| { |
| "best_metric": 0.9766666666666667, |
| "best_model_checkpoint": "/media/cse/HDD/Shawon/shawon/MY DATA/VideoMAE_BdSLW60_SR_8_kineticsFinetuned_withoutAug/checkpoint-3716", |
| "epoch": 8.050053879310346, |
| "eval_steps": 500, |
| "global_step": 8361, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 0.005387931034482759, |
| "grad_norm": 20.11014747619629, |
| "learning_rate": 2.6670258620689655e-06, |
| "loss": 16.4554, |
| "step": 100 |
| }, |
| { |
| "epoch": 0.010775862068965518, |
| "grad_norm": 26.971817016601562, |
| "learning_rate": 5.360991379310345e-06, |
| "loss": 16.1745, |
| "step": 200 |
| }, |
| { |
| "epoch": 0.016163793103448277, |
| "grad_norm": 41.51228713989258, |
| "learning_rate": 8.028017241379311e-06, |
| "loss": 15.3342, |
| "step": 300 |
| }, |
| { |
| "epoch": 0.021551724137931036, |
| "grad_norm": 51.878631591796875, |
| "learning_rate": 1.072198275862069e-05, |
| "loss": 13.6057, |
| "step": 400 |
| }, |
| { |
| "epoch": 0.02693965517241379, |
| "grad_norm": 52.984981536865234, |
| "learning_rate": 1.3415948275862069e-05, |
| "loss": 11.1741, |
| "step": 500 |
| }, |
| { |
| "epoch": 0.032327586206896554, |
| "grad_norm": 50.443359375, |
| "learning_rate": 1.610991379310345e-05, |
| "loss": 8.8003, |
| "step": 600 |
| }, |
| { |
| "epoch": 0.03771551724137931, |
| "grad_norm": 50.93900680541992, |
| "learning_rate": 1.880387931034483e-05, |
| "loss": 6.2116, |
| "step": 700 |
| }, |
| { |
| "epoch": 0.04310344827586207, |
| "grad_norm": 40.679100036621094, |
| "learning_rate": 2.149784482758621e-05, |
| "loss": 4.3287, |
| "step": 800 |
| }, |
| { |
| "epoch": 0.04849137931034483, |
| "grad_norm": 37.3469123840332, |
| "learning_rate": 2.4191810344827586e-05, |
| "loss": 2.859, |
| "step": 900 |
| }, |
| { |
| "epoch": 0.05005387931034483, |
| "eval_accuracy": 0.9183333333333333, |
| "eval_f1": 0.9087385884819826, |
| "eval_loss": 0.8475048542022705, |
| "eval_precision": 0.9240348890875206, |
| "eval_recall": 0.9183333333333333, |
| "eval_runtime": 192.0614, |
| "eval_samples_per_second": 3.124, |
| "eval_steps_per_second": 1.562, |
| "step": 929 |
| }, |
| { |
| "epoch": 1.0038254310344827, |
| "grad_norm": 19.420085906982422, |
| "learning_rate": 2.688577586206897e-05, |
| "loss": 1.5587, |
| "step": 1000 |
| }, |
| { |
| "epoch": 1.0092133620689656, |
| "grad_norm": 17.20052146911621, |
| "learning_rate": 2.9579741379310345e-05, |
| "loss": 0.9491, |
| "step": 1100 |
| }, |
| { |
| "epoch": 1.0146012931034483, |
| "grad_norm": 24.49296760559082, |
| "learning_rate": 3.2273706896551725e-05, |
| "loss": 0.6366, |
| "step": 1200 |
| }, |
| { |
| "epoch": 1.0199892241379311, |
| "grad_norm": 43.13790512084961, |
| "learning_rate": 3.496767241379311e-05, |
| "loss": 0.4445, |
| "step": 1300 |
| }, |
| { |
| "epoch": 1.0253771551724138, |
| "grad_norm": 1.7655223608016968, |
| "learning_rate": 3.7661637931034485e-05, |
| "loss": 0.2924, |
| "step": 1400 |
| }, |
| { |
| "epoch": 1.0307650862068964, |
| "grad_norm": 2.0372657775878906, |
| "learning_rate": 4.035560344827586e-05, |
| "loss": 0.1758, |
| "step": 1500 |
| }, |
| { |
| "epoch": 1.0361530172413793, |
| "grad_norm": 1.5986464023590088, |
| "learning_rate": 4.3049568965517245e-05, |
| "loss": 0.2277, |
| "step": 1600 |
| }, |
| { |
| "epoch": 1.041540948275862, |
| "grad_norm": 1.0040245056152344, |
| "learning_rate": 4.574353448275862e-05, |
| "loss": 0.1491, |
| "step": 1700 |
| }, |
| { |
| "epoch": 1.0469288793103448, |
| "grad_norm": 2.6459920406341553, |
| "learning_rate": 4.8437500000000005e-05, |
| "loss": 0.0555, |
| "step": 1800 |
| }, |
| { |
| "epoch": 1.0500538793103449, |
| "eval_accuracy": 0.9216666666666666, |
| "eval_f1": 0.9093368000230129, |
| "eval_loss": 0.30231231451034546, |
| "eval_precision": 0.9317002164502166, |
| "eval_recall": 0.9216666666666666, |
| "eval_runtime": 187.0802, |
| "eval_samples_per_second": 3.207, |
| "eval_steps_per_second": 1.604, |
| "step": 1858 |
| }, |
| { |
| "epoch": 2.002262931034483, |
| "grad_norm": 0.5840566158294678, |
| "learning_rate": 4.9874281609195405e-05, |
| "loss": 0.0647, |
| "step": 1900 |
| }, |
| { |
| "epoch": 2.0076508620689655, |
| "grad_norm": 0.16274438798427582, |
| "learning_rate": 4.95749521072797e-05, |
| "loss": 0.0725, |
| "step": 2000 |
| }, |
| { |
| "epoch": 2.013038793103448, |
| "grad_norm": 0.10402801632881165, |
| "learning_rate": 4.9275622605363985e-05, |
| "loss": 0.0139, |
| "step": 2100 |
| }, |
| { |
| "epoch": 2.0184267241379312, |
| "grad_norm": 0.0963917076587677, |
| "learning_rate": 4.897629310344828e-05, |
| "loss": 0.0206, |
| "step": 2200 |
| }, |
| { |
| "epoch": 2.023814655172414, |
| "grad_norm": 0.08095109462738037, |
| "learning_rate": 4.867696360153257e-05, |
| "loss": 0.0499, |
| "step": 2300 |
| }, |
| { |
| "epoch": 2.0292025862068965, |
| "grad_norm": 1.0376495122909546, |
| "learning_rate": 4.837763409961686e-05, |
| "loss": 0.0105, |
| "step": 2400 |
| }, |
| { |
| "epoch": 2.034590517241379, |
| "grad_norm": 0.027790505439043045, |
| "learning_rate": 4.808129789272031e-05, |
| "loss": 0.0226, |
| "step": 2500 |
| }, |
| { |
| "epoch": 2.0399784482758623, |
| "grad_norm": 0.031076578423380852, |
| "learning_rate": 4.778496168582376e-05, |
| "loss": 0.0359, |
| "step": 2600 |
| }, |
| { |
| "epoch": 2.045366379310345, |
| "grad_norm": 0.02143680490553379, |
| "learning_rate": 4.748563218390804e-05, |
| "loss": 0.0025, |
| "step": 2700 |
| }, |
| { |
| "epoch": 2.0500538793103447, |
| "eval_accuracy": 0.935, |
| "eval_f1": 0.927261710739568, |
| "eval_loss": 0.2999415695667267, |
| "eval_precision": 0.9378208874458874, |
| "eval_recall": 0.935, |
| "eval_runtime": 186.5359, |
| "eval_samples_per_second": 3.217, |
| "eval_steps_per_second": 1.608, |
| "step": 2787 |
| }, |
| { |
| "epoch": 3.000700431034483, |
| "grad_norm": 0.044046301394701004, |
| "learning_rate": 4.7186302681992336e-05, |
| "loss": 0.2036, |
| "step": 2800 |
| }, |
| { |
| "epoch": 3.0060883620689656, |
| "grad_norm": 0.02674109861254692, |
| "learning_rate": 4.688697318007663e-05, |
| "loss": 0.0227, |
| "step": 2900 |
| }, |
| { |
| "epoch": 3.011476293103448, |
| "grad_norm": 0.604284405708313, |
| "learning_rate": 4.658764367816092e-05, |
| "loss": 0.0018, |
| "step": 3000 |
| }, |
| { |
| "epoch": 3.016864224137931, |
| "grad_norm": 0.012123900465667248, |
| "learning_rate": 4.6288314176245215e-05, |
| "loss": 0.0013, |
| "step": 3100 |
| }, |
| { |
| "epoch": 3.022252155172414, |
| "grad_norm": 0.17276327311992645, |
| "learning_rate": 4.598898467432951e-05, |
| "loss": 0.0304, |
| "step": 3200 |
| }, |
| { |
| "epoch": 3.0276400862068966, |
| "grad_norm": 0.01406328845769167, |
| "learning_rate": 4.5689655172413794e-05, |
| "loss": 0.0529, |
| "step": 3300 |
| }, |
| { |
| "epoch": 3.0330280172413793, |
| "grad_norm": 0.026818769052624702, |
| "learning_rate": 4.539032567049809e-05, |
| "loss": 0.0059, |
| "step": 3400 |
| }, |
| { |
| "epoch": 3.038415948275862, |
| "grad_norm": 0.022273056209087372, |
| "learning_rate": 4.509099616858238e-05, |
| "loss": 0.0431, |
| "step": 3500 |
| }, |
| { |
| "epoch": 3.043803879310345, |
| "grad_norm": 0.008690926246345043, |
| "learning_rate": 4.4791666666666673e-05, |
| "loss": 0.001, |
| "step": 3600 |
| }, |
| { |
| "epoch": 3.0491918103448277, |
| "grad_norm": 0.009022524580359459, |
| "learning_rate": 4.449233716475096e-05, |
| "loss": 0.0009, |
| "step": 3700 |
| }, |
| { |
| "epoch": 3.0500538793103447, |
| "eval_accuracy": 0.9766666666666667, |
| "eval_f1": 0.9761724029000916, |
| "eval_loss": 0.08761004358530045, |
| "eval_precision": 0.9809054834054833, |
| "eval_recall": 0.9766666666666667, |
| "eval_runtime": 184.6188, |
| "eval_samples_per_second": 3.25, |
| "eval_steps_per_second": 1.625, |
| "step": 3716 |
| }, |
| { |
| "epoch": 4.004525862068966, |
| "grad_norm": 0.007579015102237463, |
| "learning_rate": 4.419300766283525e-05, |
| "loss": 0.0007, |
| "step": 3800 |
| }, |
| { |
| "epoch": 4.009913793103448, |
| "grad_norm": 0.010904570110142231, |
| "learning_rate": 4.3893678160919546e-05, |
| "loss": 0.0395, |
| "step": 3900 |
| }, |
| { |
| "epoch": 4.015301724137931, |
| "grad_norm": 0.006146453786641359, |
| "learning_rate": 4.359434865900383e-05, |
| "loss": 0.0006, |
| "step": 4000 |
| }, |
| { |
| "epoch": 4.020689655172414, |
| "grad_norm": 0.013002044521272182, |
| "learning_rate": 4.3295019157088125e-05, |
| "loss": 0.0011, |
| "step": 4100 |
| }, |
| { |
| "epoch": 4.026077586206896, |
| "grad_norm": 0.12733735144138336, |
| "learning_rate": 4.299568965517242e-05, |
| "loss": 0.0451, |
| "step": 4200 |
| }, |
| { |
| "epoch": 4.031465517241379, |
| "grad_norm": 11.122708320617676, |
| "learning_rate": 4.269636015325671e-05, |
| "loss": 0.305, |
| "step": 4300 |
| }, |
| { |
| "epoch": 4.0368534482758625, |
| "grad_norm": 0.13978660106658936, |
| "learning_rate": 4.2397030651341e-05, |
| "loss": 0.2215, |
| "step": 4400 |
| }, |
| { |
| "epoch": 4.042241379310345, |
| "grad_norm": 0.37221765518188477, |
| "learning_rate": 4.209770114942529e-05, |
| "loss": 0.1446, |
| "step": 4500 |
| }, |
| { |
| "epoch": 4.047629310344828, |
| "grad_norm": 20.41011619567871, |
| "learning_rate": 4.1801364942528733e-05, |
| "loss": 0.1355, |
| "step": 4600 |
| }, |
| { |
| "epoch": 4.050053879310345, |
| "eval_accuracy": 0.9266666666666666, |
| "eval_f1": 0.9146410798064111, |
| "eval_loss": 0.25398045778274536, |
| "eval_precision": 0.9182419154919156, |
| "eval_recall": 0.9266666666666666, |
| "eval_runtime": 176.3808, |
| "eval_samples_per_second": 3.402, |
| "eval_steps_per_second": 1.701, |
| "step": 4645 |
| }, |
| { |
| "epoch": 5.002963362068965, |
| "grad_norm": 0.017676187679171562, |
| "learning_rate": 4.1502035440613026e-05, |
| "loss": 0.1491, |
| "step": 4700 |
| }, |
| { |
| "epoch": 5.008351293103448, |
| "grad_norm": 1.5707539319992065, |
| "learning_rate": 4.120270593869732e-05, |
| "loss": 0.0298, |
| "step": 4800 |
| }, |
| { |
| "epoch": 5.013739224137931, |
| "grad_norm": 0.08283871412277222, |
| "learning_rate": 4.090337643678161e-05, |
| "loss": 0.1471, |
| "step": 4900 |
| }, |
| { |
| "epoch": 5.019127155172414, |
| "grad_norm": 0.0916813537478447, |
| "learning_rate": 4.0604046934865905e-05, |
| "loss": 0.1646, |
| "step": 5000 |
| }, |
| { |
| "epoch": 5.024515086206897, |
| "grad_norm": 0.012382575310766697, |
| "learning_rate": 4.03047174329502e-05, |
| "loss": 0.015, |
| "step": 5100 |
| }, |
| { |
| "epoch": 5.029903017241379, |
| "grad_norm": 0.008144189603626728, |
| "learning_rate": 4.000838122605364e-05, |
| "loss": 0.1365, |
| "step": 5200 |
| }, |
| { |
| "epoch": 5.035290948275862, |
| "grad_norm": 0.021491670981049538, |
| "learning_rate": 3.9709051724137935e-05, |
| "loss": 0.0302, |
| "step": 5300 |
| }, |
| { |
| "epoch": 5.040678879310345, |
| "grad_norm": 0.03770313411951065, |
| "learning_rate": 3.940972222222222e-05, |
| "loss": 0.1047, |
| "step": 5400 |
| }, |
| { |
| "epoch": 5.046066810344827, |
| "grad_norm": 0.0966072753071785, |
| "learning_rate": 3.9110392720306514e-05, |
| "loss": 0.099, |
| "step": 5500 |
| }, |
| { |
| "epoch": 5.050053879310345, |
| "eval_accuracy": 0.9633333333333334, |
| "eval_f1": 0.9578282624655623, |
| "eval_loss": 0.26155975461006165, |
| "eval_precision": 0.956979842282474, |
| "eval_recall": 0.9633333333333334, |
| "eval_runtime": 178.9094, |
| "eval_samples_per_second": 3.354, |
| "eval_steps_per_second": 1.677, |
| "step": 5574 |
| }, |
| { |
| "epoch": 6.001400862068966, |
| "grad_norm": 0.0034083151258528233, |
| "learning_rate": 3.881106321839081e-05, |
| "loss": 0.0444, |
| "step": 5600 |
| }, |
| { |
| "epoch": 6.006788793103448, |
| "grad_norm": 0.0034083956852555275, |
| "learning_rate": 3.851173371647509e-05, |
| "loss": 0.0439, |
| "step": 5700 |
| }, |
| { |
| "epoch": 6.012176724137931, |
| "grad_norm": 0.004689768888056278, |
| "learning_rate": 3.8212404214559386e-05, |
| "loss": 0.0007, |
| "step": 5800 |
| }, |
| { |
| "epoch": 6.017564655172414, |
| "grad_norm": 0.004528137389570475, |
| "learning_rate": 3.791307471264368e-05, |
| "loss": 0.0529, |
| "step": 5900 |
| }, |
| { |
| "epoch": 6.022952586206896, |
| "grad_norm": 0.005246564745903015, |
| "learning_rate": 3.7613745210727965e-05, |
| "loss": 0.0016, |
| "step": 6000 |
| }, |
| { |
| "epoch": 6.0283405172413795, |
| "grad_norm": 0.003063632408156991, |
| "learning_rate": 3.731441570881226e-05, |
| "loss": 0.0002, |
| "step": 6100 |
| }, |
| { |
| "epoch": 6.033728448275862, |
| "grad_norm": 0.002391215180978179, |
| "learning_rate": 3.701508620689655e-05, |
| "loss": 0.0002, |
| "step": 6200 |
| }, |
| { |
| "epoch": 6.039116379310345, |
| "grad_norm": 0.0016633198829367757, |
| "learning_rate": 3.6715756704980844e-05, |
| "loss": 0.0002, |
| "step": 6300 |
| }, |
| { |
| "epoch": 6.044504310344828, |
| "grad_norm": 0.013642443343997002, |
| "learning_rate": 3.641642720306514e-05, |
| "loss": 0.0002, |
| "step": 6400 |
| }, |
| { |
| "epoch": 6.04989224137931, |
| "grad_norm": 0.0030818418599665165, |
| "learning_rate": 3.611709770114943e-05, |
| "loss": 0.0656, |
| "step": 6500 |
| }, |
| { |
| "epoch": 6.050053879310345, |
| "eval_accuracy": 0.9616666666666667, |
| "eval_f1": 0.9597682298955686, |
| "eval_loss": 0.20607294142246246, |
| "eval_precision": 0.9674627687127687, |
| "eval_recall": 0.9616666666666667, |
| "eval_runtime": 184.33, |
| "eval_samples_per_second": 3.255, |
| "eval_steps_per_second": 1.628, |
| "step": 6503 |
| }, |
| { |
| "epoch": 7.005226293103449, |
| "grad_norm": 0.003281041979789734, |
| "learning_rate": 3.5817768199233717e-05, |
| "loss": 0.0108, |
| "step": 6600 |
| }, |
| { |
| "epoch": 7.010614224137931, |
| "grad_norm": 0.06187641620635986, |
| "learning_rate": 3.551843869731801e-05, |
| "loss": 0.0458, |
| "step": 6700 |
| }, |
| { |
| "epoch": 7.016002155172414, |
| "grad_norm": 0.6124621033668518, |
| "learning_rate": 3.52191091954023e-05, |
| "loss": 0.0931, |
| "step": 6800 |
| }, |
| { |
| "epoch": 7.021390086206897, |
| "grad_norm": 0.013154719024896622, |
| "learning_rate": 3.4919779693486596e-05, |
| "loss": 0.2289, |
| "step": 6900 |
| }, |
| { |
| "epoch": 7.026778017241379, |
| "grad_norm": 0.33358174562454224, |
| "learning_rate": 3.462045019157089e-05, |
| "loss": 0.0804, |
| "step": 7000 |
| }, |
| { |
| "epoch": 7.032165948275862, |
| "grad_norm": 0.010887747630476952, |
| "learning_rate": 3.4321120689655175e-05, |
| "loss": 0.0032, |
| "step": 7100 |
| }, |
| { |
| "epoch": 7.0375538793103445, |
| "grad_norm": 0.26187142729759216, |
| "learning_rate": 3.402179118773947e-05, |
| "loss": 0.0134, |
| "step": 7200 |
| }, |
| { |
| "epoch": 7.042941810344828, |
| "grad_norm": 0.01975095644593239, |
| "learning_rate": 3.372246168582376e-05, |
| "loss": 0.0507, |
| "step": 7300 |
| }, |
| { |
| "epoch": 7.048329741379311, |
| "grad_norm": 0.004032758995890617, |
| "learning_rate": 3.342313218390805e-05, |
| "loss": 0.0314, |
| "step": 7400 |
| }, |
| { |
| "epoch": 7.050053879310345, |
| "eval_accuracy": 0.9533333333333334, |
| "eval_f1": 0.9526073505957408, |
| "eval_loss": 0.20366963744163513, |
| "eval_precision": 0.9623829642579644, |
| "eval_recall": 0.9533333333333334, |
| "eval_runtime": 186.0707, |
| "eval_samples_per_second": 3.225, |
| "eval_steps_per_second": 1.612, |
| "step": 7432 |
| }, |
| { |
| "epoch": 8.003663793103449, |
| "grad_norm": 0.003311087377369404, |
| "learning_rate": 3.312380268199234e-05, |
| "loss": 0.0496, |
| "step": 7500 |
| }, |
| { |
| "epoch": 8.009051724137931, |
| "grad_norm": 0.03303457051515579, |
| "learning_rate": 3.282447318007663e-05, |
| "loss": 0.0004, |
| "step": 7600 |
| }, |
| { |
| "epoch": 8.014439655172414, |
| "grad_norm": 0.0018716267077252269, |
| "learning_rate": 3.252514367816092e-05, |
| "loss": 0.0002, |
| "step": 7700 |
| }, |
| { |
| "epoch": 8.019827586206896, |
| "grad_norm": 0.0035127492155879736, |
| "learning_rate": 3.222581417624521e-05, |
| "loss": 0.1371, |
| "step": 7800 |
| }, |
| { |
| "epoch": 8.02521551724138, |
| "grad_norm": 0.009622437879443169, |
| "learning_rate": 3.1926484674329505e-05, |
| "loss": 0.0445, |
| "step": 7900 |
| }, |
| { |
| "epoch": 8.030603448275862, |
| "grad_norm": 0.0022457086015492678, |
| "learning_rate": 3.163014846743295e-05, |
| "loss": 0.0073, |
| "step": 8000 |
| }, |
| { |
| "epoch": 8.035991379310344, |
| "grad_norm": 0.002641331171616912, |
| "learning_rate": 3.133081896551724e-05, |
| "loss": 0.0007, |
| "step": 8100 |
| }, |
| { |
| "epoch": 8.041379310344828, |
| "grad_norm": 0.004722919315099716, |
| "learning_rate": 3.1031489463601535e-05, |
| "loss": 0.0721, |
| "step": 8200 |
| }, |
| { |
| "epoch": 8.04676724137931, |
| "grad_norm": 0.0021432852372527122, |
| "learning_rate": 3.073215996168583e-05, |
| "loss": 0.0874, |
| "step": 8300 |
| }, |
| { |
| "epoch": 8.050053879310346, |
| "eval_accuracy": 0.95, |
| "eval_f1": 0.9457895216400736, |
| "eval_loss": 0.219703808426857, |
| "eval_precision": 0.9611800144300143, |
| "eval_recall": 0.95, |
| "eval_runtime": 186.2504, |
| "eval_samples_per_second": 3.221, |
| "eval_steps_per_second": 1.611, |
| "step": 8361 |
| }, |
| { |
| "epoch": 8.050053879310346, |
| "step": 8361, |
| "total_flos": 8.337990588259369e+19, |
| "train_loss": 1.2332628082536137, |
| "train_runtime": 25272.533, |
| "train_samples_per_second": 5.875, |
| "train_steps_per_second": 0.734 |
| } |
| ], |
| "logging_steps": 100, |
| "max_steps": 18560, |
| "num_input_tokens_seen": 0, |
| "num_train_epochs": 9223372036854775807, |
| "save_steps": 500, |
| "stateful_callbacks": { |
| "EarlyStoppingCallback": { |
| "args": { |
| "early_stopping_patience": 5, |
| "early_stopping_threshold": 0.0 |
| }, |
| "attributes": { |
| "early_stopping_patience_counter": 5 |
| } |
| }, |
| "TrainerControl": { |
| "args": { |
| "should_epoch_stop": false, |
| "should_evaluate": false, |
| "should_log": false, |
| "should_save": true, |
| "should_training_stop": true |
| }, |
| "attributes": {} |
| } |
| }, |
| "total_flos": 8.337990588259369e+19, |
| "train_batch_size": 2, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
