 1023, "train/alpha": 0.2, "train/clean_ce_loss": 0.0269775390625, "train/clean_loss_scale": 11.10136890411377, "train/clean_loss_weight": 2.220273733139038, "train/combined_loss": 0.3593851327896118, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.2424170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2994876205921173 }, { "epoch": 2.424170616113744, "step": 1023, "train/alpha": 0.2, "train/clean_ce_loss": 0.03173828125, "train/clean_loss_scale": 6.95145845413208, "train/clean_loss_weight": 1.390291690826416, "train/combined_loss": 0.2647528052330017, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.2424170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2206273376941681 }, { "epoch": 2.424170616113744, "step": 1023, "train/alpha": 0.2, "train/clean_ce_loss": 0.04345703125, "train/clean_loss_scale": 17.572418212890625, "train/clean_loss_weight": 3.514483690261841, "train/combined_loss": 0.916374146938324, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.2424170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7636451125144958 }, { "epoch": 2.424170616113744, "step": 1023, "train/alpha": 0.2, "train/clean_ce_loss": 0.0673828125, "train/clean_loss_scale": 8.286625862121582, "train/clean_loss_weight": 1.6573251485824585, "train/combined_loss": 0.6700513362884521, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.2424170616113744, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.558376133441925 }, { "epoch": 2.4265402843601898, "step": 1024, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 7.712441921234131, "train/clean_loss_weight": 1.5424884557724, "train/combined_loss": 0.7365984916687012, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.24265402843601896, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6138320565223694 }, { "epoch": 2.4265402843601898, "step": 1024, "train/alpha": 0.2, "train/clean_ce_loss": 0.13671875, "train/clean_loss_scale": 6.194351673126221, "train/clean_loss_weight": 1.23887038230896, "train/combined_loss": 1.0162608623504639, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.24265402843601896, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.84688401222229 }, { "epoch": 2.4265402843601898, "step": 1024, "train/alpha": 0.2, "train/clean_ce_loss": 0.05859375, "train/clean_loss_scale": 10.277765274047852, "train/clean_loss_weight": 2.0555531978607178, "train/combined_loss": 0.7226553559303284, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.24265402843601896, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6022127866744995 }, { "epoch": 2.4265402843601898, "step": 1024, "train/alpha": 0.2, "train/clean_ce_loss": 0.03173828125, "train/clean_loss_scale": 8.952232360839844, "train/clean_loss_weight": 1.7904465198516846, "train/combined_loss": 0.34095415472984314, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.24265402843601896, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.28412845730781555 }, { "epoch": 2.428909952606635, "grad_norm": 15.697834014892578, "learning_rate": 2e-06, "loss": 3.0934, "step": 1025 }, { "epoch": 2.428909952606635, "step": 1025, "train/alpha": 0.2, "train/clean_ce_loss": 0.0306396484375, "train/clean_loss_scale": 6.197720050811768, "train/clean_loss_weight": 1.2395440340042114, "train/combined_loss": 0.22787514328956604, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.24289099526066352, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.18989595770835876 }, { "epoch": 2.428909952606635, "step": 1025, "train/alpha": 0.2, "train/clean_ce_loss": 0.052734375, "train/clean_loss_scale": 18.260892868041992, "train/clean_loss_weight": 3.6521785259246826, "train/combined_loss": 1.1555721759796143, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.24289099526066352, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9629768133163452 }, { "epoch": 2.428909952606635, "step": 1025, "train/alpha": 0.2, "train/clean_ce_loss": 0.0703125, "train/clean_loss_scale": 7.580257892608643, "train/clean_loss_weight": 1.5160516500473022, "train/combined_loss": 0.6395842432975769, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.24289099526066352, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5329868793487549 }, { "epoch": 2.428909952606635, "step": 1025, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 7.1129150390625, "train/clean_loss_weight": 1.422582983970642, "train/combined_loss": 0.5584749579429626, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.24289099526066352, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46539580821990967 }, { "epoch": 2.4312796208530805, "step": 1026, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 6.921507835388184, "train/clean_loss_weight": 1.3843015432357788, "train/combined_loss": 0.5921133756637573, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.24312796208530807, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4934278130531311 }, { "epoch": 2.4312796208530805, "step": 1026, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 6.64512825012207, "train/clean_loss_weight": 1.3290256261825562, "train/combined_loss": 0.517852783203125, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.24312796208530807, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4315439760684967 }, { "epoch": 2.4312796208530805, "step": 1026, "train/alpha": 0.2, "train/clean_ce_loss": 0.1435546875, "train/clean_loss_scale": 7.434347152709961, "train/clean_loss_weight": 1.48686945438385, "train/combined_loss": 1.2806824445724487, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.24312796208530807, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0672353506088257 }, { "epoch": 2.4312796208530805, "step": 1026, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 6.883407115936279, "train/clean_loss_weight": 1.3766814470291138, "train/combined_loss": 0.6533858776092529, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.24312796208530807, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.544488251209259 }, { "epoch": 2.433649289099526, "step": 1027, "train/alpha": 0.2, "train/clean_ce_loss": 0.07373046875, "train/clean_loss_scale": 12.06633186340332, "train/clean_loss_weight": 2.41326642036438, "train/combined_loss": 1.0675876140594482, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.2433649289099526, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8896563053131104 }, { "epoch": 2.433649289099526, "step": 1027, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 7.262179374694824, "train/clean_loss_weight": 1.452435851097107, "train/combined_loss": 0.6170015335083008, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.2433649289099526, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5141679644584656 }, { "epoch": 2.433649289099526, "step": 1027, "train/alpha": 0.2, "train/clean_ce_loss": 0.11328125, "train/clean_loss_scale": 6.914152145385742, "train/clean_loss_weight": 1.3828305006027222, "train/combined_loss": 0.9398925304412842, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.2433649289099526, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7832437753677368 }, { "epoch": 2.433649289099526, "step": 1027, "train/alpha": 0.2, "train/clean_ce_loss": 0.083984375, "train/clean_loss_scale": 7.602493762969971, "train/clean_loss_weight": 1.5204987525939941, "train/combined_loss": 0.7661888003349304, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.2433649289099526, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6384906768798828 }, { "epoch": 2.4360189573459716, "step": 1028, "train/alpha": 0.2, "train/clean_ce_loss": 0.1220703125, "train/clean_loss_scale": 6.672047138214111, "train/clean_loss_weight": 1.334409475326538, "train/combined_loss": 0.9773507118225098, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.24360189573459715, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8144589066505432 }, { "epoch": 2.4360189573459716, "step": 1028, "train/alpha": 0.2, "train/clean_ce_loss": 0.072265625, "train/clean_loss_scale": 8.073238372802734, "train/clean_loss_weight": 1.6146477460861206, "train/combined_loss": 0.7001011371612549, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.24360189573459715, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5834175944328308 }, { "epoch": 2.4360189573459716, "step": 1028, "train/alpha": 0.2, "train/clean_ce_loss": 0.0284423828125, "train/clean_loss_scale": 7.801833629608154, "train/clean_loss_weight": 1.5603667497634888, "train/combined_loss": 0.2662833034992218, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.24360189573459715, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2219027429819107 }, { "epoch": 2.4360189573459716, "step": 1028, "train/alpha": 0.2, "train/clean_ce_loss": 0.02099609375, "train/clean_loss_scale": 12.811249732971191, "train/clean_loss_weight": 2.5622498989105225, "train/combined_loss": 0.3227834403514862, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.24360189573459715, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.26898619532585144 }, { "epoch": 2.438388625592417, "step": 1029, "train/alpha": 0.2, "train/clean_ce_loss": 0.091796875, "train/clean_loss_scale": 7.837475776672363, "train/clean_loss_weight": 1.5674952268600464, "train/combined_loss": 0.8633469343185425, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.2438388625592417, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7194557785987854 }, { "epoch": 2.438388625592417, "step": 1029, "train/alpha": 0.2, "train/clean_ce_loss": 0.054443359375, "train/clean_loss_scale": 7.077954292297363, "train/clean_loss_weight": 1.4155908823013306, "train/combined_loss": 0.4624171257019043, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.2438388625592417, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3853476047515869 }, { "epoch": 2.438388625592417, "step": 1029, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 10.535725593566895, "train/clean_loss_weight": 2.107145071029663, "train/combined_loss": 0.9012984037399292, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.2438388625592417, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7510820031166077 }, { "epoch": 2.438388625592417, "step": 1029, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 9.960052490234375, "train/clean_loss_weight": 1.992010474205017, "train/combined_loss": 0.8695592880249023, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.2438388625592417, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.724632740020752 }, { "epoch": 2.4407582938388623, "grad_norm": 56.5436897277832, "learning_rate": 2e-06, "loss": 3.2469, "step": 1030 }, { "epoch": 2.4407582938388623, "step": 1030, "train/alpha": 0.2, "train/clean_ce_loss": 0.0004119873046875, "train/clean_loss_scale": 329.684326171875, "train/clean_loss_weight": 65.9368667602539, "train/combined_loss": 0.16299089789390564, "train/pos_loss/pos0": 0.0018918007845059037, "train/pos_loss/pos1": 0.04216909781098366, "train/pos_loss/pos2": 0.1579149067401886, "train/pos_loss/pos3": 0.33865979313850403, "train/pos_loss/pos4": 1.0706994533538818, "train/pos_loss/pos5": 1.810512900352478, "train/pos_loss/pos6": 2.7601475715637207, "train/pos_loss/weight0": 2.9626855850219727, "train/pos_loss/weight1": 1.9708868265151978, "train/pos_loss/weight2": 1.1408321857452393, "train/pos_loss/weight3": 0.5764502882957458, "train/pos_loss/weight4": 0.2366788238286972, "train/pos_loss/weight5": 0.08168614655733109, "train/pos_loss/weight6": 0.03078032098710537, "train/pos_weight_progress": 0.24407582938388625, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1358257532119751 }, { "epoch": 2.4407582938388623, "step": 1030, "train/alpha": 0.2, "train/clean_ce_loss": 0.138671875, "train/clean_loss_scale": 10.13047981262207, "train/clean_loss_weight": 2.0260961055755615, "train/combined_loss": 1.6857751607894897, "train/pos_loss/pos0": 0.403174489736557, "train/pos_loss/pos1": 0.8932151794433594, "train/pos_loss/pos2": 1.6504263877868652, "train/pos_loss/pos3": 2.080981731414795, "train/pos_loss/pos4": 2.611645460128784, "train/pos_loss/pos5": 3.1965110301971436, "train/pos_loss/pos6": 3.7640604972839355, "train/pos_loss/weight0": 2.061251163482666, "train/pos_loss/weight1": 1.626855492591858, "train/pos_loss/weight2": 1.2088733911514282, "train/pos_loss/weight3": 0.8674459457397461, "train/pos_loss/weight4": 0.5996910333633423, "train/pos_loss/weight5": 0.36053916811943054, "train/pos_loss/weight6": 0.2753438353538513, "train/pos_weight_progress": 0.24407582938388625, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4048125743865967 }, { "epoch": 2.4407582938388623, "step": 1030, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 6.766241550445557, "train/clean_loss_weight": 1.3532483577728271, "train/combined_loss": 0.5907245874404907, "train/pos_loss/pos0": 0.23582857847213745, "train/pos_loss/pos1": 0.4502060115337372, "train/pos_loss/pos2": 0.8135218620300293, "train/pos_loss/pos3": 1.1606924533843994, "train/pos_loss/pos4": 1.586344838142395, "train/pos_loss/pos5": 1.8975892066955566, "train/pos_loss/pos6": 2.2786028385162354, "train/pos_loss/weight0": 3.1535611152648926, "train/pos_loss/weight1": 2.0686845779418945, "train/pos_loss/weight2": 1.1224417686462402, "train/pos_loss/weight3": 0.46389225125312805, "train/pos_loss/weight4": 0.14910690486431122, "train/pos_loss/weight5": 0.034598737955093384, "train/pos_loss/weight6": 0.007714254315942526, "train/pos_weight_progress": 0.24407582938388625, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.49227049946784973 }, { "epoch": 2.4407582938388623, "step": 1030, "train/alpha": 0.2, "train/clean_ce_loss": 0.039306640625, "train/clean_loss_scale": 10.879281997680664, "train/clean_loss_weight": 2.175856351852417, "train/combined_loss": 0.5131536722183228, "train/pos_loss/pos0": 0.11458437144756317, "train/pos_loss/pos1": 0.3080548644065857, "train/pos_loss/pos2": 0.43429142236709595, "train/pos_loss/pos3": 0.7998486161231995, "train/pos_loss/pos4": 1.1060465574264526, "train/pos_loss/pos5": 1.4931906461715698, "train/pos_loss/pos6": 1.8943288326263428, "train/pos_loss/weight0": 2.451645851135254, "train/pos_loss/weight1": 1.7945072650909424, "train/pos_loss/weight2": 1.2362470626831055, "train/pos_loss/weight3": 0.7616814970970154, "train/pos_loss/weight4": 0.4197528660297394, "train/pos_loss/weight5": 0.21682360768318176, "train/pos_loss/weight6": 0.11934191733598709, "train/pos_weight_progress": 0.24407582938388625, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4276280403137207 }, { "epoch": 2.443127962085308, "step": 1031, "train/alpha": 0.2, "train/clean_ce_loss": 0.058837890625, "train/clean_loss_scale": 24.035188674926758, "train/clean_loss_weight": 4.807037830352783, "train/combined_loss": 1.6970157623291016, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.2443127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.414179801940918 }, { "epoch": 2.443127962085308, "step": 1031, "train/alpha": 0.2, "train/clean_ce_loss": 3.7550926208496094e-06, "train/clean_loss_scale": 44247.4296875, "train/clean_loss_weight": 8849.486328125, "train/combined_loss": 0.20004317164421082, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.2443127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.16681253910064697 }, { "epoch": 2.443127962085308, "step": 1031, "train/alpha": 0.2, "train/clean_ce_loss": 0.05419921875, "train/clean_loss_scale": 12.158001899719238, "train/clean_loss_weight": 2.431600332260132, "train/combined_loss": 0.7907450199127197, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.2443127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6589542031288147 }, { "epoch": 2.443127962085308, "step": 1031, "train/alpha": 0.2, "train/clean_ce_loss": 0.1064453125, "train/clean_loss_scale": 8.667669296264648, "train/clean_loss_weight": 1.7335338592529297, "train/combined_loss": 1.1071592569351196, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.2443127962085308, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9226327538490295 }, { "epoch": 2.4454976303317535, "step": 1032, "train/alpha": 0.2, "train/clean_ce_loss": 0.0196533203125, "train/clean_loss_scale": 32.012699127197266, "train/clean_loss_weight": 6.4025397300720215, "train/combined_loss": 0.7549870014190674, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.24454976303317535, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6291558146476746 }, { "epoch": 2.4454976303317535, "step": 1032, "train/alpha": 0.2, "train/clean_ce_loss": 0.01025390625, "train/clean_loss_scale": 35.672874450683594, "train/clean_loss_weight": 7.134574890136719, "train/combined_loss": 0.4389435648918152, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.24454976303317535, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3657863140106201 }, { "epoch": 2.4454976303317535, "step": 1032, "train/alpha": 0.2, "train/clean_ce_loss": 0.07568359375, "train/clean_loss_scale": 9.53313159942627, "train/clean_loss_weight": 1.9066263437271118, "train/combined_loss": 0.865801990032196, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.24454976303317535, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7215016484260559 }, { "epoch": 2.4454976303317535, "step": 1032, "train/alpha": 0.2, "train/clean_ce_loss": 0.0263671875, "train/clean_loss_scale": 42.98479461669922, "train/clean_loss_weight": 8.596959114074707, "train/combined_loss": 1.3600658178329468, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.24454976303317535, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1333881616592407 }, { "epoch": 2.447867298578199, "step": 1033, "train/alpha": 0.2, "train/clean_ce_loss": 0.0014801025390625, "train/clean_loss_scale": 315.0921325683594, "train/clean_loss_weight": 63.018428802490234, "train/combined_loss": 0.5596423745155334, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.2447867298578199, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4663686454296112 }, { "epoch": 2.447867298578199, "step": 1033, "train/alpha": 0.2, "train/clean_ce_loss": 0.00011014938354492188, "train/clean_loss_scale": 3153.493896484375, "train/clean_loss_weight": 630.6987915039062, "train/combined_loss": 0.4168264865875244, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.2447867298578199, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3473553955554962 }, { "epoch": 2.447867298578199, "step": 1033, "train/alpha": 0.2, "train/clean_ce_loss": 0.00099945068359375, "train/clean_loss_scale": 196.47994995117188, "train/clean_loss_weight": 39.295989990234375, "train/combined_loss": 0.23564642667770386, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.2447867298578199, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.19637201726436615 }, { "epoch": 2.447867298578199, "step": 1033, "train/alpha": 0.2, "train/clean_ce_loss": 0.10009765625, "train/clean_loss_scale": 9.166740417480469, "train/clean_loss_weight": 1.8333481550216675, "train/combined_loss": 1.1010830402374268, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.2447867298578199, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9175692200660706 }, { "epoch": 2.4502369668246446, "step": 1034, "train/alpha": 0.2, "train/clean_ce_loss": 0.045166015625, "train/clean_loss_scale": 9.323260307312012, "train/clean_loss_weight": 1.8646520376205444, "train/combined_loss": 0.5053134560585022, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.24502369668246446, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.42109453678131104 }, { "epoch": 2.4502369668246446, "step": 1034, "train/alpha": 0.2, "train/clean_ce_loss": 0.02197265625, "train/clean_loss_scale": 37.18343734741211, "train/clean_loss_weight": 7.436687469482422, "train/combined_loss": 0.9804226160049438, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.24502369668246446, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8170188665390015 }, { "epoch": 2.4502369668246446, "step": 1034, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 10.546066284179688, "train/clean_loss_weight": 2.109213352203369, "train/combined_loss": 0.9021830558776855, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.24502369668246446, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7518191933631897 }, { "epoch": 2.4502369668246446, "step": 1034, "train/alpha": 0.2, "train/clean_ce_loss": 0.05419921875, "train/clean_loss_scale": 15.167840003967285, "train/clean_loss_weight": 3.0335681438446045, "train/combined_loss": 0.9865021109580994, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.24502369668246446, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8220850825309753 }, { "epoch": 2.4526066350710902, "grad_norm": 18.774736404418945, "learning_rate": 2e-06, "loss": 2.9119, "step": 1035 }, { "epoch": 2.4526066350710902, "step": 1035, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 10.918668746948242, "train/clean_loss_weight": 2.1837337017059326, "train/combined_loss": 0.8572860956192017, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.245260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7144050598144531 }, { "epoch": 2.4526066350710902, "step": 1035, "train/alpha": 0.2, "train/clean_ce_loss": 0.07470703125, "train/clean_loss_scale": 9.056962013244629, "train/clean_loss_weight": 1.8113924264907837, "train/combined_loss": 0.8119425177574158, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.245260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.676618754863739 }, { "epoch": 2.4526066350710902, "step": 1035, "train/alpha": 0.2, "train/clean_ce_loss": 0.03759765625, "train/clean_loss_scale": 11.083719253540039, "train/clean_loss_weight": 2.2167439460754395, "train/combined_loss": 0.5000662207603455, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.245260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4167218506336212 }, { "epoch": 2.4526066350710902, "step": 1035, "train/alpha": 0.2, "train/clean_ce_loss": 0.0751953125, "train/clean_loss_scale": 12.811883926391602, "train/clean_loss_weight": 2.5623767375946045, "train/combined_loss": 1.1560723781585693, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.245260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9633936285972595 }, { "epoch": 2.4549763033175354, "step": 1036, "train/alpha": 0.2, "train/clean_ce_loss": 0.0206298828125, "train/clean_loss_scale": 26.59499740600586, "train/clean_loss_weight": 5.318999767303467, "train/combined_loss": 0.6583820581436157, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.24549763033175356, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5486516952514648 }, { "epoch": 2.4549763033175354, "step": 1036, "train/alpha": 0.2, "train/clean_ce_loss": 0.1044921875, "train/clean_loss_scale": 12.074811935424805, "train/clean_loss_weight": 2.4149625301361084, "train/combined_loss": 1.5140682458877563, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.24549763033175356, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.261723518371582 }, { "epoch": 2.4549763033175354, "step": 1036, "train/alpha": 0.2, "train/clean_ce_loss": 0.000354766845703125, "train/clean_loss_scale": 1563.775146484375, "train/clean_loss_weight": 312.7550354003906, "train/combined_loss": 0.6657307147979736, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.24549763033175356, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.554775595664978 }, { "epoch": 2.4549763033175354, "step": 1036, "train/alpha": 0.2, "train/clean_ce_loss": 0.0751953125, "train/clean_loss_scale": 13.070850372314453, "train/clean_loss_weight": 2.6141700744628906, "train/combined_loss": 1.1794400215148926, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.24549763033175356, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.982866644859314 }, { "epoch": 2.457345971563981, "step": 1037, "train/alpha": 0.2, "train/clean_ce_loss": 0.0732421875, "train/clean_loss_scale": 10.751786231994629, "train/clean_loss_weight": 2.150357246398926, "train/combined_loss": 0.9449812173843384, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.24573459715639812, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.787484347820282 }, { "epoch": 2.457345971563981, "step": 1037, "train/alpha": 0.2, "train/clean_ce_loss": 0.0341796875, "train/clean_loss_scale": 28.768606185913086, "train/clean_loss_weight": 5.753721237182617, "train/combined_loss": 1.179962396621704, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.24573459715639812, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9833019375801086 }, { "epoch": 2.457345971563981, "step": 1037, "train/alpha": 0.2, "train/clean_ce_loss": 0.0004425048828125, "train/clean_loss_scale": 376.0216064453125, "train/clean_loss_weight": 75.2043228149414, "train/combined_loss": 0.19966968894004822, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.24573459715639812, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.16639140248298645 }, { "epoch": 2.457345971563981, "step": 1037, "train/alpha": 0.2, "train/clean_ce_loss": 0.044677734375, "train/clean_loss_scale": 14.462712287902832, "train/clean_loss_weight": 2.892542600631714, "train/combined_loss": 0.7753934264183044, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.24573459715639812, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6461611986160278 }, { "epoch": 2.4597156398104265, "step": 1038, "train/alpha": 0.2, "train/clean_ce_loss": 0.0927734375, "train/clean_loss_scale": 10.769213676452637, "train/clean_loss_weight": 2.1538426876068115, "train/combined_loss": 1.1989164352416992, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.24597156398104264, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9990969896316528 }, { "epoch": 2.4597156398104265, "step": 1038, "train/alpha": 0.2, "train/clean_ce_loss": 0.05322265625, "train/clean_loss_scale": 9.96279525756836, "train/clean_loss_weight": 1.9925590753555298, "train/combined_loss": 0.636295735836029, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.24597156398104264, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5302464365959167 }, { "epoch": 2.4597156398104265, "step": 1038, "train/alpha": 0.2, "train/clean_ce_loss": 0.0634765625, "train/clean_loss_scale": 10.62379264831543, "train/clean_loss_weight": 2.12475848197937, "train/combined_loss": 0.8092342019081116, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.24597156398104264, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6743618249893188 }, { "epoch": 2.4597156398104265, "step": 1038, "train/alpha": 0.2, "train/clean_ce_loss": 0.031982421875, "train/clean_loss_scale": 12.290287017822266, "train/clean_loss_weight": 2.458057403564453, "train/combined_loss": 0.47168776392936707, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.24597156398104264, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3930731415748596 }, { "epoch": 2.462085308056872, "step": 1039, "train/alpha": 0.2, "train/clean_ce_loss": 0.0732421875, "train/clean_loss_scale": 12.76905345916748, "train/clean_loss_weight": 2.5538108348846436, "train/combined_loss": 1.1222801208496094, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.2462085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9352334141731262 }, { "epoch": 2.462085308056872, "step": 1039, "train/alpha": 0.2, "train/clean_ce_loss": 0.09521484375, "train/clean_loss_scale": 10.480567932128906, "train/clean_loss_weight": 2.096113681793213, "train/combined_loss": 1.1974867582321167, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.2462085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9979056119918823 }, { "epoch": 2.462085308056872, "step": 1039, "train/alpha": 0.2, "train/clean_ce_loss": 0.05908203125, "train/clean_loss_scale": 11.623567581176758, "train/clean_loss_weight": 2.3247134685516357, "train/combined_loss": 0.8240927457809448, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.2462085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.686743974685669 }, { "epoch": 2.462085308056872, "step": 1039, "train/alpha": 0.2, "train/clean_ce_loss": 0.138671875, "train/clean_loss_scale": 9.663657188415527, "train/clean_loss_weight": 1.9327315092086792, "train/combined_loss": 1.6080929040908813, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.2462085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3400774002075195 }, { "epoch": 2.4644549763033177, "grad_norm": 15.493448257446289, "learning_rate": 2e-06, "loss": 2.8635, "step": 1040 }, { "epoch": 2.4644549763033177, "step": 1040, "train/alpha": 0.2, "train/clean_ce_loss": 0.1318359375, "train/clean_loss_scale": 7.538798809051514, "train/clean_loss_weight": 1.5077598094940186, "train/combined_loss": 1.1926615238189697, "train/pos_loss/pos0": 0.37945419549942017, "train/pos_loss/pos1": 0.6550224423408508, "train/pos_loss/pos2": 1.0356321334838867, "train/pos_loss/pos3": 1.4031144380569458, "train/pos_loss/pos4": 1.9006682634353638, "train/pos_loss/pos5": 2.2643399238586426, "train/pos_loss/pos6": 2.697038173675537, "train/pos_loss/weight0": 2.073793411254883, "train/pos_loss/weight1": 1.6510188579559326, "train/pos_loss/weight2": 1.2091057300567627, "train/pos_loss/weight3": 0.8640528321266174, "train/pos_loss/weight4": 0.5698102712631226, "train/pos_loss/weight5": 0.3767193853855133, "train/pos_loss/weight6": 0.25549957156181335, "train/pos_weight_progress": 0.24644549763033174, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9938846230506897 }, { "epoch": 2.4644549763033177, "step": 1040, "train/alpha": 0.2, "train/clean_ce_loss": 0.12060546875, "train/clean_loss_scale": 8.18826675415039, "train/clean_loss_weight": 1.6376533508300781, "train/combined_loss": 1.1850597858428955, "train/pos_loss/pos0": 0.39117392897605896, "train/pos_loss/pos1": 0.8639037013053894, "train/pos_loss/pos2": 1.3259575366973877, "train/pos_loss/pos3": 1.9243022203445435, "train/pos_loss/pos4": 2.5530312061309814, "train/pos_loss/pos5": 2.8863253593444824, "train/pos_loss/pos6": 3.266765832901001, "train/pos_loss/weight0": 2.754662275314331, "train/pos_loss/weight1": 1.9037423133850098, "train/pos_loss/weight2": 1.1896378993988037, "train/pos_loss/weight3": 0.6548906564712524, "train/pos_loss/weight4": 0.3107185959815979, "train/pos_loss/weight5": 0.12791603803634644, "train/pos_loss/weight6": 0.05843207240104675, "train/pos_weight_progress": 0.24644549763033174, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9875497817993164 }, { "epoch": 2.4644549763033177, "step": 1040, "train/alpha": 0.2, "train/clean_ce_loss": 0.038818359375, "train/clean_loss_scale": 12.515295028686523, "train/clean_loss_weight": 2.503059148788452, "train/combined_loss": 0.5829878449440002, "train/pos_loss/pos0": 0.12935110926628113, "train/pos_loss/pos1": 0.3721872866153717, "train/pos_loss/pos2": 0.737687349319458, "train/pos_loss/pos3": 1.6009491682052612, "train/pos_loss/pos4": 2.5537936687469482, "train/pos_loss/pos5": 2.7854089736938477, "train/pos_loss/pos6": 3.5520570278167725, "train/pos_loss/weight0": 3.0988528728485107, "train/pos_loss/weight1": 2.029966354370117, "train/pos_loss/weight2": 1.1409645080566406, "train/pos_loss/weight3": 0.5015349388122559, "train/pos_loss/weight4": 0.17296193540096283, "train/pos_loss/weight5": 0.043696340173482895, "train/pos_loss/weight6": 0.012023174203932285, "train/pos_weight_progress": 0.24644549763033174, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.48582321405410767 }, { "epoch": 2.4644549763033177, "step": 1040, "train/alpha": 0.2, "train/clean_ce_loss": 0.062255859375, "train/clean_loss_scale": 10.562023162841797, "train/clean_loss_weight": 2.1124045848846436, "train/combined_loss": 0.7890573740005493, "train/pos_loss/pos0": 0.18959124386310577, "train/pos_loss/pos1": 0.5310123562812805, "train/pos_loss/pos2": 0.9371584057807922, "train/pos_loss/pos3": 1.362178087234497, "train/pos_loss/pos4": 1.8512609004974365, "train/pos_loss/pos5": 2.2834625244140625, "train/pos_loss/pos6": 2.670990228652954, "train/pos_loss/weight0": 2.682318687438965, "train/pos_loss/weight1": 1.8808562755584717, "train/pos_loss/weight2": 1.2450268268585205, "train/pos_loss/weight3": 0.7249835729598999, "train/pos_loss/weight4": 0.32587891817092896, "train/pos_loss/weight5": 0.09965280443429947, "train/pos_loss/weight6": 0.041282784193754196, "train/pos_weight_progress": 0.24644549763033174, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6575478315353394 }, { "epoch": 2.466824644549763, "step": 1041, "train/alpha": 0.2, "train/clean_ce_loss": 0.1279296875, "train/clean_loss_scale": 8.212625503540039, "train/clean_loss_weight": 1.64252507686615, "train/combined_loss": 1.2607662677764893, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.2466824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0506385564804077 }, { "epoch": 2.466824644549763, "step": 1041, "train/alpha": 0.2, "train/clean_ce_loss": 0.0322265625, "train/clean_loss_scale": 9.912496566772461, "train/clean_loss_weight": 1.982499361038208, "train/combined_loss": 0.38333484530448914, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.2466824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3194456994533539 }, { "epoch": 2.466824644549763, "step": 1041, "train/alpha": 0.2, "train/clean_ce_loss": 0.091796875, "train/clean_loss_scale": 11.15866756439209, "train/clean_loss_weight": 2.231733560562134, "train/combined_loss": 1.2291970252990723, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.2466824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0243308544158936 }, { "epoch": 2.466824644549763, "step": 1041, "train/alpha": 0.2, "train/clean_ce_loss": 0.09423828125, "train/clean_loss_scale": 8.482341766357422, "train/clean_loss_weight": 1.6964683532714844, "train/combined_loss": 0.9592335224151611, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.2466824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7993612885475159 }, { "epoch": 2.4691943127962084, "step": 1042, "train/alpha": 0.2, "train/clean_ce_loss": 0.0189208984375, "train/clean_loss_scale": 12.24306869506836, "train/clean_loss_weight": 2.4486138820648193, "train/combined_loss": 0.27797985076904297, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.24691943127962085, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.23164986073970795 }, { "epoch": 2.4691943127962084, "step": 1042, "train/alpha": 0.2, "train/clean_ce_loss": 0.056884765625, "train/clean_loss_scale": 9.10642147064209, "train/clean_loss_weight": 1.821284294128418, "train/combined_loss": 0.6216199398040771, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.24691943127962085, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5180166363716125 }, { "epoch": 2.4691943127962084, "step": 1042, "train/alpha": 0.2, "train/clean_ce_loss": 0.008056640625, "train/clean_loss_scale": 27.7017765045166, "train/clean_loss_weight": 5.540355205535889, "train/combined_loss": 0.26781991124153137, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.24691943127962085, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2231832593679428 }, { "epoch": 2.4691943127962084, "step": 1042, "train/alpha": 0.2, "train/clean_ce_loss": 0.01190185546875, "train/clean_loss_scale": 22.724987030029297, "train/clean_loss_weight": 4.544997692108154, "train/combined_loss": 0.3245634138584137, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.24691943127962085, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2704695165157318 }, { "epoch": 2.471563981042654, "step": 1043, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 8.319426536560059, "train/clean_loss_weight": 1.6638853549957275, "train/combined_loss": 0.7263249754905701, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.2471563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6052708029747009 }, { "epoch": 2.471563981042654, "step": 1043, "train/alpha": 0.2, "train/clean_ce_loss": 0.053955078125, "train/clean_loss_scale": 9.354640007019043, "train/clean_loss_weight": 1.8709280490875244, "train/combined_loss": 0.6056764125823975, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.2471563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5047303438186646 }, { "epoch": 2.471563981042654, "step": 1043, "train/alpha": 0.2, "train/clean_ce_loss": 0.00634765625, "train/clean_loss_scale": 36.431949615478516, "train/clean_loss_weight": 7.2863898277282715, "train/combined_loss": 0.2775089740753174, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.2471563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.23125748336315155 }, { "epoch": 2.471563981042654, "step": 1043, "train/alpha": 0.2, "train/clean_ce_loss": 0.04736328125, "train/clean_loss_scale": 19.86680030822754, "train/clean_loss_weight": 3.973360061645508, "train/combined_loss": 1.129148244857788, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.2471563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9409568905830383 }, { "epoch": 2.4739336492890995, "step": 1044, "train/alpha": 0.2, "train/clean_ce_loss": 0.1357421875, "train/clean_loss_scale": 8.345417976379395, "train/clean_loss_weight": 1.669083595275879, "train/combined_loss": 1.3593902587890625, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.24739336492890995, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.132825255393982 }, { "epoch": 2.4739336492890995, "step": 1044, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 8.757782936096191, "train/clean_loss_weight": 1.751556634902954, "train/combined_loss": 0.6927542686462402, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.24739336492890995, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5772952437400818 }, { "epoch": 2.4739336492890995, "step": 1044, "train/alpha": 0.2, "train/clean_ce_loss": 0.0732421875, "train/clean_loss_scale": 4.764315605163574, "train/clean_loss_weight": 0.9528631567955017, "train/combined_loss": 0.4187386631965637, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.24739336492890995, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3489488959312439 }, { "epoch": 2.4739336492890995, "step": 1044, "train/alpha": 0.2, "train/clean_ce_loss": 0.0048828125, "train/clean_loss_scale": 22.753271102905273, "train/clean_loss_weight": 4.550654411315918, "train/combined_loss": 0.1333199441432953, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.24739336492890995, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1110999584197998 }, { "epoch": 2.476303317535545, "grad_norm": 21.05881690979004, "learning_rate": 2e-06, "loss": 3.2057, "step": 1045 }, { "epoch": 2.476303317535545, "step": 1045, "train/alpha": 0.2, "train/clean_ce_loss": 0.146484375, "train/clean_loss_scale": 6.422793388366699, "train/clean_loss_weight": 1.284558653831482, "train/combined_loss": 1.1290066242218018, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.2476303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9408388733863831 }, { "epoch": 2.476303317535545, "step": 1045, "train/alpha": 0.2, "train/clean_ce_loss": 0.10009765625, "train/clean_loss_scale": 9.684730529785156, "train/clean_loss_weight": 1.936946153640747, "train/combined_loss": 1.1633026599884033, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.2476303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9694188237190247 }, { "epoch": 2.476303317535545, "step": 1045, "train/alpha": 0.2, "train/clean_ce_loss": 0.0546875, "train/clean_loss_scale": 18.669830322265625, "train/clean_loss_weight": 3.733966112136841, "train/combined_loss": 1.2252075672149658, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.2476303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0210063457489014 }, { "epoch": 2.476303317535545, "step": 1045, "train/alpha": 0.2, "train/clean_ce_loss": 0.057373046875, "train/clean_loss_scale": 16.54022216796875, "train/clean_loss_weight": 3.30804443359375, "train/combined_loss": 1.1387555599212646, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.2476303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.948962926864624 }, { "epoch": 2.4786729857819907, "step": 1046, "train/alpha": 0.2, "train/clean_ce_loss": 0.16015625, "train/clean_loss_scale": 8.077916145324707, "train/clean_loss_weight": 1.6155833005905151, "train/combined_loss": 1.5524746179580688, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.24786729857819906, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2937288284301758 }, { "epoch": 2.4786729857819907, "step": 1046, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 9.514392852783203, "train/clean_loss_weight": 1.9028786420822144, "train/combined_loss": 0.9532975554466248, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.24786729857819906, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7944146394729614 }, { "epoch": 2.4786729857819907, "step": 1046, "train/alpha": 0.2, "train/clean_ce_loss": 0.0986328125, "train/clean_loss_scale": 9.233745574951172, "train/clean_loss_weight": 1.846749186515808, "train/combined_loss": 1.092900276184082, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.24786729857819906, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9107502698898315 }, { "epoch": 2.4786729857819907, "step": 1046, "train/alpha": 0.2, "train/clean_ce_loss": 0.1337890625, "train/clean_loss_scale": 7.388530254364014, "train/clean_loss_weight": 1.4777060747146606, "train/combined_loss": 1.1862053871154785, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.24786729857819906, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9885045289993286 }, { "epoch": 2.4810426540284363, "step": 1047, "train/alpha": 0.2, "train/clean_ce_loss": 0.1376953125, "train/clean_loss_scale": 7.233902931213379, "train/clean_loss_weight": 1.4467805624008179, "train/combined_loss": 1.1952893733978271, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.2481042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9960744976997375 }, { "epoch": 2.4810426540284363, "step": 1047, "train/alpha": 0.2, "train/clean_ce_loss": 0.06298828125, "train/clean_loss_scale": 2.2364261150360107, "train/clean_loss_weight": 0.4472852349281311, "train/combined_loss": 0.16904236376285553, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.2481042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1408686339855194 }, { "epoch": 2.4810426540284363, "step": 1047, "train/alpha": 0.2, "train/clean_ce_loss": 0.09521484375, "train/clean_loss_scale": 8.283279418945312, "train/clean_loss_weight": 1.6566559076309204, "train/combined_loss": 0.9464293718338013, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.2481042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7886911630630493 }, { "epoch": 2.4810426540284363, "step": 1047, "train/alpha": 0.2, "train/clean_ce_loss": 0.185546875, "train/clean_loss_scale": 6.802628517150879, "train/clean_loss_weight": 1.3605257272720337, "train/combined_loss": 1.5146477222442627, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.2481042654028436, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2622064352035522 }, { "epoch": 2.4834123222748814, "step": 1048, "train/alpha": 0.2, "train/clean_ce_loss": 0.07958984375, "train/clean_loss_scale": 8.80767822265625, "train/clean_loss_weight": 1.76153564453125, "train/combined_loss": 0.8412021398544312, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.24834123222748816, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.701001763343811 }, { "epoch": 2.4834123222748814, "step": 1048, "train/alpha": 0.2, "train/clean_ce_loss": 0.0306396484375, "train/clean_loss_scale": 9.07131576538086, "train/clean_loss_weight": 1.8142632246017456, "train/combined_loss": 0.3335303068161011, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.24834123222748816, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.27794191241264343 }, { "epoch": 2.4834123222748814, "step": 1048, "train/alpha": 0.2, "train/clean_ce_loss": 0.10888671875, "train/clean_loss_scale": 11.044677734375, "train/clean_loss_weight": 2.208935499191284, "train/combined_loss": 1.4431424140930176, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.24834123222748816, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2026187181472778 }, { "epoch": 2.4834123222748814, "step": 1048, "train/alpha": 0.2, "train/clean_ce_loss": 0.053466796875, "train/clean_loss_scale": 13.18761157989502, "train/clean_loss_weight": 2.6375224590301514, "train/combined_loss": 0.846119225025177, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.24834123222748816, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.70509934425354 }, { "epoch": 2.485781990521327, "step": 1049, "train/alpha": 0.2, "train/clean_ce_loss": 0.020751953125, "train/clean_loss_scale": 34.94868850708008, "train/clean_loss_weight": 6.9897379875183105, "train/combined_loss": 0.87030428647995, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.24857819905213271, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7252535820007324 }, { "epoch": 2.485781990521327, "step": 1049, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 7.258647918701172, "train/clean_loss_weight": 1.451729655265808, "train/combined_loss": 0.5614110827445984, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.24857819905213271, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46784254908561707 }, { "epoch": 2.485781990521327, "step": 1049, "train/alpha": 0.2, "train/clean_ce_loss": 0.1484375, "train/clean_loss_scale": 7.355400562286377, "train/clean_loss_weight": 1.4710801839828491, "train/combined_loss": 1.3101806640625, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.24857819905213271, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0918172597885132 }, { "epoch": 2.485781990521327, "step": 1049, "train/alpha": 0.2, "train/clean_ce_loss": 0.08984375, "train/clean_loss_scale": 11.37713623046875, "train/clean_loss_weight": 2.2754273414611816, "train/combined_loss": 1.226597547531128, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.24857819905213271, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0221645832061768 }, { "epoch": 2.4881516587677726, "grad_norm": 18.38526725769043, "learning_rate": 2e-06, "loss": 3.344, "step": 1050 }, { "epoch": 2.4881516587677726, "step": 1050, "train/alpha": 0.2, "train/clean_ce_loss": 0.046875, "train/clean_loss_scale": 11.321158409118652, "train/clean_loss_weight": 2.2642316818237305, "train/combined_loss": 0.6368151307106018, "train/pos_loss/pos0": 0.17655439674854279, "train/pos_loss/pos1": 0.3920776844024658, "train/pos_loss/pos2": 0.6489366292953491, "train/pos_loss/pos3": 1.0801870822906494, "train/pos_loss/pos4": 1.472734808921814, "train/pos_loss/pos5": 1.75551438331604, "train/pos_loss/pos6": 2.225606679916382, "train/pos_loss/weight0": 2.587066888809204, "train/pos_loss/weight1": 1.8983505964279175, "train/pos_loss/weight2": 1.2255921363830566, "train/pos_loss/weight3": 0.7075976729393005, "train/pos_loss/weight4": 0.35639214515686035, "train/pos_loss/weight5": 0.1519956737756729, "train/pos_loss/weight6": 0.0730045884847641, "train/pos_weight_progress": 0.24881516587677724, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5306792855262756 }, { "epoch": 2.4881516587677726, "step": 1050, "train/alpha": 0.2, "train/clean_ce_loss": 0.07568359375, "train/clean_loss_scale": 11.159476280212402, "train/clean_loss_weight": 2.2318952083587646, "train/combined_loss": 1.0135071277618408, "train/pos_loss/pos0": 0.34408554434776306, "train/pos_loss/pos1": 0.6854930520057678, "train/pos_loss/pos2": 0.9674581289291382, "train/pos_loss/pos3": 1.3514536619186401, "train/pos_loss/pos4": 1.707301378250122, "train/pos_loss/pos5": 1.9588335752487183, "train/pos_loss/pos6": 2.239103078842163, "train/pos_loss/weight0": 2.3403983116149902, "train/pos_loss/weight1": 1.7532005310058594, "train/pos_loss/weight2": 1.2340666055679321, "train/pos_loss/weight3": 0.8010407090187073, "train/pos_loss/weight4": 0.4697458744049072, "train/pos_loss/weight5": 0.2511923611164093, "train/pos_loss/weight6": 0.15035565197467804, "train/pos_weight_progress": 0.24881516587677724, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8445892930030823 }, { "epoch": 2.4881516587677726, "step": 1050, "train/alpha": 0.2, "train/clean_ce_loss": 0.123046875, "train/clean_loss_scale": 8.844758987426758, "train/clean_loss_weight": 1.7689517736434937, "train/combined_loss": 1.3059839010238647, "train/pos_loss/pos0": 0.481731116771698, "train/pos_loss/pos1": 0.921442985534668, "train/pos_loss/pos2": 1.3323004245758057, "train/pos_loss/pos3": 1.87026846408844, "train/pos_loss/pos4": 2.3745005130767822, "train/pos_loss/pos5": 2.806795597076416, "train/pos_loss/pos6": 3.2110893726348877, "train/pos_loss/weight0": 2.5660505294799805, "train/pos_loss/weight1": 1.857093334197998, "train/pos_loss/weight2": 1.1961030960083008, "train/pos_loss/weight3": 0.7171764373779297, "train/pos_loss/weight4": 0.3795526623725891, "train/pos_loss/weight5": 0.186346635222435, "train/pos_loss/weight6": 0.09767746925354004, "train/pos_weight_progress": 0.24881516587677724, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0883198976516724 }, { "epoch": 2.4881516587677726, "step": 1050, "train/alpha": 0.2, "train/clean_ce_loss": 0.0108642578125, "train/clean_loss_scale": 26.350900650024414, "train/clean_loss_weight": 5.2701802253723145, "train/combined_loss": 0.34353959560394287, "train/pos_loss/pos0": 0.034520041197538376, "train/pos_loss/pos1": 0.14573320746421814, "train/pos_loss/pos2": 0.3584170937538147, "train/pos_loss/pos3": 0.8993685245513916, "train/pos_loss/pos4": 1.6345305442810059, "train/pos_loss/pos5": 2.2989091873168945, "train/pos_loss/pos6": 2.8644001483917236, "train/pos_loss/weight0": 2.939464569091797, "train/pos_loss/weight1": 1.9820332527160645, "train/pos_loss/weight2": 1.1446657180786133, "train/pos_loss/weight3": 0.5724717974662781, "train/pos_loss/weight4": 0.24203014373779297, "train/pos_loss/weight5": 0.08682089298963547, "train/pos_loss/weight6": 0.03251342102885246, "train/pos_weight_progress": 0.24881516587677724, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2862829864025116 }, { "epoch": 2.490521327014218, "step": 1051, "train/alpha": 0.2, "train/clean_ce_loss": 0.103515625, "train/clean_loss_scale": 9.80833911895752, "train/clean_loss_weight": 1.9616678953170776, "train/combined_loss": 1.2183796167373657, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.2490521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.015316367149353 }, { "epoch": 2.490521327014218, "step": 1051, "train/alpha": 0.2, "train/clean_ce_loss": 0.05908203125, "train/clean_loss_scale": 7.658366680145264, "train/clean_loss_weight": 1.5316733121871948, "train/combined_loss": 0.5429662466049194, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.2490521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45247185230255127 }, { "epoch": 2.490521327014218, "step": 1051, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 14.289505958557129, "train/clean_loss_weight": 2.8579013347625732, "train/combined_loss": 1.121949553489685, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.2490521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9349579215049744 }, { "epoch": 2.490521327014218, "step": 1051, "train/alpha": 0.2, "train/clean_ce_loss": 0.09912109375, "train/clean_loss_scale": 9.910452842712402, "train/clean_loss_weight": 1.9820905923843384, "train/combined_loss": 1.1788018941879272, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.2490521327014218, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9823349118232727 }, { "epoch": 2.4928909952606633, "step": 1052, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 9.376382827758789, "train/clean_loss_weight": 1.8752765655517578, "train/combined_loss": 0.7306985855102539, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.24928909952606634, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6089155077934265 }, { "epoch": 2.4928909952606633, "step": 1052, "train/alpha": 0.2, "train/clean_ce_loss": 0.0181884765625, "train/clean_loss_scale": 21.805456161499023, "train/clean_loss_weight": 4.361091136932373, "train/combined_loss": 0.4759296178817749, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.24928909952606634, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.39660802483558655 }, { "epoch": 2.4928909952606633, "step": 1052, "train/alpha": 0.2, "train/clean_ce_loss": 0.138671875, "train/clean_loss_scale": 9.644495010375977, "train/clean_loss_weight": 1.9288990497589111, "train/combined_loss": 1.604904294013977, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.24928909952606634, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3374202251434326 }, { "epoch": 2.4928909952606633, "step": 1052, "train/alpha": 0.2, "train/clean_ce_loss": 0.00087738037109375, "train/clean_loss_scale": 319.1514587402344, "train/clean_loss_weight": 63.830291748046875, "train/combined_loss": 0.33602067828178406, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.24928909952606634, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.280017226934433 }, { "epoch": 2.495260663507109, "step": 1053, "train/alpha": 0.2, "train/clean_ce_loss": 0.111328125, "train/clean_loss_scale": 8.29044246673584, "train/clean_loss_weight": 1.6580885648727417, "train/combined_loss": 1.1075513362884521, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.2495260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9229594469070435 }, { "epoch": 2.495260663507109, "step": 1053, "train/alpha": 0.2, "train/clean_ce_loss": 0.09912109375, "train/clean_loss_scale": 11.116083145141602, "train/clean_loss_weight": 2.2232167720794678, "train/combined_loss": 1.3222060203552246, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.2495260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1018383502960205 }, { "epoch": 2.495260663507109, "step": 1053, "train/alpha": 0.2, "train/clean_ce_loss": 0.04345703125, "train/clean_loss_scale": 12.07813835144043, "train/clean_loss_weight": 2.4156277179718018, "train/combined_loss": 0.6298560500144958, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.2495260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.524880051612854 }, { "epoch": 2.495260663507109, "step": 1053, "train/alpha": 0.2, "train/clean_ce_loss": 0.0014190673828125, "train/clean_loss_scale": 219.1110076904297, "train/clean_loss_weight": 43.822200775146484, "train/combined_loss": 0.37311995029449463, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.2495260663507109, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.31093329191207886 }, { "epoch": 2.4976303317535544, "step": 1054, "train/alpha": 0.2, "train/clean_ce_loss": 0.000675201416015625, "train/clean_loss_scale": 325.34051513671875, "train/clean_loss_weight": 65.06810760498047, "train/combined_loss": 0.26360446214675903, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.24976303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2196703851222992 }, { "epoch": 2.4976303317535544, "step": 1054, "train/alpha": 0.2, "train/clean_ce_loss": 0.0615234375, "train/clean_loss_scale": 9.62270736694336, "train/clean_loss_weight": 1.9245414733886719, "train/combined_loss": 0.7104264497756958, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.24976303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5920220613479614 }, { "epoch": 2.4976303317535544, "step": 1054, "train/alpha": 0.2, "train/clean_ce_loss": 0.06201171875, "train/clean_loss_scale": 11.906076431274414, "train/clean_loss_weight": 2.3812153339385986, "train/combined_loss": 0.8859794735908508, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.24976303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7383162379264832 }, { "epoch": 2.4976303317535544, "step": 1054, "train/alpha": 0.2, "train/clean_ce_loss": 0.078125, "train/clean_loss_scale": 9.122467041015625, "train/clean_loss_weight": 1.824493408203125, "train/combined_loss": 0.8552312850952148, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.24976303317535545, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7126927375793457 }, { "epoch": 2.5, "grad_norm": 33.10136795043945, "learning_rate": 2e-06, "loss": 3.5504, "step": 1055 }, { "epoch": 2.5, "step": 1055, "train/alpha": 0.2, "train/clean_ce_loss": 0.043701171875, "train/clean_loss_scale": 25.77921485900879, "train/clean_loss_weight": 5.155843257904053, "train/combined_loss": 1.3518983125686646, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.25, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1265819072723389 }, { "epoch": 2.5, "step": 1055, "train/alpha": 0.2, "train/clean_ce_loss": 0.04150390625, "train/clean_loss_scale": 22.75029945373535, "train/clean_loss_weight": 4.550059795379639, "train/combined_loss": 1.1330715417861938, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.25, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9442263245582581 }, { "epoch": 2.5, "step": 1055, "train/alpha": 0.2, "train/clean_ce_loss": 0.00099945068359375, "train/clean_loss_scale": 280.66473388671875, "train/clean_loss_weight": 56.1329460144043, "train/combined_loss": 0.3366127014160156, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.25, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2805105745792389 }, { "epoch": 2.5, "step": 1055, "train/alpha": 0.2, "train/clean_ce_loss": 0.07861328125, "train/clean_loss_scale": 12.631906509399414, "train/clean_loss_weight": 2.526381254196167, "train/combined_loss": 1.1916427612304688, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.25, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.993035614490509 }, { "epoch": 2.5023696682464456, "step": 1056, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 9.357552528381348, "train/clean_loss_weight": 1.8715105056762695, "train/combined_loss": 0.795026421546936, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.2502369668246445, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.66252201795578 }, { "epoch": 2.5023696682464456, "step": 1056, "train/alpha": 0.2, "train/clean_ce_loss": 0.010986328125, "train/clean_loss_scale": 31.638626098632812, "train/clean_loss_weight": 6.327725410461426, "train/combined_loss": 0.417110800743103, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.2502369668246445, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3475923240184784 }, { "epoch": 2.5023696682464456, "step": 1056, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 9.156878471374512, "train/clean_loss_weight": 1.8313757181167603, "train/combined_loss": 0.8155344724655151, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.2502369668246445, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6796120405197144 }, { "epoch": 2.5023696682464456, "step": 1056, "train/alpha": 0.2, "train/clean_ce_loss": 0.06103515625, "train/clean_loss_scale": 15.05799388885498, "train/clean_loss_weight": 3.011598825454712, "train/combined_loss": 1.1028804779052734, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.2502369668246445, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9190670251846313 }, { "epoch": 2.504739336492891, "step": 1057, "train/alpha": 0.2, "train/clean_ce_loss": 0.03369140625, "train/clean_loss_scale": 31.094202041625977, "train/clean_loss_weight": 6.218840599060059, "train/combined_loss": 1.2571289539337158, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.2504739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.047607421875 }, { "epoch": 2.504739336492891, "step": 1057, "train/alpha": 0.2, "train/clean_ce_loss": 0.03515625, "train/clean_loss_scale": 15.38292407989502, "train/clean_loss_weight": 3.076584815979004, "train/combined_loss": 0.6489671468734741, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.2504739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5408059358596802 }, { "epoch": 2.504739336492891, "step": 1057, "train/alpha": 0.2, "train/clean_ce_loss": 0.056884765625, "train/clean_loss_scale": 12.48250961303711, "train/clean_loss_weight": 2.496501922607422, "train/combined_loss": 0.8520776033401489, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.2504739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7100646495819092 }, { "epoch": 2.504739336492891, "step": 1057, "train/alpha": 0.2, "train/clean_ce_loss": 0.087890625, "train/clean_loss_scale": 9.443026542663574, "train/clean_loss_weight": 1.8886053562164307, "train/combined_loss": 0.9959442019462585, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.2504739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8299534916877747 }, { "epoch": 2.5071090047393367, "step": 1058, "train/alpha": 0.2, "train/clean_ce_loss": 0.0033111572265625, "train/clean_loss_scale": 68.10938262939453, "train/clean_loss_weight": 13.62187671661377, "train/combined_loss": 0.2706250548362732, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.25071090047393363, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.22552087903022766 }, { "epoch": 2.5071090047393367, "step": 1058, "train/alpha": 0.2, "train/clean_ce_loss": 0.061767578125, "train/clean_loss_scale": 9.276583671569824, "train/clean_loss_weight": 1.8553167581558228, "train/combined_loss": 0.687590479850769, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.25071090047393363, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5729920864105225 }, { "epoch": 2.5071090047393367, "step": 1058, "train/alpha": 0.2, "train/clean_ce_loss": 0.087890625, "train/clean_loss_scale": 9.686273574829102, "train/clean_loss_weight": 1.937254786491394, "train/combined_loss": 1.0215991735458374, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.25071090047393363, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8513326644897461 }, { "epoch": 2.5071090047393367, "step": 1058, "train/alpha": 0.2, "train/clean_ce_loss": 0.1181640625, "train/clean_loss_scale": 8.826891899108887, "train/clean_loss_weight": 1.7653783559799194, "train/combined_loss": 1.2516257762908936, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.25071090047393363, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0430214405059814 }, { "epoch": 2.509478672985782, "step": 1059, "train/alpha": 0.2, "train/clean_ce_loss": 0.059326171875, "train/clean_loss_scale": 16.98919677734375, "train/clean_loss_weight": 3.397839307785034, "train/combined_loss": 1.2094848155975342, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.2509478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.007904052734375 }, { "epoch": 2.509478672985782, "step": 1059, "train/alpha": 0.2, "train/clean_ce_loss": 0.08740234375, "train/clean_loss_scale": 10.740714073181152, "train/clean_loss_weight": 2.1481428146362305, "train/combined_loss": 1.126516342163086, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.2509478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9387636184692383 }, { "epoch": 2.509478672985782, "step": 1059, "train/alpha": 0.2, "train/clean_ce_loss": 0.0126953125, "train/clean_loss_scale": 29.568391799926758, "train/clean_loss_weight": 5.9136786460876465, "train/combined_loss": 0.4504559636116028, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.2509478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3753799796104431 }, { "epoch": 2.509478672985782, "step": 1059, "train/alpha": 0.2, "train/clean_ce_loss": 0.11083984375, "train/clean_loss_scale": 9.254658699035645, "train/clean_loss_weight": 1.8509317636489868, "train/combined_loss": 1.2309420108795166, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.2509478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.025784969329834 }, { "epoch": 2.5118483412322274, "grad_norm": 21.07581901550293, "learning_rate": 2e-06, "loss": 3.8596, "step": 1060 }, { "epoch": 2.5118483412322274, "step": 1060, "train/alpha": 0.2, "train/clean_ce_loss": 0.02001953125, "train/clean_loss_scale": 39.39238739013672, "train/clean_loss_weight": 7.878477573394775, "train/combined_loss": 0.9463405609130859, "train/pos_loss/pos0": 0.15457767248153687, "train/pos_loss/pos1": 0.4410225749015808, "train/pos_loss/pos2": 1.0140290260314941, "train/pos_loss/pos3": 1.4779698848724365, "train/pos_loss/pos4": 1.7065671682357788, "train/pos_loss/pos5": 2.601518392562866, "train/pos_loss/pos6": 3.0547444820404053, "train/pos_loss/weight0": 2.2957754135131836, "train/pos_loss/weight1": 1.7420742511749268, "train/pos_loss/weight2": 1.251154899597168, "train/pos_loss/weight3": 0.839846134185791, "train/pos_loss/weight4": 0.4873044788837433, "train/pos_loss/weight5": 0.24524955451488495, "train/pos_loss/weight6": 0.13859543204307556, "train/pos_weight_progress": 0.25118483412322273, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7886171340942383 }, { "epoch": 2.5118483412322274, "step": 1060, "train/alpha": 0.2, "train/clean_ce_loss": 0.0595703125, "train/clean_loss_scale": 9.341279029846191, "train/clean_loss_weight": 1.868255853652954, "train/combined_loss": 0.6677554845809937, "train/pos_loss/pos0": 0.21235379576683044, "train/pos_loss/pos1": 0.4109592139720917, "train/pos_loss/pos2": 0.7061788439750671, "train/pos_loss/pos3": 1.0213004350662231, "train/pos_loss/pos4": 1.5621742010116577, "train/pos_loss/pos5": 1.8703945875167847, "train/pos_loss/pos6": 2.262457847595215, "train/pos_loss/weight0": 2.5896077156066895, "train/pos_loss/weight1": 1.8703054189682007, "train/pos_loss/weight2": 1.2852469682693481, "train/pos_loss/weight3": 0.7036606669425964, "train/pos_loss/weight4": 0.341678649187088, "train/pos_loss/weight5": 0.14454764127731323, "train/pos_loss/weight6": 0.0649532601237297, "train/pos_weight_progress": 0.25118483412322273, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5564628839492798 }, { "epoch": 2.5118483412322274, "step": 1060, "train/alpha": 0.2, "train/clean_ce_loss": 0.0673828125, "train/clean_loss_scale": 13.790416717529297, "train/clean_loss_weight": 2.7580833435058594, "train/combined_loss": 1.1150845289230347, "train/pos_loss/pos0": 0.3419688642024994, "train/pos_loss/pos1": 0.72895747423172, "train/pos_loss/pos2": 1.144867181777954, "train/pos_loss/pos3": 1.5629061460494995, "train/pos_loss/pos4": 2.020350933074951, "train/pos_loss/pos5": 2.389582872390747, "train/pos_loss/pos6": 2.7879459857940674, "train/pos_loss/weight0": 2.4213366508483887, "train/pos_loss/weight1": 1.7937535047531128, "train/pos_loss/weight2": 1.2381930351257324, "train/pos_loss/weight3": 0.7653918266296387, "train/pos_loss/weight4": 0.4347532391548157, "train/pos_loss/weight5": 0.2226467728614807, "train/pos_loss/weight6": 0.12392492592334747, "train/pos_weight_progress": 0.25118483412322273, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9292370676994324 }, { "epoch": 2.5118483412322274, "step": 1060, "train/alpha": 0.2, "train/clean_ce_loss": 0.09716796875, "train/clean_loss_scale": 10.128551483154297, "train/clean_loss_weight": 2.025710344314575, "train/combined_loss": 1.1810050010681152, "train/pos_loss/pos0": 0.4555307626724243, "train/pos_loss/pos1": 0.9116503596305847, "train/pos_loss/pos2": 1.300934910774231, "train/pos_loss/pos3": 1.8160724639892578, "train/pos_loss/pos4": 2.2208924293518066, "train/pos_loss/pos5": 2.6861112117767334, "train/pos_loss/pos6": 3.1018476486206055, "train/pos_loss/weight0": 2.755239248275757, "train/pos_loss/weight1": 1.9269599914550781, "train/pos_loss/weight2": 1.2097562551498413, "train/pos_loss/weight3": 0.6392967700958252, "train/pos_loss/weight4": 0.2965708076953888, "train/pos_loss/weight5": 0.11996015906333923, "train/pos_loss/weight6": 0.052217260003089905, "train/pos_weight_progress": 0.25118483412322273, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9841707944869995 }, { "epoch": 2.514218009478673, "step": 1061, "train/alpha": 0.2, "train/clean_ce_loss": 0.006561279296875, "train/clean_loss_scale": 21.992679595947266, "train/clean_loss_weight": 4.398536205291748, "train/combined_loss": 0.1731601357460022, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.2514218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.14430011808872223 }, { "epoch": 2.514218009478673, "step": 1061, "train/alpha": 0.2, "train/clean_ce_loss": 0.019287109375, "train/clean_loss_scale": 6.04779052734375, "train/clean_loss_weight": 1.209558129310608, "train/combined_loss": 0.13997328281402588, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.2514218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.11664439737796783 }, { "epoch": 2.514218009478673, "step": 1061, "train/alpha": 0.2, "train/clean_ce_loss": 0.0771484375, "train/clean_loss_scale": 10.602259635925293, "train/clean_loss_weight": 2.1204519271850586, "train/combined_loss": 0.9815372824668884, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.2514218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8179477453231812 }, { "epoch": 2.514218009478673, "step": 1061, "train/alpha": 0.2, "train/clean_ce_loss": 0.049072265625, "train/clean_loss_scale": 7.355406284332275, "train/clean_loss_weight": 1.471081256866455, "train/combined_loss": 0.4331357479095459, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.2514218009478673, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3609464466571808 }, { "epoch": 2.5165876777251186, "step": 1062, "train/alpha": 0.2, "train/clean_ce_loss": 8.106231689453125e-06, "train/clean_loss_scale": 8731.130859375, "train/clean_loss_weight": 1746.2261962890625, "train/combined_loss": 0.084931880235672, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.25165876777251184, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.07077656686306 }, { "epoch": 2.5165876777251186, "step": 1062, "train/alpha": 0.2, "train/clean_ce_loss": 0.00010728836059570312, "train/clean_loss_scale": 1315.2911376953125, "train/clean_loss_weight": 263.0582275390625, "train/combined_loss": 0.16933850944042206, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.25165876777251184, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.14111542701721191 }, { "epoch": 2.5165876777251186, "step": 1062, "train/alpha": 0.2, "train/clean_ce_loss": 0.0537109375, "train/clean_loss_scale": 7.964334011077881, "train/clean_loss_weight": 1.5928667783737183, "train/combined_loss": 0.5133262276649475, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.25165876777251184, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.42777183651924133 }, { "epoch": 2.5165876777251186, "step": 1062, "train/alpha": 0.2, "train/clean_ce_loss": 0.003387451171875, "train/clean_loss_scale": 36.809104919433594, "train/clean_loss_weight": 7.361821174621582, "train/combined_loss": 0.14962686598300934, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.25165876777251184, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.12468905001878738 }, { "epoch": 2.5189573459715637, "step": 1063, "train/alpha": 0.2, "train/clean_ce_loss": 0.04833984375, "train/clean_loss_scale": 10.383232116699219, "train/clean_loss_weight": 2.076646566390991, "train/combined_loss": 0.6023085713386536, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.2518957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5019237995147705 }, { "epoch": 2.5189573459715637, "step": 1063, "train/alpha": 0.2, "train/clean_ce_loss": 0.06005859375, "train/clean_loss_scale": 7.095283508300781, "train/clean_loss_weight": 1.41905677318573, "train/combined_loss": 0.5113592743873596, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.2518957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4261327385902405 }, { "epoch": 2.5189573459715637, "step": 1063, "train/alpha": 0.2, "train/clean_ce_loss": 0.011474609375, "train/clean_loss_scale": 19.835805892944336, "train/clean_loss_weight": 3.967161178588867, "train/combined_loss": 0.27312976121902466, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.2518957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.22760812938213348 }, { "epoch": 2.5189573459715637, "step": 1063, "train/alpha": 0.2, "train/clean_ce_loss": 0.022216796875, "train/clean_loss_scale": 8.734505653381348, "train/clean_loss_weight": 1.7469011545181274, "train/combined_loss": 0.23286329209804535, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.2518957345971564, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.19405274093151093 }, { "epoch": 2.5213270142180093, "step": 1064, "train/alpha": 0.2, "train/clean_ce_loss": 0.078125, "train/clean_loss_scale": 10.737794876098633, "train/clean_loss_weight": 2.1475589275360107, "train/combined_loss": 1.006668210029602, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.25213270142180094, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8388901948928833 }, { "epoch": 2.5213270142180093, "step": 1064, "train/alpha": 0.2, "train/clean_ce_loss": 0.130859375, "train/clean_loss_scale": 6.395988941192627, "train/clean_loss_weight": 1.2791978120803833, "train/combined_loss": 1.004370093345642, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.25213270142180094, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.83697509765625 }, { "epoch": 2.5213270142180093, "step": 1064, "train/alpha": 0.2, "train/clean_ce_loss": 0.006744384765625, "train/clean_loss_scale": 9.223373413085938, "train/clean_loss_weight": 1.8446747064590454, "train/combined_loss": 0.07464717328548431, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.25213270142180094, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.06220598146319389 }, { "epoch": 2.5213270142180093, "step": 1064, "train/alpha": 0.2, "train/clean_ce_loss": 0.050537109375, "train/clean_loss_scale": 15.700704574584961, "train/clean_loss_weight": 3.140141010284424, "train/combined_loss": 0.9521619081497192, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.25213270142180094, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7934682369232178 }, { "epoch": 2.523696682464455, "grad_norm": 13.124574661254883, "learning_rate": 2e-06, "loss": 2.9896, "step": 1065 }, { "epoch": 2.523696682464455, "step": 1065, "train/alpha": 0.2, "train/clean_ce_loss": 0.05029296875, "train/clean_loss_scale": 4.326066970825195, "train/clean_loss_weight": 0.8652133941650391, "train/combined_loss": 0.2610849142074585, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.2523696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.21757075190544128 }, { "epoch": 2.523696682464455, "step": 1065, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 6.910017490386963, "train/clean_loss_weight": 1.3820035457611084, "train/combined_loss": 0.534446656703949, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.2523696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.44537222385406494 }, { "epoch": 2.523696682464455, "step": 1065, "train/alpha": 0.2, "train/clean_ce_loss": 0.0947265625, "train/clean_loss_scale": 7.259444236755371, "train/clean_loss_weight": 1.451888918876648, "train/combined_loss": 0.8251945972442627, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.2523696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6876621842384338 }, { "epoch": 2.523696682464455, "step": 1065, "train/alpha": 0.2, "train/clean_ce_loss": 0.095703125, "train/clean_loss_scale": 6.706770896911621, "train/clean_loss_weight": 1.341354250907898, "train/combined_loss": 0.7702307105064392, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.2523696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6418589353561401 }, { "epoch": 2.5260663507109005, "step": 1066, "train/alpha": 0.2, "train/clean_ce_loss": 0.083984375, "train/clean_loss_scale": 10.648150444030762, "train/clean_loss_weight": 2.1296300888061523, "train/combined_loss": 1.073133945465088, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.25260663507109005, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8942782878875732 }, { "epoch": 2.5260663507109005, "step": 1066, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 8.326395988464355, "train/clean_loss_weight": 1.6652792692184448, "train/combined_loss": 0.8098721504211426, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.25260663507109005, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6748934388160706 }, { "epoch": 2.5260663507109005, "step": 1066, "train/alpha": 0.2, "train/clean_ce_loss": 0.11279296875, "train/clean_loss_scale": 6.164907455444336, "train/clean_loss_weight": 1.232981562614441, "train/combined_loss": 0.8344298601150513, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.25260663507109005, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6953582167625427 }, { "epoch": 2.5260663507109005, "step": 1066, "train/alpha": 0.2, "train/clean_ce_loss": 0.019775390625, "train/clean_loss_scale": 7.5430216789245605, "train/clean_loss_weight": 1.5086044073104858, "train/combined_loss": 0.17899943888187408, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.25260663507109005, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.14916619658470154 }, { "epoch": 2.528436018957346, "step": 1067, "train/alpha": 0.2, "train/clean_ce_loss": 0.01409912109375, "train/clean_loss_scale": 12.27204704284668, "train/clean_loss_weight": 2.45440936088562, "train/combined_loss": 0.20763008296489716, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.25284360189573457, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.17302507162094116 }, { "epoch": 2.528436018957346, "step": 1067, "train/alpha": 0.2, "train/clean_ce_loss": 0.0419921875, "train/clean_loss_scale": 19.940160751342773, "train/clean_loss_weight": 3.988032102584839, "train/combined_loss": 1.0047972202301025, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.25284360189573457, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8373309969902039 }, { "epoch": 2.528436018957346, "step": 1067, "train/alpha": 0.2, "train/clean_ce_loss": 0.119140625, "train/clean_loss_scale": 8.829804420471191, "train/clean_loss_weight": 1.765960931777954, "train/combined_loss": 1.2623860836029053, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.25284360189573457, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0519883632659912 }, { "epoch": 2.528436018957346, "step": 1067, "train/alpha": 0.2, "train/clean_ce_loss": 0.1005859375, "train/clean_loss_scale": 6.815918922424316, "train/clean_loss_weight": 1.363183856010437, "train/combined_loss": 0.8227027654647827, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.25284360189573457, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.685585618019104 }, { "epoch": 2.5308056872037916, "step": 1068, "train/alpha": 0.2, "train/clean_ce_loss": 0.07373046875, "train/clean_loss_scale": 11.690526962280273, "train/clean_loss_weight": 2.3381054401397705, "train/combined_loss": 1.0343376398086548, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.25308056872037915, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8619480133056641 }, { "epoch": 2.5308056872037916, "step": 1068, "train/alpha": 0.2, "train/clean_ce_loss": 0.004364013671875, "train/clean_loss_scale": 22.081398010253906, "train/clean_loss_weight": 4.4162797927856445, "train/combined_loss": 0.11563622951507568, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.25308056872037915, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.09636352211236954 }, { "epoch": 2.5308056872037916, "step": 1068, "train/alpha": 0.2, "train/clean_ce_loss": 0.052001953125, "train/clean_loss_scale": 7.484936714172363, "train/clean_loss_weight": 1.4969873428344727, "train/combined_loss": 0.4670775830745697, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.25308056872037915, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3892313241958618 }, { "epoch": 2.5308056872037916, "step": 1068, "train/alpha": 0.2, "train/clean_ce_loss": 0.0032196044921875, "train/clean_loss_scale": 63.647621154785156, "train/clean_loss_weight": 12.729524612426758, "train/combined_loss": 0.24590420722961426, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.25308056872037915, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.20492017269134521 }, { "epoch": 2.533175355450237, "step": 1069, "train/alpha": 0.2, "train/clean_ce_loss": 0.10595703125, "train/clean_loss_scale": 8.292828559875488, "train/clean_loss_weight": 1.6585657596588135, "train/combined_loss": 1.0544201135635376, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.2533175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8786834478378296 }, { "epoch": 2.533175355450237, "step": 1069, "train/alpha": 0.2, "train/clean_ce_loss": 0.0478515625, "train/clean_loss_scale": 10.047097206115723, "train/clean_loss_weight": 2.0094194412231445, "train/combined_loss": 0.5769231915473938, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.2533175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4807693064212799 }, { "epoch": 2.533175355450237, "step": 1069, "train/alpha": 0.2, "train/clean_ce_loss": 0.10107421875, "train/clean_loss_scale": 7.695292949676514, "train/clean_loss_weight": 1.5390585660934448, "train/combined_loss": 0.9333548545837402, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.2533175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7777957320213318 }, { "epoch": 2.533175355450237, "step": 1069, "train/alpha": 0.2, "train/clean_ce_loss": 0.050537109375, "train/clean_loss_scale": 3.908717632293701, "train/clean_loss_weight": 0.7817435264587402, "train/combined_loss": 0.2370423525571823, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.2533175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.19753529131412506 }, { "epoch": 2.5355450236966823, "grad_norm": 29.75287628173828, "learning_rate": 2e-06, "loss": 3.0114, "step": 1070 }, { "epoch": 2.5355450236966823, "step": 1070, "train/alpha": 0.2, "train/clean_ce_loss": 0.035400390625, "train/clean_loss_scale": 7.458056449890137, "train/clean_loss_weight": 1.491611361503601, "train/combined_loss": 0.31682175397872925, "train/pos_loss/pos0": 0.04488889127969742, "train/pos_loss/pos1": 0.1928994506597519, "train/pos_loss/pos2": 0.38364899158477783, "train/pos_loss/pos3": 0.7802473306655884, "train/pos_loss/pos4": 1.1660966873168945, "train/pos_loss/pos5": 1.614699125289917, "train/pos_loss/pos6": 2.062398910522461, "train/pos_loss/weight0": 2.960486650466919, "train/pos_loss/weight1": 1.9976664781570435, "train/pos_loss/weight2": 1.1455821990966797, "train/pos_loss/weight3": 0.5587866902351379, "train/pos_loss/weight4": 0.22942635416984558, "train/pos_loss/weight5": 0.07914704084396362, "train/pos_loss/weight6": 0.028903957456350327, "train/pos_weight_progress": 0.25355450236966826, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.26401811838150024 }, { "epoch": 2.5355450236966823, "step": 1070, "train/alpha": 0.2, "train/clean_ce_loss": 0.0849609375, "train/clean_loss_scale": 9.414140701293945, "train/clean_loss_weight": 1.8828281164169312, "train/combined_loss": 0.9598010778427124, "train/pos_loss/pos0": 0.27756765484809875, "train/pos_loss/pos1": 0.5162116289138794, "train/pos_loss/pos2": 0.8336219191551208, "train/pos_loss/pos3": 1.222876787185669, "train/pos_loss/pos4": 1.5797685384750366, "train/pos_loss/pos5": 1.9520198106765747, "train/pos_loss/pos6": 2.1112844944000244, "train/pos_loss/weight0": 2.09263014793396, "train/pos_loss/weight1": 1.6583445072174072, "train/pos_loss/weight2": 1.2209138870239258, "train/pos_loss/weight3": 0.8772717118263245, "train/pos_loss/weight4": 0.5671937465667725, "train/pos_loss/weight5": 0.34851810336112976, "train/pos_loss/weight6": 0.23512791097164154, "train/pos_weight_progress": 0.25355450236966826, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7998342514038086 }, { "epoch": 2.5355450236966823, "step": 1070, "train/alpha": 0.2, "train/clean_ce_loss": 0.060546875, "train/clean_loss_scale": 9.077061653137207, "train/clean_loss_weight": 1.8154124021530151, "train/combined_loss": 0.6595052480697632, "train/pos_loss/pos0": 0.17146652936935425, "train/pos_loss/pos1": 0.35836532711982727, "train/pos_loss/pos2": 0.6163486242294312, "train/pos_loss/pos3": 0.9578933119773865, "train/pos_loss/pos4": 1.433430790901184, "train/pos_loss/pos5": 1.7451088428497314, "train/pos_loss/pos6": 2.0973386764526367, "train/pos_loss/weight0": 2.4189109802246094, "train/pos_loss/weight1": 1.7874271869659424, "train/pos_loss/weight2": 1.2275856733322144, "train/pos_loss/weight3": 0.7719400525093079, "train/pos_loss/weight4": 0.43003049492836, "train/pos_loss/weight5": 0.2356196939945221, "train/pos_loss/weight6": 0.1284855753183365, "train/pos_weight_progress": 0.25355450236966826, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5495877265930176 }, { "epoch": 2.5355450236966823, "step": 1070, "train/alpha": 0.2, "train/clean_ce_loss": 0.00555419921875, "train/clean_loss_scale": 45.25175094604492, "train/clean_loss_weight": 9.050350189208984, "train/combined_loss": 0.301604688167572, "train/pos_loss/pos0": 0.012835354544222355, "train/pos_loss/pos1": 0.030837377533316612, "train/pos_loss/pos2": 0.14002542197704315, "train/pos_loss/pos3": 0.38364458084106445, "train/pos_loss/pos4": 0.9673059582710266, "train/pos_loss/pos5": 1.6055657863616943, "train/pos_loss/pos6": 2.060197591781616, "train/pos_loss/weight0": 2.313725233078003, "train/pos_loss/weight1": 1.7666248083114624, "train/pos_loss/weight2": 1.2297098636627197, "train/pos_loss/weight3": 0.7990201711654663, "train/pos_loss/weight4": 0.47603586316108704, "train/pos_loss/weight5": 0.25998184084892273, "train/pos_loss/weight6": 0.15490204095840454, "train/pos_weight_progress": 0.25355450236966826, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2513372302055359 }, { "epoch": 2.537914691943128, "step": 1071, "train/alpha": 0.2, "train/clean_ce_loss": 0.1572265625, "train/clean_loss_scale": 7.73554801940918, "train/clean_loss_weight": 1.547109603881836, "train/combined_loss": 1.4594802856445312, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.2537914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2162336111068726 }, { "epoch": 2.537914691943128, "step": 1071, "train/alpha": 0.2, "train/clean_ce_loss": 0.0191650390625, "train/clean_loss_scale": 27.089035034179688, "train/clean_loss_weight": 5.417807102203369, "train/combined_loss": 0.6229948997497559, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.2537914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5191624164581299 }, { "epoch": 2.537914691943128, "step": 1071, "train/alpha": 0.2, "train/clean_ce_loss": 0.0576171875, "train/clean_loss_scale": 12.924569129943848, "train/clean_loss_weight": 2.584913969039917, "train/combined_loss": 0.8936127424240112, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.2537914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7446773052215576 }, { "epoch": 2.537914691943128, "step": 1071, "train/alpha": 0.2, "train/clean_ce_loss": 0.09619140625, "train/clean_loss_scale": 10.146121978759766, "train/clean_loss_weight": 2.029224395751953, "train/combined_loss": 1.1711636781692505, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.2537914691943128, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9759697318077087 }, { "epoch": 2.5402843601895735, "step": 1072, "train/alpha": 0.2, "train/clean_ce_loss": 0.05419921875, "train/clean_loss_scale": 18.826324462890625, "train/clean_loss_weight": 3.7652649879455566, "train/combined_loss": 1.224446415901184, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.25402843601895736, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0203720331192017 }, { "epoch": 2.5402843601895735, "step": 1072, "train/alpha": 0.2, "train/clean_ce_loss": 0.08544921875, "train/clean_loss_scale": 2.765273094177246, "train/clean_loss_weight": 0.5530546307563782, "train/combined_loss": 0.283548504114151, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.25402843601895736, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.23629042506217957 }, { "epoch": 2.5402843601895735, "step": 1072, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 13.208256721496582, "train/clean_loss_weight": 2.6416513919830322, "train/combined_loss": 1.2847093343734741, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.25402843601895736, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0705910921096802 }, { "epoch": 2.5402843601895735, "step": 1072, "train/alpha": 0.2, "train/clean_ce_loss": 0.11376953125, "train/clean_loss_scale": 10.654318809509277, "train/clean_loss_weight": 2.130863904953003, "train/combined_loss": 1.4545642137527466, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.25402843601895736, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2121368646621704 }, { "epoch": 2.542654028436019, "step": 1073, "train/alpha": 0.2, "train/clean_ce_loss": 0.078125, "train/clean_loss_scale": 9.18342113494873, "train/clean_loss_weight": 1.836684226989746, "train/combined_loss": 0.8609457612037659, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.2542654028436019, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7174547910690308 }, { "epoch": 2.542654028436019, "step": 1073, "train/alpha": 0.2, "train/clean_ce_loss": 0.056884765625, "train/clean_loss_scale": 10.583521842956543, "train/clean_loss_weight": 2.1167044639587402, "train/combined_loss": 0.7224494218826294, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.2542654028436019, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6020411849021912 }, { "epoch": 2.542654028436019, "step": 1073, "train/alpha": 0.2, "train/clean_ce_loss": 0.052490234375, "train/clean_loss_scale": 14.786216735839844, "train/clean_loss_weight": 2.9572434425354004, "train/combined_loss": 0.9313583970069885, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.2542654028436019, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7761319875717163 }, { "epoch": 2.542654028436019, "step": 1073, "train/alpha": 0.2, "train/clean_ce_loss": 0.154296875, "train/clean_loss_scale": 10.646611213684082, "train/clean_loss_weight": 2.1293222904205322, "train/combined_loss": 1.9712865352630615, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.2542654028436019, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.6427388191223145 }, { "epoch": 2.545023696682464, "step": 1074, "train/alpha": 0.2, "train/clean_ce_loss": 0.11767578125, "train/clean_loss_scale": 11.588228225708008, "train/clean_loss_weight": 2.317645788192749, "train/combined_loss": 1.6363844871520996, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.25450236966824646, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3636537790298462 }, { "epoch": 2.545023696682464, "step": 1074, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 12.4074125289917, "train/clean_loss_weight": 2.48148250579834, "train/combined_loss": 1.2140846252441406, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.25450236966824646, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0117372274398804 }, { "epoch": 2.545023696682464, "step": 1074, "train/alpha": 0.2, "train/clean_ce_loss": 0.10107421875, "train/clean_loss_scale": 16.87169647216797, "train/clean_loss_weight": 3.3743393421173096, "train/combined_loss": 2.0463521480560303, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.25450236966824646, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.7052935361862183 }, { "epoch": 2.545023696682464, "step": 1074, "train/alpha": 0.2, "train/clean_ce_loss": 0.146484375, "train/clean_loss_scale": 8.955242156982422, "train/clean_loss_weight": 1.7910484075546265, "train/combined_loss": 1.5741636753082275, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.25450236966824646, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3118031024932861 }, { "epoch": 2.5473933649289098, "grad_norm": 27.1143741607666, "learning_rate": 2e-06, "loss": 3.1431, "step": 1075 }, { "epoch": 2.5473933649289098, "step": 1075, "train/alpha": 0.2, "train/clean_ce_loss": 0.11083984375, "train/clean_loss_scale": 12.551252365112305, "train/clean_loss_weight": 2.5102505683898926, "train/combined_loss": 1.6694146394729614, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.254739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.391178846359253 }, { "epoch": 2.5473933649289098, "step": 1075, "train/alpha": 0.2, "train/clean_ce_loss": 0.06396484375, "train/clean_loss_scale": 11.792166709899902, "train/clean_loss_weight": 2.358433485031128, "train/combined_loss": 0.9051408767700195, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.254739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7542840838432312 }, { "epoch": 2.5473933649289098, "step": 1075, "train/alpha": 0.2, "train/clean_ce_loss": 0.048828125, "train/clean_loss_scale": 6.976357936859131, "train/clean_loss_weight": 1.3952716588974, "train/combined_loss": 0.40877097845077515, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.254739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3406424820423126 }, { "epoch": 2.5473933649289098, "step": 1075, "train/alpha": 0.2, "train/clean_ce_loss": 0.002227783203125, "train/clean_loss_scale": 75.98423767089844, "train/clean_loss_weight": 15.196847915649414, "train/combined_loss": 0.20313170552253723, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.254739336492891, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1692764163017273 }, { "epoch": 2.5497630331753554, "step": 1076, "train/alpha": 0.2, "train/clean_ce_loss": 0.08984375, "train/clean_loss_scale": 12.461638450622559, "train/clean_loss_weight": 2.4923276901245117, "train/combined_loss": 1.3435204029083252, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.25497630331753557, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1196002960205078 }, { "epoch": 2.5497630331753554, "step": 1076, "train/alpha": 0.2, "train/clean_ce_loss": 0.08447265625, "train/clean_loss_scale": 11.264430046081543, "train/clean_loss_weight": 2.2528860569000244, "train/combined_loss": 1.1418436765670776, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.25497630331753557, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9515363574028015 }, { "epoch": 2.5497630331753554, "step": 1076, "train/alpha": 0.2, "train/clean_ce_loss": 0.0186767578125, "train/clean_loss_scale": 19.200544357299805, "train/clean_loss_weight": 3.840108871459961, "train/combined_loss": 0.4303247034549713, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.25497630331753557, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.35860392451286316 }, { "epoch": 2.5497630331753554, "step": 1076, "train/alpha": 0.2, "train/clean_ce_loss": 0.08056640625, "train/clean_loss_scale": 11.507585525512695, "train/clean_loss_weight": 2.3015172481536865, "train/combined_loss": 1.1125497817993164, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.25497630331753557, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9271247982978821 }, { "epoch": 2.552132701421801, "step": 1077, "train/alpha": 0.2, "train/clean_ce_loss": 0.0260009765625, "train/clean_loss_scale": 15.276689529418945, "train/clean_loss_weight": 3.055337905883789, "train/combined_loss": 0.476650595664978, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.2552132701421801, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3972088396549225 }, { "epoch": 2.552132701421801, "step": 1077, "train/alpha": 0.2, "train/clean_ce_loss": 0.0908203125, "train/clean_loss_scale": 13.222371101379395, "train/clean_loss_weight": 2.6444742679595947, "train/combined_loss": 1.4410319328308105, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.2552132701421801, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2008599042892456 }, { "epoch": 2.552132701421801, "step": 1077, "train/alpha": 0.2, "train/clean_ce_loss": 0.03955078125, "train/clean_loss_scale": 9.789812088012695, "train/clean_loss_weight": 1.9579623937606812, "train/combined_loss": 0.46463367342948914, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.2552132701421801, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3871947228908539 }, { "epoch": 2.552132701421801, "step": 1077, "train/alpha": 0.2, "train/clean_ce_loss": 0.10302734375, "train/clean_loss_scale": 12.233721733093262, "train/clean_loss_weight": 2.446744441986084, "train/combined_loss": 1.5124893188476562, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.2552132701421801, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.26040780544281 }, { "epoch": 2.5545023696682465, "step": 1078, "train/alpha": 0.2, "train/clean_ce_loss": 0.1591796875, "train/clean_loss_scale": 9.816882133483887, "train/clean_loss_weight": 1.9633764028549194, "train/combined_loss": 1.8751779794692993, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.2554502369668246, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5626482963562012 }, { "epoch": 2.5545023696682465, "step": 1078, "train/alpha": 0.2, "train/clean_ce_loss": 0.03466796875, "train/clean_loss_scale": 31.449411392211914, "train/clean_loss_weight": 6.289882183074951, "train/combined_loss": 1.3083446025848389, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.2554502369668246, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.090287208557129 }, { "epoch": 2.5545023696682465, "step": 1078, "train/alpha": 0.2, "train/clean_ce_loss": 0.0380859375, "train/clean_loss_scale": 12.752421379089355, "train/clean_loss_weight": 2.5504844188690186, "train/combined_loss": 0.5828254818916321, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.2554502369668246, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.48568791151046753 }, { "epoch": 2.5545023696682465, "step": 1078, "train/alpha": 0.2, "train/clean_ce_loss": 0.1513671875, "train/clean_loss_scale": 10.096039772033691, "train/clean_loss_weight": 2.0192079544067383, "train/combined_loss": 1.8338508605957031, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.2554502369668246, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5282090902328491 }, { "epoch": 2.556872037914692, "step": 1079, "train/alpha": 0.2, "train/clean_ce_loss": 0.154296875, "train/clean_loss_scale": 8.884313583374023, "train/clean_loss_weight": 1.7768627405166626, "train/combined_loss": 1.6449861526489258, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.2556872037914692, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3708218336105347 }, { "epoch": 2.556872037914692, "step": 1079, "train/alpha": 0.2, "train/clean_ce_loss": 0.1279296875, "train/clean_loss_scale": 10.105059623718262, "train/clean_loss_weight": 2.0210120677948, "train/combined_loss": 1.5512845516204834, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.2556872037914692, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2927371263504028 }, { "epoch": 2.556872037914692, "step": 1079, "train/alpha": 0.2, "train/clean_ce_loss": 0.06884765625, "train/clean_loss_scale": 10.387439727783203, "train/clean_loss_weight": 2.0774879455566406, "train/combined_loss": 0.8581810593605042, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.2556872037914692, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7151508927345276 }, { "epoch": 2.556872037914692, "step": 1079, "train/alpha": 0.2, "train/clean_ce_loss": 0.11474609375, "train/clean_loss_scale": 12.405494689941406, "train/clean_loss_weight": 2.4810988903045654, "train/combined_loss": 1.7081785202026367, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.2556872037914692, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.423482060432434 }, { "epoch": 2.5592417061611377, "grad_norm": 16.291545867919922, "learning_rate": 2e-06, "loss": 3.3799, "step": 1080 }, { "epoch": 2.5592417061611377, "step": 1080, "train/alpha": 0.2, "train/clean_ce_loss": 0.0020751953125, "train/clean_loss_scale": 114.88518524169922, "train/clean_loss_weight": 22.97703742980957, "train/combined_loss": 0.28609102964401245, "train/pos_loss/pos0": 0.0002879572566598654, "train/pos_loss/pos1": 0.01651155576109886, "train/pos_loss/pos2": 0.05716977268457413, "train/pos_loss/pos3": 0.2590167224407196, "train/pos_loss/pos4": 0.6713982224464417, "train/pos_loss/pos5": 0.9553906917572021, "train/pos_loss/pos6": 1.7255644798278809, "train/pos_loss/weight0": 1.9919041395187378, "train/pos_loss/weight1": 1.5668708086013794, "train/pos_loss/weight2": 1.2061532735824585, "train/pos_loss/weight3": 0.8832272291183472, "train/pos_loss/weight4": 0.6320658326148987, "train/pos_loss/weight5": 0.4115302264690399, "train/pos_loss/weight6": 0.3082488179206848, "train/pos_weight_progress": 0.2559241706161137, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.23840919137001038 }, { "epoch": 2.5592417061611377, "step": 1080, "train/alpha": 0.2, "train/clean_ce_loss": 0.03857421875, "train/clean_loss_scale": 17.75836944580078, "train/clean_loss_weight": 3.5516738891601562, "train/combined_loss": 0.8220182657241821, "train/pos_loss/pos0": 0.14426691830158234, "train/pos_loss/pos1": 0.3159116506576538, "train/pos_loss/pos2": 0.5466848611831665, "train/pos_loss/pos3": 0.9238346815109253, "train/pos_loss/pos4": 1.2282540798187256, "train/pos_loss/pos5": 1.5496975183486938, "train/pos_loss/pos6": 1.8238669633865356, "train/pos_loss/weight0": 1.7350951433181763, "train/pos_loss/weight1": 1.4334598779678345, "train/pos_loss/weight2": 1.1542226076126099, "train/pos_loss/weight3": 0.9274107813835144, "train/pos_loss/weight4": 0.7303842306137085, "train/pos_loss/weight5": 0.5466011166572571, "train/pos_loss/weight6": 0.4728260934352875, "train/pos_weight_progress": 0.2559241706161137, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6850152015686035 }, { "epoch": 2.5592417061611377, "step": 1080, "train/alpha": 0.2, "train/clean_ce_loss": 0.05322265625, "train/clean_loss_scale": 10.2811279296875, "train/clean_loss_weight": 2.056225538253784, "train/combined_loss": 0.6566267013549805, "train/pos_loss/pos0": 0.0979422777891159, "train/pos_loss/pos1": 0.2781074643135071, "train/pos_loss/pos2": 0.6133461594581604, "train/pos_loss/pos3": 0.9093823432922363, "train/pos_loss/pos4": 1.2877132892608643, "train/pos_loss/pos5": 1.7629188299179077, "train/pos_loss/pos6": 2.2323081493377686, "train/pos_loss/weight0": 2.2117366790771484, "train/pos_loss/weight1": 1.7133514881134033, "train/pos_loss/weight2": 1.22318434715271, "train/pos_loss/weight3": 0.8421358466148376, "train/pos_loss/weight4": 0.5155445337295532, "train/pos_loss/weight5": 0.3056771755218506, "train/pos_loss/weight6": 0.1883702278137207, "train/pos_weight_progress": 0.2559241706161137, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.547188937664032 }, { "epoch": 2.5592417061611377, "step": 1080, "train/alpha": 0.2, "train/clean_ce_loss": 0.050537109375, "train/clean_loss_scale": 8.709207534790039, "train/clean_loss_weight": 1.7418415546417236, "train/combined_loss": 0.5281658172607422, "train/pos_loss/pos0": 0.04692170396447182, "train/pos_loss/pos1": 0.11712506413459778, "train/pos_loss/pos2": 0.3668488562107086, "train/pos_loss/pos3": 0.7428482174873352, "train/pos_loss/pos4": 1.6003235578536987, "train/pos_loss/pos5": 2.132889747619629, "train/pos_loss/pos6": 2.604762077331543, "train/pos_loss/weight0": 2.289665699005127, "train/pos_loss/weight1": 1.7604202032089233, "train/pos_loss/weight2": 1.255386471748352, "train/pos_loss/weight3": 0.8108896017074585, "train/pos_loss/weight4": 0.47127318382263184, "train/pos_loss/weight5": 0.2600664794445038, "train/pos_loss/weight6": 0.15229882299900055, "train/pos_weight_progress": 0.2559241706161137, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.44013819098472595 }, { "epoch": 2.561611374407583, "step": 1081, "train/alpha": 0.2, "train/clean_ce_loss": 0.0986328125, "train/clean_loss_scale": 10.574851036071777, "train/clean_loss_weight": 2.1149702072143555, "train/combined_loss": 1.2516326904296875, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.2561611374407583, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0430272817611694 }, { "epoch": 2.561611374407583, "step": 1081, "train/alpha": 0.2, "train/clean_ce_loss": 0.0038604736328125, "train/clean_loss_scale": 84.40699768066406, "train/clean_loss_weight": 16.881399154663086, "train/combined_loss": 0.391021192073822, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.2561611374407583, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3258509933948517 }, { "epoch": 2.561611374407583, "step": 1081, "train/alpha": 0.2, "train/clean_ce_loss": 0.059814453125, "train/clean_loss_scale": 10.084248542785645, "train/clean_loss_weight": 2.0168497562408447, "train/combined_loss": 0.7238205671310425, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.2561611374407583, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6031838059425354 }, { "epoch": 2.561611374407583, "step": 1081, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 11.587677955627441, "train/clean_loss_weight": 2.317535638809204, "train/combined_loss": 0.9844999313354492, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.2561611374407583, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8204166293144226 }, { "epoch": 2.5639810426540284, "step": 1082, "train/alpha": 0.2, "train/clean_ce_loss": 3.218650817871094e-05, "train/clean_loss_scale": 8016.263671875, "train/clean_loss_weight": 1603.2528076171875, "train/combined_loss": 0.30961865186691284, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.2563981042654028, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.25801554322242737 }, { "epoch": 2.5639810426540284, "step": 1082, "train/alpha": 0.2, "train/clean_ce_loss": 0.08984375, "train/clean_loss_scale": 8.213339805603027, "train/clean_loss_weight": 1.6426680088043213, "train/combined_loss": 0.8855006694793701, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.2563981042654028, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7379172444343567 }, { "epoch": 2.5639810426540284, "step": 1082, "train/alpha": 0.2, "train/clean_ce_loss": 0.01068115234375, "train/clean_loss_scale": 20.710336685180664, "train/clean_loss_weight": 4.1420674324035645, "train/combined_loss": 0.2654523253440857, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.2563981042654028, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2212102711200714 }, { "epoch": 2.5639810426540284, "step": 1082, "train/alpha": 0.2, "train/clean_ce_loss": 0.0498046875, "train/clean_loss_scale": 13.590898513793945, "train/clean_loss_weight": 2.718179702758789, "train/combined_loss": 0.8122684955596924, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.2563981042654028, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6768904328346252 }, { "epoch": 2.566350710900474, "step": 1083, "train/alpha": 0.2, "train/clean_ce_loss": 0.06982421875, "train/clean_loss_scale": 9.27955436706543, "train/clean_loss_weight": 1.8559108972549438, "train/combined_loss": 0.7775251865386963, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.2566350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6479376554489136 }, { "epoch": 2.566350710900474, "step": 1083, "train/alpha": 0.2, "train/clean_ce_loss": 0.034912109375, "train/clean_loss_scale": 12.748238563537598, "train/clean_loss_weight": 2.549647808074951, "train/combined_loss": 0.5340814590454102, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.2566350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4450678825378418 }, { "epoch": 2.566350710900474, "step": 1083, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 10.912891387939453, "train/clean_loss_weight": 2.1825783252716064, "train/combined_loss": 0.8504382371902466, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.2566350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7086985111236572 }, { "epoch": 2.566350710900474, "step": 1083, "train/alpha": 0.2, "train/clean_ce_loss": 0.025390625, "train/clean_loss_scale": 18.734012603759766, "train/clean_loss_weight": 3.746802568435669, "train/combined_loss": 0.5708019733428955, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.2566350710900474, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4756683111190796 }, { "epoch": 2.5687203791469195, "step": 1084, "train/alpha": 0.2, "train/clean_ce_loss": 0.015869140625, "train/clean_loss_scale": 17.25621795654297, "train/clean_loss_weight": 3.4512436389923096, "train/combined_loss": 0.3286096155643463, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.25687203791469193, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.27384135127067566 }, { "epoch": 2.5687203791469195, "step": 1084, "train/alpha": 0.2, "train/clean_ce_loss": 0.000270843505859375, "train/clean_loss_scale": 854.4268798828125, "train/clean_loss_weight": 170.8853759765625, "train/combined_loss": 0.27769917249679565, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.25687203791469193, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.23141597211360931 }, { "epoch": 2.5687203791469195, "step": 1084, "train/alpha": 0.2, "train/clean_ce_loss": 0.00701904296875, "train/clean_loss_scale": 31.60095977783203, "train/clean_loss_weight": 6.320191860198975, "train/combined_loss": 0.2661702036857605, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.25687203791469193, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.22180849313735962 }, { "epoch": 2.5687203791469195, "step": 1084, "train/alpha": 0.2, "train/clean_ce_loss": 0.04931640625, "train/clean_loss_scale": 9.969913482666016, "train/clean_loss_weight": 1.9939826726913452, "train/combined_loss": 0.5900163650512695, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.25687203791469193, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4916803240776062 }, { "epoch": 2.5710900473933647, "grad_norm": 33.03076934814453, "learning_rate": 2e-06, "loss": 3.3174, "step": 1085 }, { "epoch": 2.5710900473933647, "step": 1085, "train/alpha": 0.2, "train/clean_ce_loss": 0.06201171875, "train/clean_loss_scale": 8.905149459838867, "train/clean_loss_weight": 1.7810299396514893, "train/combined_loss": 0.6626683473587036, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.2571090047393365, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5522236227989197 }, { "epoch": 2.5710900473933647, "step": 1085, "train/alpha": 0.2, "train/clean_ce_loss": 0.1767578125, "train/clean_loss_scale": 8.329545021057129, "train/clean_loss_weight": 1.6659090518951416, "train/combined_loss": 1.7667745351791382, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.2571090047393365, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.472312092781067 }, { "epoch": 2.5710900473933647, "step": 1085, "train/alpha": 0.2, "train/clean_ce_loss": 0.044677734375, "train/clean_loss_scale": 3.2403724193573, "train/clean_loss_weight": 0.6480745077133179, "train/combined_loss": 0.17372700572013855, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.2571090047393365, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1447724997997284 }, { "epoch": 2.5710900473933647, "step": 1085, "train/alpha": 0.2, "train/clean_ce_loss": 0.0546875, "train/clean_loss_scale": 11.824155807495117, "train/clean_loss_weight": 2.3648312091827393, "train/combined_loss": 0.7759602069854736, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.2571090047393365, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.646633505821228 }, { "epoch": 2.5734597156398102, "step": 1086, "train/alpha": 0.2, "train/clean_ce_loss": 0.10302734375, "train/clean_loss_scale": 10.545296669006348, "train/clean_loss_weight": 2.1090593338012695, "train/combined_loss": 1.3037446737289429, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.25734597156398104, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.086453914642334 }, { "epoch": 2.5734597156398102, "step": 1086, "train/alpha": 0.2, "train/clean_ce_loss": 0.13671875, "train/clean_loss_scale": 11.34814167022705, "train/clean_loss_weight": 2.2696282863616943, "train/combined_loss": 1.861804485321045, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.25734597156398104, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5515037775039673 }, { "epoch": 2.5734597156398102, "step": 1086, "train/alpha": 0.2, "train/clean_ce_loss": 0.029296875, "train/clean_loss_scale": 17.665090560913086, "train/clean_loss_weight": 3.533018112182617, "train/combined_loss": 0.6210383176803589, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.25734597156398104, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5175319314002991 }, { "epoch": 2.5734597156398102, "step": 1086, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 10.520363807678223, "train/clean_loss_weight": 2.1040728092193604, "train/combined_loss": 1.0109412670135498, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.25734597156398104, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8424510359764099 }, { "epoch": 2.575829383886256, "step": 1087, "train/alpha": 0.2, "train/clean_ce_loss": 0.07666015625, "train/clean_loss_scale": 9.703417778015137, "train/clean_loss_weight": 1.9406836032867432, "train/combined_loss": 0.8926386833190918, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.2575829383886256, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7438655495643616 }, { "epoch": 2.575829383886256, "step": 1087, "train/alpha": 0.2, "train/clean_ce_loss": 0.0101318359375, "train/clean_loss_scale": 84.88695526123047, "train/clean_loss_weight": 16.977392196655273, "train/combined_loss": 1.0320727825164795, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.2575829383886256, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8600606918334961 }, { "epoch": 2.575829383886256, "step": 1087, "train/alpha": 0.2, "train/clean_ce_loss": 0.060302734375, "train/clean_loss_scale": 10.792108535766602, "train/clean_loss_weight": 2.158421754837036, "train/combined_loss": 0.7809523940086365, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.2575829383886256, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6507936716079712 }, { "epoch": 2.575829383886256, "step": 1087, "train/alpha": 0.2, "train/clean_ce_loss": 0.07568359375, "train/clean_loss_scale": 11.97612190246582, "train/clean_loss_weight": 2.3952243328094482, "train/combined_loss": 1.0876750946044922, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.2575829383886256, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9063959717750549 }, { "epoch": 2.5781990521327014, "step": 1088, "train/alpha": 0.2, "train/clean_ce_loss": 0.0145263671875, "train/clean_loss_scale": 18.569679260253906, "train/clean_loss_weight": 3.7139358520507812, "train/combined_loss": 0.323699951171875, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.25781990521327014, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2697499692440033 }, { "epoch": 2.5781990521327014, "step": 1088, "train/alpha": 0.2, "train/clean_ce_loss": 0.06689453125, "train/clean_loss_scale": 11.6300048828125, "train/clean_loss_weight": 2.326000928878784, "train/combined_loss": 0.9335804581642151, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.25781990521327014, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7779837250709534 }, { "epoch": 2.5781990521327014, "step": 1088, "train/alpha": 0.2, "train/clean_ce_loss": 0.1201171875, "train/clean_loss_scale": 9.36488151550293, "train/clean_loss_weight": 1.872976303100586, "train/combined_loss": 1.3498598337173462, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.25781990521327014, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1248831748962402 }, { "epoch": 2.5781990521327014, "step": 1088, "train/alpha": 0.2, "train/clean_ce_loss": 0.00885009765625, "train/clean_loss_scale": 45.82394027709961, "train/clean_loss_weight": 9.164788246154785, "train/combined_loss": 0.486655592918396, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.25781990521327014, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4055463373661041 }, { "epoch": 2.580568720379147, "step": 1089, "train/alpha": 0.2, "train/clean_ce_loss": 0.04443359375, "train/clean_loss_scale": 15.841365814208984, "train/clean_loss_weight": 3.1682732105255127, "train/combined_loss": 0.844666600227356, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.2580568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7038888335227966 }, { "epoch": 2.580568720379147, "step": 1089, "train/alpha": 0.2, "train/clean_ce_loss": 0.11083984375, "train/clean_loss_scale": 11.58410358428955, "train/clean_loss_weight": 2.3168208599090576, "train/combined_loss": 1.540776252746582, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.2580568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2839802503585815 }, { "epoch": 2.580568720379147, "step": 1089, "train/alpha": 0.2, "train/clean_ce_loss": 0.076171875, "train/clean_loss_scale": 15.66323471069336, "train/clean_loss_weight": 3.1326470375061035, "train/combined_loss": 1.4317175149917603, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.2580568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1930979490280151 }, { "epoch": 2.580568720379147, "step": 1089, "train/alpha": 0.2, "train/clean_ce_loss": 0.1591796875, "train/clean_loss_scale": 9.221884727478027, "train/clean_loss_weight": 1.8443769216537476, "train/combined_loss": 1.7615240812301636, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.2580568720379147, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4679367542266846 }, { "epoch": 2.5829383886255926, "grad_norm": 37.99967575073242, "learning_rate": 2e-06, "loss": 3.7239, "step": 1090 }, { "epoch": 2.5829383886255926, "step": 1090, "train/alpha": 0.2, "train/clean_ce_loss": 0.04736328125, "train/clean_loss_scale": 9.359016418457031, "train/clean_loss_weight": 1.8718032836914062, "train/combined_loss": 0.5319284796714783, "train/pos_loss/pos0": 0.036025866866111755, "train/pos_loss/pos1": 0.20711922645568848, "train/pos_loss/pos2": 0.28917941451072693, "train/pos_loss/pos3": 0.6076405644416809, "train/pos_loss/pos4": 1.4483164548873901, "train/pos_loss/pos5": 2.0712766647338867, "train/pos_loss/pos6": 2.551846504211426, "train/pos_loss/weight0": 2.2283263206481934, "train/pos_loss/weight1": 1.701249361038208, "train/pos_loss/weight2": 1.237164855003357, "train/pos_loss/weight3": 0.8459165096282959, "train/pos_loss/weight4": 0.5292280912399292, "train/pos_loss/weight5": 0.2855454683303833, "train/pos_loss/weight6": 0.17256946861743927, "train/pos_weight_progress": 0.25829383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.44327372312545776 }, { "epoch": 2.5829383886255926, "step": 1090, "train/alpha": 0.2, "train/clean_ce_loss": 0.08984375, "train/clean_loss_scale": 12.097742080688477, "train/clean_loss_weight": 2.419548511505127, "train/combined_loss": 1.3042879104614258, "train/pos_loss/pos0": 0.224154993891716, "train/pos_loss/pos1": 0.5430693030357361, "train/pos_loss/pos2": 1.0933119058609009, "train/pos_loss/pos3": 2.0113182067871094, "train/pos_loss/pos4": 2.761748790740967, "train/pos_loss/pos5": 2.7007827758789062, "train/pos_loss/pos6": 3.179020643234253, "train/pos_loss/weight0": 2.128265857696533, "train/pos_loss/weight1": 1.6592750549316406, "train/pos_loss/weight2": 1.2309027910232544, "train/pos_loss/weight3": 0.8570252060890198, "train/pos_loss/weight4": 0.573452353477478, "train/pos_loss/weight5": 0.34084752202033997, "train/pos_loss/weight6": 0.21023130416870117, "train/pos_weight_progress": 0.25829383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0869065523147583 }, { "epoch": 2.5829383886255926, "step": 1090, "train/alpha": 0.2, "train/clean_ce_loss": 0.09521484375, "train/clean_loss_scale": 9.050263404846191, "train/clean_loss_weight": 1.810052752494812, "train/combined_loss": 1.0340633392333984, "train/pos_loss/pos0": 0.1851259469985962, "train/pos_loss/pos1": 0.7701084017753601, "train/pos_loss/pos2": 1.4375584125518799, "train/pos_loss/pos3": 2.1123287677764893, "train/pos_loss/pos4": 2.6511597633361816, "train/pos_loss/pos5": 3.1945130825042725, "train/pos_loss/pos6": 3.5688986778259277, "train/pos_loss/weight0": 2.848343849182129, "train/pos_loss/weight1": 1.9971930980682373, "train/pos_loss/weight2": 1.2109410762786865, "train/pos_loss/weight3": 0.617606520652771, "train/pos_loss/weight4": 0.2354608178138733, "train/pos_loss/weight5": 0.06728282570838928, "train/pos_loss/weight6": 0.023172158747911453, "train/pos_weight_progress": 0.25829383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8617194294929504 }, { "epoch": 2.5829383886255926, "step": 1090, "train/alpha": 0.2, "train/clean_ce_loss": 0.05908203125, "train/clean_loss_scale": 11.61170482635498, "train/clean_loss_weight": 2.322340965270996, "train/combined_loss": 0.8232517242431641, "train/pos_loss/pos0": 0.21866880357265472, "train/pos_loss/pos1": 0.6350032091140747, "train/pos_loss/pos2": 1.0290268659591675, "train/pos_loss/pos3": 1.9058573246002197, "train/pos_loss/pos4": 2.5096747875213623, "train/pos_loss/pos5": 2.819333076477051, "train/pos_loss/pos6": 3.4521501064300537, "train/pos_loss/weight0": 2.9399280548095703, "train/pos_loss/weight1": 2.130134344100952, "train/pos_loss/weight2": 1.1726274490356445, "train/pos_loss/weight3": 0.5271251201629639, "train/pos_loss/weight4": 0.17589451372623444, "train/pos_loss/weight5": 0.0436415933072567, "train/pos_loss/weight6": 0.010649014264345169, "train/pos_weight_progress": 0.25829383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6860430836677551 }, { "epoch": 2.585308056872038, "step": 1091, "train/alpha": 0.2, "train/clean_ce_loss": 0.0027618408203125, "train/clean_loss_scale": 71.11216735839844, "train/clean_loss_weight": 14.222434043884277, "train/combined_loss": 0.23568059504032135, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.25853080568720377, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1964004933834076 }, { "epoch": 2.585308056872038, "step": 1091, "train/alpha": 0.2, "train/clean_ce_loss": 0.03662109375, "train/clean_loss_scale": 5.735336780548096, "train/clean_loss_weight": 1.1470674276351929, "train/combined_loss": 0.25204116106033325, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.25853080568720377, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2100343108177185 }, { "epoch": 2.585308056872038, "step": 1091, "train/alpha": 0.2, "train/clean_ce_loss": 0.033447265625, "train/clean_loss_scale": 4.721275806427002, "train/clean_loss_weight": 0.9442551732063293, "train/combined_loss": 0.18949651718139648, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.25853080568720377, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.15791375935077667 }, { "epoch": 2.585308056872038, "step": 1091, "train/alpha": 0.2, "train/clean_ce_loss": 0.11572265625, "train/clean_loss_scale": 6.910329818725586, "train/clean_loss_weight": 1.382066011428833, "train/combined_loss": 0.959618091583252, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.25853080568720377, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7996817231178284 }, { "epoch": 2.5876777251184833, "step": 1092, "train/alpha": 0.2, "train/clean_ce_loss": 0.1328125, "train/clean_loss_scale": 7.081966400146484, "train/clean_loss_weight": 1.4163932800292969, "train/combined_loss": 1.1286883354187012, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.25876777251184835, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9405736327171326 }, { "epoch": 2.5876777251184833, "step": 1092, "train/alpha": 0.2, "train/clean_ce_loss": 0.052001953125, "train/clean_loss_scale": 8.326894760131836, "train/clean_loss_weight": 1.6653789281845093, "train/combined_loss": 0.5196177959442139, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.25876777251184835, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43301481008529663 }, { "epoch": 2.5876777251184833, "step": 1092, "train/alpha": 0.2, "train/clean_ce_loss": 0.0002536773681640625, "train/clean_loss_scale": 315.4805908203125, "train/clean_loss_weight": 63.09611892700195, "train/combined_loss": 0.09603634476661682, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.25876777251184835, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.08003028482198715 }, { "epoch": 2.5876777251184833, "step": 1092, "train/alpha": 0.2, "train/clean_ce_loss": 0.05224609375, "train/clean_loss_scale": 7.426908016204834, "train/clean_loss_weight": 1.4853816032409668, "train/combined_loss": 0.4656323194503784, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.25876777251184835, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3880269229412079 }, { "epoch": 2.590047393364929, "step": 1093, "train/alpha": 0.2, "train/clean_ce_loss": 0.0020599365234375, "train/clean_loss_scale": 29.121793746948242, "train/clean_loss_weight": 5.824358940124512, "train/combined_loss": 0.0719868540763855, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.2590047393364929, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.059989046305418015 }, { "epoch": 2.590047393364929, "step": 1093, "train/alpha": 0.2, "train/clean_ce_loss": 0.03125, "train/clean_loss_scale": 18.42015266418457, "train/clean_loss_weight": 3.684030532836914, "train/combined_loss": 0.6907557249069214, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.2590047393364929, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5756297707557678 }, { "epoch": 2.590047393364929, "step": 1093, "train/alpha": 0.2, "train/clean_ce_loss": 0.058349609375, "train/clean_loss_scale": 4.808014869689941, "train/clean_loss_weight": 0.9616029858589172, "train/combined_loss": 0.3366549611091614, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.2590047393364929, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.28054580092430115 }, { "epoch": 2.590047393364929, "step": 1093, "train/alpha": 0.2, "train/clean_ce_loss": 0.01458740234375, "train/clean_loss_scale": 18.469301223754883, "train/clean_loss_weight": 3.6938602924346924, "train/combined_loss": 0.3233029544353485, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.2590047393364929, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2694191336631775 }, { "epoch": 2.5924170616113744, "step": 1094, "train/alpha": 0.2, "train/clean_ce_loss": 0.00469970703125, "train/clean_loss_scale": 42.931209564208984, "train/clean_loss_weight": 8.586241722106934, "train/combined_loss": 0.24211692810058594, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.25924170616113745, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.20176410675048828 }, { "epoch": 2.5924170616113744, "step": 1094, "train/alpha": 0.2, "train/clean_ce_loss": 0.12451171875, "train/clean_loss_scale": 7.7080979347229, "train/clean_loss_weight": 1.5416196584701538, "train/combined_loss": 1.1516982316970825, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.25924170616113745, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9597485065460205 }, { "epoch": 2.5924170616113744, "step": 1094, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 7.995746612548828, "train/clean_loss_weight": 1.5991493463516235, "train/combined_loss": 0.8011363744735718, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.25924170616113745, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6676136255264282 }, { "epoch": 2.5924170616113744, "step": 1094, "train/alpha": 0.2, "train/clean_ce_loss": 0.091796875, "train/clean_loss_scale": 6.897144794464111, "train/clean_loss_weight": 1.3794289827346802, "train/combined_loss": 0.7597635984420776, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.25924170616113745, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6331363320350647 }, { "epoch": 2.59478672985782, "grad_norm": 31.788955688476562, "learning_rate": 2e-06, "loss": 2.87, "step": 1095 }, { "epoch": 2.59478672985782, "step": 1095, "train/alpha": 0.2, "train/clean_ce_loss": 0.0306396484375, "train/clean_loss_scale": 8.388426780700684, "train/clean_loss_weight": 1.6776853799819946, "train/combined_loss": 0.30842214822769165, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.259478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.25701844692230225 }, { "epoch": 2.59478672985782, "step": 1095, "train/alpha": 0.2, "train/clean_ce_loss": 0.10302734375, "train/clean_loss_scale": 7.0804362297058105, "train/clean_loss_weight": 1.41608726978302, "train/combined_loss": 0.8753742575645447, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.259478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7294785380363464 }, { "epoch": 2.59478672985782, "step": 1095, "train/alpha": 0.2, "train/clean_ce_loss": 0.083984375, "train/clean_loss_scale": 7.290513038635254, "train/clean_loss_weight": 1.4581025838851929, "train/combined_loss": 0.734747052192688, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.259478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6122891902923584 }, { "epoch": 2.59478672985782, "step": 1095, "train/alpha": 0.2, "train/clean_ce_loss": 0.0262451171875, "train/clean_loss_scale": 28.85171890258789, "train/clean_loss_weight": 5.770343780517578, "train/combined_loss": 0.9086601138114929, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.259478672985782, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.75721675157547 }, { "epoch": 2.597156398104265, "step": 1096, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 8.661334991455078, "train/clean_loss_weight": 1.7322670221328735, "train/combined_loss": 0.8323001861572266, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.25971563981042656, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6935834884643555 }, { "epoch": 2.597156398104265, "step": 1096, "train/alpha": 0.2, "train/clean_ce_loss": 0.0240478515625, "train/clean_loss_scale": 31.217609405517578, "train/clean_loss_weight": 6.2435221672058105, "train/combined_loss": 0.9008597135543823, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.25971563981042656, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7507164478302002 }, { "epoch": 2.597156398104265, "step": 1096, "train/alpha": 0.2, "train/clean_ce_loss": 0.058837890625, "train/clean_loss_scale": 8.254122734069824, "train/clean_loss_weight": 1.6508245468139648, "train/combined_loss": 0.5827862024307251, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.25971563981042656, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.48565515875816345 }, { "epoch": 2.597156398104265, "step": 1096, "train/alpha": 0.2, "train/clean_ce_loss": 0.061279296875, "train/clean_loss_scale": 6.32750129699707, "train/clean_loss_weight": 1.2655003070831299, "train/combined_loss": 0.465293824672699, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.25971563981042656, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.38774484395980835 }, { "epoch": 2.5995260663507107, "step": 1097, "train/alpha": 0.2, "train/clean_ce_loss": 0.08203125, "train/clean_loss_scale": 6.708093643188477, "train/clean_loss_weight": 1.3416187763214111, "train/combined_loss": 0.6603279709815979, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.2599526066350711, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5502732992172241 }, { "epoch": 2.5995260663507107, "step": 1097, "train/alpha": 0.2, "train/clean_ce_loss": 0.015625, "train/clean_loss_scale": 26.408586502075195, "train/clean_loss_weight": 5.281717300415039, "train/combined_loss": 0.4951609969139099, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.2599526066350711, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4126341640949249 }, { "epoch": 2.5995260663507107, "step": 1097, "train/alpha": 0.2, "train/clean_ce_loss": 0.07568359375, "train/clean_loss_scale": 9.500096321105957, "train/clean_loss_weight": 1.9000192880630493, "train/combined_loss": 0.8628016710281372, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.2599526066350711, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7190014123916626 }, { "epoch": 2.5995260663507107, "step": 1097, "train/alpha": 0.2, "train/clean_ce_loss": 0.0145263671875, "train/clean_loss_scale": 11.75854778289795, "train/clean_loss_weight": 2.3517096042633057, "train/combined_loss": 0.20497077703475952, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.2599526066350711, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.17080898582935333 }, { "epoch": 2.6018957345971563, "step": 1098, "train/alpha": 0.2, "train/clean_ce_loss": 0.049560546875, "train/clean_loss_scale": 6.29826545715332, "train/clean_loss_weight": 1.259653091430664, "train/combined_loss": 0.37457457184791565, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.26018957345971566, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.312145471572876 }, { "epoch": 2.6018957345971563, "step": 1098, "train/alpha": 0.2, "train/clean_ce_loss": 0.08544921875, "train/clean_loss_scale": 10.31233024597168, "train/clean_loss_weight": 2.0624661445617676, "train/combined_loss": 1.0574166774749756, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.26018957345971566, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8811805844306946 }, { "epoch": 2.6018957345971563, "step": 1098, "train/alpha": 0.2, "train/clean_ce_loss": 0.032958984375, "train/clean_loss_scale": 16.116134643554688, "train/clean_loss_weight": 3.223227024078369, "train/combined_loss": 0.6374057531356812, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.26018957345971566, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.531171441078186 }, { "epoch": 2.6018957345971563, "step": 1098, "train/alpha": 0.2, "train/clean_ce_loss": 0.043701171875, "train/clean_loss_scale": 7.942251682281494, "train/clean_loss_weight": 1.588450312614441, "train/combined_loss": 0.41650286316871643, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.26018957345971566, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.34708571434020996 }, { "epoch": 2.604265402843602, "step": 1099, "train/alpha": 0.2, "train/clean_ce_loss": 0.0703125, "train/clean_loss_scale": 7.494121074676514, "train/clean_loss_weight": 1.4988242387771606, "train/combined_loss": 0.6323164701461792, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.2604265402843602, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5269303917884827 }, { "epoch": 2.604265402843602, "step": 1099, "train/alpha": 0.2, "train/clean_ce_loss": 0.0032501220703125, "train/clean_loss_scale": 51.30048751831055, "train/clean_loss_weight": 10.26009750366211, "train/combined_loss": 0.20007941126823425, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.2604265402843602, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.16673284769058228 }, { "epoch": 2.604265402843602, "step": 1099, "train/alpha": 0.2, "train/clean_ce_loss": 0.003326416015625, "train/clean_loss_scale": 57.27086639404297, "train/clean_loss_weight": 11.45417308807373, "train/combined_loss": 0.22860807180404663, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.2604265402843602, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1905067265033722 }, { "epoch": 2.604265402843602, "step": 1099, "train/alpha": 0.2, "train/clean_ce_loss": 0.0047607421875, "train/clean_loss_scale": 29.731958389282227, "train/clean_loss_weight": 5.946391582489014, "train/combined_loss": 0.16985543072223663, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.2604265402843602, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.14154618978500366 }, { "epoch": 2.6066350710900474, "grad_norm": 19.969972610473633, "learning_rate": 2e-06, "loss": 2.9711, "step": 1100 }, { "epoch": 2.6066350710900474, "step": 1100, "train/alpha": 0.2, "train/clean_ce_loss": 0.1328125, "train/clean_loss_scale": 6.522012710571289, "train/clean_loss_weight": 1.3044025897979736, "train/combined_loss": 1.0394457578659058, "train/pos_loss/pos0": 0.3266201615333557, "train/pos_loss/pos1": 0.7703732848167419, "train/pos_loss/pos2": 1.2490952014923096, "train/pos_loss/pos3": 2.0101680755615234, "train/pos_loss/pos4": 2.5674500465393066, "train/pos_loss/pos5": 2.893458127975464, "train/pos_loss/pos6": 3.2401061058044434, "train/pos_loss/weight0": 2.8505101203918457, "train/pos_loss/weight1": 2.016309976577759, "train/pos_loss/weight2": 1.210806965827942, "train/pos_loss/weight3": 0.598982572555542, "train/pos_loss/weight4": 0.23669114708900452, "train/pos_loss/weight5": 0.06508970260620117, "train/pos_loss/weight6": 0.021609380841255188, "train/pos_weight_progress": 0.26066350710900477, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8662047982215881 }, { "epoch": 2.6066350710900474, "step": 1100, "train/alpha": 0.2, "train/clean_ce_loss": 0.04296875, "train/clean_loss_scale": 9.044490814208984, "train/clean_loss_weight": 1.8088982105255127, "train/combined_loss": 0.4663565456867218, "train/pos_loss/pos0": 0.12853911519050598, "train/pos_loss/pos1": 0.3205752968788147, "train/pos_loss/pos2": 0.6379857063293457, "train/pos_loss/pos3": 0.9533864855766296, "train/pos_loss/pos4": 1.413595199584961, "train/pos_loss/pos5": 1.7470256090164185, "train/pos_loss/pos6": 2.1809093952178955, "train/pos_loss/weight0": 2.985690116882324, "train/pos_loss/weight1": 2.019146203994751, "train/pos_loss/weight2": 1.1703770160675049, "train/pos_loss/weight3": 0.5534799098968506, "train/pos_loss/weight4": 0.19709496200084686, "train/pos_loss/weight5": 0.05711796507239342, "train/pos_loss/weight6": 0.01709396205842495, "train/pos_weight_progress": 0.26066350710900477, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3886304497718811 }, { "epoch": 2.6066350710900474, "step": 1100, "train/alpha": 0.2, "train/clean_ce_loss": 0.09912109375, "train/clean_loss_scale": 9.61415958404541, "train/clean_loss_weight": 1.9228318929672241, "train/combined_loss": 1.143559217453003, "train/pos_loss/pos0": 0.32490235567092896, "train/pos_loss/pos1": 0.6012585759162903, "train/pos_loss/pos2": 1.0394970178604126, "train/pos_loss/pos3": 1.5612995624542236, "train/pos_loss/pos4": 2.1806726455688477, "train/pos_loss/pos5": 2.868115186691284, "train/pos_loss/pos6": 3.4549686908721924, "train/pos_loss/weight0": 2.3048648834228516, "train/pos_loss/weight1": 1.7482514381408691, "train/pos_loss/weight2": 1.2539825439453125, "train/pos_loss/weight3": 0.8041337728500366, "train/pos_loss/weight4": 0.4741766154766083, "train/pos_loss/weight5": 0.2593698799610138, "train/pos_loss/weight6": 0.15522116422653198, "train/pos_weight_progress": 0.26066350710900477, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.952966034412384 }, { "epoch": 2.6066350710900474, "step": 1100, "train/alpha": 0.2, "train/clean_ce_loss": 0.0052490234375, "train/clean_loss_scale": 29.21076774597168, "train/clean_loss_weight": 5.842153549194336, "train/combined_loss": 0.1839936077594757, "train/pos_loss/pos0": 0.0018432897049933672, "train/pos_loss/pos1": 0.03965635225176811, "train/pos_loss/pos2": 0.1768624186515808, "train/pos_loss/pos3": 0.4699736535549164, "train/pos_loss/pos4": 1.0477373600006104, "train/pos_loss/pos5": 1.6912657022476196, "train/pos_loss/pos6": 2.3314273357391357, "train/pos_loss/weight0": 2.8764309883117676, "train/pos_loss/weight1": 1.9301958084106445, "train/pos_loss/weight2": 1.2049342393875122, "train/pos_loss/weight3": 0.6103902459144592, "train/pos_loss/weight4": 0.25966864824295044, "train/pos_loss/weight5": 0.0880962684750557, "train/pos_loss/weight6": 0.030283521860837936, "train/pos_weight_progress": 0.26066350710900477, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.15332800149917603 }, { "epoch": 2.609004739336493, "step": 1101, "train/alpha": 0.2, "train/clean_ce_loss": 0.09814453125, "train/clean_loss_scale": 14.382022857666016, "train/clean_loss_weight": 2.8764045238494873, "train/combined_loss": 1.6938202381134033, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2609004739336493, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4115169048309326 }, { "epoch": 2.609004739336493, "step": 1101, "train/alpha": 0.2, "train/clean_ce_loss": 0.0703125, "train/clean_loss_scale": 16.27741813659668, "train/clean_loss_weight": 3.255483627319336, "train/combined_loss": 1.3734071254730225, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2609004739336493, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1445059776306152 }, { "epoch": 2.609004739336493, "step": 1101, "train/alpha": 0.2, "train/clean_ce_loss": 0.060302734375, "train/clean_loss_scale": 12.950942039489746, "train/clean_loss_weight": 2.590188503265381, "train/combined_loss": 0.9371726512908936, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2609004739336493, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.780977189540863 }, { "epoch": 2.609004739336493, "step": 1101, "train/alpha": 0.2, "train/clean_ce_loss": 0.134765625, "train/clean_loss_scale": 8.531744003295898, "train/clean_loss_weight": 1.7063487768173218, "train/combined_loss": 1.3797428607940674, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2609004739336493, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1497857570648193 }, { "epoch": 2.6113744075829386, "step": 1102, "train/alpha": 0.2, "train/clean_ce_loss": 0.078125, "train/clean_loss_scale": 11.625287055969238, "train/clean_loss_weight": 2.3250575065612793, "train/combined_loss": 1.0898706912994385, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2611374407582938, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9082255363464355 }, { "epoch": 2.6113744075829386, "step": 1102, "train/alpha": 0.2, "train/clean_ce_loss": 0.142578125, "train/clean_loss_scale": 8.44375228881836, "train/clean_loss_weight": 1.6887505054473877, "train/combined_loss": 1.4446732997894287, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2611374407582938, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2038943767547607 }, { "epoch": 2.6113744075829386, "step": 1102, "train/alpha": 0.2, "train/clean_ce_loss": 0.0693359375, "train/clean_loss_scale": 15.4883451461792, "train/clean_loss_weight": 3.0976691246032715, "train/combined_loss": 1.2886786460876465, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2611374407582938, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0738989114761353 }, { "epoch": 2.6113744075829386, "step": 1102, "train/alpha": 0.2, "train/clean_ce_loss": 0.043701171875, "train/clean_loss_scale": 18.752836227416992, "train/clean_loss_weight": 3.7505671977996826, "train/combined_loss": 0.9834251403808594, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2611374407582938, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8195209503173828 }, { "epoch": 2.6137440758293837, "step": 1103, "train/alpha": 0.2, "train/clean_ce_loss": 0.0888671875, "train/clean_loss_scale": 16.926698684692383, "train/clean_loss_weight": 3.3853397369384766, "train/combined_loss": 1.8050737380981445, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2613744075829384, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.504228115081787 }, { "epoch": 2.6137440758293837, "step": 1103, "train/alpha": 0.2, "train/clean_ce_loss": 0.0147705078125, "train/clean_loss_scale": 23.262006759643555, "train/clean_loss_weight": 4.652401447296143, "train/combined_loss": 0.41231000423431396, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2613744075829384, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3435916602611542 }, { "epoch": 2.6137440758293837, "step": 1103, "train/alpha": 0.2, "train/clean_ce_loss": 0.024658203125, "train/clean_loss_scale": 16.34971046447754, "train/clean_loss_weight": 3.269942045211792, "train/combined_loss": 0.48378536105155945, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2613744075829384, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4031544625759125 }, { "epoch": 2.6137440758293837, "step": 1103, "train/alpha": 0.2, "train/clean_ce_loss": 0.1572265625, "train/clean_loss_scale": 8.571833610534668, "train/clean_loss_weight": 1.7143667936325073, "train/combined_loss": 1.617263913154602, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2613744075829384, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3477199077606201 }, { "epoch": 2.6161137440758293, "step": 1104, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 10.55953311920166, "train/clean_loss_weight": 2.1119067668914795, "train/combined_loss": 0.8352755904197693, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2616113744075829, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6960629820823669 }, { "epoch": 2.6161137440758293, "step": 1104, "train/alpha": 0.2, "train/clean_ce_loss": 0.1123046875, "train/clean_loss_scale": 8.659515380859375, "train/clean_loss_weight": 1.731903076171875, "train/combined_loss": 1.167004942893982, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2616113744075829, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9725041389465332 }, { "epoch": 2.6161137440758293, "step": 1104, "train/alpha": 0.2, "train/clean_ce_loss": 0.03759765625, "train/clean_loss_scale": 11.491905212402344, "train/clean_loss_weight": 2.2983810901641846, "train/combined_loss": 0.5184824466705322, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2616113744075829, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43206870555877686 }, { "epoch": 2.6161137440758293, "step": 1104, "train/alpha": 0.2, "train/clean_ce_loss": 0.12890625, "train/clean_loss_scale": 11.749201774597168, "train/clean_loss_weight": 2.3498404026031494, "train/combined_loss": 1.8174546957015991, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2616113744075829, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5145455598831177 }, { "epoch": 2.618483412322275, "grad_norm": 20.949968338012695, "learning_rate": 2e-06, "loss": 3.1266, "step": 1105 }, { "epoch": 2.618483412322275, "step": 1105, "train/alpha": 0.2, "train/clean_ce_loss": 0.0830078125, "train/clean_loss_scale": 10.864799499511719, "train/clean_loss_weight": 2.172960042953491, "train/combined_loss": 1.0822358131408691, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2618483412322275, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9018632173538208 }, { "epoch": 2.618483412322275, "step": 1105, "train/alpha": 0.2, "train/clean_ce_loss": 0.09375, "train/clean_loss_scale": 8.860334396362305, "train/clean_loss_weight": 1.7720669507980347, "train/combined_loss": 0.9967876076698303, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2618483412322275, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8306563496589661 }, { "epoch": 2.618483412322275, "step": 1105, "train/alpha": 0.2, "train/clean_ce_loss": 0.04638671875, "train/clean_loss_scale": 22.326416015625, "train/clean_loss_weight": 4.465283393859863, "train/combined_loss": 1.242779016494751, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2618483412322275, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0356491804122925 }, { "epoch": 2.618483412322275, "step": 1105, "train/alpha": 0.2, "train/clean_ce_loss": 0.1513671875, "train/clean_loss_scale": 10.08547592163086, "train/clean_loss_weight": 2.0170953273773193, "train/combined_loss": 1.8319321870803833, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2618483412322275, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5266101360321045 }, { "epoch": 2.6208530805687205, "step": 1106, "train/alpha": 0.2, "train/clean_ce_loss": 0.06103515625, "train/clean_loss_scale": 17.15464210510254, "train/clean_loss_weight": 3.4309284687042236, "train/combined_loss": 1.2564435005187988, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.262085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0470362901687622 }, { "epoch": 2.6208530805687205, "step": 1106, "train/alpha": 0.2, "train/clean_ce_loss": 0.076171875, "train/clean_loss_scale": 9.861292839050293, "train/clean_loss_weight": 1.9722585678100586, "train/combined_loss": 0.9013838171958923, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.262085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7511531710624695 }, { "epoch": 2.6208530805687205, "step": 1106, "train/alpha": 0.2, "train/clean_ce_loss": 0.08154296875, "train/clean_loss_scale": 14.014835357666016, "train/clean_loss_weight": 2.802967071533203, "train/combined_loss": 1.3713735342025757, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.262085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1428112983703613 }, { "epoch": 2.6208530805687205, "step": 1106, "train/alpha": 0.2, "train/clean_ce_loss": 0.04052734375, "train/clean_loss_scale": 22.887840270996094, "train/clean_loss_weight": 4.577568054199219, "train/combined_loss": 1.1131000518798828, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.262085308056872, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9275833964347839 }, { "epoch": 2.623222748815166, "step": 1107, "train/alpha": 0.2, "train/clean_ce_loss": 0.1337890625, "train/clean_loss_scale": 10.196464538574219, "train/clean_loss_weight": 2.039293050765991, "train/combined_loss": 1.6370105743408203, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2623222748815166, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3641754388809204 }, { "epoch": 2.623222748815166, "step": 1107, "train/alpha": 0.2, "train/clean_ce_loss": 0.05078125, "train/clean_loss_scale": 11.080862045288086, "train/clean_loss_weight": 2.216172456741333, "train/combined_loss": 0.6752400398254395, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2623222748815166, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5627000331878662 }, { "epoch": 2.623222748815166, "step": 1107, "train/alpha": 0.2, "train/clean_ce_loss": 0.146484375, "train/clean_loss_scale": 8.352324485778809, "train/clean_loss_weight": 1.6704648733139038, "train/combined_loss": 1.4681819677352905, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2623222748815166, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.223484992980957 }, { "epoch": 2.623222748815166, "step": 1107, "train/alpha": 0.2, "train/clean_ce_loss": 0.005767822265625, "train/clean_loss_scale": 48.419795989990234, "train/clean_loss_weight": 9.683959007263184, "train/combined_loss": 0.3351321518421173, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2623222748815166, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2792767882347107 }, { "epoch": 2.625592417061611, "step": 1108, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 13.872479438781738, "train/clean_loss_weight": 2.774495840072632, "train/combined_loss": 1.0729495286941528, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.26255924170616113, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8941246271133423 }, { "epoch": 2.625592417061611, "step": 1108, "train/alpha": 0.2, "train/clean_ce_loss": 0.005218505859375, "train/clean_loss_scale": 52.94446563720703, "train/clean_loss_weight": 10.588892936706543, "train/combined_loss": 0.3315492272377014, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.26255924170616113, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.27629101276397705 }, { "epoch": 2.625592417061611, "step": 1108, "train/alpha": 0.2, "train/clean_ce_loss": 0.1494140625, "train/clean_loss_scale": 7.704211711883545, "train/clean_loss_weight": 1.5408424139022827, "train/combined_loss": 1.3813410997390747, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.26255924170616113, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1511175632476807 }, { "epoch": 2.625592417061611, "step": 1108, "train/alpha": 0.2, "train/clean_ce_loss": 0.08251953125, "train/clean_loss_scale": 16.000070571899414, "train/clean_loss_weight": 3.200014114379883, "train/combined_loss": 1.5843820571899414, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.26255924170616113, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.320318341255188 }, { "epoch": 2.6279620853080567, "step": 1109, "train/alpha": 0.2, "train/clean_ce_loss": 0.060791015625, "train/clean_loss_scale": 16.5083065032959, "train/clean_loss_weight": 3.301661252975464, "train/combined_loss": 1.2042680978775024, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2627962085308057, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.003556728363037 }, { "epoch": 2.6279620853080567, "step": 1109, "train/alpha": 0.2, "train/clean_ce_loss": 0.06298828125, "train/clean_loss_scale": 10.520339012145996, "train/clean_loss_weight": 2.104067802429199, "train/combined_loss": 0.7951897382736206, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2627962085308057, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6626580953598022 }, { "epoch": 2.6279620853080567, "step": 1109, "train/alpha": 0.2, "train/clean_ce_loss": 0.08056640625, "train/clean_loss_scale": 10.958123207092285, "train/clean_loss_weight": 2.191624641418457, "train/combined_loss": 1.0594279766082764, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2627962085308057, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8828566074371338 }, { "epoch": 2.6279620853080567, "step": 1109, "train/alpha": 0.2, "train/clean_ce_loss": 0.012939453125, "train/clean_loss_scale": 14.253429412841797, "train/clean_loss_weight": 2.8506858348846436, "train/combined_loss": 0.22131790220737457, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.2627962085308057, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.18443158268928528 }, { "epoch": 2.6303317535545023, "grad_norm": 15.218462944030762, "learning_rate": 2e-06, "loss": 2.9454, "step": 1110 }, { "epoch": 2.6303317535545023, "step": 1110, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 18.892473220825195, "train/clean_loss_weight": 3.7784945964813232, "train/combined_loss": 1.4612146615982056, "train/pos_loss/pos0": 0.33131298422813416, "train/pos_loss/pos1": 0.5630459785461426, "train/pos_loss/pos2": 1.4026397466659546, "train/pos_loss/pos3": 1.784084677696228, "train/pos_loss/pos4": 2.4229373931884766, "train/pos_loss/pos5": 2.9034886360168457, "train/pos_loss/pos6": 3.710409164428711, "train/pos_loss/weight0": 2.0144519805908203, "train/pos_loss/weight1": 1.5957567691802979, "train/pos_loss/weight2": 1.2117195129394531, "train/pos_loss/weight3": 0.8843592405319214, "train/pos_loss/weight4": 0.6158087849617004, "train/pos_loss/weight5": 0.39375412464141846, "train/pos_loss/weight6": 0.2841495871543884, "train/pos_weight_progress": 0.26303317535545023, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2176789045333862 }, { "epoch": 2.6303317535545023, "step": 1110, "train/alpha": 0.2, "train/clean_ce_loss": 0.00872802734375, "train/clean_loss_scale": 14.423815727233887, "train/clean_loss_weight": 2.884763240814209, "train/combined_loss": 0.1510697603225708, "train/pos_loss/pos0": 0.004746522754430771, "train/pos_loss/pos1": 0.040940575301647186, "train/pos_loss/pos2": 0.1358194202184677, "train/pos_loss/pos3": 0.4892466068267822, "train/pos_loss/pos4": 0.9518393278121948, "train/pos_loss/pos5": 1.63869047164917, "train/pos_loss/pos6": 2.0903449058532715, "train/pos_loss/weight0": 2.964301109313965, "train/pos_loss/weight1": 1.9503881931304932, "train/pos_loss/weight2": 1.2151384353637695, "train/pos_loss/weight3": 0.5761157274246216, "train/pos_loss/weight4": 0.21699222922325134, "train/pos_loss/weight5": 0.05893885716795921, "train/pos_loss/weight6": 0.018125813454389572, "train/pos_weight_progress": 0.26303317535545023, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.12589146196842194 }, { "epoch": 2.6303317535545023, "step": 1110, "train/alpha": 0.2, "train/clean_ce_loss": 0.00150299072265625, "train/clean_loss_scale": 190.95181274414062, "train/clean_loss_weight": 38.190364837646484, "train/combined_loss": 0.344398558139801, "train/pos_loss/pos0": 0.0007833485142327845, "train/pos_loss/pos1": 0.0058599174953997135, "train/pos_loss/pos2": 0.04250538349151611, "train/pos_loss/pos3": 0.22007502615451813, "train/pos_loss/pos4": 0.7851592302322388, "train/pos_loss/pos5": 1.4306696653366089, "train/pos_loss/pos6": 1.9227875471115112, "train/pos_loss/weight0": 1.9457807540893555, "train/pos_loss/weight1": 1.5579503774642944, "train/pos_loss/weight2": 1.1907817125320435, "train/pos_loss/weight3": 0.9018174409866333, "train/pos_loss/weight4": 0.6456687450408936, "train/pos_loss/weight5": 0.437896728515625, "train/pos_loss/weight6": 0.3201039731502533, "train/pos_weight_progress": 0.26303317535545023, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.28699880838394165 }, { "epoch": 2.6303317535545023, "step": 1110, "train/alpha": 0.2, "train/clean_ce_loss": 0.083984375, "train/clean_loss_scale": 11.347577095031738, "train/clean_loss_weight": 2.2695155143737793, "train/combined_loss": 1.1436229944229126, "train/pos_loss/pos0": 0.27313706278800964, "train/pos_loss/pos1": 0.5601752996444702, "train/pos_loss/pos2": 0.8949799537658691, "train/pos_loss/pos3": 1.2710449695587158, "train/pos_loss/pos4": 1.7276188135147095, "train/pos_loss/pos5": 2.1203746795654297, "train/pos_loss/pos6": 2.409127712249756, "train/pos_loss/weight0": 1.8756598234176636, "train/pos_loss/weight1": 1.5116028785705566, "train/pos_loss/weight2": 1.1744962930679321, "train/pos_loss/weight3": 0.883065938949585, "train/pos_loss/weight4": 0.6780267357826233, "train/pos_loss/weight5": 0.5002008080482483, "train/pos_loss/weight6": 0.3769475817680359, "train/pos_weight_progress": 0.26303317535545023, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9530192017555237 }, { "epoch": 2.632701421800948, "step": 1111, "train/alpha": 0.2, "train/clean_ce_loss": 0.03076171875, "train/clean_loss_scale": 11.358509063720703, "train/clean_loss_weight": 2.2717018127441406, "train/combined_loss": 0.419288694858551, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.2632701421800948, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.34940725564956665 }, { "epoch": 2.632701421800948, "step": 1111, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 7.996165752410889, "train/clean_loss_weight": 1.5992331504821777, "train/combined_loss": 0.7777520418167114, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.2632701421800948, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6481267213821411 }, { "epoch": 2.632701421800948, "step": 1111, "train/alpha": 0.2, "train/clean_ce_loss": 0.06884765625, "train/clean_loss_scale": 9.852595329284668, "train/clean_loss_weight": 1.9705190658569336, "train/combined_loss": 0.8139936923980713, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.2632701421800948, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6783280968666077 }, { "epoch": 2.632701421800948, "step": 1111, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 13.826838493347168, "train/clean_loss_weight": 2.7653677463531494, "train/combined_loss": 1.3286727666854858, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.2632701421800948, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1072273254394531 }, { "epoch": 2.6350710900473935, "step": 1112, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 8.919658660888672, "train/clean_loss_weight": 1.7839317321777344, "train/combined_loss": 0.6951062083244324, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26350710900473934, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5792551636695862 }, { "epoch": 2.6350710900473935, "step": 1112, "train/alpha": 0.2, "train/clean_ce_loss": 0.046630859375, "train/clean_loss_scale": 8.68336009979248, "train/clean_loss_weight": 1.736672043800354, "train/combined_loss": 0.485895037651062, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26350710900473934, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.404912531375885 }, { "epoch": 2.6350710900473935, "step": 1112, "train/alpha": 0.2, "train/clean_ce_loss": 0.0751953125, "train/clean_loss_scale": 8.733016967773438, "train/clean_loss_weight": 1.7466033697128296, "train/combined_loss": 0.7880183458328247, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26350710900473934, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6566819548606873 }, { "epoch": 2.6350710900473935, "step": 1112, "train/alpha": 0.2, "train/clean_ce_loss": 0.020751953125, "train/clean_loss_scale": 13.908205032348633, "train/clean_loss_weight": 2.7816410064697266, "train/combined_loss": 0.3463468849658966, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26350710900473934, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2886224091053009 }, { "epoch": 2.637440758293839, "step": 1113, "train/alpha": 0.2, "train/clean_ce_loss": 0.06982421875, "train/clean_loss_scale": 13.191742897033691, "train/clean_loss_weight": 2.6383485794067383, "train/combined_loss": 1.1053237915039062, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26374407582938386, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9211031198501587 }, { "epoch": 2.637440758293839, "step": 1113, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 8.174908638000488, "train/clean_loss_weight": 1.6349817514419556, "train/combined_loss": 0.7951376438140869, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26374407582938386, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6626147031784058 }, { "epoch": 2.637440758293839, "step": 1113, "train/alpha": 0.2, "train/clean_ce_loss": 0.17578125, "train/clean_loss_scale": 6.939872741699219, "train/clean_loss_weight": 1.3879746198654175, "train/combined_loss": 1.4638794660568237, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26374407582938386, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2198995351791382 }, { "epoch": 2.637440758293839, "step": 1113, "train/alpha": 0.2, "train/clean_ce_loss": 0.0167236328125, "train/clean_loss_scale": 12.473722457885742, "train/clean_loss_weight": 2.4947445392608643, "train/combined_loss": 0.25032714009284973, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26374407582938386, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.20860596001148224 }, { "epoch": 2.639810426540284, "step": 1114, "train/alpha": 0.2, "train/clean_ce_loss": 0.1640625, "train/clean_loss_scale": 7.474132061004639, "train/clean_loss_weight": 1.4948264360427856, "train/combined_loss": 1.471469759941101, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26398104265402844, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2262247800827026 }, { "epoch": 2.639810426540284, "step": 1114, "train/alpha": 0.2, "train/clean_ce_loss": 0.05810546875, "train/clean_loss_scale": 17.255229949951172, "train/clean_loss_weight": 3.4510459899902344, "train/combined_loss": 1.2031478881835938, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26398104265402844, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.002623200416565 }, { "epoch": 2.639810426540284, "step": 1114, "train/alpha": 0.2, "train/clean_ce_loss": 0.0196533203125, "train/clean_loss_scale": 18.176267623901367, "train/clean_loss_weight": 3.635253667831421, "train/combined_loss": 0.4286688268184662, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26398104265402844, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3572240173816681 }, { "epoch": 2.639810426540284, "step": 1114, "train/alpha": 0.2, "train/clean_ce_loss": 0.080078125, "train/clean_loss_scale": 10.334779739379883, "train/clean_loss_weight": 2.066956043243408, "train/combined_loss": 0.9931077361106873, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26398104265402844, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.827589750289917 }, { "epoch": 2.6421800947867298, "grad_norm": 16.1204776763916, "learning_rate": 2e-06, "loss": 3.2399, "step": 1115 }, { "epoch": 2.6421800947867298, "step": 1115, "train/alpha": 0.2, "train/clean_ce_loss": 0.05078125, "train/clean_loss_scale": 11.809282302856445, "train/clean_loss_weight": 2.361856460571289, "train/combined_loss": 0.7196281552314758, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26421800947867297, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5996901392936707 }, { "epoch": 2.6421800947867298, "step": 1115, "train/alpha": 0.2, "train/clean_ce_loss": 0.1015625, "train/clean_loss_scale": 10.464158058166504, "train/clean_loss_weight": 2.092831611633301, "train/combined_loss": 1.2753193378448486, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26421800947867297, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0627660751342773 }, { "epoch": 2.6421800947867298, "step": 1115, "train/alpha": 0.2, "train/clean_ce_loss": 0.07861328125, "train/clean_loss_scale": 9.49680233001709, "train/clean_loss_weight": 1.8993605375289917, "train/combined_loss": 0.8958897590637207, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26421800947867297, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7465748190879822 }, { "epoch": 2.6421800947867298, "step": 1115, "train/alpha": 0.2, "train/clean_ce_loss": 0.115234375, "train/clean_loss_scale": 9.972892761230469, "train/clean_loss_weight": 1.9945785999298096, "train/combined_loss": 1.3790640830993652, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26421800947867297, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1492201089859009 }, { "epoch": 2.6445497630331753, "step": 1116, "train/alpha": 0.2, "train/clean_ce_loss": 0.047607421875, "train/clean_loss_scale": 11.305960655212402, "train/clean_loss_weight": 2.2611920833587646, "train/combined_loss": 0.6458971500396729, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26445497630331755, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5382476449012756 }, { "epoch": 2.6445497630331753, "step": 1116, "train/alpha": 0.2, "train/clean_ce_loss": 0.051513671875, "train/clean_loss_scale": 4.184605121612549, "train/clean_loss_weight": 0.8369210362434387, "train/combined_loss": 0.25867724418640137, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26445497630331755, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.21556437015533447 }, { "epoch": 2.6445497630331753, "step": 1116, "train/alpha": 0.2, "train/clean_ce_loss": 0.06787109375, "train/clean_loss_scale": 17.757823944091797, "train/clean_loss_weight": 3.551564931869507, "train/combined_loss": 1.4462915658950806, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26445497630331755, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2052429914474487 }, { "epoch": 2.6445497630331753, "step": 1116, "train/alpha": 0.2, "train/clean_ce_loss": 0.1474609375, "train/clean_loss_scale": 7.7344584465026855, "train/clean_loss_weight": 1.546891689300537, "train/combined_loss": 1.3686366081237793, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26445497630331755, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1405304670333862 }, { "epoch": 2.646919431279621, "step": 1117, "train/alpha": 0.2, "train/clean_ce_loss": 0.00152587890625, "train/clean_loss_scale": 59.826602935791016, "train/clean_loss_weight": 11.965320587158203, "train/combined_loss": 0.10954578220844269, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26469194312796207, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.09128814935684204 }, { "epoch": 2.646919431279621, "step": 1117, "train/alpha": 0.2, "train/clean_ce_loss": 0.103515625, "train/clean_loss_scale": 7.428395748138428, "train/clean_loss_weight": 1.4856791496276855, "train/combined_loss": 0.9227460622787476, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26469194312796207, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7689550518989563 }, { "epoch": 2.646919431279621, "step": 1117, "train/alpha": 0.2, "train/clean_ce_loss": 0.08251953125, "train/clean_loss_scale": 7.023064613342285, "train/clean_loss_weight": 1.4046128988265991, "train/combined_loss": 0.6954480409622192, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26469194312796207, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5795400142669678 }, { "epoch": 2.646919431279621, "step": 1117, "train/alpha": 0.2, "train/clean_ce_loss": 0.030517578125, "train/clean_loss_scale": 7.464288711547852, "train/clean_loss_weight": 1.492857813835144, "train/combined_loss": 0.27335041761398315, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26469194312796207, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.22779200971126556 }, { "epoch": 2.6492890995260665, "step": 1118, "train/alpha": 0.2, "train/clean_ce_loss": 0.09375, "train/clean_loss_scale": 9.219755172729492, "train/clean_loss_weight": 1.8439511060714722, "train/combined_loss": 1.0372225046157837, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26492890995260665, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8643520474433899 }, { "epoch": 2.6492890995260665, "step": 1118, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 8.858036994934082, "train/clean_loss_weight": 1.7716073989868164, "train/combined_loss": 0.8408214449882507, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26492890995260665, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7006845474243164 }, { "epoch": 2.6492890995260665, "step": 1118, "train/alpha": 0.2, "train/clean_ce_loss": 0.0888671875, "train/clean_loss_scale": 6.937292098999023, "train/clean_loss_weight": 1.3874584436416626, "train/combined_loss": 0.7397971749305725, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26492890995260665, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6164976358413696 }, { "epoch": 2.6492890995260665, "step": 1118, "train/alpha": 0.2, "train/clean_ce_loss": 0.0162353515625, "train/clean_loss_scale": 17.2021427154541, "train/clean_loss_weight": 3.4404284954071045, "train/combined_loss": 0.3351393938064575, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26492890995260665, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2792828381061554 }, { "epoch": 2.6516587677725116, "step": 1119, "train/alpha": 0.2, "train/clean_ce_loss": 0.06982421875, "train/clean_loss_scale": 8.813199996948242, "train/clean_loss_weight": 1.7626399993896484, "train/combined_loss": 0.73844975233078, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.2651658767772512, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6153748035430908 }, { "epoch": 2.6516587677725116, "step": 1119, "train/alpha": 0.2, "train/clean_ce_loss": 0.072265625, "train/clean_loss_scale": 9.447244644165039, "train/clean_loss_weight": 1.8894490003585815, "train/combined_loss": 0.8192532062530518, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.2651658767772512, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6827110052108765 }, { "epoch": 2.6516587677725116, "step": 1119, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 8.59967041015625, "train/clean_loss_weight": 1.719934105873108, "train/combined_loss": 0.6852862238883972, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.2651658767772512, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5710718631744385 }, { "epoch": 2.6516587677725116, "step": 1119, "train/alpha": 0.2, "train/clean_ce_loss": 0.00732421875, "train/clean_loss_scale": 19.367549896240234, "train/clean_loss_weight": 3.8735101222991943, "train/combined_loss": 0.17022261023521423, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.2651658767772512, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.14185217022895813 }, { "epoch": 2.654028436018957, "grad_norm": 48.820335388183594, "learning_rate": 2e-06, "loss": 3.2356, "step": 1120 }, { "epoch": 2.654028436018957, "step": 1120, "train/alpha": 0.2, "train/clean_ce_loss": 0.11181640625, "train/clean_loss_scale": 7.948376178741455, "train/clean_loss_weight": 1.5896753072738647, "train/combined_loss": 1.0665106773376465, "train/pos_loss/pos0": 0.16493606567382812, "train/pos_loss/pos1": 0.6980727910995483, "train/pos_loss/pos2": 1.204934000968933, "train/pos_loss/pos3": 1.8457965850830078, "train/pos_loss/pos4": 2.465472459793091, "train/pos_loss/pos5": 3.1915578842163086, "train/pos_loss/pos6": 3.604649066925049, "train/pos_loss/weight0": 2.5913572311401367, "train/pos_loss/weight1": 1.863584041595459, "train/pos_loss/weight2": 1.2259727716445923, "train/pos_loss/weight3": 0.7150009870529175, "train/pos_loss/weight4": 0.361477255821228, "train/pos_loss/weight5": 0.16338713467121124, "train/pos_loss/weight6": 0.07922031730413437, "train/pos_weight_progress": 0.26540284360189575, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8887588381767273 }, { "epoch": 2.654028436018957, "step": 1120, "train/alpha": 0.2, "train/clean_ce_loss": 0.10791015625, "train/clean_loss_scale": 6.434071063995361, "train/clean_loss_weight": 1.2868142127990723, "train/combined_loss": 0.8331619501113892, "train/pos_loss/pos0": 0.2707018554210663, "train/pos_loss/pos1": 0.6139633059501648, "train/pos_loss/pos2": 1.0066395998001099, "train/pos_loss/pos3": 1.5204015970230103, "train/pos_loss/pos4": 1.9449206590652466, "train/pos_loss/pos5": 2.395235061645508, "train/pos_loss/pos6": 2.857450485229492, "train/pos_loss/weight0": 2.8072903156280518, "train/pos_loss/weight1": 2.035586357116699, "train/pos_loss/weight2": 1.218058466911316, "train/pos_loss/weight3": 0.6056774854660034, "train/pos_loss/weight4": 0.231199249625206, "train/pos_loss/weight5": 0.07698045670986176, "train/pos_loss/weight6": 0.025208014994859695, "train/pos_weight_progress": 0.26540284360189575, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6943016052246094 }, { "epoch": 2.654028436018957, "step": 1120, "train/alpha": 0.2, "train/clean_ce_loss": 0.058349609375, "train/clean_loss_scale": 11.042052268981934, "train/clean_loss_weight": 2.2084105014801025, "train/combined_loss": 0.7731593251228333, "train/pos_loss/pos0": 0.3486618995666504, "train/pos_loss/pos1": 0.5201306939125061, "train/pos_loss/pos2": 0.8660206198692322, "train/pos_loss/pos3": 1.1763060092926025, "train/pos_loss/pos4": 1.442482829093933, "train/pos_loss/pos5": 1.718738317489624, "train/pos_loss/pos6": 2.0519940853118896, "train/pos_loss/weight0": 2.7415242195129395, "train/pos_loss/weight1": 1.936370611190796, "train/pos_loss/weight2": 1.1949067115783691, "train/pos_loss/weight3": 0.6580475568771362, "train/pos_loss/weight4": 0.29986488819122314, "train/pos_loss/weight5": 0.1190408319234848, "train/pos_loss/weight6": 0.050245314836502075, "train/pos_weight_progress": 0.26540284360189575, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6442994475364685 }, { "epoch": 2.654028436018957, "step": 1120, "train/alpha": 0.2, "train/clean_ce_loss": 0.05908203125, "train/clean_loss_scale": 13.999008178710938, "train/clean_loss_weight": 2.7998015880584717, "train/combined_loss": 0.992507815361023, "train/pos_loss/pos0": 0.2339341789484024, "train/pos_loss/pos1": 0.4036562144756317, "train/pos_loss/pos2": 1.017124891281128, "train/pos_loss/pos3": 1.7359042167663574, "train/pos_loss/pos4": 2.150972604751587, "train/pos_loss/pos5": 2.695366382598877, "train/pos_loss/pos6": 3.3904473781585693, "train/pos_loss/weight0": 2.452361822128296, "train/pos_loss/weight1": 1.7921984195709229, "train/pos_loss/weight2": 1.2255017757415771, "train/pos_loss/weight3": 0.7651167511940002, "train/pos_loss/weight4": 0.41928231716156006, "train/pos_loss/weight5": 0.22115302085876465, "train/pos_loss/weight6": 0.12438595294952393, "train/pos_weight_progress": 0.26540284360189575, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8270898461341858 }, { "epoch": 2.656398104265403, "step": 1121, "train/alpha": 0.2, "train/clean_ce_loss": 0.057861328125, "train/clean_loss_scale": 8.149847030639648, "train/clean_loss_weight": 1.6299694776535034, "train/combined_loss": 0.5658732056617737, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2656398104265403, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4715609848499298 }, { "epoch": 2.656398104265403, "step": 1121, "train/alpha": 0.2, "train/clean_ce_loss": 0.06640625, "train/clean_loss_scale": 6.897204399108887, "train/clean_loss_weight": 1.3794409036636353, "train/combined_loss": 0.5496209859848022, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2656398104265403, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4580174684524536 }, { "epoch": 2.656398104265403, "step": 1121, "train/alpha": 0.2, "train/clean_ce_loss": 0.0283203125, "train/clean_loss_scale": 10.003135681152344, "train/clean_loss_weight": 2.000627279281616, "train/combined_loss": 0.3399503231048584, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2656398104265403, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.28329193592071533 }, { "epoch": 2.656398104265403, "step": 1121, "train/alpha": 0.2, "train/clean_ce_loss": 0.01953125, "train/clean_loss_scale": 12.09440803527832, "train/clean_loss_weight": 2.41888165473938, "train/combined_loss": 0.2834627032279968, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2656398104265403, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2362189143896103 }, { "epoch": 2.6587677725118484, "step": 1122, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 10.31956958770752, "train/clean_loss_weight": 2.0639140605926514, "train/combined_loss": 0.9795528650283813, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.26587677725118486, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8162940740585327 }, { "epoch": 2.6587677725118484, "step": 1122, "train/alpha": 0.2, "train/clean_ce_loss": 0.0751953125, "train/clean_loss_scale": 8.560731887817383, "train/clean_loss_weight": 1.7121464014053345, "train/combined_loss": 0.7724723219871521, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.26587677725118486, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6437269449234009 }, { "epoch": 2.6587677725118484, "step": 1122, "train/alpha": 0.2, "train/clean_ce_loss": 2.4437904357910156e-05, "train/clean_loss_scale": 4919.861328125, "train/clean_loss_weight": 983.9722900390625, "train/combined_loss": 0.14427731931209564, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.26587677725118486, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.12023109942674637 }, { "epoch": 2.6587677725118484, "step": 1122, "train/alpha": 0.2, "train/clean_ce_loss": 0.08837890625, "train/clean_loss_scale": 8.214411735534668, "train/clean_loss_weight": 1.6428823471069336, "train/combined_loss": 0.8711768388748169, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.26587677725118486, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7259806990623474 }, { "epoch": 2.661137440758294, "step": 1123, "train/alpha": 0.2, "train/clean_ce_loss": 0.037841796875, "train/clean_loss_scale": 9.338604927062988, "train/clean_loss_weight": 1.8677209615707397, "train/combined_loss": 0.42406749725341797, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2661137440758294, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.35338959097862244 }, { "epoch": 2.661137440758294, "step": 1123, "train/alpha": 0.2, "train/clean_ce_loss": 0.0419921875, "train/clean_loss_scale": 17.68194580078125, "train/clean_loss_weight": 3.536389112472534, "train/combined_loss": 0.8910043239593506, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2661137440758294, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7425035834312439 }, { "epoch": 2.661137440758294, "step": 1123, "train/alpha": 0.2, "train/clean_ce_loss": 0.10498046875, "train/clean_loss_scale": 6.720486164093018, "train/clean_loss_weight": 1.3440972566604614, "train/combined_loss": 0.8466237783432007, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2661137440758294, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7055197954177856 }, { "epoch": 2.661137440758294, "step": 1123, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 9.847013473510742, "train/clean_loss_weight": 1.9694026708602905, "train/combined_loss": 0.7616049647331238, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2661137440758294, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6346707940101624 }, { "epoch": 2.6635071090047395, "step": 1124, "train/alpha": 0.2, "train/clean_ce_loss": 0.01348876953125, "train/clean_loss_scale": 15.755703926086426, "train/clean_loss_weight": 3.1511409282684326, "train/combined_loss": 0.2550300657749176, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2663507109004739, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.21252505481243134 }, { "epoch": 2.6635071090047395, "step": 1124, "train/alpha": 0.2, "train/clean_ce_loss": 0.1025390625, "train/clean_loss_scale": 9.692111015319824, "train/clean_loss_weight": 1.9384222030639648, "train/combined_loss": 1.1925840377807617, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2663507109004739, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9938200116157532 }, { "epoch": 2.6635071090047395, "step": 1124, "train/alpha": 0.2, "train/clean_ce_loss": 0.0157470703125, "train/clean_loss_scale": 11.49083137512207, "train/clean_loss_weight": 2.298166275024414, "train/combined_loss": 0.21713632345199585, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2663507109004739, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1809469312429428 }, { "epoch": 2.6635071090047395, "step": 1124, "train/alpha": 0.2, "train/clean_ce_loss": 0.1015625, "train/clean_loss_scale": 11.035801887512207, "train/clean_loss_weight": 2.207160472869873, "train/combined_loss": 1.3449883460998535, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2663507109004739, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.120823621749878 }, { "epoch": 2.665876777251185, "grad_norm": 14.871132850646973, "learning_rate": 2e-06, "loss": 3.0409, "step": 1125 }, { "epoch": 2.665876777251185, "step": 1125, "train/alpha": 0.2, "train/clean_ce_loss": 0.11083984375, "train/clean_loss_scale": 6.142586708068848, "train/clean_loss_weight": 1.2285174131393433, "train/combined_loss": 0.8170120120048523, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2665876777251185, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6808433532714844 }, { "epoch": 2.665876777251185, "step": 1125, "train/alpha": 0.2, "train/clean_ce_loss": 0.05810546875, "train/clean_loss_scale": 1.2207562923431396, "train/clean_loss_weight": 0.2441512644290924, "train/combined_loss": 0.08511914312839508, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2665876777251185, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.07093261927366257 }, { "epoch": 2.665876777251185, "step": 1125, "train/alpha": 0.2, "train/clean_ce_loss": 0.00127410888671875, "train/clean_loss_scale": 94.10862731933594, "train/clean_loss_weight": 18.821725845336914, "train/combined_loss": 0.1438855677843094, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2665876777251185, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.11990463733673096 }, { "epoch": 2.665876777251185, "step": 1125, "train/alpha": 0.2, "train/clean_ce_loss": 0.10107421875, "train/clean_loss_scale": 8.199307441711426, "train/clean_loss_weight": 1.6398614645004272, "train/combined_loss": 0.9944862723350525, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2665876777251185, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8287385702133179 }, { "epoch": 2.6682464454976302, "step": 1126, "train/alpha": 0.2, "train/clean_ce_loss": 0.0283203125, "train/clean_loss_scale": 24.94271469116211, "train/clean_loss_weight": 4.9885430335998535, "train/combined_loss": 0.8476625680923462, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.266824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7063854932785034 }, { "epoch": 2.6682464454976302, "step": 1126, "train/alpha": 0.2, "train/clean_ce_loss": 0.06396484375, "train/clean_loss_scale": 13.198894500732422, "train/clean_loss_weight": 2.6397788524627686, "train/combined_loss": 1.0131182670593262, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.266824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8442652225494385 }, { "epoch": 2.6682464454976302, "step": 1126, "train/alpha": 0.2, "train/clean_ce_loss": 0.00604248046875, "train/clean_loss_scale": 20.444150924682617, "train/clean_loss_weight": 4.088830471038818, "train/combined_loss": 0.14824005961418152, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.266824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.12353338301181793 }, { "epoch": 2.6682464454976302, "step": 1126, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 10.743627548217773, "train/clean_loss_weight": 2.1487255096435547, "train/combined_loss": 0.8498377799987793, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.266824644549763, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7081981301307678 }, { "epoch": 2.670616113744076, "step": 1127, "train/alpha": 0.2, "train/clean_ce_loss": 0.09521484375, "train/clean_loss_scale": 7.769372940063477, "train/clean_loss_weight": 1.5538746118545532, "train/combined_loss": 0.8877115249633789, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2670616113744076, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.739759624004364 }, { "epoch": 2.670616113744076, "step": 1127, "train/alpha": 0.2, "train/clean_ce_loss": 0.11279296875, "train/clean_loss_scale": 9.048230171203613, "train/clean_loss_weight": 1.8096460103988647, "train/combined_loss": 1.2246921062469482, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2670616113744076, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0205767154693604 }, { "epoch": 2.670616113744076, "step": 1127, "train/alpha": 0.2, "train/clean_ce_loss": 0.1298828125, "train/clean_loss_scale": 8.36700439453125, "train/clean_loss_weight": 1.67340087890625, "train/combined_loss": 1.3040761947631836, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2670616113744076, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0867301225662231 }, { "epoch": 2.670616113744076, "step": 1127, "train/alpha": 0.2, "train/clean_ce_loss": 0.0130615234375, "train/clean_loss_scale": 29.070716857910156, "train/clean_loss_weight": 5.814143657684326, "train/combined_loss": 0.45564940571784973, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2670616113744076, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.37970784306526184 }, { "epoch": 2.6729857819905214, "step": 1128, "train/alpha": 0.2, "train/clean_ce_loss": 0.03125, "train/clean_loss_scale": 21.555438995361328, "train/clean_loss_weight": 4.3110880851745605, "train/combined_loss": 0.8083289861679077, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2672985781990521, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6736074686050415 }, { "epoch": 2.6729857819905214, "step": 1128, "train/alpha": 0.2, "train/clean_ce_loss": 0.0022125244140625, "train/clean_loss_scale": 49.82944107055664, "train/clean_loss_weight": 9.965888023376465, "train/combined_loss": 0.13229861855506897, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2672985781990521, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1102488562464714 }, { "epoch": 2.6729857819905214, "step": 1128, "train/alpha": 0.2, "train/clean_ce_loss": 0.00193023681640625, "train/clean_loss_scale": 41.41334915161133, "train/clean_loss_weight": 8.282670021057129, "train/combined_loss": 0.09592508524656296, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2672985781990521, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.07993756979703903 }, { "epoch": 2.6729857819905214, "step": 1128, "train/alpha": 0.2, "train/clean_ce_loss": 0.0120849609375, "train/clean_loss_scale": 12.836173057556152, "train/clean_loss_weight": 2.567234754562378, "train/combined_loss": 0.18614958226680756, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2672985781990521, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.15512464940547943 }, { "epoch": 2.675355450236967, "step": 1129, "train/alpha": 0.2, "train/clean_ce_loss": 0.045166015625, "train/clean_loss_scale": 7.970343589782715, "train/clean_loss_weight": 1.5940687656402588, "train/combined_loss": 0.43198639154434204, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2675355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.35998865962028503 }, { "epoch": 2.675355450236967, "step": 1129, "train/alpha": 0.2, "train/clean_ce_loss": 0.0693359375, "train/clean_loss_scale": 11.704315185546875, "train/clean_loss_weight": 2.340862989425659, "train/combined_loss": 0.9738355875015259, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2675355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8115296363830566 }, { "epoch": 2.675355450236967, "step": 1129, "train/alpha": 0.2, "train/clean_ce_loss": 0.023193359375, "train/clean_loss_scale": 13.642327308654785, "train/clean_loss_weight": 2.7284655570983887, "train/combined_loss": 0.3796936869621277, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2675355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.31641140580177307 }, { "epoch": 2.675355450236967, "step": 1129, "train/alpha": 0.2, "train/clean_ce_loss": 0.056396484375, "train/clean_loss_scale": 14.902050971984863, "train/clean_loss_weight": 2.98041033744812, "train/combined_loss": 1.0085079669952393, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2675355450236967, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8404232859611511 }, { "epoch": 2.677725118483412, "grad_norm": 25.920164108276367, "learning_rate": 2e-06, "loss": 3.2015, "step": 1130 }, { "epoch": 2.677725118483412, "step": 1130, "train/alpha": 0.2, "train/clean_ce_loss": 0.0201416015625, "train/clean_loss_scale": 8.149580001831055, "train/clean_loss_weight": 1.6299160718917847, "train/combined_loss": 0.1969747096300125, "train/pos_loss/pos0": 0.025361070409417152, "train/pos_loss/pos1": 0.07178305089473724, "train/pos_loss/pos2": 0.1589965522289276, "train/pos_loss/pos3": 0.41078275442123413, "train/pos_loss/pos4": 0.9342851042747498, "train/pos_loss/pos5": 1.445488691329956, "train/pos_loss/pos6": 1.9502274990081787, "train/pos_loss/weight0": 2.767239809036255, "train/pos_loss/weight1": 1.9053148031234741, "train/pos_loss/weight2": 1.247576117515564, "train/pos_loss/weight3": 0.6715697050094604, "train/pos_loss/weight4": 0.27531418204307556, "train/pos_loss/weight5": 0.09520789980888367, "train/pos_loss/weight6": 0.037777841091156006, "train/pos_weight_progress": 0.2677725118483412, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1641455888748169 }, { "epoch": 2.677725118483412, "step": 1130, "train/alpha": 0.2, "train/clean_ce_loss": 0.0830078125, "train/clean_loss_scale": 12.040498733520508, "train/clean_loss_weight": 2.408099889755249, "train/combined_loss": 1.1993465423583984, "train/pos_loss/pos0": 0.351847767829895, "train/pos_loss/pos1": 0.6439214944839478, "train/pos_loss/pos2": 0.9769613146781921, "train/pos_loss/pos3": 1.3552438020706177, "train/pos_loss/pos4": 1.719546914100647, "train/pos_loss/pos5": 2.2706573009490967, "train/pos_loss/pos6": 2.506450653076172, "train/pos_loss/weight0": 1.9451749324798584, "train/pos_loss/weight1": 1.5481951236724854, "train/pos_loss/weight2": 1.1945372819900513, "train/pos_loss/weight3": 0.8953208923339844, "train/pos_loss/weight4": 0.6486861109733582, "train/pos_loss/weight5": 0.4463939368724823, "train/pos_loss/weight6": 0.32169151306152344, "train/pos_weight_progress": 0.2677725118483412, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.999455451965332 }, { "epoch": 2.677725118483412, "step": 1130, "train/alpha": 0.2, "train/clean_ce_loss": 0.10888671875, "train/clean_loss_scale": 7.79273796081543, "train/clean_loss_weight": 1.5585476160049438, "train/combined_loss": 1.0182307958602905, "train/pos_loss/pos0": 0.36389783024787903, "train/pos_loss/pos1": 0.702212393283844, "train/pos_loss/pos2": 1.0517573356628418, "train/pos_loss/pos3": 1.5751723051071167, "train/pos_loss/pos4": 2.0278573036193848, "train/pos_loss/pos5": 2.4539051055908203, "train/pos_loss/pos6": 2.843993902206421, "train/pos_loss/weight0": 2.6490275859832764, "train/pos_loss/weight1": 1.8673113584518433, "train/pos_loss/weight2": 1.219813346862793, "train/pos_loss/weight3": 0.6820496916770935, "train/pos_loss/weight4": 0.3487485945224762, "train/pos_loss/weight5": 0.16016025841236115, "train/pos_loss/weight6": 0.0728895291686058, "train/pos_weight_progress": 0.2677725118483412, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8485256433486938 }, { "epoch": 2.677725118483412, "step": 1130, "train/alpha": 0.2, "train/clean_ce_loss": 0.06689453125, "train/clean_loss_scale": 8.60347843170166, "train/clean_loss_weight": 1.7206957340240479, "train/combined_loss": 0.6906307935714722, "train/pos_loss/pos0": 0.16898828744888306, "train/pos_loss/pos1": 0.38901206851005554, "train/pos_loss/pos2": 0.7589352130889893, "train/pos_loss/pos3": 1.407606840133667, "train/pos_loss/pos4": 2.0766031742095947, "train/pos_loss/pos5": 2.46474552154541, "train/pos_loss/pos6": 2.9759299755096436, "train/pos_loss/weight0": 2.806312084197998, "train/pos_loss/weight1": 1.924714207649231, "train/pos_loss/weight2": 1.1833844184875488, "train/pos_loss/weight3": 0.6387308835983276, "train/pos_loss/weight4": 0.29052436351776123, "train/pos_loss/weight5": 0.11068715900182724, "train/pos_loss/weight6": 0.04564672335982323, "train/pos_weight_progress": 0.2677725118483412, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5755256414413452 }, { "epoch": 2.6800947867298577, "step": 1131, "train/alpha": 0.2, "train/clean_ce_loss": 0.038330078125, "train/clean_loss_scale": 22.426424026489258, "train/clean_loss_weight": 4.485284805297852, "train/combined_loss": 1.0315278768539429, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.2680094786729858, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8596065640449524 }, { "epoch": 2.6800947867298577, "step": 1131, "train/alpha": 0.2, "train/clean_ce_loss": 0.01190185546875, "train/clean_loss_scale": 14.03594970703125, "train/clean_loss_weight": 2.80718994140625, "train/combined_loss": 0.2004646211862564, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.2680094786729858, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.16705384850502014 }, { "epoch": 2.6800947867298577, "step": 1131, "train/alpha": 0.2, "train/clean_ce_loss": 0.00787353515625, "train/clean_loss_scale": 11.776392936706543, "train/clean_loss_weight": 2.355278730392456, "train/combined_loss": 0.11126621067523956, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.2680094786729858, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.0927218422293663 }, { "epoch": 2.6800947867298577, "step": 1131, "train/alpha": 0.2, "train/clean_ce_loss": 0.10302734375, "train/clean_loss_scale": 7.936154842376709, "train/clean_loss_weight": 1.5872310400009155, "train/combined_loss": 0.9811691641807556, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.2680094786729858, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8176409602165222 }, { "epoch": 2.6824644549763033, "step": 1132, "train/alpha": 0.2, "train/clean_ce_loss": 0.0012969970703125, "train/clean_loss_scale": 122.50447845458984, "train/clean_loss_weight": 24.500896453857422, "train/combined_loss": 0.19066554307937622, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.2682464454976303, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.15888795256614685 }, { "epoch": 2.6824644549763033, "step": 1132, "train/alpha": 0.2, "train/clean_ce_loss": 0.006256103515625, "train/clean_loss_scale": 26.74325942993164, "train/clean_loss_weight": 5.348651885986328, "train/combined_loss": 0.2007703185081482, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.2682464454976303, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.16730859875679016 }, { "epoch": 2.6824644549763033, "step": 1132, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 9.274862289428711, "train/clean_loss_weight": 1.8549724817276, "train/combined_loss": 0.9021252989768982, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.2682464454976303, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.751771092414856 }, { "epoch": 2.6824644549763033, "step": 1132, "train/alpha": 0.2, "train/clean_ce_loss": 0.031494140625, "train/clean_loss_scale": 3.6749391555786133, "train/clean_loss_weight": 0.7349878549575806, "train/combined_loss": 0.13888685405254364, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.2682464454976303, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.11573904752731323 }, { "epoch": 2.684834123222749, "step": 1133, "train/alpha": 0.2, "train/clean_ce_loss": 0.08837890625, "train/clean_loss_scale": 9.433603286743164, "train/clean_loss_weight": 1.8867206573486328, "train/combined_loss": 1.0004777908325195, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.2684834123222749, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8337315320968628 }, { "epoch": 2.684834123222749, "step": 1133, "train/alpha": 0.2, "train/clean_ce_loss": 0.134765625, "train/clean_loss_scale": 8.899576187133789, "train/clean_loss_weight": 1.7799152135849, "train/combined_loss": 1.4392282962799072, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.2684834123222749, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1993569135665894 }, { "epoch": 2.684834123222749, "step": 1133, "train/alpha": 0.2, "train/clean_ce_loss": 0.07568359375, "train/clean_loss_scale": 7.322606563568115, "train/clean_loss_weight": 1.4645212888717651, "train/combined_loss": 0.665041446685791, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.2684834123222749, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5542011857032776 }, { "epoch": 2.684834123222749, "step": 1133, "train/alpha": 0.2, "train/clean_ce_loss": 0.00347900390625, "train/clean_loss_scale": 45.98478698730469, "train/clean_loss_weight": 9.1969575881958, "train/combined_loss": 0.19197750091552734, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.2684834123222749, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.15998125076293945 }, { "epoch": 2.6872037914691944, "step": 1134, "train/alpha": 0.2, "train/clean_ce_loss": 0.0286865234375, "train/clean_loss_scale": 11.09219741821289, "train/clean_loss_weight": 2.2184395790100098, "train/combined_loss": 0.3818359076976776, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.26872037914691943, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3181965947151184 }, { "epoch": 2.6872037914691944, "step": 1134, "train/alpha": 0.2, "train/clean_ce_loss": 0.068359375, "train/clean_loss_scale": 10.166881561279297, "train/clean_loss_weight": 2.033376455307007, "train/combined_loss": 0.8340020179748535, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.26872037914691943, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6950016617774963 }, { "epoch": 2.6872037914691944, "step": 1134, "train/alpha": 0.2, "train/clean_ce_loss": 0.036865234375, "train/clean_loss_scale": 7.93256950378418, "train/clean_loss_weight": 1.586513876914978, "train/combined_loss": 0.35092324018478394, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.26872037914691943, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.29243603348731995 }, { "epoch": 2.6872037914691944, "step": 1134, "train/alpha": 0.2, "train/clean_ce_loss": 0.0166015625, "train/clean_loss_scale": 12.07728099822998, "train/clean_loss_weight": 2.4154562950134277, "train/combined_loss": 0.24060209095478058, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.26872037914691943, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.20050173997879028 }, { "epoch": 2.68957345971564, "grad_norm": 28.18712615966797, "learning_rate": 2e-06, "loss": 2.9977, "step": 1135 }, { "epoch": 2.68957345971564, "step": 1135, "train/alpha": 0.2, "train/clean_ce_loss": 0.0257568359375, "train/clean_loss_scale": 10.236604690551758, "train/clean_loss_weight": 2.047321081161499, "train/combined_loss": 0.3163950443267822, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.26895734597156395, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.26366254687309265 }, { "epoch": 2.68957345971564, "step": 1135, "train/alpha": 0.2, "train/clean_ce_loss": 0.036376953125, "train/clean_loss_scale": 11.762553215026855, "train/clean_loss_weight": 2.352510690689087, "train/combined_loss": 0.513463020324707, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.26895734597156395, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.42788586020469666 }, { "epoch": 2.68957345971564, "step": 1135, "train/alpha": 0.2, "train/clean_ce_loss": 0.041015625, "train/clean_loss_scale": 21.255586624145508, "train/clean_loss_weight": 4.25111722946167, "train/combined_loss": 1.046173334121704, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.26895734597156395, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8718111515045166 }, { "epoch": 2.68957345971564, "step": 1135, "train/alpha": 0.2, "train/clean_ce_loss": 0.0025482177734375, "train/clean_loss_scale": 27.780323028564453, "train/clean_loss_weight": 5.556064605712891, "train/combined_loss": 0.08494837582111359, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.26895734597156395, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.07079031318426132 }, { "epoch": 2.6919431279620856, "step": 1136, "train/alpha": 0.2, "train/clean_ce_loss": 0.010986328125, "train/clean_loss_scale": 15.505510330200195, "train/clean_loss_weight": 3.101102113723755, "train/combined_loss": 0.2044183611869812, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.26919431279620853, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1703486293554306 }, { "epoch": 2.6919431279620856, "step": 1136, "train/alpha": 0.2, "train/clean_ce_loss": 0.1728515625, "train/clean_loss_scale": 6.8876214027404785, "train/clean_loss_weight": 1.3775242567062378, "train/combined_loss": 1.4286433458328247, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.26919431279620853, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1905361413955688 }, { "epoch": 2.6919431279620856, "step": 1136, "train/alpha": 0.2, "train/clean_ce_loss": 0.07373046875, "train/clean_loss_scale": 9.434781074523926, "train/clean_loss_weight": 1.8869562149047852, "train/combined_loss": 0.8347570300102234, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.26919431279620853, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6956308484077454 }, { "epoch": 2.6919431279620856, "step": 1136, "train/alpha": 0.2, "train/clean_ce_loss": 0.09130859375, "train/clean_loss_scale": 9.929361343383789, "train/clean_loss_weight": 1.9858722686767578, "train/combined_loss": 1.0879632234573364, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.26919431279620853, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9066359996795654 }, { "epoch": 2.6943127962085307, "step": 1137, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 8.67117691040039, "train/clean_loss_weight": 1.734235405921936, "train/combined_loss": 0.7722766995429993, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.26943127962085306, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6435639262199402 }, { "epoch": 2.6943127962085307, "step": 1137, "train/alpha": 0.2, "train/clean_ce_loss": 0.09033203125, "train/clean_loss_scale": 7.06310510635376, "train/clean_loss_weight": 1.412621021270752, "train/combined_loss": 0.7656295299530029, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.26943127962085306, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.638024628162384 }, { "epoch": 2.6943127962085307, "step": 1137, "train/alpha": 0.2, "train/clean_ce_loss": 0.0517578125, "train/clean_loss_scale": 8.470976829528809, "train/clean_loss_weight": 1.6941953897476196, "train/combined_loss": 0.5261270999908447, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.26943127962085306, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4384392499923706 }, { "epoch": 2.6943127962085307, "step": 1137, "train/alpha": 0.2, "train/clean_ce_loss": 0.0537109375, "train/clean_loss_scale": 8.491595268249512, "train/clean_loss_weight": 1.6983190774917603, "train/combined_loss": 0.5473098754882812, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.26943127962085306, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4560915529727936 }, { "epoch": 2.6966824644549763, "step": 1138, "train/alpha": 0.2, "train/clean_ce_loss": 0.10205078125, "train/clean_loss_scale": 11.70067310333252, "train/clean_loss_weight": 2.340134620666504, "train/combined_loss": 1.432875394821167, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.26966824644549764, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1940628290176392 }, { "epoch": 2.6966824644549763, "step": 1138, "train/alpha": 0.2, "train/clean_ce_loss": 0.017822265625, "train/clean_loss_scale": 23.01679229736328, "train/clean_loss_weight": 4.603358745574951, "train/combined_loss": 0.4922536611557007, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.26966824644549764, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.41021138429641724 }, { "epoch": 2.6966824644549763, "step": 1138, "train/alpha": 0.2, "train/clean_ce_loss": 0.0029754638671875, "train/clean_loss_scale": 54.66333770751953, "train/clean_loss_weight": 10.93266773223877, "train/combined_loss": 0.1951785385608673, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.26966824644549764, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1626487821340561 }, { "epoch": 2.6966824644549763, "step": 1138, "train/alpha": 0.2, "train/clean_ce_loss": 0.01446533203125, "train/clean_loss_scale": 14.778829574584961, "train/clean_loss_weight": 2.955765962600708, "train/combined_loss": 0.2565368115901947, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.26966824644549764, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.21378067135810852 }, { "epoch": 2.699052132701422, "step": 1139, "train/alpha": 0.2, "train/clean_ce_loss": 0.11572265625, "train/clean_loss_scale": 7.814483165740967, "train/clean_loss_weight": 1.5628966093063354, "train/combined_loss": 1.0851752758026123, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.26990521327014216, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9043127298355103 }, { "epoch": 2.699052132701422, "step": 1139, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 9.03548526763916, "train/clean_loss_weight": 1.80709707736969, "train/combined_loss": 0.709426760673523, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.26990521327014216, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5911889672279358 }, { "epoch": 2.699052132701422, "step": 1139, "train/alpha": 0.2, "train/clean_ce_loss": 0.056640625, "train/clean_loss_scale": 6.841160774230957, "train/clean_loss_weight": 1.3682321310043335, "train/combined_loss": 0.4649851322174072, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.26990521327014216, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.38748762011528015 }, { "epoch": 2.699052132701422, "step": 1139, "train/alpha": 0.2, "train/clean_ce_loss": 0.019287109375, "train/clean_loss_scale": 7.716756820678711, "train/clean_loss_weight": 1.543351411819458, "train/combined_loss": 0.1786007136106491, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.26990521327014216, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.14883393049240112 }, { "epoch": 2.7014218009478674, "grad_norm": 26.82090950012207, "learning_rate": 2e-06, "loss": 3.1387, "step": 1140 }, { "epoch": 2.7014218009478674, "step": 1140, "train/alpha": 0.2, "train/clean_ce_loss": 0.07275390625, "train/clean_loss_scale": 4.752410888671875, "train/clean_loss_weight": 0.950482189655304, "train/combined_loss": 0.4149077534675598, "train/pos_loss/pos0": 0.04634704440832138, "train/pos_loss/pos1": 0.13681389391422272, "train/pos_loss/pos2": 0.44902023673057556, "train/pos_loss/pos3": 0.869705080986023, "train/pos_loss/pos4": 1.429443359375, "train/pos_loss/pos5": 1.9203537702560425, "train/pos_loss/pos6": 2.3949971199035645, "train/pos_loss/weight0": 2.610532760620117, "train/pos_loss/weight1": 1.908565640449524, "train/pos_loss/weight2": 1.247268557548523, "train/pos_loss/weight3": 0.7149556279182434, "train/pos_loss/weight4": 0.33660635352134705, "train/pos_loss/weight5": 0.1260453760623932, "train/pos_loss/weight6": 0.05602569505572319, "train/pos_weight_progress": 0.27014218009478674, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.345756471157074 }, { "epoch": 2.7014218009478674, "step": 1140, "train/alpha": 0.2, "train/clean_ce_loss": 0.048828125, "train/clean_loss_scale": 22.521284103393555, "train/clean_loss_weight": 4.504256725311279, "train/combined_loss": 1.3196065425872803, "train/pos_loss/pos0": 0.24162845313549042, "train/pos_loss/pos1": 0.7473803758621216, "train/pos_loss/pos2": 1.2170220613479614, "train/pos_loss/pos3": 2.0470967292785645, "train/pos_loss/pos4": 2.5031378269195557, "train/pos_loss/pos5": 2.977993965148926, "train/pos_loss/pos6": 3.418365001678467, "train/pos_loss/weight0": 2.2574286460876465, "train/pos_loss/weight1": 1.7040605545043945, "train/pos_loss/weight2": 1.253340244293213, "train/pos_loss/weight3": 0.846929669380188, "train/pos_loss/weight4": 0.5075162053108215, "train/pos_loss/weight5": 0.2683596611022949, "train/pos_loss/weight6": 0.16236472129821777, "train/pos_weight_progress": 0.27014218009478674, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0996720790863037 }, { "epoch": 2.7014218009478674, "step": 1140, "train/alpha": 0.2, "train/clean_ce_loss": 0.017822265625, "train/clean_loss_scale": 21.117462158203125, "train/clean_loss_weight": 4.223492622375488, "train/combined_loss": 0.4516332149505615, "train/pos_loss/pos0": 0.03026210144162178, "train/pos_loss/pos1": 0.2871137857437134, "train/pos_loss/pos2": 0.5695310235023499, "train/pos_loss/pos3": 1.0594884157180786, "train/pos_loss/pos4": 1.9870083332061768, "train/pos_loss/pos5": 2.6574196815490723, "train/pos_loss/pos6": 3.353855609893799, "train/pos_loss/weight0": 2.87155818939209, "train/pos_loss/weight1": 2.0407164096832275, "train/pos_loss/weight2": 1.2003029584884644, "train/pos_loss/weight3": 0.5870805978775024, "train/pos_loss/weight4": 0.21736344695091248, "train/pos_loss/weight5": 0.06321188807487488, "train/pos_loss/weight6": 0.019766654819250107, "train/pos_weight_progress": 0.27014218009478674, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.37636101245880127 }, { "epoch": 2.7014218009478674, "step": 1140, "train/alpha": 0.2, "train/clean_ce_loss": 0.00408935546875, "train/clean_loss_scale": 59.175025939941406, "train/clean_loss_weight": 11.835005760192871, "train/combined_loss": 0.29038527607917786, "train/pos_loss/pos0": 0.028297865763306618, "train/pos_loss/pos1": 0.07682514190673828, "train/pos_loss/pos2": 0.24375545978546143, "train/pos_loss/pos3": 0.6077759861946106, "train/pos_loss/pos4": 1.0362098217010498, "train/pos_loss/pos5": 1.7274136543273926, "train/pos_loss/pos6": 2.1838626861572266, "train/pos_loss/weight0": 2.5366294384002686, "train/pos_loss/weight1": 1.8945426940917969, "train/pos_loss/weight2": 1.2914910316467285, "train/pos_loss/weight3": 0.743334949016571, "train/pos_loss/weight4": 0.345781147480011, "train/pos_loss/weight5": 0.13035902380943298, "train/pos_loss/weight6": 0.05786140263080597, "train/pos_weight_progress": 0.27014218009478674, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.24198772013187408 }, { "epoch": 2.7037914691943126, "step": 1141, "train/alpha": 0.2, "train/clean_ce_loss": 0.01458740234375, "train/clean_loss_scale": 35.38419723510742, "train/clean_loss_weight": 7.076839447021484, "train/combined_loss": 0.6193962097167969, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.27037914691943127, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5161635279655457 }, { "epoch": 2.7037914691943126, "step": 1141, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 13.654184341430664, "train/clean_loss_weight": 2.730836868286133, "train/combined_loss": 1.2160757780075073, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.27037914691943127, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0133965015411377 }, { "epoch": 2.7037914691943126, "step": 1141, "train/alpha": 0.2, "train/clean_ce_loss": 0.058349609375, "train/clean_loss_scale": 13.729043960571289, "train/clean_loss_weight": 2.7458088397979736, "train/combined_loss": 0.9613012075424194, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.27037914691943127, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8010843396186829 }, { "epoch": 2.7037914691943126, "step": 1141, "train/alpha": 0.2, "train/clean_ce_loss": 0.060546875, "train/clean_loss_scale": 11.865446090698242, "train/clean_loss_weight": 2.37308931350708, "train/combined_loss": 0.8620988130569458, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.27037914691943127, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7184156775474548 }, { "epoch": 2.706161137440758, "step": 1142, "train/alpha": 0.2, "train/clean_ce_loss": 0.02685546875, "train/clean_loss_scale": 42.92129898071289, "train/clean_loss_weight": 8.584259986877441, "train/combined_loss": 1.3832058906555176, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.27061611374407585, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1526715755462646 }, { "epoch": 2.706161137440758, "step": 1142, "train/alpha": 0.2, "train/clean_ce_loss": 0.00946044921875, "train/clean_loss_scale": 32.17131042480469, "train/clean_loss_weight": 6.434262275695801, "train/combined_loss": 0.36522606015205383, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.27061611374407585, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.30435505509376526 }, { "epoch": 2.706161137440758, "step": 1142, "train/alpha": 0.2, "train/clean_ce_loss": 0.04931640625, "train/clean_loss_scale": 22.024532318115234, "train/clean_loss_weight": 4.404906749725342, "train/combined_loss": 1.3034049272537231, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.27061611374407585, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0861707925796509 }, { "epoch": 2.706161137440758, "step": 1142, "train/alpha": 0.2, "train/clean_ce_loss": 0.059814453125, "train/clean_loss_scale": 11.424458503723145, "train/clean_loss_weight": 2.2848918437957764, "train/combined_loss": 0.8200173377990723, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.27061611374407585, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.683347761631012 }, { "epoch": 2.7085308056872037, "step": 1143, "train/alpha": 0.2, "train/clean_ce_loss": 0.021240234375, "train/clean_loss_scale": 21.47336769104004, "train/clean_loss_weight": 4.294673442840576, "train/combined_loss": 0.547319233417511, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.2708530805687204, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45609936118125916 }, { "epoch": 2.7085308056872037, "step": 1143, "train/alpha": 0.2, "train/clean_ce_loss": 0.0712890625, "train/clean_loss_scale": 11.055222511291504, "train/clean_loss_weight": 2.2110445499420166, "train/combined_loss": 0.94573974609375, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.2708530805687204, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.788116455078125 }, { "epoch": 2.7085308056872037, "step": 1143, "train/alpha": 0.2, "train/clean_ce_loss": 0.0012359619140625, "train/clean_loss_scale": 173.9188995361328, "train/clean_loss_weight": 34.78377914428711, "train/combined_loss": 0.25794854760169983, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.2708530805687204, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.214957132935524 }, { "epoch": 2.7085308056872037, "step": 1143, "train/alpha": 0.2, "train/clean_ce_loss": 0.07421875, "train/clean_loss_scale": 10.907379150390625, "train/clean_loss_weight": 2.181475877761841, "train/combined_loss": 0.971438467502594, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.2708530805687204, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8095320463180542 }, { "epoch": 2.7109004739336493, "step": 1144, "train/alpha": 0.2, "train/clean_ce_loss": 0.04150390625, "train/clean_loss_scale": 14.557490348815918, "train/clean_loss_weight": 2.9114980697631836, "train/combined_loss": 0.7250312566757202, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.27109004739336495, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6041927337646484 }, { "epoch": 2.7109004739336493, "step": 1144, "train/alpha": 0.2, "train/clean_ce_loss": 0.057861328125, "train/clean_loss_scale": 11.917123794555664, "train/clean_loss_weight": 2.383424758911133, "train/combined_loss": 0.8274487257003784, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.27109004739336495, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6895406246185303 }, { "epoch": 2.7109004739336493, "step": 1144, "train/alpha": 0.2, "train/clean_ce_loss": 0.052978515625, "train/clean_loss_scale": 12.003296852111816, "train/clean_loss_weight": 2.4006593227386475, "train/combined_loss": 0.7631002068519592, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.27109004739336495, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6359168291091919 }, { "epoch": 2.7109004739336493, "step": 1144, "train/alpha": 0.2, "train/clean_ce_loss": 0.00885009765625, "train/clean_loss_scale": 43.10026931762695, "train/clean_loss_weight": 8.620054244995117, "train/combined_loss": 0.45772990584373474, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.27109004739336495, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.381441593170166 }, { "epoch": 2.713270142180095, "grad_norm": 32.40397644042969, "learning_rate": 2e-06, "loss": 3.0662, "step": 1145 }, { "epoch": 2.713270142180095, "step": 1145, "train/alpha": 0.2, "train/clean_ce_loss": 0.010498046875, "train/clean_loss_scale": 23.128055572509766, "train/clean_loss_weight": 4.625611305236816, "train/combined_loss": 0.2913593053817749, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.2713270142180095, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.24279941618442535 }, { "epoch": 2.713270142180095, "step": 1145, "train/alpha": 0.2, "train/clean_ce_loss": 0.08935546875, "train/clean_loss_scale": 11.954401969909668, "train/clean_loss_weight": 2.3908803462982178, "train/combined_loss": 1.2818293571472168, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.2713270142180095, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0681911706924438 }, { "epoch": 2.713270142180095, "step": 1145, "train/alpha": 0.2, "train/clean_ce_loss": 0.01312255859375, "train/clean_loss_scale": 82.33866882324219, "train/clean_loss_weight": 16.46773338317871, "train/combined_loss": 1.2965928316116333, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.2713270142180095, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0804940462112427 }, { "epoch": 2.713270142180095, "step": 1145, "train/alpha": 0.2, "train/clean_ce_loss": 0.0267333984375, "train/clean_loss_scale": 45.077392578125, "train/clean_loss_weight": 9.01547908782959, "train/combined_loss": 1.4460862874984741, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.2713270142180095, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2050719261169434 }, { "epoch": 2.7156398104265405, "step": 1146, "train/alpha": 0.2, "train/clean_ce_loss": 0.072265625, "train/clean_loss_scale": 12.310981750488281, "train/clean_loss_weight": 2.4621963500976562, "train/combined_loss": 1.067592978477478, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.271563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8896607756614685 }, { "epoch": 2.7156398104265405, "step": 1146, "train/alpha": 0.2, "train/clean_ce_loss": 0.0264892578125, "train/clean_loss_scale": 30.96821403503418, "train/clean_loss_weight": 6.193643093109131, "train/combined_loss": 0.9843900203704834, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.271563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8203250169754028 }, { "epoch": 2.7156398104265405, "step": 1146, "train/alpha": 0.2, "train/clean_ce_loss": 0.043212890625, "train/clean_loss_scale": 12.860722541809082, "train/clean_loss_weight": 2.5721445083618164, "train/combined_loss": 0.666898787021637, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.271563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5557489991188049 }, { "epoch": 2.7156398104265405, "step": 1146, "train/alpha": 0.2, "train/clean_ce_loss": 0.0008392333984375, "train/clean_loss_scale": 303.3705749511719, "train/clean_loss_weight": 60.67411422729492, "train/combined_loss": 0.3055184483528137, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.271563981042654, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2545987069606781 }, { "epoch": 2.718009478672986, "step": 1147, "train/alpha": 0.2, "train/clean_ce_loss": 0.14453125, "train/clean_loss_scale": 9.576711654663086, "train/clean_loss_weight": 1.9153423309326172, "train/combined_loss": 1.6609609127044678, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.2718009478672986, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.38413405418396 }, { "epoch": 2.718009478672986, "step": 1147, "train/alpha": 0.2, "train/clean_ce_loss": 0.04150390625, "train/clean_loss_scale": 27.661075592041016, "train/clean_loss_weight": 5.532215118408203, "train/combined_loss": 1.3776512145996094, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.2718009478672986, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1480426788330078 }, { "epoch": 2.718009478672986, "step": 1147, "train/alpha": 0.2, "train/clean_ce_loss": 0.01953125, "train/clean_loss_scale": 29.091638565063477, "train/clean_loss_weight": 5.818327903747559, "train/combined_loss": 0.6818352937698364, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.2718009478672986, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5681960582733154 }, { "epoch": 2.718009478672986, "step": 1147, "train/alpha": 0.2, "train/clean_ce_loss": 0.01031494140625, "train/clean_loss_scale": 29.901899337768555, "train/clean_loss_weight": 5.980380058288574, "train/combined_loss": 0.37012359499931335, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.2718009478672986, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3084363341331482 }, { "epoch": 2.720379146919431, "step": 1148, "train/alpha": 0.2, "train/clean_ce_loss": 0.072265625, "train/clean_loss_scale": 13.763529777526855, "train/clean_loss_weight": 2.7527060508728027, "train/combined_loss": 1.1935560703277588, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.2720379146919431, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9946300983428955 }, { "epoch": 2.720379146919431, "step": 1148, "train/alpha": 0.2, "train/clean_ce_loss": 0.06884765625, "train/clean_loss_scale": 12.85273265838623, "train/clean_loss_weight": 2.5705466270446777, "train/combined_loss": 1.0618566274642944, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.2720379146919431, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.884880542755127 }, { "epoch": 2.720379146919431, "step": 1148, "train/alpha": 0.2, "train/clean_ce_loss": 0.08251953125, "train/clean_loss_scale": 12.953832626342773, "train/clean_loss_weight": 2.590766668319702, "train/combined_loss": 1.2827330827713013, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.2720379146919431, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0689442157745361 }, { "epoch": 2.720379146919431, "step": 1148, "train/alpha": 0.2, "train/clean_ce_loss": 0.11865234375, "train/clean_loss_scale": 9.143943786621094, "train/clean_loss_weight": 1.8287887573242188, "train/combined_loss": 1.3019404411315918, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.2720379146919431, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.08495032787323 }, { "epoch": 2.7227488151658767, "step": 1149, "train/alpha": 0.2, "train/clean_ce_loss": 0.0031585693359375, "train/clean_loss_scale": 71.01560974121094, "train/clean_loss_weight": 14.20312213897705, "train/combined_loss": 0.26916927099227905, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.2722748815165877, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.22430771589279175 }, { "epoch": 2.7227488151658767, "step": 1149, "train/alpha": 0.2, "train/clean_ce_loss": 0.033203125, "train/clean_loss_scale": 4.506535530090332, "train/clean_loss_weight": 0.9013071060180664, "train/combined_loss": 0.17955727875232697, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.2722748815165877, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.149631068110466 }, { "epoch": 2.7227488151658767, "step": 1149, "train/alpha": 0.2, "train/clean_ce_loss": 0.123046875, "train/clean_loss_scale": 11.99878215789795, "train/clean_loss_weight": 2.39975643157959, "train/combined_loss": 1.7716951370239258, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.2722748815165877, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.4764126539230347 }, { "epoch": 2.7227488151658767, "step": 1149, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 11.06523609161377, "train/clean_loss_weight": 2.2130472660064697, "train/combined_loss": 0.8752774596214294, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.2722748815165877, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7293978929519653 }, { "epoch": 2.7251184834123223, "grad_norm": 21.41611099243164, "learning_rate": 2e-06, "loss": 3.2565, "step": 1150 }, { "epoch": 2.7251184834123223, "step": 1150, "train/alpha": 0.2, "train/clean_ce_loss": 0.10595703125, "train/clean_loss_scale": 8.880449295043945, "train/clean_loss_weight": 1.7760899066925049, "train/combined_loss": 1.129135251045227, "train/pos_loss/pos0": 0.30759719014167786, "train/pos_loss/pos1": 0.6427289843559265, "train/pos_loss/pos2": 0.9263681173324585, "train/pos_loss/pos3": 1.2849249839782715, "train/pos_loss/pos4": 1.7663029432296753, "train/pos_loss/pos5": 2.211146593093872, "train/pos_loss/pos6": 2.545888662338257, "train/pos_loss/weight0": 2.015404462814331, "train/pos_loss/weight1": 1.5873124599456787, "train/pos_loss/weight2": 1.2103732824325562, "train/pos_loss/weight3": 0.9046567678451538, "train/pos_loss/weight4": 0.6026468873023987, "train/pos_loss/weight5": 0.3912908732891083, "train/pos_loss/weight6": 0.2883148193359375, "train/pos_weight_progress": 0.2725118483412322, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9409460425376892 }, { "epoch": 2.7251184834123223, "step": 1150, "train/alpha": 0.2, "train/clean_ce_loss": 0.0035858154296875, "train/clean_loss_scale": 45.53627014160156, "train/clean_loss_weight": 9.107254028320312, "train/combined_loss": 0.19594159722328186, "train/pos_loss/pos0": 0.0030002936255186796, "train/pos_loss/pos1": 0.052546825259923935, "train/pos_loss/pos2": 0.1350315660238266, "train/pos_loss/pos3": 0.4694994390010834, "train/pos_loss/pos4": 0.8003491759300232, "train/pos_loss/pos5": 1.382153034210205, "train/pos_loss/pos6": 1.8962546586990356, "train/pos_loss/weight0": 2.6976542472839355, "train/pos_loss/weight1": 1.921107530593872, "train/pos_loss/weight2": 1.1966192722320557, "train/pos_loss/weight3": 0.6646685004234314, "train/pos_loss/weight4": 0.3248216509819031, "train/pos_loss/weight5": 0.13471157848834991, "train/pos_loss/weight6": 0.06041738763451576, "train/pos_weight_progress": 0.2725118483412322, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.16328465938568115 }, { "epoch": 2.7251184834123223, "step": 1150, "train/alpha": 0.2, "train/clean_ce_loss": 0.01348876953125, "train/clean_loss_scale": 29.4316463470459, "train/clean_loss_weight": 5.886329174041748, "train/combined_loss": 0.4763960540294647, "train/pos_loss/pos0": 0.0016065717209130526, "train/pos_loss/pos1": 0.028605207800865173, "train/pos_loss/pos2": 0.19444993138313293, "train/pos_loss/pos3": 0.4034484326839447, "train/pos_loss/pos4": 1.058925986289978, "train/pos_loss/pos5": 1.4529627561569214, "train/pos_loss/pos6": 2.092024803161621, "train/pos_loss/weight0": 1.8849644660949707, "train/pos_loss/weight1": 1.528200387954712, "train/pos_loss/weight2": 1.1852400302886963, "train/pos_loss/weight3": 0.9192855954170227, "train/pos_loss/weight4": 0.6449984908103943, "train/pos_loss/weight5": 0.47191402316093445, "train/pos_loss/weight6": 0.3653970956802368, "train/pos_weight_progress": 0.2725118483412322, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.39699670672416687 }, { "epoch": 2.7251184834123223, "step": 1150, "train/alpha": 0.2, "train/clean_ce_loss": 0.062255859375, "train/clean_loss_scale": 12.066245079040527, "train/clean_loss_weight": 2.4132490158081055, "train/combined_loss": 0.9014333486557007, "train/pos_loss/pos0": 0.1755094677209854, "train/pos_loss/pos1": 0.4170472025871277, "train/pos_loss/pos2": 0.7593088746070862, "train/pos_loss/pos3": 1.056806206703186, "train/pos_loss/pos4": 1.4238436222076416, "train/pos_loss/pos5": 1.8564847707748413, "train/pos_loss/pos6": 2.233203649520874, "train/pos_loss/weight0": 1.9634636640548706, "train/pos_loss/weight1": 1.5613189935684204, "train/pos_loss/weight2": 1.2076689004898071, "train/pos_loss/weight3": 0.8824124336242676, "train/pos_loss/weight4": 0.645993173122406, "train/pos_loss/weight5": 0.41596612334251404, "train/pos_loss/weight6": 0.3231765925884247, "train/pos_weight_progress": 0.2725118483412322, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7511944770812988 }, { "epoch": 2.727488151658768, "step": 1151, "train/alpha": 0.2, "train/clean_ce_loss": 0.04443359375, "train/clean_loss_scale": 15.964783668518066, "train/clean_loss_weight": 3.1929566860198975, "train/combined_loss": 0.851247251033783, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.2727488151658768, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7093726992607117 }, { "epoch": 2.727488151658768, "step": 1151, "train/alpha": 0.2, "train/clean_ce_loss": 0.0308837890625, "train/clean_loss_scale": 17.93380355834961, "train/clean_loss_weight": 3.5867607593536377, "train/combined_loss": 0.6646366119384766, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.2727488151658768, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5538638234138489 }, { "epoch": 2.727488151658768, "step": 1151, "train/alpha": 0.2, "train/clean_ce_loss": 0.072265625, "train/clean_loss_scale": 8.773198127746582, "train/clean_loss_weight": 1.7546396255493164, "train/combined_loss": 0.7608007788658142, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.2727488151658768, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6340006589889526 }, { "epoch": 2.727488151658768, "step": 1151, "train/alpha": 0.2, "train/clean_ce_loss": 0.0140380859375, "train/clean_loss_scale": 10.282857894897461, "train/clean_loss_weight": 2.0565717220306396, "train/combined_loss": 0.17322197556495667, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.2727488151658768, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.14435164630413055 }, { "epoch": 2.729857819905213, "step": 1152, "train/alpha": 0.2, "train/clean_ce_loss": 0.0966796875, "train/clean_loss_scale": 8.171722412109375, "train/clean_loss_weight": 1.634344458580017, "train/combined_loss": 0.9480475187301636, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.2729857819905213, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.790039598941803 }, { "epoch": 2.729857819905213, "step": 1152, "train/alpha": 0.2, "train/clean_ce_loss": 0.09228515625, "train/clean_loss_scale": 8.581443786621094, "train/clean_loss_weight": 1.7162888050079346, "train/combined_loss": 0.9503278136253357, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.2729857819905213, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7919398546218872 }, { "epoch": 2.729857819905213, "step": 1152, "train/alpha": 0.2, "train/clean_ce_loss": 0.107421875, "train/clean_loss_scale": 8.414790153503418, "train/clean_loss_weight": 1.6829580068588257, "train/combined_loss": 1.0847190618515015, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.2729857819905213, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9039325714111328 }, { "epoch": 2.729857819905213, "step": 1152, "train/alpha": 0.2, "train/clean_ce_loss": 0.07470703125, "train/clean_loss_scale": 8.096844673156738, "train/clean_loss_weight": 1.6193689107894897, "train/combined_loss": 0.7258694767951965, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.2729857819905213, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6048912405967712 }, { "epoch": 2.7322274881516586, "step": 1153, "train/alpha": 0.2, "train/clean_ce_loss": 0.052001953125, "train/clean_loss_scale": 12.924800872802734, "train/clean_loss_weight": 2.5849602222442627, "train/combined_loss": 0.8065378665924072, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.2732227488151659, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.672114908695221 }, { "epoch": 2.7322274881516586, "step": 1153, "train/alpha": 0.2, "train/clean_ce_loss": 0.043701171875, "train/clean_loss_scale": 17.870569229125977, "train/clean_loss_weight": 3.5741138458251953, "train/combined_loss": 0.9371577501296997, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.2732227488151659, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7809647917747498 }, { "epoch": 2.7322274881516586, "step": 1153, "train/alpha": 0.2, "train/clean_ce_loss": 0.08203125, "train/clean_loss_scale": 13.642781257629395, "train/clean_loss_weight": 2.7285563945770264, "train/combined_loss": 1.342961311340332, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.2732227488151659, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1191344261169434 }, { "epoch": 2.7322274881516586, "step": 1153, "train/alpha": 0.2, "train/clean_ce_loss": 0.001373291015625, "train/clean_loss_scale": 185.74610900878906, "train/clean_loss_weight": 37.14922332763672, "train/combined_loss": 0.30610015988349915, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.2732227488151659, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.25508347153663635 }, { "epoch": 2.734597156398104, "step": 1154, "train/alpha": 0.2, "train/clean_ce_loss": 0.0087890625, "train/clean_loss_scale": 17.980613708496094, "train/clean_loss_weight": 3.5961227416992188, "train/combined_loss": 0.18963930010795593, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.2734597156398104, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.15803274512290955 }, { "epoch": 2.734597156398104, "step": 1154, "train/alpha": 0.2, "train/clean_ce_loss": 0.0966796875, "train/clean_loss_scale": 11.332659721374512, "train/clean_loss_weight": 2.2665319442749023, "train/combined_loss": 1.3147656917572021, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.2734597156398104, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0956380367279053 }, { "epoch": 2.734597156398104, "step": 1154, "train/alpha": 0.2, "train/clean_ce_loss": 0.01531982421875, "train/clean_loss_scale": 35.73907470703125, "train/clean_loss_weight": 7.147815227508545, "train/combined_loss": 0.6570196151733398, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.2734597156398104, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5475163459777832 }, { "epoch": 2.734597156398104, "step": 1154, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 9.07156753540039, "train/clean_loss_weight": 1.814313530921936, "train/combined_loss": 0.8610902428627014, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.2734597156398104, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.717575192451477 }, { "epoch": 2.7369668246445498, "grad_norm": 24.565034866333008, "learning_rate": 2e-06, "loss": 3.2753, "step": 1155 }, { "epoch": 2.7369668246445498, "step": 1155, "train/alpha": 0.2, "train/clean_ce_loss": 0.12158203125, "train/clean_loss_scale": 9.12220287322998, "train/clean_loss_weight": 1.824440598487854, "train/combined_loss": 1.330915093421936, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.273696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1090959310531616 }, { "epoch": 2.7369668246445498, "step": 1155, "train/alpha": 0.2, "train/clean_ce_loss": 0.068359375, "train/clean_loss_scale": 14.570113182067871, "train/clean_loss_weight": 2.91402268409729, "train/combined_loss": 1.1952046155929565, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.273696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9960038065910339 }, { "epoch": 2.7369668246445498, "step": 1155, "train/alpha": 0.2, "train/clean_ce_loss": 0.08447265625, "train/clean_loss_scale": 8.611209869384766, "train/clean_loss_weight": 1.722241997718811, "train/combined_loss": 0.8728941082954407, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.273696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7274117469787598 }, { "epoch": 2.7369668246445498, "step": 1155, "train/alpha": 0.2, "train/clean_ce_loss": 0.041015625, "train/clean_loss_scale": 9.99842643737793, "train/clean_loss_weight": 1.999685287475586, "train/combined_loss": 0.49211007356643677, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.273696682464455, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.41009172797203064 }, { "epoch": 2.7393364928909953, "step": 1156, "train/alpha": 0.2, "train/clean_ce_loss": 0.048095703125, "train/clean_loss_scale": 6.440868377685547, "train/clean_loss_weight": 1.2881736755371094, "train/combined_loss": 0.37173372507095337, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.2739336492890995, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.309778094291687 }, { "epoch": 2.7393364928909953, "step": 1156, "train/alpha": 0.2, "train/clean_ce_loss": 0.1123046875, "train/clean_loss_scale": 8.030327796936035, "train/clean_loss_weight": 1.6060656309127808, "train/combined_loss": 1.082212209701538, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.2739336492890995, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9018434882164001 }, { "epoch": 2.7393364928909953, "step": 1156, "train/alpha": 0.2, "train/clean_ce_loss": 0.0849609375, "train/clean_loss_scale": 12.252232551574707, "train/clean_loss_weight": 2.450446605682373, "train/combined_loss": 1.2491533756256104, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.2739336492890995, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0409611463546753 }, { "epoch": 2.7393364928909953, "step": 1156, "train/alpha": 0.2, "train/clean_ce_loss": 0.08056640625, "train/clean_loss_scale": 9.244787216186523, "train/clean_loss_weight": 1.8489574193954468, "train/combined_loss": 0.8937831521034241, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.2739336492890995, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7448192834854126 }, { "epoch": 2.741706161137441, "step": 1157, "train/alpha": 0.2, "train/clean_ce_loss": 0.047119140625, "train/clean_loss_scale": 22.000051498413086, "train/clean_loss_weight": 4.400010585784912, "train/combined_loss": 1.243948221206665, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.2741706161137441, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.036623477935791 }, { "epoch": 2.741706161137441, "step": 1157, "train/alpha": 0.2, "train/clean_ce_loss": 0.0556640625, "train/clean_loss_scale": 11.159049987792969, "train/clean_loss_weight": 2.2318100929260254, "train/combined_loss": 0.7453896999359131, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.2741706161137441, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6211580634117126 }, { "epoch": 2.741706161137441, "step": 1157, "train/alpha": 0.2, "train/clean_ce_loss": 0.07666015625, "train/clean_loss_scale": 9.532388687133789, "train/clean_loss_weight": 1.9064778089523315, "train/combined_loss": 0.8769053220748901, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.2741706161137441, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7307544350624084 }, { "epoch": 2.741706161137441, "step": 1157, "train/alpha": 0.2, "train/clean_ce_loss": 0.087890625, "train/clean_loss_scale": 7.521326541900635, "train/clean_loss_weight": 1.504265308380127, "train/combined_loss": 0.7932648658752441, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.2741706161137441, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6610540747642517 }, { "epoch": 2.7440758293838865, "step": 1158, "train/alpha": 0.2, "train/clean_ce_loss": 0.047607421875, "train/clean_loss_scale": 8.498522758483887, "train/clean_loss_weight": 1.6997045278549194, "train/combined_loss": 0.48551130294799805, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.27440758293838863, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.40459275245666504 }, { "epoch": 2.7440758293838865, "step": 1158, "train/alpha": 0.2, "train/clean_ce_loss": 0.04248046875, "train/clean_loss_scale": 29.206029891967773, "train/clean_loss_weight": 5.841206073760986, "train/combined_loss": 1.4888229370117188, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.27440758293838863, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2406858205795288 }, { "epoch": 2.7440758293838865, "step": 1158, "train/alpha": 0.2, "train/clean_ce_loss": 0.083984375, "train/clean_loss_scale": 7.887923717498779, "train/clean_loss_weight": 1.5775847434997559, "train/combined_loss": 0.7949548363685608, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.27440758293838863, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6624623537063599 }, { "epoch": 2.7440758293838865, "step": 1158, "train/alpha": 0.2, "train/clean_ce_loss": 0.004364013671875, "train/clean_loss_scale": 48.37555694580078, "train/clean_loss_weight": 9.675111770629883, "train/combined_loss": 0.2533339262008667, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.27440758293838863, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.21111159026622772 }, { "epoch": 2.7464454976303316, "step": 1159, "train/alpha": 0.2, "train/clean_ce_loss": 0.0238037109375, "train/clean_loss_scale": 25.625749588012695, "train/clean_loss_weight": 5.125150203704834, "train/combined_loss": 0.7319855093955994, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.27464454976303315, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6099879145622253 }, { "epoch": 2.7464454976303316, "step": 1159, "train/alpha": 0.2, "train/clean_ce_loss": 0.059814453125, "train/clean_loss_scale": 12.130725860595703, "train/clean_loss_weight": 2.426145315170288, "train/combined_loss": 0.8707112669944763, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.27464454976303315, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7255927324295044 }, { "epoch": 2.7464454976303316, "step": 1159, "train/alpha": 0.2, "train/clean_ce_loss": 0.10205078125, "train/clean_loss_scale": 10.246430397033691, "train/clean_loss_weight": 2.049286127090454, "train/combined_loss": 1.2547874450683594, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.27464454976303315, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0456562042236328 }, { "epoch": 2.7464454976303316, "step": 1159, "train/alpha": 0.2, "train/clean_ce_loss": 0.03173828125, "train/clean_loss_scale": 11.991693496704102, "train/clean_loss_weight": 2.398338794708252, "train/combined_loss": 0.4567148983478546, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.27464454976303315, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.38059574365615845 }, { "epoch": 2.748815165876777, "grad_norm": 24.48335838317871, "learning_rate": 2e-06, "loss": 3.2328, "step": 1160 }, { "epoch": 2.748815165876777, "step": 1160, "train/alpha": 0.2, "train/clean_ce_loss": 0.059326171875, "train/clean_loss_scale": 16.807329177856445, "train/clean_loss_weight": 3.3614659309387207, "train/combined_loss": 1.1965373754501343, "train/pos_loss/pos0": 0.22774240374565125, "train/pos_loss/pos1": 0.5697579979896545, "train/pos_loss/pos2": 1.2508563995361328, "train/pos_loss/pos3": 2.0297954082489014, "train/pos_loss/pos4": 2.2619924545288086, "train/pos_loss/pos5": 3.251398801803589, "train/pos_loss/pos6": 3.986649513244629, "train/pos_loss/weight0": 2.379021406173706, "train/pos_loss/weight1": 1.784549593925476, "train/pos_loss/weight2": 1.2485871315002441, "train/pos_loss/weight3": 0.7760399580001831, "train/pos_loss/weight4": 0.44297727942466736, "train/pos_loss/weight5": 0.23596467077732086, "train/pos_loss/weight6": 0.13286007940769196, "train/pos_weight_progress": 0.27488151658767773, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9971144795417786 }, { "epoch": 2.748815165876777, "step": 1160, "train/alpha": 0.2, "train/clean_ce_loss": 0.007049560546875, "train/clean_loss_scale": 25.39811134338379, "train/clean_loss_weight": 5.079622268676758, "train/combined_loss": 0.2148546278476715, "train/pos_loss/pos0": 0.00680741248652339, "train/pos_loss/pos1": 0.04152004420757294, "train/pos_loss/pos2": 0.25432300567626953, "train/pos_loss/pos3": 0.599449098110199, "train/pos_loss/pos4": 1.0919189453125, "train/pos_loss/pos5": 1.6544607877731323, "train/pos_loss/pos6": 2.2259368896484375, "train/pos_loss/weight0": 2.777179479598999, "train/pos_loss/weight1": 1.977739691734314, "train/pos_loss/weight2": 1.276255488395691, "train/pos_loss/weight3": 0.6213087439537048, "train/pos_loss/weight4": 0.24017353355884552, "train/pos_loss/weight5": 0.07951180636882782, "train/pos_loss/weight6": 0.027830729261040688, "train/pos_weight_progress": 0.27488151658767773, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.17904552817344666 }, { "epoch": 2.748815165876777, "step": 1160, "train/alpha": 0.2, "train/clean_ce_loss": 0.043701171875, "train/clean_loss_scale": 21.367759704589844, "train/clean_loss_weight": 4.273551940917969, "train/combined_loss": 1.1205554008483887, "train/pos_loss/pos0": 0.14800302684307098, "train/pos_loss/pos1": 0.30171626806259155, "train/pos_loss/pos2": 0.8321762681007385, "train/pos_loss/pos3": 1.2380220890045166, "train/pos_loss/pos4": 2.4147751331329346, "train/pos_loss/pos5": 3.174852132797241, "train/pos_loss/pos6": 3.2179501056671143, "train/pos_loss/weight0": 2.01365327835083, "train/pos_loss/weight1": 1.581332802772522, "train/pos_loss/weight2": 1.2232407331466675, "train/pos_loss/weight3": 0.892235279083252, "train/pos_loss/weight4": 0.6105491518974304, "train/pos_loss/weight5": 0.3957032859325409, "train/pos_loss/weight6": 0.28328564763069153, "train/pos_weight_progress": 0.27488151658767773, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9337961077690125 }, { "epoch": 2.748815165876777, "step": 1160, "train/alpha": 0.2, "train/clean_ce_loss": 0.099609375, "train/clean_loss_scale": 7.435662746429443, "train/clean_loss_weight": 1.4871325492858887, "train/combined_loss": 0.8887940645217896, "train/pos_loss/pos0": 0.3171662390232086, "train/pos_loss/pos1": 0.6215772032737732, "train/pos_loss/pos2": 1.0800607204437256, "train/pos_loss/pos3": 1.3389743566513062, "train/pos_loss/pos4": 1.8063772916793823, "train/pos_loss/pos5": 2.1043505668640137, "train/pos_loss/pos6": 2.5148513317108154, "train/pos_loss/weight0": 2.68670654296875, "train/pos_loss/weight1": 1.915137529373169, "train/pos_loss/weight2": 1.2613263130187988, "train/pos_loss/weight3": 0.6948851346969604, "train/pos_loss/weight4": 0.3101130425930023, "train/pos_loss/weight5": 0.0977872759103775, "train/pos_loss/weight6": 0.03404463455080986, "train/pos_weight_progress": 0.27488151658767773, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7406617403030396 }, { "epoch": 2.751184834123223, "step": 1161, "train/alpha": 0.2, "train/clean_ce_loss": 0.06005859375, "train/clean_loss_scale": 7.685551166534424, "train/clean_loss_weight": 1.5371102094650269, "train/combined_loss": 0.5539000630378723, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27511848341232226, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4615834057331085 }, { "epoch": 2.751184834123223, "step": 1161, "train/alpha": 0.2, "train/clean_ce_loss": 0.059814453125, "train/clean_loss_scale": 9.417243003845215, "train/clean_loss_weight": 1.883448600769043, "train/combined_loss": 0.6759446859359741, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27511848341232226, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5632872581481934 }, { "epoch": 2.751184834123223, "step": 1161, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 14.767400741577148, "train/clean_loss_weight": 2.9534802436828613, "train/combined_loss": 1.1594717502593994, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27511848341232226, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9662263989448547 }, { "epoch": 2.751184834123223, "step": 1161, "train/alpha": 0.2, "train/clean_ce_loss": 0.031982421875, "train/clean_loss_scale": 13.031041145324707, "train/clean_loss_weight": 2.606208324432373, "train/combined_loss": 0.5001171231269836, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27511848341232226, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4167642593383789 }, { "epoch": 2.7535545023696684, "step": 1162, "train/alpha": 0.2, "train/clean_ce_loss": 0.036865234375, "train/clean_loss_scale": 22.95271110534668, "train/clean_loss_weight": 4.590542316436768, "train/combined_loss": 1.0153884887695312, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27535545023696684, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8461570739746094 }, { "epoch": 2.7535545023696684, "step": 1162, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 8.228813171386719, "train/clean_loss_weight": 1.6457626819610596, "train/combined_loss": 0.8003805875778198, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27535545023696684, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6669838428497314 }, { "epoch": 2.7535545023696684, "step": 1162, "train/alpha": 0.2, "train/clean_ce_loss": 0.06591796875, "train/clean_loss_scale": 8.92314624786377, "train/clean_loss_weight": 1.784629225730896, "train/combined_loss": 0.7058348059654236, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27535545023696684, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5881956815719604 }, { "epoch": 2.7535545023696684, "step": 1162, "train/alpha": 0.2, "train/clean_ce_loss": 0.126953125, "train/clean_loss_scale": 9.371308326721191, "train/clean_loss_weight": 1.874261736869812, "train/combined_loss": 1.427660346031189, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27535545023696684, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1897169351577759 }, { "epoch": 2.7559241706161135, "step": 1163, "train/alpha": 0.2, "train/clean_ce_loss": 0.000820159912109375, "train/clean_loss_scale": 157.51638793945312, "train/clean_loss_weight": 31.503278732299805, "train/combined_loss": 0.15502634644508362, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27559241706161136, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1291886270046234 }, { "epoch": 2.7559241706161135, "step": 1163, "train/alpha": 0.2, "train/clean_ce_loss": 0.0634765625, "train/clean_loss_scale": 8.391011238098145, "train/clean_loss_weight": 1.6782022714614868, "train/combined_loss": 0.6391590237617493, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27559241706161136, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5326325297355652 }, { "epoch": 2.7559241706161135, "step": 1163, "train/alpha": 0.2, "train/clean_ce_loss": 0.1083984375, "train/clean_loss_scale": 8.595767974853516, "train/clean_loss_weight": 1.719153642654419, "train/combined_loss": 1.1181213855743408, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27559241706161136, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9317678213119507 }, { "epoch": 2.7559241706161135, "step": 1163, "train/alpha": 0.2, "train/clean_ce_loss": 0.06494140625, "train/clean_loss_scale": 8.177668571472168, "train/clean_loss_weight": 1.6355336904525757, "train/combined_loss": 0.6372831463813782, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27559241706161136, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.531069278717041 }, { "epoch": 2.758293838862559, "step": 1164, "train/alpha": 0.2, "train/clean_ce_loss": 0.05908203125, "train/clean_loss_scale": 3.6396398544311523, "train/clean_loss_weight": 0.7279279828071594, "train/combined_loss": 0.2580447793006897, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27582938388625594, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.21503731608390808 }, { "epoch": 2.758293838862559, "step": 1164, "train/alpha": 0.2, "train/clean_ce_loss": 0.028076171875, "train/clean_loss_scale": 18.772165298461914, "train/clean_loss_weight": 3.7544331550598145, "train/combined_loss": 0.6324606537818909, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27582938388625594, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5270505547523499 }, { "epoch": 2.758293838862559, "step": 1164, "train/alpha": 0.2, "train/clean_ce_loss": 0.015869140625, "train/clean_loss_scale": 9.167877197265625, "train/clean_loss_weight": 1.8335754871368408, "train/combined_loss": 0.17458361387252808, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27582938388625594, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.14548633992671967 }, { "epoch": 2.758293838862559, "step": 1164, "train/alpha": 0.2, "train/clean_ce_loss": 0.032958984375, "train/clean_loss_scale": 26.333433151245117, "train/clean_loss_weight": 5.266686916351318, "train/combined_loss": 1.0415078401565552, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27582938388625594, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8679232001304626 }, { "epoch": 2.7606635071090047, "grad_norm": 19.99915313720703, "learning_rate": 2e-06, "loss": 2.9686, "step": 1165 }, { "epoch": 2.7606635071090047, "step": 1165, "train/alpha": 0.2, "train/clean_ce_loss": 0.06396484375, "train/clean_loss_scale": 6.388425350189209, "train/clean_loss_weight": 1.2776850461959839, "train/combined_loss": 0.4903615713119507, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27606635071090047, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.40863463282585144 }, { "epoch": 2.7606635071090047, "step": 1165, "train/alpha": 0.2, "train/clean_ce_loss": 0.060791015625, "train/clean_loss_scale": 10.171810150146484, "train/clean_loss_weight": 2.0343620777130127, "train/combined_loss": 0.74202561378479, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27606635071090047, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6183546781539917 }, { "epoch": 2.7606635071090047, "step": 1165, "train/alpha": 0.2, "train/clean_ce_loss": 0.0234375, "train/clean_loss_scale": 29.054841995239258, "train/clean_loss_weight": 5.810968399047852, "train/combined_loss": 0.8171674609184265, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27606635071090047, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6809728741645813 }, { "epoch": 2.7606635071090047, "step": 1165, "train/alpha": 0.2, "train/clean_ce_loss": 0.040283203125, "train/clean_loss_scale": 11.84170150756836, "train/clean_loss_weight": 2.368340253829956, "train/combined_loss": 0.5724260210990906, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27606635071090047, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4770216643810272 }, { "epoch": 2.7630331753554502, "step": 1166, "train/alpha": 0.2, "train/clean_ce_loss": 0.060546875, "train/clean_loss_scale": 5.85807991027832, "train/clean_loss_weight": 1.1716159582138062, "train/combined_loss": 0.42562612891197205, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27630331753554505, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.354688435792923 }, { "epoch": 2.7630331753554502, "step": 1166, "train/alpha": 0.2, "train/clean_ce_loss": 0.00165557861328125, "train/clean_loss_scale": 127.7889633178711, "train/clean_loss_weight": 25.55779266357422, "train/combined_loss": 0.2538776099681854, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27630331753554505, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.21156467497348785 }, { "epoch": 2.7630331753554502, "step": 1166, "train/alpha": 0.2, "train/clean_ce_loss": 0.0673828125, "train/clean_loss_scale": 11.75107479095459, "train/clean_loss_weight": 2.350214958190918, "train/combined_loss": 0.9501845836639404, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27630331753554505, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7918204665184021 }, { "epoch": 2.7630331753554502, "step": 1166, "train/alpha": 0.2, "train/clean_ce_loss": 0.125, "train/clean_loss_scale": 8.822545051574707, "train/clean_loss_weight": 1.7645090818405151, "train/combined_loss": 1.323381781578064, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27630331753554505, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1028181314468384 }, { "epoch": 2.765402843601896, "step": 1167, "train/alpha": 0.2, "train/clean_ce_loss": 0.0277099609375, "train/clean_loss_scale": 14.2749662399292, "train/clean_loss_weight": 2.8549933433532715, "train/combined_loss": 0.47467049956321716, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27654028436018957, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.39555874466896057 }, { "epoch": 2.765402843601896, "step": 1167, "train/alpha": 0.2, "train/clean_ce_loss": 0.11865234375, "train/clean_loss_scale": 6.478666305541992, "train/clean_loss_weight": 1.2957333326339722, "train/combined_loss": 0.9224507212638855, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27654028436018957, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7687089443206787 }, { "epoch": 2.765402843601896, "step": 1167, "train/alpha": 0.2, "train/clean_ce_loss": 0.011962890625, "train/clean_loss_scale": 15.584245681762695, "train/clean_loss_weight": 3.116849184036255, "train/combined_loss": 0.22371914982795715, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27654028436018957, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1864326298236847 }, { "epoch": 2.765402843601896, "step": 1167, "train/alpha": 0.2, "train/clean_ce_loss": 0.017822265625, "train/clean_loss_scale": 13.29236888885498, "train/clean_loss_weight": 2.6584737300872803, "train/combined_loss": 0.2842801511287689, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27654028436018957, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.23690013587474823 }, { "epoch": 2.7677725118483414, "step": 1168, "train/alpha": 0.2, "train/clean_ce_loss": 0.006011962890625, "train/clean_loss_scale": 34.71700668334961, "train/clean_loss_weight": 6.943401336669922, "train/combined_loss": 0.25046080350875854, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27677725118483415, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.20871734619140625 }, { "epoch": 2.7677725118483414, "step": 1168, "train/alpha": 0.2, "train/clean_ce_loss": 0.027099609375, "train/clean_loss_scale": 32.71136474609375, "train/clean_loss_weight": 6.542273044586182, "train/combined_loss": 1.0637582540512085, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27677725118483415, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8864652514457703 }, { "epoch": 2.7677725118483414, "step": 1168, "train/alpha": 0.2, "train/clean_ce_loss": 0.06298828125, "train/clean_loss_scale": 7.7385945320129395, "train/clean_loss_weight": 1.54771888256073, "train/combined_loss": 0.5849289298057556, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27677725118483415, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4874407649040222 }, { "epoch": 2.7677725118483414, "step": 1168, "train/alpha": 0.2, "train/clean_ce_loss": 0.025390625, "train/clean_loss_scale": 9.70136833190918, "train/clean_loss_weight": 1.940273642539978, "train/combined_loss": 0.29558855295181274, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.27677725118483415, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.24632380902767181 }, { "epoch": 2.770142180094787, "step": 1169, "train/alpha": 0.2, "train/clean_ce_loss": 0.10693359375, "train/clean_loss_scale": 9.517375946044922, "train/clean_loss_weight": 1.9034751653671265, "train/combined_loss": 1.2212727069854736, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.2770142180094787, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.017727255821228 }, { "epoch": 2.770142180094787, "step": 1169, "train/alpha": 0.2, "train/clean_ce_loss": 0.013427734375, "train/clean_loss_scale": 13.176892280578613, "train/clean_loss_weight": 2.63537859916687, "train/combined_loss": 0.21232296526432037, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.2770142180094787, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1769358068704605 }, { "epoch": 2.770142180094787, "step": 1169, "train/alpha": 0.2, "train/clean_ce_loss": 0.10205078125, "train/clean_loss_scale": 6.0843186378479, "train/clean_loss_weight": 1.216863751411438, "train/combined_loss": 0.7450913190841675, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.2770142180094787, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6209094524383545 }, { "epoch": 2.770142180094787, "step": 1169, "train/alpha": 0.2, "train/clean_ce_loss": 0.08056640625, "train/clean_loss_scale": 9.695919036865234, "train/clean_loss_weight": 1.9391838312149048, "train/combined_loss": 0.9373984336853027, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.2770142180094787, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.781165361404419 }, { "epoch": 2.772511848341232, "grad_norm": 16.973373413085938, "learning_rate": 2e-06, "loss": 2.7442, "step": 1170 }, { "epoch": 2.772511848341232, "step": 1170, "train/alpha": 0.2, "train/clean_ce_loss": 0.1357421875, "train/clean_loss_scale": 6.547343730926514, "train/clean_loss_weight": 1.3094687461853027, "train/combined_loss": 1.0665009021759033, "train/pos_loss/pos0": 0.38091641664505005, "train/pos_loss/pos1": 0.7150898575782776, "train/pos_loss/pos2": 1.2008342742919922, "train/pos_loss/pos3": 1.629723072052002, "train/pos_loss/pos4": 2.031923532485962, "train/pos_loss/pos5": 2.464763641357422, "train/pos_loss/pos6": 2.7591500282287598, "train/pos_loss/weight0": 2.645421028137207, "train/pos_loss/weight1": 1.8646210432052612, "train/pos_loss/weight2": 1.2291340827941895, "train/pos_loss/weight3": 0.6857522130012512, "train/pos_loss/weight4": 0.3461723029613495, "train/pos_loss/weight5": 0.15452423691749573, "train/pos_loss/weight6": 0.07437469065189362, "train/pos_weight_progress": 0.2772511848341232, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8887507319450378 }, { "epoch": 2.772511848341232, "step": 1170, "train/alpha": 0.2, "train/clean_ce_loss": 0.08251953125, "train/clean_loss_scale": 8.133320808410645, "train/clean_loss_weight": 1.626664161682129, "train/combined_loss": 0.805389404296875, "train/pos_loss/pos0": 0.26163017749786377, "train/pos_loss/pos1": 0.5720108151435852, "train/pos_loss/pos2": 0.9102228283882141, "train/pos_loss/pos3": 1.522926926612854, "train/pos_loss/pos4": 1.8815195560455322, "train/pos_loss/pos5": 2.5655863285064697, "train/pos_loss/pos6": 2.7396960258483887, "train/pos_loss/weight0": 2.8763582706451416, "train/pos_loss/weight1": 1.9413589239120483, "train/pos_loss/weight2": 1.1827175617218018, "train/pos_loss/weight3": 0.6052981019020081, "train/pos_loss/weight4": 0.2601737976074219, "train/pos_loss/weight5": 0.09706585109233856, "train/pos_loss/weight6": 0.03702762350440025, "train/pos_weight_progress": 0.2772511848341232, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6711578369140625 }, { "epoch": 2.772511848341232, "step": 1170, "train/alpha": 0.2, "train/clean_ce_loss": 0.05126953125, "train/clean_loss_scale": 6.5009765625, "train/clean_loss_weight": 1.300195336341858, "train/combined_loss": 0.3999624252319336, "train/pos_loss/pos0": 0.06024358421564102, "train/pos_loss/pos1": 0.16675181686878204, "train/pos_loss/pos2": 0.3439197242259979, "train/pos_loss/pos3": 0.8491615653038025, "train/pos_loss/pos4": 1.2703279256820679, "train/pos_loss/pos5": 1.9135404825210571, "train/pos_loss/pos6": 2.3734323978424072, "train/pos_loss/weight0": 2.657200336456299, "train/pos_loss/weight1": 1.8745694160461426, "train/pos_loss/weight2": 1.2332795858383179, "train/pos_loss/weight3": 0.6963675022125244, "train/pos_loss/weight4": 0.33285069465637207, "train/pos_loss/weight5": 0.14396555721759796, "train/pos_loss/weight6": 0.06176695227622986, "train/pos_weight_progress": 0.2772511848341232, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.33330202102661133 }, { "epoch": 2.772511848341232, "step": 1170, "train/alpha": 0.2, "train/clean_ce_loss": 0.0257568359375, "train/clean_loss_scale": 42.0388298034668, "train/clean_loss_weight": 8.407766342163086, "train/combined_loss": 1.299344778060913, "train/pos_loss/pos0": 0.164284810423851, "train/pos_loss/pos1": 0.4699791371822357, "train/pos_loss/pos2": 1.1943508386611938, "train/pos_loss/pos3": 1.9329921007156372, "train/pos_loss/pos4": 2.4287467002868652, "train/pos_loss/pos5": 3.3745200634002686, "train/pos_loss/pos6": 4.154601097106934, "train/pos_loss/weight0": 2.151707649230957, "train/pos_loss/weight1": 1.6635204553604126, "train/pos_loss/weight2": 1.2230815887451172, "train/pos_loss/weight3": 0.8782122731208801, "train/pos_loss/weight4": 0.5481761693954468, "train/pos_loss/weight5": 0.32850995659828186, "train/pos_loss/weight6": 0.206792414188385, "train/pos_weight_progress": 0.2772511848341232, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.082787275314331 }, { "epoch": 2.7748815165876777, "step": 1171, "train/alpha": 0.2, "train/clean_ce_loss": 0.0869140625, "train/clean_loss_scale": 7.529857158660889, "train/clean_loss_weight": 1.5059714317321777, "train/combined_loss": 0.7853405475616455, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2774881516587678, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6544504761695862 }, { "epoch": 2.7748815165876777, "step": 1171, "train/alpha": 0.2, "train/clean_ce_loss": 0.1533203125, "train/clean_loss_scale": 5.999431610107422, "train/clean_loss_weight": 1.1998863220214844, "train/combined_loss": 1.1038017272949219, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2774881516587678, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9198347330093384 }, { "epoch": 2.7748815165876777, "step": 1171, "train/alpha": 0.2, "train/clean_ce_loss": 9.489059448242188e-05, "train/clean_loss_scale": 430.865966796875, "train/clean_loss_weight": 86.1731948852539, "train/combined_loss": 0.049062155187129974, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2774881516587678, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.04088512808084488 }, { "epoch": 2.7748815165876777, "step": 1171, "train/alpha": 0.2, "train/clean_ce_loss": 0.06201171875, "train/clean_loss_scale": 6.904083728790283, "train/clean_loss_weight": 1.3808168172836304, "train/combined_loss": 0.5137609243392944, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2774881516587678, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.42813408374786377 }, { "epoch": 2.7772511848341233, "step": 1172, "train/alpha": 0.2, "train/clean_ce_loss": 0.032470703125, "train/clean_loss_scale": 4.23785924911499, "train/clean_loss_weight": 0.847571849822998, "train/combined_loss": 0.16512751579284668, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2777251184834123, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.13760626316070557 }, { "epoch": 2.7772511848341233, "step": 1172, "train/alpha": 0.2, "train/clean_ce_loss": 0.01129150390625, "train/clean_loss_scale": 14.345579147338867, "train/clean_loss_weight": 2.8691158294677734, "train/combined_loss": 0.1943797916173935, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2777251184834123, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1619831621646881 }, { "epoch": 2.7772511848341233, "step": 1172, "train/alpha": 0.2, "train/clean_ce_loss": 0.05224609375, "train/clean_loss_scale": 8.34604263305664, "train/clean_loss_weight": 1.6692085266113281, "train/combined_loss": 0.5232577323913574, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2777251184834123, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.43604812026023865 }, { "epoch": 2.7772511848341233, "step": 1172, "train/alpha": 0.2, "train/clean_ce_loss": 0.01025390625, "train/clean_loss_scale": 26.420654296875, "train/clean_loss_weight": 5.284131050109863, "train/combined_loss": 0.3250978887081146, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2777251184834123, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2709149122238159 }, { "epoch": 2.779620853080569, "step": 1173, "train/alpha": 0.2, "train/clean_ce_loss": 0.07470703125, "train/clean_loss_scale": 12.338942527770996, "train/clean_loss_weight": 2.4677884578704834, "train/combined_loss": 1.1061668395996094, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2779620853080569, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.921805739402771 }, { "epoch": 2.779620853080569, "step": 1173, "train/alpha": 0.2, "train/clean_ce_loss": 0.00030517578125, "train/clean_loss_scale": 369.26605224609375, "train/clean_loss_weight": 73.85321044921875, "train/combined_loss": 0.13522925972938538, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2779620853080569, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.11269105225801468 }, { "epoch": 2.779620853080569, "step": 1173, "train/alpha": 0.2, "train/clean_ce_loss": 0.0888671875, "train/clean_loss_scale": 7.304088115692139, "train/clean_loss_weight": 1.4608176946640015, "train/combined_loss": 0.7789125442504883, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2779620853080569, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.649093747138977 }, { "epoch": 2.779620853080569, "step": 1173, "train/alpha": 0.2, "train/clean_ce_loss": 0.04248046875, "train/clean_loss_scale": 9.552120208740234, "train/clean_loss_weight": 1.9104241132736206, "train/combined_loss": 0.48693424463272095, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2779620853080569, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.40577852725982666 }, { "epoch": 2.781990521327014, "step": 1174, "train/alpha": 0.2, "train/clean_ce_loss": 0.0242919921875, "train/clean_loss_scale": 13.36719036102295, "train/clean_loss_weight": 2.67343807220459, "train/combined_loss": 0.3896588087081909, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2781990521327014, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.32471567392349243 }, { "epoch": 2.781990521327014, "step": 1174, "train/alpha": 0.2, "train/clean_ce_loss": 0.12255859375, "train/clean_loss_scale": 5.780490875244141, "train/clean_loss_weight": 1.1560982465744019, "train/combined_loss": 0.8501386046409607, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2781990521327014, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7084488272666931 }, { "epoch": 2.781990521327014, "step": 1174, "train/alpha": 0.2, "train/clean_ce_loss": 0.08203125, "train/clean_loss_scale": 6.847894668579102, "train/clean_loss_weight": 1.3695789575576782, "train/combined_loss": 0.6740896105766296, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2781990521327014, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5617413520812988 }, { "epoch": 2.781990521327014, "step": 1174, "train/alpha": 0.2, "train/clean_ce_loss": 0.052490234375, "train/clean_loss_scale": 13.604384422302246, "train/clean_loss_weight": 2.720876932144165, "train/combined_loss": 0.8569167852401733, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2781990521327014, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7140973210334778 }, { "epoch": 2.7843601895734595, "grad_norm": 19.74150276184082, "learning_rate": 2e-06, "loss": 2.8381, "step": 1175 }, { "epoch": 2.7843601895734595, "step": 1175, "train/alpha": 0.2, "train/clean_ce_loss": 0.06884765625, "train/clean_loss_scale": 7.246562957763672, "train/clean_loss_weight": 1.4493125677108765, "train/combined_loss": 0.5986906290054321, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.278436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4989088773727417 }, { "epoch": 2.7843601895734595, "step": 1175, "train/alpha": 0.2, "train/clean_ce_loss": 0.0791015625, "train/clean_loss_scale": 7.861639976501465, "train/clean_loss_weight": 1.572327971458435, "train/combined_loss": 0.7462416291236877, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.278436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6218680143356323 }, { "epoch": 2.7843601895734595, "step": 1175, "train/alpha": 0.2, "train/clean_ce_loss": 0.08544921875, "train/clean_loss_scale": 7.30857515335083, "train/clean_loss_weight": 1.4617151021957397, "train/combined_loss": 0.7494144439697266, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.278436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6245120167732239 }, { "epoch": 2.7843601895734595, "step": 1175, "train/alpha": 0.2, "train/clean_ce_loss": 0.107421875, "train/clean_loss_scale": 6.377852439880371, "train/clean_loss_weight": 1.2755705118179321, "train/combined_loss": 0.8221450448036194, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.278436018957346, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6851208806037903 }, { "epoch": 2.786729857819905, "step": 1176, "train/alpha": 0.2, "train/clean_ce_loss": 0.13671875, "train/clean_loss_scale": 8.265535354614258, "train/clean_loss_weight": 1.6531070470809937, "train/combined_loss": 1.3560643196105957, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2786729857819905, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.1300536394119263 }, { "epoch": 2.786729857819905, "step": 1176, "train/alpha": 0.2, "train/clean_ce_loss": 0.05224609375, "train/clean_loss_scale": 4.8351945877075195, "train/clean_loss_weight": 0.9670389294624329, "train/combined_loss": 0.30314403772354126, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2786729857819905, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.25262004137039185 }, { "epoch": 2.786729857819905, "step": 1176, "train/alpha": 0.2, "train/clean_ce_loss": 0.050048828125, "train/clean_loss_scale": 7.011318683624268, "train/clean_loss_weight": 1.4022637605667114, "train/combined_loss": 0.42108994722366333, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2786729857819905, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3509082794189453 }, { "epoch": 2.786729857819905, "step": 1176, "train/alpha": 0.2, "train/clean_ce_loss": 0.0625, "train/clean_loss_scale": 8.061124801635742, "train/clean_loss_weight": 1.6122249364852905, "train/combined_loss": 0.6045843362808228, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2786729857819905, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5038203001022339 }, { "epoch": 2.7890995260663507, "step": 1177, "train/alpha": 0.2, "train/clean_ce_loss": 0.018798828125, "train/clean_loss_scale": 5.466707706451416, "train/clean_loss_weight": 1.093341588973999, "train/combined_loss": 0.12332123517990112, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2789099526066351, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1027676984667778 }, { "epoch": 2.7890995260663507, "step": 1177, "train/alpha": 0.2, "train/clean_ce_loss": 0.0595703125, "train/clean_loss_scale": 16.164003372192383, "train/clean_loss_weight": 3.2328007221221924, "train/combined_loss": 1.1554737091064453, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2789099526066351, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9628947377204895 }, { "epoch": 2.7890995260663507, "step": 1177, "train/alpha": 0.2, "train/clean_ce_loss": 0.04150390625, "train/clean_loss_scale": 23.642290115356445, "train/clean_loss_weight": 4.728457927703857, "train/combined_loss": 1.1774969100952148, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2789099526066351, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9812474250793457 }, { "epoch": 2.7890995260663507, "step": 1177, "train/alpha": 0.2, "train/clean_ce_loss": 0.11767578125, "train/clean_loss_scale": 7.310888767242432, "train/clean_loss_weight": 1.4621777534484863, "train/combined_loss": 1.0323774814605713, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2789099526066351, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8603145480155945 }, { "epoch": 2.7914691943127963, "step": 1178, "train/alpha": 0.2, "train/clean_ce_loss": 0.08349609375, "train/clean_loss_scale": 8.786186218261719, "train/clean_loss_weight": 1.7572373151779175, "train/combined_loss": 0.8803346753120422, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2791469194312796, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7336122393608093 }, { "epoch": 2.7914691943127963, "step": 1178, "train/alpha": 0.2, "train/clean_ce_loss": 0.044677734375, "train/clean_loss_scale": 5.868124008178711, "train/clean_loss_weight": 1.173624873161316, "train/combined_loss": 0.3146093785762787, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2791469194312796, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.26217448711395264 }, { "epoch": 2.7914691943127963, "step": 1178, "train/alpha": 0.2, "train/clean_ce_loss": 0.00982666015625, "train/clean_loss_scale": 12.916159629821777, "train/clean_loss_weight": 2.5832319259643555, "train/combined_loss": 0.15230725705623627, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2791469194312796, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.12692271173000336 }, { "epoch": 2.7914691943127963, "step": 1178, "train/alpha": 0.2, "train/clean_ce_loss": 0.02392578125, "train/clean_loss_scale": 27.607162475585938, "train/clean_loss_weight": 5.521432399749756, "train/combined_loss": 0.7926275134086609, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2791469194312796, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6605229377746582 }, { "epoch": 2.793838862559242, "step": 1179, "train/alpha": 0.2, "train/clean_ce_loss": 0.000331878662109375, "train/clean_loss_scale": 310.0898742675781, "train/clean_loss_weight": 62.017974853515625, "train/combined_loss": 0.12349466234445572, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2793838862559242, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.10291221737861633 }, { "epoch": 2.793838862559242, "step": 1179, "train/alpha": 0.2, "train/clean_ce_loss": 0.16015625, "train/clean_loss_scale": 7.520425319671631, "train/clean_loss_weight": 1.5040850639343262, "train/combined_loss": 1.4453316926956177, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2793838862559242, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.204443097114563 }, { "epoch": 2.793838862559242, "step": 1179, "train/alpha": 0.2, "train/clean_ce_loss": 0.01348876953125, "train/clean_loss_scale": 7.877641201019287, "train/clean_loss_weight": 1.5755282640457153, "train/combined_loss": 0.1275116205215454, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2793838862559242, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.10625968873500824 }, { "epoch": 2.793838862559242, "step": 1179, "train/alpha": 0.2, "train/clean_ce_loss": 0.000492095947265625, "train/clean_loss_scale": 249.4529266357422, "train/clean_loss_weight": 49.890586853027344, "train/combined_loss": 0.14730572700500488, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2793838862559242, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1227547749876976 }, { "epoch": 2.7962085308056874, "grad_norm": 26.485322952270508, "learning_rate": 2e-06, "loss": 2.9674, "step": 1180 }, { "epoch": 2.7962085308056874, "step": 1180, "train/alpha": 0.2, "train/clean_ce_loss": 0.006805419921875, "train/clean_loss_scale": 17.909618377685547, "train/clean_loss_weight": 3.581923723220825, "train/combined_loss": 0.14625896513462067, "train/pos_loss/pos0": 0.0017712379340082407, "train/pos_loss/pos1": 0.015013912692666054, "train/pos_loss/pos2": 0.06157282367348671, "train/pos_loss/pos3": 0.3582725524902344, "train/pos_loss/pos4": 1.192393183708191, "train/pos_loss/pos5": 2.063924551010132, "train/pos_loss/pos6": 2.7463459968566895, "train/pos_loss/weight0": 2.778440475463867, "train/pos_loss/weight1": 1.9486846923828125, "train/pos_loss/weight2": 1.2927038669586182, "train/pos_loss/weight3": 0.6381596922874451, "train/pos_loss/weight4": 0.24332109093666077, "train/pos_loss/weight5": 0.07532934844493866, "train/pos_loss/weight6": 0.023360682651400566, "train/pos_weight_progress": 0.2796208530805687, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.12188247591257095 }, { "epoch": 2.7962085308056874, "step": 1180, "train/alpha": 0.2, "train/clean_ce_loss": 0.0908203125, "train/clean_loss_scale": 14.483189582824707, "train/clean_loss_weight": 2.8966379165649414, "train/combined_loss": 1.5784413814544678, "train/pos_loss/pos0": 0.4215666949748993, "train/pos_loss/pos1": 0.8787373304367065, "train/pos_loss/pos2": 1.4121873378753662, "train/pos_loss/pos3": 1.991158127784729, "train/pos_loss/pos4": 2.506547212600708, "train/pos_loss/pos5": 2.6800363063812256, "train/pos_loss/pos6": 3.631662368774414, "train/pos_loss/weight0": 2.0612680912017822, "train/pos_loss/weight1": 1.6178252696990967, "train/pos_loss/weight2": 1.1991751194000244, "train/pos_loss/weight3": 0.8723132610321045, "train/pos_loss/weight4": 0.600751519203186, "train/pos_loss/weight5": 0.38766443729400635, "train/pos_loss/weight6": 0.2610018849372864, "train/pos_weight_progress": 0.2796208530805687, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.3153678178787231 }, { "epoch": 2.7962085308056874, "step": 1180, "train/alpha": 0.2, "train/clean_ce_loss": 0.032470703125, "train/clean_loss_scale": 2.43794846534729, "train/clean_loss_weight": 0.48758968710899353, "train/combined_loss": 0.09499427676200867, "train/pos_loss/pos0": 0.0007754136458970606, "train/pos_loss/pos1": 0.001151207834482193, "train/pos_loss/pos2": 0.046266697347164154, "train/pos_loss/pos3": 0.2228531390428543, "train/pos_loss/pos4": 0.7567535638809204, "train/pos_loss/pos5": 1.2858538627624512, "train/pos_loss/pos6": 1.986629605293274, "train/pos_loss/weight0": 2.8554768562316895, "train/pos_loss/weight1": 1.968109369277954, "train/pos_loss/weight2": 1.2163068056106567, "train/pos_loss/weight3": 0.5939247608184814, "train/pos_loss/weight4": 0.2535453140735626, "train/pos_loss/weight5": 0.0791737511754036, "train/pos_loss/weight6": 0.033463265746831894, "train/pos_weight_progress": 0.2796208530805687, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.07916189730167389 }, { "epoch": 2.7962085308056874, "step": 1180, "train/alpha": 0.2, "train/clean_ce_loss": 0.046875, "train/clean_loss_scale": 15.517566680908203, "train/clean_loss_weight": 3.103513479232788, "train/combined_loss": 0.8728631138801575, "train/pos_loss/pos0": 0.17159229516983032, "train/pos_loss/pos1": 0.3469087779521942, "train/pos_loss/pos2": 0.6216729879379272, "train/pos_loss/pos3": 0.9627484083175659, "train/pos_loss/pos4": 1.342366099357605, "train/pos_loss/pos5": 1.7498613595962524, "train/pos_loss/pos6": 2.125819683074951, "train/pos_loss/weight0": 1.8491649627685547, "train/pos_loss/weight1": 1.4935493469238281, "train/pos_loss/weight2": 1.173736333847046, "train/pos_loss/weight3": 0.8852199912071228, "train/pos_loss/weight4": 0.6806501150131226, "train/pos_loss/weight5": 0.5015558004379272, "train/pos_loss/weight6": 0.4161234200000763, "train/pos_weight_progress": 0.2796208530805687, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.727385938167572 }, { "epoch": 2.7985781990521326, "step": 1181, "train/alpha": 0.2, "train/clean_ce_loss": 0.006317138671875, "train/clean_loss_scale": 28.067623138427734, "train/clean_loss_weight": 5.613524913787842, "train/combined_loss": 0.21276848018169403, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.27985781990521325, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.17730706930160522 }, { "epoch": 2.7985781990521326, "step": 1181, "train/alpha": 0.2, "train/clean_ce_loss": 0.14453125, "train/clean_loss_scale": 8.878817558288574, "train/clean_loss_weight": 1.7757635116577148, "train/combined_loss": 1.5399198532104492, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.27985781990521325, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.283266544342041 }, { "epoch": 2.7985781990521326, "step": 1181, "train/alpha": 0.2, "train/clean_ce_loss": 0.09521484375, "train/clean_loss_scale": 16.190351486206055, "train/clean_loss_weight": 3.238070249557495, "train/combined_loss": 1.8498742580413818, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.27985781990521325, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.5415618419647217 }, { "epoch": 2.7985781990521326, "step": 1181, "train/alpha": 0.2, "train/clean_ce_loss": 0.0830078125, "train/clean_loss_scale": 11.425703048706055, "train/clean_loss_weight": 2.2851407527923584, "train/combined_loss": 1.138107180595398, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.27985781990521325, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9484226107597351 }, { "epoch": 2.800947867298578, "step": 1182, "train/alpha": 0.2, "train/clean_ce_loss": 0.12060546875, "train/clean_loss_scale": 8.560674667358398, "train/clean_loss_weight": 1.7121349573135376, "train/combined_loss": 1.2389569282531738, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.2800947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0324641466140747 }, { "epoch": 2.800947867298578, "step": 1182, "train/alpha": 0.2, "train/clean_ce_loss": 0.004425048828125, "train/clean_loss_scale": 63.10712432861328, "train/clean_loss_weight": 12.621424674987793, "train/combined_loss": 0.33510252833366394, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.2800947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2792521119117737 }, { "epoch": 2.800947867298578, "step": 1182, "train/alpha": 0.2, "train/clean_ce_loss": 0.06005859375, "train/clean_loss_scale": 9.649686813354492, "train/clean_loss_weight": 1.9299373626708984, "train/combined_loss": 0.6954559683799744, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.2800947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5795466303825378 }, { "epoch": 2.800947867298578, "step": 1182, "train/alpha": 0.2, "train/clean_ce_loss": 0.00848388671875, "train/clean_loss_scale": 24.313486099243164, "train/clean_loss_weight": 4.862697124481201, "train/combined_loss": 0.24752742052078247, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.2800947867298578, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.20627285540103912 }, { "epoch": 2.8033175355450237, "step": 1183, "train/alpha": 0.2, "train/clean_ce_loss": 0.0751953125, "train/clean_loss_scale": 11.200784683227539, "train/clean_loss_weight": 2.240156888961792, "train/combined_loss": 1.0106958150863647, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28033175355450235, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8422464728355408 }, { "epoch": 2.8033175355450237, "step": 1183, "train/alpha": 0.2, "train/clean_ce_loss": 0.053955078125, "train/clean_loss_scale": 11.785151481628418, "train/clean_loss_weight": 2.3570303916931152, "train/combined_loss": 0.7630425691604614, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28033175355450235, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6358687877655029 }, { "epoch": 2.8033175355450237, "step": 1183, "train/alpha": 0.2, "train/clean_ce_loss": 0.001983642578125, "train/clean_loss_scale": 141.100830078125, "train/clean_loss_weight": 28.22016716003418, "train/combined_loss": 0.3358723223209381, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28033175355450235, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2798936069011688 }, { "epoch": 2.8033175355450237, "step": 1183, "train/alpha": 0.2, "train/clean_ce_loss": 0.10595703125, "train/clean_loss_scale": 9.740072250366211, "train/clean_loss_weight": 1.948014497756958, "train/combined_loss": 1.2384350299835205, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28033175355450235, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.032029151916504 }, { "epoch": 2.8056872037914693, "step": 1184, "train/alpha": 0.2, "train/clean_ce_loss": 0.0849609375, "train/clean_loss_scale": 11.079754829406738, "train/clean_loss_weight": 2.2159509658813477, "train/combined_loss": 1.1296156644821167, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28056872037914693, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9413463473320007 }, { "epoch": 2.8056872037914693, "step": 1184, "train/alpha": 0.2, "train/clean_ce_loss": 0.0002460479736328125, "train/clean_loss_scale": 904.986083984375, "train/clean_loss_weight": 180.99722290039062, "train/combined_loss": 0.2672039866447449, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28056872037914693, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.22266998887062073 }, { "epoch": 2.8056872037914693, "step": 1184, "train/alpha": 0.2, "train/clean_ce_loss": 0.064453125, "train/clean_loss_scale": 18.207019805908203, "train/clean_loss_weight": 3.641403913497925, "train/combined_loss": 1.4081991910934448, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28056872037914693, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.173499345779419 }, { "epoch": 2.8056872037914693, "step": 1184, "train/alpha": 0.2, "train/clean_ce_loss": 0.0693359375, "train/clean_loss_scale": 10.411036491394043, "train/clean_loss_weight": 2.082207441329956, "train/combined_loss": 0.8662307858467102, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28056872037914693, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7218589782714844 }, { "epoch": 2.8080568720379144, "grad_norm": 38.6883544921875, "learning_rate": 2e-06, "loss": 3.4649, "step": 1185 }, { "epoch": 2.8080568720379144, "step": 1185, "train/alpha": 0.2, "train/clean_ce_loss": 0.0208740234375, "train/clean_loss_scale": 6.973996162414551, "train/clean_loss_weight": 1.3947992324829102, "train/combined_loss": 0.17469042539596558, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28080568720379145, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1455753594636917 }, { "epoch": 2.8080568720379144, "step": 1185, "train/alpha": 0.2, "train/clean_ce_loss": 0.05224609375, "train/clean_loss_scale": 10.905715942382812, "train/clean_loss_weight": 2.181143283843994, "train/combined_loss": 0.6837372779846191, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28080568720379145, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5697810649871826 }, { "epoch": 2.8080568720379144, "step": 1185, "train/alpha": 0.2, "train/clean_ce_loss": 0.0869140625, "train/clean_loss_scale": 8.224039077758789, "train/clean_loss_weight": 1.6448078155517578, "train/combined_loss": 0.8577415347099304, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28080568720379145, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7147846221923828 }, { "epoch": 2.8080568720379144, "step": 1185, "train/alpha": 0.2, "train/clean_ce_loss": 0.0279541015625, "train/clean_loss_scale": 8.742506980895996, "train/clean_loss_weight": 1.7485014200210571, "train/combined_loss": 0.2932667136192322, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28080568720379145, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.24438892304897308 }, { "epoch": 2.81042654028436, "step": 1186, "train/alpha": 0.2, "train/clean_ce_loss": 0.03466796875, "train/clean_loss_scale": 28.229774475097656, "train/clean_loss_weight": 5.6459550857543945, "train/combined_loss": 1.1744027137756348, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28104265402843603, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9786689281463623 }, { "epoch": 2.81042654028436, "step": 1186, "train/alpha": 0.2, "train/clean_ce_loss": 0.00104522705078125, "train/clean_loss_scale": 166.4994659423828, "train/clean_loss_weight": 33.29989242553711, "train/combined_loss": 0.20883570611476898, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28104265402843603, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.17402975261211395 }, { "epoch": 2.81042654028436, "step": 1186, "train/alpha": 0.2, "train/clean_ce_loss": 0.024658203125, "train/clean_loss_scale": 14.233244895935059, "train/clean_loss_weight": 2.846648931503296, "train/combined_loss": 0.4211595058441162, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28104265402843603, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.3509662449359894 }, { "epoch": 2.81042654028436, "step": 1186, "train/alpha": 0.2, "train/clean_ce_loss": 0.029541015625, "train/clean_loss_scale": 13.115944862365723, "train/clean_loss_weight": 2.6231889724731445, "train/combined_loss": 0.46494999527931213, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28104265402843603, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.38745832443237305 }, { "epoch": 2.8127962085308056, "step": 1187, "train/alpha": 0.2, "train/clean_ce_loss": 0.1455078125, "train/clean_loss_scale": 7.1514105796813965, "train/clean_loss_weight": 1.4302821159362793, "train/combined_loss": 1.2487033605575562, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28127962085308056, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0405861139297485 }, { "epoch": 2.8127962085308056, "step": 1187, "train/alpha": 0.2, "train/clean_ce_loss": 0.09228515625, "train/clean_loss_scale": 11.433300971984863, "train/clean_loss_weight": 2.2866601943969727, "train/combined_loss": 1.2661486864089966, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28127962085308056, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0551239252090454 }, { "epoch": 2.8127962085308056, "step": 1187, "train/alpha": 0.2, "train/clean_ce_loss": 0.030517578125, "train/clean_loss_scale": 5.351130962371826, "train/clean_loss_weight": 1.0702261924743652, "train/combined_loss": 0.1959642618894577, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28127962085308056, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.16330355405807495 }, { "epoch": 2.8127962085308056, "step": 1187, "train/alpha": 0.2, "train/clean_ce_loss": 0.0810546875, "train/clean_loss_scale": 7.568455696105957, "train/clean_loss_weight": 1.5136911869049072, "train/combined_loss": 0.7361505627632141, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28127962085308056, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6134588122367859 }, { "epoch": 2.815165876777251, "step": 1188, "train/alpha": 0.2, "train/clean_ce_loss": 0.09619140625, "train/clean_loss_scale": 8.877164840698242, "train/clean_loss_weight": 1.7754329442977905, "train/combined_loss": 1.0246883630752563, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28151658767772514, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8539069890975952 }, { "epoch": 2.815165876777251, "step": 1188, "train/alpha": 0.2, "train/clean_ce_loss": 0.00909423828125, "train/clean_loss_scale": 30.40777015686035, "train/clean_loss_weight": 6.081553936004639, "train/combined_loss": 0.3318426012992859, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28151658767772514, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.2765355110168457 }, { "epoch": 2.815165876777251, "step": 1188, "train/alpha": 0.2, "train/clean_ce_loss": 0.1708984375, "train/clean_loss_scale": 7.553890705108643, "train/clean_loss_weight": 1.5107781887054443, "train/combined_loss": 1.549137830734253, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28151658767772514, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2909481525421143 }, { "epoch": 2.815165876777251, "step": 1188, "train/alpha": 0.2, "train/clean_ce_loss": 0.03564453125, "train/clean_loss_scale": 3.7900390625, "train/clean_loss_weight": 0.758007824420929, "train/combined_loss": 0.1621129959821701, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28151658767772514, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.13509416580200195 }, { "epoch": 2.8175355450236967, "step": 1189, "train/alpha": 0.2, "train/clean_ce_loss": 0.01446533203125, "train/clean_loss_scale": 23.490480422973633, "train/clean_loss_weight": 4.69809627532959, "train/combined_loss": 0.40775710344314575, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28175355450236966, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.33979758620262146 }, { "epoch": 2.8175355450236967, "step": 1189, "train/alpha": 0.2, "train/clean_ce_loss": 0.033935546875, "train/clean_loss_scale": 12.651132583618164, "train/clean_loss_weight": 2.530226469039917, "train/combined_loss": 0.5151877403259277, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28175355450236966, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.42932310700416565 }, { "epoch": 2.8175355450236967, "step": 1189, "train/alpha": 0.2, "train/clean_ce_loss": 0.052734375, "train/clean_loss_scale": 2.6941335201263428, "train/clean_loss_weight": 0.5388267040252686, "train/combined_loss": 0.1704881489276886, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28175355450236966, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.14207345247268677 }, { "epoch": 2.8175355450236967, "step": 1189, "train/alpha": 0.2, "train/clean_ce_loss": 0.0185546875, "train/clean_loss_scale": 39.59444808959961, "train/clean_loss_weight": 7.91888952255249, "train/combined_loss": 0.8815951347351074, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28175355450236966, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7346625924110413 }, { "epoch": 2.8199052132701423, "grad_norm": 20.34628677368164, "learning_rate": 2e-06, "loss": 3.6408, "step": 1190 }, { "epoch": 2.8199052132701423, "step": 1190, "train/alpha": 0.2, "train/clean_ce_loss": 8.535385131835938e-05, "train/clean_loss_scale": 1991.365966796875, "train/clean_loss_weight": 398.273193359375, "train/combined_loss": 0.20396490395069122, "train/pos_loss/pos0": 0.000610874209087342, "train/pos_loss/pos1": 0.00561969680711627, "train/pos_loss/pos2": 0.040079742670059204, "train/pos_loss/pos3": 0.22326017916202545, "train/pos_loss/pos4": 0.5477797389030457, "train/pos_loss/pos5": 1.2827779054641724, "train/pos_loss/pos6": 2.0803356170654297, "train/pos_loss/weight0": 2.3027939796447754, "train/pos_loss/weight1": 1.7383785247802734, "train/pos_loss/weight2": 1.2384533882141113, "train/pos_loss/weight3": 0.8041070103645325, "train/pos_loss/weight4": 0.48200342059135437, "train/pos_loss/weight5": 0.271797239780426, "train/pos_loss/weight6": 0.16246652603149414, "train/pos_weight_progress": 0.28199052132701424, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.16997075080871582 }, { "epoch": 2.8199052132701423, "step": 1190, "train/alpha": 0.2, "train/clean_ce_loss": 0.00933837890625, "train/clean_loss_scale": 55.80771255493164, "train/clean_loss_weight": 11.161542892456055, "train/combined_loss": 0.6253842711448669, "train/pos_loss/pos0": 0.03710482642054558, "train/pos_loss/pos1": 0.10672017931938171, "train/pos_loss/pos2": 0.2570638954639435, "train/pos_loss/pos3": 0.6614202857017517, "train/pos_loss/pos4": 1.2015588283538818, "train/pos_loss/pos5": 1.5504652261734009, "train/pos_loss/pos6": 2.0512619018554688, "train/pos_loss/weight0": 1.8182694911956787, "train/pos_loss/weight1": 1.48306405544281, "train/pos_loss/weight2": 1.162360668182373, "train/pos_loss/weight3": 0.8979480862617493, "train/pos_loss/weight4": 0.6778573989868164, "train/pos_loss/weight5": 0.5068057775497437, "train/pos_loss/weight6": 0.4536946713924408, "train/pos_weight_progress": 0.28199052132701424, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5211535692214966 }, { "epoch": 2.8199052132701423, "step": 1190, "train/alpha": 0.2, "train/clean_ce_loss": 0.0654296875, "train/clean_loss_scale": 18.607698440551758, "train/clean_loss_weight": 3.7215397357940674, "train/combined_loss": 1.460995078086853, "train/pos_loss/pos0": 0.20021596550941467, "train/pos_loss/pos1": 0.7383726835250854, "train/pos_loss/pos2": 1.4419875144958496, "train/pos_loss/pos3": 2.006178140640259, "train/pos_loss/pos4": 2.412412166595459, "train/pos_loss/pos5": 2.953498363494873, "train/pos_loss/pos6": 3.3553006649017334, "train/pos_loss/weight0": 2.0587239265441895, "train/pos_loss/weight1": 1.619879961013794, "train/pos_loss/weight2": 1.2147278785705566, "train/pos_loss/weight3": 0.8784217238426208, "train/pos_loss/weight4": 0.5992117524147034, "train/pos_loss/weight5": 0.36956945061683655, "train/pos_loss/weight6": 0.2594650685787201, "train/pos_weight_progress": 0.28199052132701424, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2174959182739258 }, { "epoch": 2.8199052132701423, "step": 1190, "train/alpha": 0.2, "train/clean_ce_loss": 0.1884765625, "train/clean_loss_scale": 6.607340335845947, "train/clean_loss_weight": 1.3214681148529053, "train/combined_loss": 1.4943945407867432, "train/pos_loss/pos0": 0.5407499670982361, "train/pos_loss/pos1": 1.2058132886886597, "train/pos_loss/pos2": 1.7718948125839233, "train/pos_loss/pos3": 2.3919670581817627, "train/pos_loss/pos4": 3.1857192516326904, "train/pos_loss/pos5": 3.41237735748291, "train/pos_loss/pos6": 3.589008092880249, "train/pos_loss/weight0": 2.7936513423919678, "train/pos_loss/weight1": 2.026235342025757, "train/pos_loss/weight2": 1.2189462184906006, "train/pos_loss/weight3": 0.6132214665412903, "train/pos_loss/weight4": 0.24075478315353394, "train/pos_loss/weight5": 0.07965933531522751, "train/pos_loss/weight6": 0.027531592175364494, "train/pos_weight_progress": 0.28199052132701424, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.2453287839889526 }, { "epoch": 2.822274881516588, "step": 1191, "train/alpha": 0.2, "train/clean_ce_loss": 0.00016307830810546875, "train/clean_loss_scale": 261.8929443359375, "train/clean_loss_weight": 52.37858963012695, "train/combined_loss": 0.051250871270895004, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.28222748815165877, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.04270906001329422 }, { "epoch": 2.822274881516588, "step": 1191, "train/alpha": 0.2, "train/clean_ce_loss": 0.0034027099609375, "train/clean_loss_scale": 18.584936141967773, "train/clean_loss_weight": 3.716987371444702, "train/combined_loss": 0.07588697969913483, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.28222748815165877, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.06323914974927902 }, { "epoch": 2.822274881516588, "step": 1191, "train/alpha": 0.2, "train/clean_ce_loss": 0.08056640625, "train/clean_loss_scale": 5.367517471313477, "train/clean_loss_weight": 1.0735034942626953, "train/combined_loss": 0.5189298987388611, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.28222748815165877, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4324415922164917 }, { "epoch": 2.822274881516588, "step": 1191, "train/alpha": 0.2, "train/clean_ce_loss": 0.232421875, "train/clean_loss_scale": 3.7683119773864746, "train/clean_loss_weight": 0.7536624073982239, "train/combined_loss": 1.0510058403015137, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.28222748815165877, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8758381605148315 }, { "epoch": 2.824644549763033, "step": 1192, "train/alpha": 0.2, "train/clean_ce_loss": 0.060791015625, "train/clean_loss_scale": 9.569479942321777, "train/clean_loss_weight": 1.9138959646224976, "train/combined_loss": 0.6980860829353333, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.2824644549763033, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5817384123802185 }, { "epoch": 2.824644549763033, "step": 1192, "train/alpha": 0.2, "train/clean_ce_loss": 0.0034027099609375, "train/clean_loss_scale": 39.189517974853516, "train/clean_loss_weight": 7.8379034996032715, "train/combined_loss": 0.16002067923545837, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.2824644549763033, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.13335056602954865 }, { "epoch": 2.824644549763033, "step": 1192, "train/alpha": 0.2, "train/clean_ce_loss": 0.10498046875, "train/clean_loss_scale": 5.916665077209473, "train/clean_loss_weight": 1.1833330392837524, "train/combined_loss": 0.7453611493110657, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.2824644549763033, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6211342811584473 }, { "epoch": 2.824644549763033, "step": 1192, "train/alpha": 0.2, "train/clean_ce_loss": 0.1015625, "train/clean_loss_scale": 8.727588653564453, "train/clean_loss_weight": 1.7455177307128906, "train/combined_loss": 1.0636749267578125, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.2824644549763033, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8863957524299622 }, { "epoch": 2.8270142180094786, "step": 1193, "train/alpha": 0.2, "train/clean_ce_loss": 0.03369140625, "train/clean_loss_scale": 6.232419013977051, "train/clean_loss_weight": 1.2464838027954102, "train/combined_loss": 0.25197476148605347, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.28270142180094787, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.20997895300388336 }, { "epoch": 2.8270142180094786, "step": 1193, "train/alpha": 0.2, "train/clean_ce_loss": 0.099609375, "train/clean_loss_scale": 7.502401828765869, "train/clean_loss_weight": 1.5004804134368896, "train/combined_loss": 0.8967714905738831, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.28270142180094787, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7473095655441284 }, { "epoch": 2.8270142180094786, "step": 1193, "train/alpha": 0.2, "train/clean_ce_loss": 0.0093994140625, "train/clean_loss_scale": 8.473188400268555, "train/clean_loss_weight": 1.694637656211853, "train/combined_loss": 0.0955716073513031, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.28270142180094787, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.07964300364255905 }, { "epoch": 2.8270142180094786, "step": 1193, "train/alpha": 0.2, "train/clean_ce_loss": 0.08544921875, "train/clean_loss_scale": 6.606820106506348, "train/clean_loss_weight": 1.3213640451431274, "train/combined_loss": 0.6774570941925049, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.28270142180094787, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.564547598361969 }, { "epoch": 2.829383886255924, "step": 1194, "train/alpha": 0.2, "train/clean_ce_loss": 0.062255859375, "train/clean_loss_scale": 9.796517372131348, "train/clean_loss_weight": 1.9593034982681274, "train/combined_loss": 0.7318686842918396, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.2829383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6098905801773071 }, { "epoch": 2.829383886255924, "step": 1194, "train/alpha": 0.2, "train/clean_ce_loss": 0.0732421875, "train/clean_loss_scale": 6.367971420288086, "train/clean_loss_weight": 1.2735942602157593, "train/combined_loss": 0.5596849918365479, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.2829383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.46640413999557495 }, { "epoch": 2.829383886255924, "step": 1194, "train/alpha": 0.2, "train/clean_ce_loss": 0.00165557861328125, "train/clean_loss_scale": 30.49410629272461, "train/clean_loss_weight": 6.09882116317749, "train/combined_loss": 0.060582470148801804, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.2829383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.05048539116978645 }, { "epoch": 2.829383886255924, "step": 1194, "train/alpha": 0.2, "train/clean_ce_loss": 0.11279296875, "train/clean_loss_scale": 6.182296276092529, "train/clean_loss_weight": 1.2364592552185059, "train/combined_loss": 0.8367834687232971, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.2829383886255924, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6973195672035217 }, { "epoch": 2.8317535545023698, "grad_norm": 22.315946578979492, "learning_rate": 2e-06, "loss": 2.9377, "step": 1195 }, { "epoch": 2.8317535545023698, "step": 1195, "train/alpha": 0.2, "train/clean_ce_loss": 0.1005859375, "train/clean_loss_scale": 8.149330139160156, "train/clean_loss_weight": 1.6298660039901733, "train/combined_loss": 0.9836496710777283, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.283175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8197080492973328 }, { "epoch": 2.8317535545023698, "step": 1195, "train/alpha": 0.2, "train/clean_ce_loss": 0.0693359375, "train/clean_loss_scale": 7.197807312011719, "train/clean_loss_weight": 1.4395614862442017, "train/combined_loss": 0.5988800525665283, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.283175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.49906671047210693 }, { "epoch": 2.8317535545023698, "step": 1195, "train/alpha": 0.2, "train/clean_ce_loss": 0.035888671875, "train/clean_loss_scale": 3.6357433795928955, "train/clean_loss_weight": 0.727148711681366, "train/combined_loss": 0.1565784066915512, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.283175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.1304820030927658 }, { "epoch": 2.8317535545023698, "step": 1195, "train/alpha": 0.2, "train/clean_ce_loss": 0.04248046875, "train/clean_loss_scale": 12.817588806152344, "train/clean_loss_weight": 2.5635178089141846, "train/combined_loss": 0.6533966064453125, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.283175355450237, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5444971919059753 }, { "epoch": 2.834123222748815, "step": 1196, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 6.491938591003418, "train/clean_loss_weight": 1.2983877658843994, "train/combined_loss": 0.551561176776886, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.2834123222748815, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.45963433384895325 }, { "epoch": 2.834123222748815, "step": 1196, "train/alpha": 0.2, "train/clean_ce_loss": 0.08251953125, "train/clean_loss_scale": 6.213635444641113, "train/clean_loss_weight": 1.2427271604537964, "train/combined_loss": 0.6152955293655396, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.2834123222748815, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.512746274471283 }, { "epoch": 2.834123222748815, "step": 1196, "train/alpha": 0.2, "train/clean_ce_loss": 0.0771484375, "train/clean_loss_scale": 7.123230934143066, "train/clean_loss_weight": 1.424646258354187, "train/combined_loss": 0.6594553589820862, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.2834123222748815, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5495461225509644 }, { "epoch": 2.834123222748815, "step": 1196, "train/alpha": 0.2, "train/clean_ce_loss": 0.06689453125, "train/clean_loss_scale": 7.92313289642334, "train/clean_loss_weight": 1.58462655544281, "train/combined_loss": 0.6360171437263489, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.2834123222748815, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5300142765045166 }, { "epoch": 2.8364928909952605, "step": 1197, "train/alpha": 0.2, "train/clean_ce_loss": 0.09912109375, "train/clean_loss_scale": 10.710678100585938, "train/clean_loss_weight": 2.1421356201171875, "train/combined_loss": 1.2739849090576172, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.2836492890995261, "train/pos_weight_scheme": 1.0, "train/task_loss": 1.0616540908813477 }, { "epoch": 2.8364928909952605, "step": 1197, "train/alpha": 0.2, "train/clean_ce_loss": 0.08935546875, "train/clean_loss_scale": 5.112609386444092, "train/clean_loss_weight": 1.0225218534469604, "train/combined_loss": 0.5482075214385986, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.2836492890995261, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4568396210670471 }, { "epoch": 2.8364928909952605, "step": 1197, "train/alpha": 0.2, "train/clean_ce_loss": 0.11962890625, "train/clean_loss_scale": 7.7460503578186035, "train/clean_loss_weight": 1.5492100715637207, "train/combined_loss": 1.1119818687438965, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.2836492890995261, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.9266515374183655 }, { "epoch": 2.8364928909952605, "step": 1197, "train/alpha": 0.2, "train/clean_ce_loss": 0.036865234375, "train/clean_loss_scale": 8.840333938598633, "train/clean_loss_weight": 1.7680667638778687, "train/combined_loss": 0.39108115434646606, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.2836492890995261, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.32590097188949585 }, { "epoch": 2.838862559241706, "step": 1198, "train/alpha": 0.2, "train/clean_ce_loss": 0.09375, "train/clean_loss_scale": 6.210536479949951, "train/clean_loss_weight": 1.2421072721481323, "train/combined_loss": 0.698685348033905, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.2838862559241706, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5822377800941467 }, { "epoch": 2.838862559241706, "step": 1198, "train/alpha": 0.2, "train/clean_ce_loss": 0.033203125, "train/clean_loss_scale": 13.820962905883789, "train/clean_loss_weight": 2.764192581176758, "train/combined_loss": 0.5506789684295654, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.2838862559241706, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4588991701602936 }, { "epoch": 2.838862559241706, "step": 1198, "train/alpha": 0.2, "train/clean_ce_loss": 0.134765625, "train/clean_loss_scale": 5.5290703773498535, "train/clean_loss_weight": 1.1058140993118286, "train/combined_loss": 0.8941543698310852, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.2838862559241706, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.7451286315917969 }, { "epoch": 2.838862559241706, "step": 1198, "train/alpha": 0.2, "train/clean_ce_loss": 0.11328125, "train/clean_loss_scale": 7.521058559417725, "train/clean_loss_weight": 1.5042117834091187, "train/combined_loss": 1.0223939418792725, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.2838862559241706, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.8519949316978455 }, { "epoch": 2.8412322274881516, "step": 1199, "train/alpha": 0.2, "train/clean_ce_loss": 0.07080078125, "train/clean_loss_scale": 6.395112037658691, "train/clean_loss_weight": 1.279022455215454, "train/combined_loss": 0.5433347225189209, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.2841232227488152, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.4527789354324341 }, { "epoch": 2.8412322274881516, "step": 1199, "train/alpha": 0.2, "train/clean_ce_loss": 0.022216796875, "train/clean_loss_scale": 27.346555709838867, "train/clean_loss_weight": 5.469311237335205, "train/combined_loss": 0.7290634512901306, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.2841232227488152, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.6075528860092163 }, { "epoch": 2.8412322274881516, "step": 1199, "train/alpha": 0.2, "train/clean_ce_loss": 0.02001953125, "train/clean_loss_scale": 26.303503036499023, "train/clean_loss_weight": 5.260700702667236, "train/combined_loss": 0.6319005489349365, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.2841232227488152, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5265837907791138 }, { "epoch": 2.8412322274881516, "step": 1199, "train/alpha": 0.2, "train/clean_ce_loss": 0.0272216796875, "train/clean_loss_scale": 21.470436096191406, "train/clean_loss_weight": 4.2940874099731445, "train/combined_loss": 0.7013536095619202, "train/pos_loss/weight0": 3.3532185554504395, "train/pos_loss/weight1": 2.0551209449768066, "train/pos_loss/weight2": 1.0432041883468628, "train/pos_loss/weight3": 0.4040808379650116, "train/pos_loss/weight4": 0.11551930755376816, "train/pos_loss/weight5": 0.023991668596863747, "train/pos_loss/weight6": 0.004864179063588381, "train/pos_weight_progress": 0.2841232227488152, "train/pos_weight_scheme": 1.0, "train/task_loss": 0.5844613313674927 }, { "epoch": 2.843601895734597, "grad_norm": 24.14377212524414, "learning_rate": 2e-06, "loss": 3.0911, "step": 1200 } ], "logging_steps": 5, "max_steps": 4220, "num_input_tokens_seen": 0, "num_train_epochs": 10, "save_steps": 400, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": true, "should_training_stop": false }, "attributes": {} } }, "total_flos": 4.10831276218096e+18, "train_batch_size": 1, "trial_name": null, "trial_params": null }