{ "best_global_step": null, "best_metric": null, "best_model_checkpoint": null, "epoch": 6.1731066460587325, "eval_steps": 250, "global_step": 2000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.07727975270479134, "grad_norm": 50.397064208984375, "learning_rate": 3e-06, "loss": 14.117890625, "step": 25 }, { "epoch": 0.1545595054095827, "grad_norm": 47.54403305053711, "learning_rate": 6.125e-06, "loss": 9.375330810546876, "step": 50 }, { "epoch": 0.23183925811437403, "grad_norm": 53.90830993652344, "learning_rate": 9.25e-06, "loss": 7.50126708984375, "step": 75 }, { "epoch": 0.3091190108191654, "grad_norm": 88.0354232788086, "learning_rate": 1.2375000000000001e-05, "loss": 6.569615478515625, "step": 100 }, { "epoch": 0.38639876352395675, "grad_norm": 90.9751205444336, "learning_rate": 1.55e-05, "loss": 5.774239501953125, "step": 125 }, { "epoch": 0.46367851622874806, "grad_norm": 54.41282272338867, "learning_rate": 1.8625000000000002e-05, "loss": 4.550369873046875, "step": 150 }, { "epoch": 0.5409582689335394, "grad_norm": 56.61479949951172, "learning_rate": 2.175e-05, "loss": 3.0210330200195314, "step": 175 }, { "epoch": 0.6182380216383307, "grad_norm": 53.15815734863281, "learning_rate": 2.4875e-05, "loss": 1.905018310546875, "step": 200 }, { "epoch": 0.6955177743431221, "grad_norm": 23.92143440246582, "learning_rate": 2.498903537623573e-05, "loss": 1.2014694213867188, "step": 225 }, { "epoch": 0.7727975270479135, "grad_norm": 18.9036865234375, "learning_rate": 2.495431620238633e-05, "loss": 0.9354300689697266, "step": 250 }, { "epoch": 0.7727975270479135, "eval_cer": 14.2451, "eval_chrf": 74.6466, "eval_der": 18.4991, "eval_ier": 1.3881, "eval_loss": 0.10310772806406021, "eval_rtf": 0.1069, "eval_runtime": 802.3284, "eval_samples_per_second": 2.867, "eval_ser": 97.6997, "eval_steps_per_second": 0.179, "eval_wer": 60.5565, "step": 250 }, { "epoch": 0.8500772797527048, "grad_norm": 24.115568161010742, "learning_rate": 2.489588952978863e-05, "loss": 0.7913719177246094, "step": 275 }, { "epoch": 0.9273570324574961, "grad_norm": 13.435430526733398, "learning_rate": 2.4813866576934676e-05, "loss": 0.7055984497070312, "step": 300 }, { "epoch": 1.0030911901081916, "grad_norm": 11.58161449432373, "learning_rate": 2.47084034791771e-05, "loss": 0.6120201873779297, "step": 325 }, { "epoch": 1.080370942812983, "grad_norm": 13.320256233215332, "learning_rate": 2.457970099151662e-05, "loss": 0.5029972457885742, "step": 350 }, { "epoch": 1.1576506955177743, "grad_norm": 13.035049438476562, "learning_rate": 2.4428004106453462e-05, "loss": 0.47681015014648437, "step": 375 }, { "epoch": 1.2349304482225656, "grad_norm": 9.085765838623047, "learning_rate": 2.4253601587630236e-05, "loss": 0.4498796081542969, "step": 400 }, { "epoch": 1.312210200927357, "grad_norm": 10.488614082336426, "learning_rate": 2.4056825420153917e-05, "loss": 0.4376938629150391, "step": 425 }, { "epoch": 1.3894899536321483, "grad_norm": 12.869024276733398, "learning_rate": 2.3838050178643312e-05, "loss": 0.4292305374145508, "step": 450 }, { "epoch": 1.4667697063369398, "grad_norm": 7.625246524810791, "learning_rate": 2.3597692314205016e-05, "loss": 0.3776122283935547, "step": 475 }, { "epoch": 1.544049459041731, "grad_norm": 8.058732032775879, "learning_rate": 2.3336209361695035e-05, "loss": 0.3638324356079102, "step": 500 }, { "epoch": 1.544049459041731, "eval_cer": 6.4586, "eval_chrf": 88.1309, "eval_der": 4.8453, "eval_ier": 4.2226, "eval_loss": 0.049306295812129974, "eval_rtf": 0.1087, "eval_runtime": 815.7783, "eval_samples_per_second": 2.819, "eval_ser": 82.1615, "eval_steps_per_second": 0.177, "eval_wer": 33.982, "step": 500 }, { "epoch": 1.6213292117465223, "grad_norm": 10.343655586242676, "learning_rate": 2.305409906877519e-05, "loss": 0.3691394805908203, "step": 525 }, { "epoch": 1.6986089644513136, "grad_norm": 7.9086761474609375, "learning_rate": 2.2751898448422155e-05, "loss": 0.35427001953125, "step": 550 }, { "epoch": 1.775888717156105, "grad_norm": 6.961205005645752, "learning_rate": 2.2430182756692728e-05, "loss": 0.3289776611328125, "step": 575 }, { "epoch": 1.8531684698608966, "grad_norm": 7.977571964263916, "learning_rate": 2.208956439769125e-05, "loss": 0.3318151092529297, "step": 600 }, { "epoch": 1.9304482225656878, "grad_norm": 7.199207782745361, "learning_rate": 2.1730691757823553e-05, "loss": 0.298912410736084, "step": 625 }, { "epoch": 2.006182380216383, "grad_norm": 5.2214226722717285, "learning_rate": 2.135424797155661e-05, "loss": 0.2817891311645508, "step": 650 }, { "epoch": 2.0834621329211744, "grad_norm": 4.931973457336426, "learning_rate": 2.096094962103326e-05, "loss": 0.18875328063964844, "step": 675 }, { "epoch": 2.160741885625966, "grad_norm": 5.099040985107422, "learning_rate": 2.0551545372017332e-05, "loss": 0.1921413803100586, "step": 700 }, { "epoch": 2.2380216383307574, "grad_norm": 4.408350467681885, "learning_rate": 2.012681454876585e-05, "loss": 0.2022266387939453, "step": 725 }, { "epoch": 2.3153013910355487, "grad_norm": 4.918865203857422, "learning_rate": 1.9687565650540984e-05, "loss": 0.18742366790771484, "step": 750 }, { "epoch": 2.3153013910355487, "eval_cer": 5.1436, "eval_chrf": 91.0868, "eval_der": 4.9361, "eval_ier": 3.0291, "eval_loss": 0.03869618475437164, "eval_rtf": 0.1092, "eval_runtime": 819.3445, "eval_samples_per_second": 2.807, "eval_ser": 71.6146, "eval_steps_per_second": 0.176, "eval_wer": 28.0599, "step": 750 }, { "epoch": 2.39258114374034, "grad_norm": 9.961092948913574, "learning_rate": 1.9234634812585788e-05, "loss": 0.19167911529541015, "step": 775 }, { "epoch": 2.469860896445131, "grad_norm": 5.683217525482178, "learning_rate": 1.8768884214493217e-05, "loss": 0.18798019409179687, "step": 800 }, { "epoch": 2.547140649149923, "grad_norm": 5.087718963623047, "learning_rate": 1.8291200438998273e-05, "loss": 0.19169319152832032, "step": 825 }, { "epoch": 2.624420401854714, "grad_norm": 6.890854358673096, "learning_rate": 1.7802492784317377e-05, "loss": 0.1817630958557129, "step": 850 }, { "epoch": 2.7017001545595054, "grad_norm": 7.042235851287842, "learning_rate": 1.7303691533247546e-05, "loss": 0.17658203125, "step": 875 }, { "epoch": 2.7789799072642967, "grad_norm": 4.715079307556152, "learning_rate": 1.6795746182320205e-05, "loss": 0.17304580688476562, "step": 900 }, { "epoch": 2.856259659969088, "grad_norm": 5.154935836791992, "learning_rate": 1.6279623634380557e-05, "loss": 0.17773818969726562, "step": 925 }, { "epoch": 2.9335394126738796, "grad_norm": 6.3147053718566895, "learning_rate": 1.575630635803311e-05, "loss": 0.17373537063598632, "step": 950 }, { "epoch": 3.009273570324575, "grad_norm": 3.0092267990112305, "learning_rate": 1.5226790517456785e-05, "loss": 0.1598484420776367, "step": 975 }, { "epoch": 3.0865533230293662, "grad_norm": 2.872157573699951, "learning_rate": 1.469208407614984e-05, "loss": 0.08689453125, "step": 1000 }, { "epoch": 3.0865533230293662, "eval_cer": 3.8985, "eval_chrf": 93.5402, "eval_der": 3.8788, "eval_ier": 3.2691, "eval_loss": 0.033831637352705, "eval_rtf": 0.1094, "eval_runtime": 820.4358, "eval_samples_per_second": 2.803, "eval_ser": 65.5382, "eval_steps_per_second": 0.176, "eval_wer": 23.8243, "step": 1000 }, { "epoch": 3.1638330757341575, "grad_norm": 7.512609004974365, "learning_rate": 1.4153204878214032e-05, "loss": 0.0933732795715332, "step": 1025 }, { "epoch": 3.2411128284389488, "grad_norm": 4.315701007843018, "learning_rate": 1.3611178710830521e-05, "loss": 0.09186286926269531, "step": 1050 }, { "epoch": 3.3183925811437405, "grad_norm": 3.3317482471466064, "learning_rate": 1.3067037351615672e-05, "loss": 0.09010017395019532, "step": 1075 }, { "epoch": 3.3956723338485317, "grad_norm": 3.593557596206665, "learning_rate": 1.252181660457373e-05, "loss": 0.0851977825164795, "step": 1100 }, { "epoch": 3.472952086553323, "grad_norm": 3.8656005859375, "learning_rate": 1.1976554328385008e-05, "loss": 0.09125997543334961, "step": 1125 }, { "epoch": 3.5502318392581143, "grad_norm": 5.493981838226318, "learning_rate": 1.1432288460782909e-05, "loss": 0.09132560729980468, "step": 1150 }, { "epoch": 3.6275115919629055, "grad_norm": 4.440531253814697, "learning_rate": 1.0890055042780467e-05, "loss": 0.08915603637695313, "step": 1175 }, { "epoch": 3.704791344667697, "grad_norm": 4.907797813415527, "learning_rate": 1.035088624650738e-05, "loss": 0.08034603118896484, "step": 1200 }, { "epoch": 3.7820710973724885, "grad_norm": 5.263486862182617, "learning_rate": 9.81580841041171e-06, "loss": 0.08620176315307618, "step": 1225 }, { "epoch": 3.8593508500772797, "grad_norm": 3.7332587242126465, "learning_rate": 9.2858400855663e-06, "loss": 0.08419341087341309, "step": 1250 }, { "epoch": 3.8593508500772797, "eval_cer": 3.6986, "eval_chrf": 93.9971, "eval_der": 5.4615, "eval_ier": 2.16, "eval_loss": 0.03135475888848305, "eval_rtf": 0.1089, "eval_runtime": 817.1653, "eval_samples_per_second": 2.815, "eval_ser": 64.974, "eval_steps_per_second": 0.176, "eval_wer": 23.4352, "step": 1250 }, { "epoch": 3.936630602782071, "grad_norm": 3.414971351623535, "learning_rate": 8.761990096798918e-06, "loss": 0.07733444690704346, "step": 1275 }, { "epoch": 4.012364760432766, "grad_norm": 2.145824909210205, "learning_rate": 8.245255622336876e-06, "loss": 0.06913675308227539, "step": 1300 }, { "epoch": 4.089644513137558, "grad_norm": 2.0813205242156982, "learning_rate": 7.736620295621574e-06, "loss": 0.03386144876480102, "step": 1325 }, { "epoch": 4.166924265842349, "grad_norm": 3.1392276287078857, "learning_rate": 7.237052332906363e-06, "loss": 0.03620212078094483, "step": 1350 }, { "epoch": 4.244204018547141, "grad_norm": 2.0920841693878174, "learning_rate": 6.747502690201855e-06, "loss": 0.03203610420227051, "step": 1375 }, { "epoch": 4.321483771251932, "grad_norm": 2.0931801795959473, "learning_rate": 6.268903253077122e-06, "loss": 0.03280980348587036, "step": 1400 }, { "epoch": 4.398763523956723, "grad_norm": 2.1217074394226074, "learning_rate": 5.8021650627625395e-06, "loss": 0.03292776584625244, "step": 1425 }, { "epoch": 4.476043276661515, "grad_norm": 2.0681726932525635, "learning_rate": 5.34817658193105e-06, "loss": 0.0360551381111145, "step": 1450 }, { "epoch": 4.553323029366306, "grad_norm": 2.6600136756896973, "learning_rate": 4.9078020034589195e-06, "loss": 0.03164109468460083, "step": 1475 }, { "epoch": 4.630602782071097, "grad_norm": 2.0437726974487305, "learning_rate": 4.481879605385526e-06, "loss": 0.03186525821685791, "step": 1500 }, { "epoch": 4.630602782071097, "eval_cer": 3.3703, "eval_chrf": 94.583, "eval_der": 3.3145, "eval_ier": 3.4702, "eval_loss": 0.03388617932796478, "eval_rtf": 0.1091, "eval_runtime": 818.5955, "eval_samples_per_second": 2.81, "eval_ser": 60.3299, "eval_steps_per_second": 0.176, "eval_wer": 21.2947, "step": 1500 }, { "epoch": 4.707882534775889, "grad_norm": 1.9571408033370972, "learning_rate": 4.071220155203462e-06, "loss": 0.026449871063232423, "step": 1525 }, { "epoch": 4.78516228748068, "grad_norm": 1.9489446878433228, "learning_rate": 3.6766053665164875e-06, "loss": 0.029815149307250977, "step": 1550 }, { "epoch": 4.8624420401854715, "grad_norm": 1.925317645072937, "learning_rate": 3.2987864110033212e-06, "loss": 0.029331560134887694, "step": 1575 }, { "epoch": 4.939721792890262, "grad_norm": 2.1763415336608887, "learning_rate": 2.9384824885196034e-06, "loss": 0.026700735092163086, "step": 1600 }, { "epoch": 5.015455950540958, "grad_norm": 0.8997424244880676, "learning_rate": 2.596379458060118e-06, "loss": 0.021269240379333497, "step": 1625 }, { "epoch": 5.09273570324575, "grad_norm": 1.2491432428359985, "learning_rate": 2.273128532187205e-06, "loss": 0.008204447031021118, "step": 1650 }, { "epoch": 5.170015455950541, "grad_norm": 1.0970476865768433, "learning_rate": 1.9693450374106565e-06, "loss": 0.008956260681152343, "step": 1675 }, { "epoch": 5.247295208655332, "grad_norm": 1.0404987335205078, "learning_rate": 1.6856072428786965e-06, "loss": 0.00936429262161255, "step": 1700 }, { "epoch": 5.324574961360124, "grad_norm": 1.6842068433761597, "learning_rate": 1.4224552596098158e-06, "loss": 0.008714500665664673, "step": 1725 }, { "epoch": 5.401854714064915, "grad_norm": 1.1528061628341675, "learning_rate": 1.1803900123607576e-06, "loss": 0.007954384088516236, "step": 1750 }, { "epoch": 5.401854714064915, "eval_cer": 3.2476, "eval_chrf": 94.7964, "eval_der": 3.8723, "eval_ier": 2.9059, "eval_loss": 0.03716501593589783, "eval_rtf": 0.1087, "eval_runtime": 815.1538, "eval_samples_per_second": 2.822, "eval_ser": 58.9844, "eval_steps_per_second": 0.177, "eval_wer": 20.9639, "step": 1750 }, { "epoch": 5.479134466769707, "grad_norm": 0.7014907598495483, "learning_rate": 9.598722860877388e-07, "loss": 0.00877517819404602, "step": 1775 }, { "epoch": 5.556414219474497, "grad_norm": 1.1765307188034058, "learning_rate": 7.613218488161363e-07, "loss": 0.006946336030960083, "step": 1800 }, { "epoch": 5.633693972179289, "grad_norm": 0.9923800230026245, "learning_rate": 5.851166525881636e-07, "loss": 0.007716541886329651, "step": 1825 }, { "epoch": 5.710973724884081, "grad_norm": 2.2408559322357178, "learning_rate": 4.315921140096879e-07, "loss": 0.008162953853607179, "step": 1850 }, { "epoch": 5.788253477588872, "grad_norm": 1.339606761932373, "learning_rate": 3.010404757656571e-07, "loss": 0.007456899881362915, "step": 1875 }, { "epoch": 5.865533230293663, "grad_norm": 1.7567073106765747, "learning_rate": 1.937102503195426e-07, "loss": 0.006572351455688476, "step": 1900 }, { "epoch": 5.942812982998454, "grad_norm": 0.6343240737915039, "learning_rate": 1.0980574685572897e-07, "loss": 0.007703549265861511, "step": 1925 }, { "epoch": 6.01854714064915, "grad_norm": 0.4855692684650421, "learning_rate": 4.948668236538084e-08, "loss": 0.005484968423843384, "step": 1950 }, { "epoch": 6.095826893353942, "grad_norm": 0.7014145851135254, "learning_rate": 1.2867877616067059e-08, "loss": 0.004488465785980225, "step": 1975 }, { "epoch": 6.1731066460587325, "grad_norm": 0.3288016617298126, "learning_rate": 1.9038583903840945e-11, "loss": 0.004280143082141876, "step": 2000 }, { "epoch": 6.1731066460587325, "eval_cer": 3.2053, "eval_chrf": 94.8599, "eval_der": 3.8723, "eval_ier": 2.7826, "eval_loss": 0.03753867372870445, "eval_rtf": 0.109, "eval_runtime": 817.5818, "eval_samples_per_second": 2.813, "eval_ser": 58.5938, "eval_steps_per_second": 0.176, "eval_wer": 20.6979, "step": 2000 }, { "epoch": 6.1731066460587325, "step": 2000, "total_flos": 3.6883523136097616e+19, "train_loss": 0.8259544283039868, "train_runtime": 17069.0783, "train_samples_per_second": 7.499, "train_steps_per_second": 0.117 } ], "logging_steps": 25, "max_steps": 2000, "num_input_tokens_seen": 0, "num_train_epochs": 7, "save_steps": 500, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": false, "should_training_stop": false }, "attributes": {} } }, "total_flos": 3.6883523136097616e+19, "train_batch_size": 8, "trial_name": null, "trial_params": null }