diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,10183 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.4795918367346939, + "eval_steps": 500, + "global_step": 725, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0020408163265306124, + "grad_norm": 4.2167510986328125, + "learning_rate": 0.0, + "loss": 1.313438892364502, + "memory/device_reserved (GiB)": 75.76, + "memory/max_active (GiB)": 74.61, + "memory/max_allocated (GiB)": 74.61, + "ppl": 3.71894, + "step": 1, + "tokens/total": 249728, + "tokens/trainable": 1992 + }, + { + "epoch": 0.004081632653061225, + "grad_norm": 3.8650901317596436, + "learning_rate": 3.448275862068966e-06, + "loss": 1.2385082244873047, + "memory/device_reserved (GiB)": 75.77, + "memory/max_active (GiB)": 74.28, + "memory/max_allocated (GiB)": 74.28, + "ppl": 3.45046, + "step": 2, + "tokens/total": 483392, + "tokens/train_per_sec_per_gpu": 12.03, + "tokens/trainable": 3913 + }, + { + "epoch": 0.006122448979591836, + "grad_norm": 3.7181951999664307, + "learning_rate": 6.896551724137932e-06, + "loss": 1.2442840337753296, + "memory/device_reserved (GiB)": 76.4, + "memory/max_active (GiB)": 75.5, + "memory/max_allocated (GiB)": 75.5, + "ppl": 3.47045, + "step": 3, + "tokens/total": 726144, + "tokens/train_per_sec_per_gpu": 12.05, + "tokens/trainable": 5903 + }, + { + "epoch": 0.00816326530612245, + "grad_norm": 44.8486328125, + "learning_rate": 1.0344827586206897e-05, + "loss": 1.3412562608718872, + "memory/device_reserved (GiB)": 76.4, + "memory/max_active (GiB)": 72.18, + "memory/max_allocated (GiB)": 72.18, + "ppl": 3.82384, + "step": 4, + "tokens/total": 943168, + "tokens/train_per_sec_per_gpu": 15.39, + "tokens/trainable": 7867 + }, + { + "epoch": 0.01020408163265306, + "grad_norm": 3.929412841796875, + "learning_rate": 1.3793103448275863e-05, + "loss": 1.3407814502716064, + "memory/device_reserved (GiB)": 76.4, + "memory/max_active (GiB)": 75.42, + "memory/max_allocated (GiB)": 75.42, + "ppl": 3.82203, + "step": 5, + "tokens/total": 1169664, + "tokens/train_per_sec_per_gpu": 12.41, + "tokens/trainable": 9745 + }, + { + "epoch": 0.012244897959183673, + "grad_norm": 3.892376661300659, + "learning_rate": 1.7241379310344828e-05, + "loss": 1.2880868911743164, + "memory/device_reserved (GiB)": 76.4, + "memory/max_active (GiB)": 73.05, + "memory/max_allocated (GiB)": 73.05, + "ppl": 3.62584, + "step": 6, + "tokens/total": 1406336, + "tokens/train_per_sec_per_gpu": 13.98, + "tokens/trainable": 11650 + }, + { + "epoch": 0.014285714285714285, + "grad_norm": 4.0600128173828125, + "learning_rate": 2.0689655172413793e-05, + "loss": 1.3101682662963867, + "memory/device_reserved (GiB)": 76.4, + "memory/max_active (GiB)": 72.54, + "memory/max_allocated (GiB)": 72.54, + "ppl": 3.7068, + "step": 7, + "tokens/total": 1647104, + "tokens/train_per_sec_per_gpu": 13.02, + "tokens/trainable": 13571 + }, + { + "epoch": 0.0163265306122449, + "grad_norm": 3.508254289627075, + "learning_rate": 2.413793103448276e-05, + "loss": 1.1288416385650635, + "memory/device_reserved (GiB)": 76.4, + "memory/max_active (GiB)": 74.44, + "memory/max_allocated (GiB)": 74.44, + "ppl": 3.09207, + "step": 8, + "tokens/total": 1913472, + "tokens/train_per_sec_per_gpu": 11.37, + "tokens/trainable": 15743 + }, + { + "epoch": 0.018367346938775512, + "grad_norm": 4.13547945022583, + "learning_rate": 2.7586206896551727e-05, + "loss": 1.1055421829223633, + "memory/device_reserved (GiB)": 76.4, + "memory/max_active (GiB)": 75.26, + "memory/max_allocated (GiB)": 75.26, + "ppl": 3.02086, + "step": 9, + "tokens/total": 2159104, + "tokens/train_per_sec_per_gpu": 13.39, + "tokens/trainable": 17993 + }, + { + "epoch": 0.02040816326530612, + "grad_norm": 3.8564958572387695, + "learning_rate": 3.103448275862069e-05, + "loss": 1.1748199462890625, + "memory/device_reserved (GiB)": 76.61, + "memory/max_active (GiB)": 75.69, + "memory/max_allocated (GiB)": 75.69, + "ppl": 3.23756, + "step": 10, + "tokens/total": 2406336, + "tokens/train_per_sec_per_gpu": 9.87, + "tokens/trainable": 19603 + }, + { + "epoch": 0.022448979591836733, + "grad_norm": 2.662527322769165, + "learning_rate": 3.4482758620689657e-05, + "loss": 1.0491664409637451, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 76.81, + "memory/max_allocated (GiB)": 76.81, + "ppl": 2.85527, + "step": 11, + "tokens/total": 2637440, + "tokens/train_per_sec_per_gpu": 13.08, + "tokens/trainable": 21674 + }, + { + "epoch": 0.024489795918367346, + "grad_norm": 2.4675278663635254, + "learning_rate": 3.793103448275862e-05, + "loss": 0.7866945266723633, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 73.65, + "memory/max_allocated (GiB)": 73.65, + "ppl": 2.19613, + "step": 12, + "tokens/total": 2872896, + "tokens/train_per_sec_per_gpu": 10.39, + "tokens/trainable": 23481 + }, + { + "epoch": 0.026530612244897958, + "grad_norm": 2.2947044372558594, + "learning_rate": 4.1379310344827587e-05, + "loss": 0.7748140692710876, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 74.77, + "memory/max_allocated (GiB)": 74.77, + "ppl": 2.17019, + "step": 13, + "tokens/total": 3128768, + "tokens/train_per_sec_per_gpu": 11.12, + "tokens/trainable": 25355 + }, + { + "epoch": 0.02857142857142857, + "grad_norm": 1.5860671997070312, + "learning_rate": 4.482758620689655e-05, + "loss": 0.7104668021202087, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 75.83, + "memory/max_allocated (GiB)": 75.83, + "ppl": 2.03494, + "step": 14, + "tokens/total": 3354624, + "tokens/train_per_sec_per_gpu": 13.4, + "tokens/trainable": 27379 + }, + { + "epoch": 0.030612244897959183, + "grad_norm": 1.2193500995635986, + "learning_rate": 4.827586206896552e-05, + "loss": 0.6342951059341431, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 76.32, + "memory/max_allocated (GiB)": 76.32, + "ppl": 1.88569, + "step": 15, + "tokens/total": 3632576, + "tokens/train_per_sec_per_gpu": 10.45, + "tokens/trainable": 29348 + }, + { + "epoch": 0.0326530612244898, + "grad_norm": 0.8882229924201965, + "learning_rate": 5.172413793103449e-05, + "loss": 0.4914810061454773, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 76.86, + "memory/max_allocated (GiB)": 76.86, + "ppl": 1.63474, + "step": 16, + "tokens/total": 3893632, + "tokens/train_per_sec_per_gpu": 11.24, + "tokens/trainable": 31409 + }, + { + "epoch": 0.03469387755102041, + "grad_norm": 0.750299870967865, + "learning_rate": 5.517241379310345e-05, + "loss": 0.533316969871521, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 73.6, + "memory/max_allocated (GiB)": 73.6, + "ppl": 1.70458, + "step": 17, + "tokens/total": 4120000, + "tokens/train_per_sec_per_gpu": 16.43, + "tokens/trainable": 33895 + }, + { + "epoch": 0.036734693877551024, + "grad_norm": 0.9324926137924194, + "learning_rate": 5.862068965517241e-05, + "loss": 0.4928280711174011, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 73.19, + "memory/max_allocated (GiB)": 73.19, + "ppl": 1.63694, + "step": 18, + "tokens/total": 4334208, + "tokens/train_per_sec_per_gpu": 13.05, + "tokens/trainable": 35668 + }, + { + "epoch": 0.03877551020408163, + "grad_norm": 0.4430578649044037, + "learning_rate": 6.206896551724138e-05, + "loss": 0.4595395028591156, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 75.47, + "memory/max_allocated (GiB)": 75.47, + "ppl": 1.58334, + "step": 19, + "tokens/total": 4595520, + "tokens/train_per_sec_per_gpu": 15.09, + "tokens/trainable": 38006 + }, + { + "epoch": 0.04081632653061224, + "grad_norm": 0.6463696360588074, + "learning_rate": 6.551724137931034e-05, + "loss": 0.35258132219314575, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 73.57, + "memory/max_allocated (GiB)": 73.57, + "ppl": 1.42274, + "step": 20, + "tokens/total": 4823424, + "tokens/train_per_sec_per_gpu": 15.34, + "tokens/trainable": 40245 + }, + { + "epoch": 0.04285714285714286, + "grad_norm": 0.5994691252708435, + "learning_rate": 6.896551724137931e-05, + "loss": 0.4857128858566284, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 74.82, + "memory/max_allocated (GiB)": 74.82, + "ppl": 1.62533, + "step": 21, + "tokens/total": 5095296, + "tokens/train_per_sec_per_gpu": 12.55, + "tokens/trainable": 42288 + }, + { + "epoch": 0.044897959183673466, + "grad_norm": 0.37300923466682434, + "learning_rate": 7.241379310344828e-05, + "loss": 0.3811933100223541, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 76.37, + "memory/max_allocated (GiB)": 76.37, + "ppl": 1.46403, + "step": 22, + "tokens/total": 5361536, + "tokens/train_per_sec_per_gpu": 12.0, + "tokens/trainable": 44219 + }, + { + "epoch": 0.04693877551020408, + "grad_norm": 0.3530994951725006, + "learning_rate": 7.586206896551724e-05, + "loss": 0.47953876852989197, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 76.18, + "memory/max_allocated (GiB)": 76.18, + "ppl": 1.61533, + "step": 23, + "tokens/total": 5611776, + "tokens/train_per_sec_per_gpu": 11.94, + "tokens/trainable": 46265 + }, + { + "epoch": 0.04897959183673469, + "grad_norm": 0.4046308696269989, + "learning_rate": 7.931034482758621e-05, + "loss": 0.4454658031463623, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 75.09, + "memory/max_allocated (GiB)": 75.09, + "ppl": 1.56122, + "step": 24, + "tokens/total": 5842944, + "tokens/train_per_sec_per_gpu": 13.15, + "tokens/trainable": 48369 + }, + { + "epoch": 0.05102040816326531, + "grad_norm": 0.36315539479255676, + "learning_rate": 8.275862068965517e-05, + "loss": 0.31761664152145386, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 75.72, + "memory/max_allocated (GiB)": 75.72, + "ppl": 1.37385, + "step": 25, + "tokens/total": 6081152, + "tokens/train_per_sec_per_gpu": 13.54, + "tokens/trainable": 50389 + }, + { + "epoch": 0.053061224489795916, + "grad_norm": 0.5031322240829468, + "learning_rate": 8.620689655172413e-05, + "loss": 0.4392084777355194, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 73.22, + "memory/max_allocated (GiB)": 73.22, + "ppl": 1.55148, + "step": 26, + "tokens/total": 6343808, + "tokens/train_per_sec_per_gpu": 7.89, + "tokens/trainable": 52264 + }, + { + "epoch": 0.05510204081632653, + "grad_norm": 0.33433303236961365, + "learning_rate": 8.96551724137931e-05, + "loss": 0.3558434844017029, + "memory/device_reserved (GiB)": 76.65, + "memory/max_active (GiB)": 75.86, + "memory/max_allocated (GiB)": 75.86, + "ppl": 1.42738, + "step": 27, + "tokens/total": 6588608, + "tokens/train_per_sec_per_gpu": 10.7, + "tokens/trainable": 54059 + }, + { + "epoch": 0.05714285714285714, + "grad_norm": 0.39968281984329224, + "learning_rate": 9.310344827586207e-05, + "loss": 0.32551220059394836, + "memory/device_reserved (GiB)": 76.65, + "memory/max_active (GiB)": 73.19, + "memory/max_allocated (GiB)": 73.19, + "ppl": 1.38474, + "step": 28, + "tokens/total": 6826560, + "tokens/train_per_sec_per_gpu": 10.04, + "tokens/trainable": 55763 + }, + { + "epoch": 0.05918367346938776, + "grad_norm": 0.4150298833847046, + "learning_rate": 9.655172413793105e-05, + "loss": 0.40454256534576416, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 76.86, + "memory/max_allocated (GiB)": 76.86, + "ppl": 1.49862, + "step": 29, + "tokens/total": 7088000, + "tokens/train_per_sec_per_gpu": 12.62, + "tokens/trainable": 57812 + }, + { + "epoch": 0.061224489795918366, + "grad_norm": 0.6367387771606445, + "learning_rate": 0.0001, + "loss": 0.42207133769989014, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 75.91, + "memory/max_allocated (GiB)": 75.91, + "ppl": 1.52512, + "step": 30, + "tokens/total": 7351744, + "tokens/train_per_sec_per_gpu": 11.4, + "tokens/trainable": 59788 + }, + { + "epoch": 0.06326530612244897, + "grad_norm": 0.27459612488746643, + "learning_rate": 9.999972717866783e-05, + "loss": 0.26368996500968933, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 75.28, + "memory/max_allocated (GiB)": 75.28, + "ppl": 1.30172, + "step": 31, + "tokens/total": 7589888, + "tokens/train_per_sec_per_gpu": 11.98, + "tokens/trainable": 61698 + }, + { + "epoch": 0.0653061224489796, + "grad_norm": 0.3257054090499878, + "learning_rate": 9.999890871764856e-05, + "loss": 0.32186996936798096, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 76.75, + "memory/max_allocated (GiB)": 76.75, + "ppl": 1.37971, + "step": 32, + "tokens/total": 7812352, + "tokens/train_per_sec_per_gpu": 11.42, + "tokens/trainable": 63730 + }, + { + "epoch": 0.0673469387755102, + "grad_norm": 0.43009933829307556, + "learning_rate": 9.999754462587395e-05, + "loss": 0.3165552020072937, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 74.25, + "memory/max_allocated (GiB)": 74.25, + "ppl": 1.37239, + "step": 33, + "tokens/total": 8090112, + "tokens/train_per_sec_per_gpu": 12.14, + "tokens/trainable": 65782 + }, + { + "epoch": 0.06938775510204082, + "grad_norm": 0.29606345295906067, + "learning_rate": 9.999563491823011e-05, + "loss": 0.3038017749786377, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 76.51, + "memory/max_allocated (GiB)": 76.51, + "ppl": 1.355, + "step": 34, + "tokens/total": 8340224, + "tokens/train_per_sec_per_gpu": 10.36, + "tokens/trainable": 67558 + }, + { + "epoch": 0.07142857142857142, + "grad_norm": 0.26481109857559204, + "learning_rate": 9.999317961555746e-05, + "loss": 0.2744923532009125, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 74.41, + "memory/max_allocated (GiB)": 74.41, + "ppl": 1.31586, + "step": 35, + "tokens/total": 8599168, + "tokens/train_per_sec_per_gpu": 9.93, + "tokens/trainable": 69256 + }, + { + "epoch": 0.07346938775510205, + "grad_norm": 0.2818477153778076, + "learning_rate": 9.999017874465026e-05, + "loss": 0.32188448309898376, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 75.12, + "memory/max_allocated (GiB)": 75.12, + "ppl": 1.37973, + "step": 36, + "tokens/total": 8879872, + "tokens/train_per_sec_per_gpu": 11.11, + "tokens/trainable": 71213 + }, + { + "epoch": 0.07551020408163266, + "grad_norm": 0.31210315227508545, + "learning_rate": 9.998663233825667e-05, + "loss": 0.35000020265579224, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 72.67, + "memory/max_allocated (GiB)": 72.67, + "ppl": 1.41907, + "step": 37, + "tokens/total": 9092480, + "tokens/train_per_sec_per_gpu": 15.52, + "tokens/trainable": 73212 + }, + { + "epoch": 0.07755102040816327, + "grad_norm": 0.4802961051464081, + "learning_rate": 9.998254043507804e-05, + "loss": 0.3628653287887573, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 73.13, + "memory/max_allocated (GiB)": 73.13, + "ppl": 1.43744, + "step": 38, + "tokens/total": 9338944, + "tokens/train_per_sec_per_gpu": 10.57, + "tokens/trainable": 75027 + }, + { + "epoch": 0.07959183673469387, + "grad_norm": 0.3178337812423706, + "learning_rate": 9.997790307976873e-05, + "loss": 0.3139835596084595, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 72.86, + "memory/max_allocated (GiB)": 72.86, + "ppl": 1.36887, + "step": 39, + "tokens/total": 9552064, + "tokens/train_per_sec_per_gpu": 13.09, + "tokens/trainable": 76940 + }, + { + "epoch": 0.08163265306122448, + "grad_norm": 0.2966010868549347, + "learning_rate": 9.997272032293553e-05, + "loss": 0.34238967299461365, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 75.53, + "memory/max_allocated (GiB)": 75.53, + "ppl": 1.40831, + "step": 40, + "tokens/total": 9780480, + "tokens/train_per_sec_per_gpu": 12.68, + "tokens/trainable": 78924 + }, + { + "epoch": 0.0836734693877551, + "grad_norm": 0.3346101641654968, + "learning_rate": 9.996699222113709e-05, + "loss": 0.2840029001235962, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 74.11, + "memory/max_allocated (GiB)": 74.11, + "ppl": 1.32844, + "step": 41, + "tokens/total": 10032896, + "tokens/train_per_sec_per_gpu": 10.74, + "tokens/trainable": 80796 + }, + { + "epoch": 0.08571428571428572, + "grad_norm": 0.3669697940349579, + "learning_rate": 9.996071883688334e-05, + "loss": 0.38213422894477844, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 75.45, + "memory/max_allocated (GiB)": 75.45, + "ppl": 1.46541, + "step": 42, + "tokens/total": 10258880, + "tokens/train_per_sec_per_gpu": 13.46, + "tokens/trainable": 82842 + }, + { + "epoch": 0.08775510204081632, + "grad_norm": 0.3180081248283386, + "learning_rate": 9.995390023863483e-05, + "loss": 0.3140525817871094, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 74.41, + "memory/max_allocated (GiB)": 74.41, + "ppl": 1.36896, + "step": 43, + "tokens/total": 10509120, + "tokens/train_per_sec_per_gpu": 13.79, + "tokens/trainable": 84934 + }, + { + "epoch": 0.08979591836734693, + "grad_norm": 0.26853886246681213, + "learning_rate": 9.994653650080188e-05, + "loss": 0.3507641553878784, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 72.23, + "memory/max_allocated (GiB)": 72.23, + "ppl": 1.42015, + "step": 44, + "tokens/total": 10749568, + "tokens/train_per_sec_per_gpu": 12.49, + "tokens/trainable": 86784 + }, + { + "epoch": 0.09183673469387756, + "grad_norm": 0.26175931096076965, + "learning_rate": 9.99386277037439e-05, + "loss": 0.27149611711502075, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 74.66, + "memory/max_allocated (GiB)": 74.66, + "ppl": 1.31193, + "step": 45, + "tokens/total": 10991040, + "tokens/train_per_sec_per_gpu": 12.04, + "tokens/trainable": 88613 + }, + { + "epoch": 0.09387755102040816, + "grad_norm": 0.28962087631225586, + "learning_rate": 9.993017393376844e-05, + "loss": 0.34222665429115295, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 72.51, + "memory/max_allocated (GiB)": 72.51, + "ppl": 1.40808, + "step": 46, + "tokens/total": 11252224, + "tokens/train_per_sec_per_gpu": 12.59, + "tokens/trainable": 90773 + }, + { + "epoch": 0.09591836734693877, + "grad_norm": 0.41601240634918213, + "learning_rate": 9.992117528313025e-05, + "loss": 0.41904228925704956, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 74.3, + "memory/max_allocated (GiB)": 74.3, + "ppl": 1.5205, + "step": 47, + "tokens/total": 11496768, + "tokens/train_per_sec_per_gpu": 13.52, + "tokens/trainable": 93011 + }, + { + "epoch": 0.09795918367346938, + "grad_norm": 0.2995956838130951, + "learning_rate": 9.991163185003029e-05, + "loss": 0.32476919889450073, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 76.81, + "memory/max_allocated (GiB)": 76.81, + "ppl": 1.38371, + "step": 48, + "tokens/total": 11754816, + "tokens/train_per_sec_per_gpu": 11.42, + "tokens/trainable": 94980 + }, + { + "epoch": 0.1, + "grad_norm": 0.35347771644592285, + "learning_rate": 9.990154373861462e-05, + "loss": 0.3826776444911957, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 72.81, + "memory/max_allocated (GiB)": 72.81, + "ppl": 1.46621, + "step": 49, + "tokens/total": 11997568, + "tokens/train_per_sec_per_gpu": 14.25, + "tokens/trainable": 97237 + }, + { + "epoch": 0.10204081632653061, + "grad_norm": 0.37251126766204834, + "learning_rate": 9.989091105897335e-05, + "loss": 0.2627924084663391, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 73.71, + "memory/max_allocated (GiB)": 73.71, + "ppl": 1.30056, + "step": 50, + "tokens/total": 12199616, + "tokens/train_per_sec_per_gpu": 13.62, + "tokens/trainable": 99077 + }, + { + "epoch": 0.10408163265306122, + "grad_norm": 0.4113054871559143, + "learning_rate": 9.987973392713932e-05, + "loss": 0.26638543605804443, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 75.09, + "memory/max_allocated (GiB)": 75.09, + "ppl": 1.30524, + "step": 51, + "tokens/total": 12458944, + "tokens/train_per_sec_per_gpu": 8.67, + "tokens/trainable": 100904 + }, + { + "epoch": 0.10612244897959183, + "grad_norm": 0.27514827251434326, + "learning_rate": 9.986801246508695e-05, + "loss": 0.3189134895801544, + "memory/device_reserved (GiB)": 76.16, + "memory/max_active (GiB)": 75.12, + "memory/max_allocated (GiB)": 75.12, + "ppl": 1.37563, + "step": 52, + "tokens/total": 12717056, + "tokens/train_per_sec_per_gpu": 9.76, + "tokens/trainable": 102655 + }, + { + "epoch": 0.10816326530612246, + "grad_norm": 0.2949848473072052, + "learning_rate": 9.985574680073085e-05, + "loss": 0.2989092767238617, + "memory/device_reserved (GiB)": 76.16, + "memory/max_active (GiB)": 74.2, + "memory/max_allocated (GiB)": 74.2, + "ppl": 1.34839, + "step": 53, + "tokens/total": 12979520, + "tokens/train_per_sec_per_gpu": 10.41, + "tokens/trainable": 104528 + }, + { + "epoch": 0.11020408163265306, + "grad_norm": 0.2654709815979004, + "learning_rate": 9.984293706792438e-05, + "loss": 0.3334200382232666, + "memory/device_reserved (GiB)": 76.16, + "memory/max_active (GiB)": 74.44, + "memory/max_allocated (GiB)": 74.44, + "ppl": 1.39573, + "step": 54, + "tokens/total": 13256256, + "tokens/train_per_sec_per_gpu": 12.32, + "tokens/trainable": 106680 + }, + { + "epoch": 0.11224489795918367, + "grad_norm": 0.36602842807769775, + "learning_rate": 9.982958340645831e-05, + "loss": 0.39756059646606445, + "memory/device_reserved (GiB)": 76.16, + "memory/max_active (GiB)": 72.13, + "memory/max_allocated (GiB)": 72.13, + "ppl": 1.48819, + "step": 55, + "tokens/total": 13483968, + "tokens/train_per_sec_per_gpu": 13.09, + "tokens/trainable": 108647 + }, + { + "epoch": 0.11428571428571428, + "grad_norm": 0.2748600244522095, + "learning_rate": 9.981568596205917e-05, + "loss": 0.3055681586265564, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 76.7, + "memory/max_allocated (GiB)": 76.7, + "ppl": 1.3574, + "step": 56, + "tokens/total": 13749568, + "tokens/train_per_sec_per_gpu": 11.05, + "tokens/trainable": 110620 + }, + { + "epoch": 0.11632653061224489, + "grad_norm": 0.32185810804367065, + "learning_rate": 9.980124488638774e-05, + "loss": 0.37074923515319824, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 72.1, + "memory/max_allocated (GiB)": 72.1, + "ppl": 1.44882, + "step": 57, + "tokens/total": 13985344, + "tokens/train_per_sec_per_gpu": 16.08, + "tokens/trainable": 112743 + }, + { + "epoch": 0.11836734693877551, + "grad_norm": 0.3911129832267761, + "learning_rate": 9.978626033703732e-05, + "loss": 0.3393978774547577, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 73.43, + "memory/max_allocated (GiB)": 73.43, + "ppl": 1.4041, + "step": 58, + "tokens/total": 14196480, + "tokens/train_per_sec_per_gpu": 13.31, + "tokens/trainable": 114572 + }, + { + "epoch": 0.12040816326530612, + "grad_norm": 0.3574847877025604, + "learning_rate": 9.977073247753218e-05, + "loss": 0.3169441223144531, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 74.69, + "memory/max_allocated (GiB)": 74.69, + "ppl": 1.37293, + "step": 59, + "tokens/total": 14435008, + "tokens/train_per_sec_per_gpu": 13.0, + "tokens/trainable": 116707 + }, + { + "epoch": 0.12244897959183673, + "grad_norm": 0.30045002698898315, + "learning_rate": 9.97546614773255e-05, + "loss": 0.2758142650127411, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 74.39, + "memory/max_allocated (GiB)": 74.39, + "ppl": 1.3176, + "step": 60, + "tokens/total": 14648576, + "tokens/train_per_sec_per_gpu": 11.21, + "tokens/trainable": 118447 + }, + { + "epoch": 0.12448979591836734, + "grad_norm": 0.25962135195732117, + "learning_rate": 9.973804751179779e-05, + "loss": 0.27344828844070435, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 75.67, + "memory/max_allocated (GiB)": 75.67, + "ppl": 1.31449, + "step": 61, + "tokens/total": 14898368, + "tokens/train_per_sec_per_gpu": 12.14, + "tokens/trainable": 120410 + }, + { + "epoch": 0.12653061224489795, + "grad_norm": 0.2542338967323303, + "learning_rate": 9.972089076225477e-05, + "loss": 0.2752627432346344, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 75.34, + "memory/max_allocated (GiB)": 75.34, + "ppl": 1.31688, + "step": 62, + "tokens/total": 15121792, + "tokens/train_per_sec_per_gpu": 11.59, + "tokens/trainable": 122325 + }, + { + "epoch": 0.12857142857142856, + "grad_norm": 0.3297300338745117, + "learning_rate": 9.97031914159256e-05, + "loss": 0.3394104242324829, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 74.49, + "memory/max_allocated (GiB)": 74.49, + "ppl": 1.40412, + "step": 63, + "tokens/total": 15389312, + "tokens/train_per_sec_per_gpu": 12.41, + "tokens/trainable": 124224 + }, + { + "epoch": 0.1306122448979592, + "grad_norm": 0.28276708722114563, + "learning_rate": 9.968494966596058e-05, + "loss": 0.31572309136390686, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 73.95, + "memory/max_allocated (GiB)": 73.95, + "ppl": 1.37125, + "step": 64, + "tokens/total": 15630144, + "tokens/train_per_sec_per_gpu": 11.28, + "tokens/trainable": 126004 + }, + { + "epoch": 0.1326530612244898, + "grad_norm": 0.31448298692703247, + "learning_rate": 9.966616571142931e-05, + "loss": 0.333346962928772, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 76.32, + "memory/max_allocated (GiB)": 76.32, + "ppl": 1.39563, + "step": 65, + "tokens/total": 15912896, + "tokens/train_per_sec_per_gpu": 10.31, + "tokens/trainable": 128003 + }, + { + "epoch": 0.1346938775510204, + "grad_norm": 0.31218236684799194, + "learning_rate": 9.964683975731828e-05, + "loss": 0.3107539415359497, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 75.94, + "memory/max_allocated (GiB)": 75.94, + "ppl": 1.36445, + "step": 66, + "tokens/total": 16185728, + "tokens/train_per_sec_per_gpu": 12.49, + "tokens/trainable": 130139 + }, + { + "epoch": 0.13673469387755102, + "grad_norm": 0.35470691323280334, + "learning_rate": 9.962697201452881e-05, + "loss": 0.3213728368282318, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 76.32, + "memory/max_allocated (GiB)": 76.32, + "ppl": 1.37902, + "step": 67, + "tokens/total": 16449728, + "tokens/train_per_sec_per_gpu": 12.06, + "tokens/trainable": 132201 + }, + { + "epoch": 0.13877551020408163, + "grad_norm": 0.34666892886161804, + "learning_rate": 9.96065626998747e-05, + "loss": 0.37185078859329224, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 76.48, + "memory/max_allocated (GiB)": 76.48, + "ppl": 1.45042, + "step": 68, + "tokens/total": 16719488, + "tokens/train_per_sec_per_gpu": 9.9, + "tokens/trainable": 134160 + }, + { + "epoch": 0.14081632653061224, + "grad_norm": 0.32578787207603455, + "learning_rate": 9.958561203607975e-05, + "loss": 0.3230299651622772, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 73.57, + "memory/max_allocated (GiB)": 73.57, + "ppl": 1.38131, + "step": 69, + "tokens/total": 16953152, + "tokens/train_per_sec_per_gpu": 13.59, + "tokens/trainable": 136020 + }, + { + "epoch": 0.14285714285714285, + "grad_norm": 0.5893422961235046, + "learning_rate": 9.956412025177551e-05, + "loss": 0.35762453079223633, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 76.73, + "memory/max_allocated (GiB)": 76.73, + "ppl": 1.42993, + "step": 70, + "tokens/total": 17219072, + "tokens/train_per_sec_per_gpu": 11.73, + "tokens/trainable": 137981 + }, + { + "epoch": 0.14489795918367346, + "grad_norm": 0.2695072293281555, + "learning_rate": 9.954208758149867e-05, + "loss": 0.23460917174816132, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 73.92, + "memory/max_allocated (GiB)": 73.92, + "ppl": 1.26441, + "step": 71, + "tokens/total": 17490944, + "tokens/train_per_sec_per_gpu": 10.35, + "tokens/trainable": 139847 + }, + { + "epoch": 0.1469387755102041, + "grad_norm": 0.2971181869506836, + "learning_rate": 9.951951426568851e-05, + "loss": 0.2664044499397278, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 72.34, + "memory/max_allocated (GiB)": 72.34, + "ppl": 1.30526, + "step": 72, + "tokens/total": 17732032, + "tokens/train_per_sec_per_gpu": 13.7, + "tokens/trainable": 141879 + }, + { + "epoch": 0.1489795918367347, + "grad_norm": 0.3867640495300293, + "learning_rate": 9.949640055068432e-05, + "loss": 0.4099368751049042, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 72.29, + "memory/max_allocated (GiB)": 72.29, + "ppl": 1.50672, + "step": 73, + "tokens/total": 17969088, + "tokens/train_per_sec_per_gpu": 12.42, + "tokens/trainable": 143793 + }, + { + "epoch": 0.1510204081632653, + "grad_norm": 0.28894367814064026, + "learning_rate": 9.94727466887227e-05, + "loss": 0.34861448407173157, + "memory/device_reserved (GiB)": 77.77, + "memory/max_active (GiB)": 76.84, + "memory/max_allocated (GiB)": 76.84, + "ppl": 1.4171, + "step": 74, + "tokens/total": 18238784, + "tokens/train_per_sec_per_gpu": 11.22, + "tokens/trainable": 145834 + }, + { + "epoch": 0.15306122448979592, + "grad_norm": 0.2793218493461609, + "learning_rate": 9.944855293793475e-05, + "loss": 0.2703927457332611, + "memory/device_reserved (GiB)": 77.77, + "memory/max_active (GiB)": 76.51, + "memory/max_allocated (GiB)": 76.51, + "ppl": 1.31048, + "step": 75, + "tokens/total": 18495168, + "tokens/train_per_sec_per_gpu": 11.85, + "tokens/trainable": 147888 + }, + { + "epoch": 0.15510204081632653, + "grad_norm": 0.2610492408275604, + "learning_rate": 9.942381956234335e-05, + "loss": 0.2886396050453186, + "memory/device_reserved (GiB)": 77.77, + "memory/max_active (GiB)": 76.4, + "memory/max_allocated (GiB)": 76.4, + "ppl": 1.33461, + "step": 76, + "tokens/total": 18734016, + "tokens/train_per_sec_per_gpu": 9.23, + "tokens/trainable": 149764 + }, + { + "epoch": 0.15714285714285714, + "grad_norm": 0.31321877241134644, + "learning_rate": 9.939854683186016e-05, + "loss": 0.32045096158981323, + "memory/device_reserved (GiB)": 77.37, + "memory/max_active (GiB)": 73.62, + "memory/max_allocated (GiB)": 73.62, + "ppl": 1.37775, + "step": 77, + "tokens/total": 18972672, + "tokens/train_per_sec_per_gpu": 12.79, + "tokens/trainable": 151857 + }, + { + "epoch": 0.15918367346938775, + "grad_norm": 0.34637442231178284, + "learning_rate": 9.937273502228281e-05, + "loss": 0.36426636576652527, + "memory/device_reserved (GiB)": 77.37, + "memory/max_active (GiB)": 75.8, + "memory/max_allocated (GiB)": 75.8, + "ppl": 1.43946, + "step": 78, + "tokens/total": 19195392, + "tokens/train_per_sec_per_gpu": 13.15, + "tokens/trainable": 153856 + }, + { + "epoch": 0.16122448979591836, + "grad_norm": 0.34072041511535645, + "learning_rate": 9.934638441529179e-05, + "loss": 0.34764987230300903, + "memory/device_reserved (GiB)": 77.37, + "memory/max_active (GiB)": 74.06, + "memory/max_allocated (GiB)": 74.06, + "ppl": 1.41574, + "step": 79, + "tokens/total": 19472064, + "tokens/train_per_sec_per_gpu": 10.01, + "tokens/trainable": 155739 + }, + { + "epoch": 0.16326530612244897, + "grad_norm": 0.4057715833187103, + "learning_rate": 9.93194952984474e-05, + "loss": 0.38709670305252075, + "memory/device_reserved (GiB)": 77.37, + "memory/max_active (GiB)": 74.06, + "memory/max_allocated (GiB)": 74.06, + "ppl": 1.4727, + "step": 80, + "tokens/total": 19724544, + "tokens/train_per_sec_per_gpu": 13.56, + "tokens/trainable": 158060 + }, + { + "epoch": 0.1653061224489796, + "grad_norm": 0.3533332347869873, + "learning_rate": 9.929206796518662e-05, + "loss": 0.3303971588611603, + "memory/device_reserved (GiB)": 77.37, + "memory/max_active (GiB)": 71.64, + "memory/max_allocated (GiB)": 71.64, + "ppl": 1.39152, + "step": 81, + "tokens/total": 19941440, + "tokens/train_per_sec_per_gpu": 14.34, + "tokens/trainable": 159968 + }, + { + "epoch": 0.1673469387755102, + "grad_norm": 0.576310932636261, + "learning_rate": 9.926410271481994e-05, + "loss": 0.29558199644088745, + "memory/device_reserved (GiB)": 77.37, + "memory/max_active (GiB)": 74.52, + "memory/max_allocated (GiB)": 74.52, + "ppl": 1.34391, + "step": 82, + "tokens/total": 20194560, + "tokens/train_per_sec_per_gpu": 11.0, + "tokens/trainable": 161784 + }, + { + "epoch": 0.16938775510204082, + "grad_norm": 0.2885003685951233, + "learning_rate": 9.9235599852528e-05, + "loss": 0.2939549684524536, + "memory/device_reserved (GiB)": 77.37, + "memory/max_active (GiB)": 75.09, + "memory/max_allocated (GiB)": 75.09, + "ppl": 1.34172, + "step": 83, + "tokens/total": 20438272, + "tokens/train_per_sec_per_gpu": 12.77, + "tokens/trainable": 163745 + }, + { + "epoch": 0.17142857142857143, + "grad_norm": 0.27577269077301025, + "learning_rate": 9.920655968935838e-05, + "loss": 0.3108862042427063, + "memory/device_reserved (GiB)": 77.37, + "memory/max_active (GiB)": 73.62, + "memory/max_allocated (GiB)": 73.62, + "ppl": 1.36463, + "step": 84, + "tokens/total": 20647488, + "tokens/train_per_sec_per_gpu": 12.17, + "tokens/trainable": 165491 + }, + { + "epoch": 0.17346938775510204, + "grad_norm": 0.2597094178199768, + "learning_rate": 9.91769825422221e-05, + "loss": 0.29625654220581055, + "memory/device_reserved (GiB)": 77.37, + "memory/max_active (GiB)": 73.54, + "memory/max_allocated (GiB)": 73.54, + "ppl": 1.34482, + "step": 85, + "tokens/total": 20908352, + "tokens/train_per_sec_per_gpu": 12.03, + "tokens/trainable": 167486 + }, + { + "epoch": 0.17551020408163265, + "grad_norm": 0.31459540128707886, + "learning_rate": 9.914686873389025e-05, + "loss": 0.3925963342189789, + "memory/device_reserved (GiB)": 77.37, + "memory/max_active (GiB)": 73.95, + "memory/max_allocated (GiB)": 73.95, + "ppl": 1.48082, + "step": 86, + "tokens/total": 21166656, + "tokens/train_per_sec_per_gpu": 12.74, + "tokens/trainable": 169501 + }, + { + "epoch": 0.17755102040816326, + "grad_norm": 0.3207368552684784, + "learning_rate": 9.91162185929904e-05, + "loss": 0.36425894498825073, + "memory/device_reserved (GiB)": 77.53, + "memory/max_active (GiB)": 76.35, + "memory/max_allocated (GiB)": 76.35, + "ppl": 1.43945, + "step": 87, + "tokens/total": 21438912, + "tokens/train_per_sec_per_gpu": 11.2, + "tokens/trainable": 171575 + }, + { + "epoch": 0.17959183673469387, + "grad_norm": 0.30726850032806396, + "learning_rate": 9.908503245400302e-05, + "loss": 0.363973468542099, + "memory/device_reserved (GiB)": 77.53, + "memory/max_active (GiB)": 75.56, + "memory/max_allocated (GiB)": 75.56, + "ppl": 1.43904, + "step": 88, + "tokens/total": 21722176, + "tokens/train_per_sec_per_gpu": 12.12, + "tokens/trainable": 173659 + }, + { + "epoch": 0.1816326530612245, + "grad_norm": 0.3307989835739136, + "learning_rate": 9.90533106572579e-05, + "loss": 0.37638866901397705, + "memory/device_reserved (GiB)": 77.53, + "memory/max_active (GiB)": 74.11, + "memory/max_allocated (GiB)": 74.11, + "ppl": 1.45701, + "step": 89, + "tokens/total": 21958848, + "tokens/train_per_sec_per_gpu": 13.93, + "tokens/trainable": 175607 + }, + { + "epoch": 0.1836734693877551, + "grad_norm": 0.2957727313041687, + "learning_rate": 9.90210535489303e-05, + "loss": 0.2558442950248718, + "memory/device_reserved (GiB)": 77.53, + "memory/max_active (GiB)": 72.89, + "memory/max_allocated (GiB)": 72.89, + "ppl": 1.29155, + "step": 90, + "tokens/total": 22210112, + "tokens/train_per_sec_per_gpu": 13.12, + "tokens/trainable": 177797 + }, + { + "epoch": 0.18571428571428572, + "grad_norm": 0.2863942086696625, + "learning_rate": 9.898826148103737e-05, + "loss": 0.3638124465942383, + "memory/device_reserved (GiB)": 77.53, + "memory/max_active (GiB)": 75.34, + "memory/max_allocated (GiB)": 75.34, + "ppl": 1.4388, + "step": 91, + "tokens/total": 22450496, + "tokens/train_per_sec_per_gpu": 11.32, + "tokens/trainable": 179665 + }, + { + "epoch": 0.18775510204081633, + "grad_norm": 0.2795948088169098, + "learning_rate": 9.895493481143411e-05, + "loss": 0.31911370158195496, + "memory/device_reserved (GiB)": 77.53, + "memory/max_active (GiB)": 74.03, + "memory/max_allocated (GiB)": 74.03, + "ppl": 1.37591, + "step": 92, + "tokens/total": 22683776, + "tokens/train_per_sec_per_gpu": 13.26, + "tokens/trainable": 181685 + }, + { + "epoch": 0.18979591836734694, + "grad_norm": 0.25272026658058167, + "learning_rate": 9.892107390380958e-05, + "loss": 0.2508898377418518, + "memory/device_reserved (GiB)": 77.53, + "memory/max_active (GiB)": 72.37, + "memory/max_allocated (GiB)": 72.37, + "ppl": 1.28517, + "step": 93, + "tokens/total": 22917888, + "tokens/train_per_sec_per_gpu": 12.28, + "tokens/trainable": 183472 + }, + { + "epoch": 0.19183673469387755, + "grad_norm": 0.35054582357406616, + "learning_rate": 9.888667912768288e-05, + "loss": 0.35517990589141846, + "memory/device_reserved (GiB)": 77.53, + "memory/max_active (GiB)": 74.39, + "memory/max_allocated (GiB)": 74.39, + "ppl": 1.42644, + "step": 94, + "tokens/total": 23187520, + "tokens/train_per_sec_per_gpu": 10.77, + "tokens/trainable": 185479 + }, + { + "epoch": 0.19387755102040816, + "grad_norm": 0.26934683322906494, + "learning_rate": 9.885175085839919e-05, + "loss": 0.31262412667274475, + "memory/device_reserved (GiB)": 77.53, + "memory/max_active (GiB)": 74.47, + "memory/max_allocated (GiB)": 74.47, + "ppl": 1.36701, + "step": 95, + "tokens/total": 23443008, + "tokens/train_per_sec_per_gpu": 13.83, + "tokens/trainable": 187567 + }, + { + "epoch": 0.19591836734693877, + "grad_norm": 0.5132613778114319, + "learning_rate": 9.881628947712556e-05, + "loss": 0.273453950881958, + "memory/device_reserved (GiB)": 77.53, + "memory/max_active (GiB)": 73.79, + "memory/max_allocated (GiB)": 73.79, + "ppl": 1.3145, + "step": 96, + "tokens/total": 23698432, + "tokens/train_per_sec_per_gpu": 10.01, + "tokens/trainable": 189220 + }, + { + "epoch": 0.19795918367346937, + "grad_norm": 0.4519254267215729, + "learning_rate": 9.878029537084684e-05, + "loss": 0.28466570377349854, + "memory/device_reserved (GiB)": 77.53, + "memory/max_active (GiB)": 74.82, + "memory/max_allocated (GiB)": 74.82, + "ppl": 1.32932, + "step": 97, + "tokens/total": 23947968, + "tokens/train_per_sec_per_gpu": 10.12, + "tokens/trainable": 190943 + }, + { + "epoch": 0.2, + "grad_norm": 0.2954767048358917, + "learning_rate": 9.874376893236145e-05, + "loss": 0.33792048692703247, + "memory/device_reserved (GiB)": 77.53, + "memory/max_active (GiB)": 73.79, + "memory/max_allocated (GiB)": 73.79, + "ppl": 1.40203, + "step": 98, + "tokens/total": 24192704, + "tokens/train_per_sec_per_gpu": 13.04, + "tokens/trainable": 192967 + }, + { + "epoch": 0.20204081632653062, + "grad_norm": 0.3651937246322632, + "learning_rate": 9.870671056027705e-05, + "loss": 0.30805450677871704, + "memory/device_reserved (GiB)": 77.53, + "memory/max_active (GiB)": 73.3, + "memory/max_allocated (GiB)": 73.3, + "ppl": 1.36078, + "step": 99, + "tokens/total": 24420672, + "tokens/train_per_sec_per_gpu": 13.96, + "tokens/trainable": 194997 + }, + { + "epoch": 0.20408163265306123, + "grad_norm": 0.2739502489566803, + "learning_rate": 9.866912065900621e-05, + "loss": 0.2902035117149353, + "memory/device_reserved (GiB)": 77.53, + "memory/max_active (GiB)": 76.51, + "memory/max_allocated (GiB)": 76.51, + "ppl": 1.3367, + "step": 100, + "tokens/total": 24663168, + "tokens/train_per_sec_per_gpu": 13.35, + "tokens/trainable": 197125 + }, + { + "epoch": 0.20612244897959184, + "grad_norm": 0.2788029611110687, + "learning_rate": 9.863099963876201e-05, + "loss": 0.3034919500350952, + "memory/device_reserved (GiB)": 77.53, + "memory/max_active (GiB)": 75.58, + "memory/max_allocated (GiB)": 75.58, + "ppl": 1.35458, + "step": 101, + "tokens/total": 24901056, + "tokens/train_per_sec_per_gpu": 10.13, + "tokens/trainable": 199053 + }, + { + "epoch": 0.20816326530612245, + "grad_norm": 0.2763316333293915, + "learning_rate": 9.859234791555355e-05, + "loss": 0.2486642301082611, + "memory/device_reserved (GiB)": 77.12, + "memory/max_active (GiB)": 76.29, + "memory/max_allocated (GiB)": 76.29, + "ppl": 1.28231, + "step": 102, + "tokens/total": 25157504, + "tokens/train_per_sec_per_gpu": 10.61, + "tokens/trainable": 200967 + }, + { + "epoch": 0.21020408163265306, + "grad_norm": 0.24316149950027466, + "learning_rate": 9.855316591118142e-05, + "loss": 0.2835119366645813, + "memory/device_reserved (GiB)": 77.12, + "memory/max_active (GiB)": 75.42, + "memory/max_allocated (GiB)": 75.42, + "ppl": 1.32778, + "step": 103, + "tokens/total": 25423872, + "tokens/train_per_sec_per_gpu": 11.78, + "tokens/trainable": 202962 + }, + { + "epoch": 0.21224489795918366, + "grad_norm": 0.29139262437820435, + "learning_rate": 9.851345405323307e-05, + "loss": 0.29183459281921387, + "memory/device_reserved (GiB)": 77.12, + "memory/max_active (GiB)": 74.03, + "memory/max_allocated (GiB)": 74.03, + "ppl": 1.33888, + "step": 104, + "tokens/total": 25640512, + "tokens/train_per_sec_per_gpu": 13.7, + "tokens/trainable": 204885 + }, + { + "epoch": 0.21428571428571427, + "grad_norm": 0.2823760211467743, + "learning_rate": 9.84732127750782e-05, + "loss": 0.22634169459342957, + "memory/device_reserved (GiB)": 77.14, + "memory/max_active (GiB)": 76.13, + "memory/max_allocated (GiB)": 76.13, + "ppl": 1.254, + "step": 105, + "tokens/total": 25880832, + "tokens/train_per_sec_per_gpu": 12.23, + "tokens/trainable": 206643 + }, + { + "epoch": 0.2163265306122449, + "grad_norm": 0.22927753627300262, + "learning_rate": 9.843244251586397e-05, + "loss": 0.1902308166027069, + "memory/device_reserved (GiB)": 77.14, + "memory/max_active (GiB)": 76.1, + "memory/max_allocated (GiB)": 76.1, + "ppl": 1.20953, + "step": 106, + "tokens/total": 26132928, + "tokens/train_per_sec_per_gpu": 11.31, + "tokens/trainable": 208388 + }, + { + "epoch": 0.21836734693877552, + "grad_norm": 0.268435537815094, + "learning_rate": 9.839114372051024e-05, + "loss": 0.2938929796218872, + "memory/device_reserved (GiB)": 77.14, + "memory/max_active (GiB)": 74.0, + "memory/max_allocated (GiB)": 74.0, + "ppl": 1.34164, + "step": 107, + "tokens/total": 26360768, + "tokens/train_per_sec_per_gpu": 13.07, + "tokens/trainable": 210308 + }, + { + "epoch": 0.22040816326530613, + "grad_norm": 0.2757108509540558, + "learning_rate": 9.834931683970468e-05, + "loss": 0.2682138979434967, + "memory/device_reserved (GiB)": 77.14, + "memory/max_active (GiB)": 72.15, + "memory/max_allocated (GiB)": 72.15, + "ppl": 1.30763, + "step": 108, + "tokens/total": 26586944, + "tokens/train_per_sec_per_gpu": 12.24, + "tokens/trainable": 212251 + }, + { + "epoch": 0.22244897959183674, + "grad_norm": 0.2802518904209137, + "learning_rate": 9.830696232989793e-05, + "loss": 0.2823032736778259, + "memory/device_reserved (GiB)": 77.14, + "memory/max_active (GiB)": 74.12, + "memory/max_allocated (GiB)": 74.12, + "ppl": 1.32618, + "step": 109, + "tokens/total": 26843840, + "tokens/train_per_sec_per_gpu": 13.56, + "tokens/trainable": 214312 + }, + { + "epoch": 0.22448979591836735, + "grad_norm": 0.5594452023506165, + "learning_rate": 9.826408065329853e-05, + "loss": 0.38349130749702454, + "memory/device_reserved (GiB)": 77.14, + "memory/max_active (GiB)": 76.18, + "memory/max_allocated (GiB)": 76.18, + "ppl": 1.4674, + "step": 110, + "tokens/total": 27089600, + "tokens/train_per_sec_per_gpu": 12.47, + "tokens/trainable": 216567 + }, + { + "epoch": 0.22653061224489796, + "grad_norm": 0.2607867419719696, + "learning_rate": 9.822067227786793e-05, + "loss": 0.31956565380096436, + "memory/device_reserved (GiB)": 77.14, + "memory/max_active (GiB)": 74.22, + "memory/max_allocated (GiB)": 74.22, + "ppl": 1.37653, + "step": 111, + "tokens/total": 27325696, + "tokens/train_per_sec_per_gpu": 15.4, + "tokens/trainable": 218793 + }, + { + "epoch": 0.22857142857142856, + "grad_norm": 0.6517570614814758, + "learning_rate": 9.817673767731535e-05, + "loss": 0.20926529169082642, + "memory/device_reserved (GiB)": 77.14, + "memory/max_active (GiB)": 73.62, + "memory/max_allocated (GiB)": 73.62, + "ppl": 1.23277, + "step": 112, + "tokens/total": 27571328, + "tokens/train_per_sec_per_gpu": 9.82, + "tokens/trainable": 220384 + }, + { + "epoch": 0.23061224489795917, + "grad_norm": 0.2895031273365021, + "learning_rate": 9.813227733109268e-05, + "loss": 0.3075171113014221, + "memory/device_reserved (GiB)": 77.14, + "memory/max_active (GiB)": 76.02, + "memory/max_allocated (GiB)": 76.02, + "ppl": 1.36004, + "step": 113, + "tokens/total": 27822464, + "tokens/train_per_sec_per_gpu": 12.46, + "tokens/trainable": 222598 + }, + { + "epoch": 0.23265306122448978, + "grad_norm": 0.2864657938480377, + "learning_rate": 9.808729172438908e-05, + "loss": 0.371175616979599, + "memory/device_reserved (GiB)": 77.14, + "memory/max_active (GiB)": 73.87, + "memory/max_allocated (GiB)": 73.87, + "ppl": 1.44944, + "step": 114, + "tokens/total": 28060800, + "tokens/train_per_sec_per_gpu": 14.41, + "tokens/trainable": 224757 + }, + { + "epoch": 0.23469387755102042, + "grad_norm": 0.2823830842971802, + "learning_rate": 9.804178134812596e-05, + "loss": 0.32755061984062195, + "memory/device_reserved (GiB)": 77.14, + "memory/max_active (GiB)": 73.92, + "memory/max_allocated (GiB)": 73.92, + "ppl": 1.38757, + "step": 115, + "tokens/total": 28311232, + "tokens/train_per_sec_per_gpu": 12.9, + "tokens/trainable": 226762 + }, + { + "epoch": 0.23673469387755103, + "grad_norm": 0.31970539689064026, + "learning_rate": 9.799574669895133e-05, + "loss": 0.4158673882484436, + "memory/device_reserved (GiB)": 77.14, + "memory/max_active (GiB)": 75.42, + "memory/max_allocated (GiB)": 75.42, + "ppl": 1.51568, + "step": 116, + "tokens/total": 28578496, + "tokens/train_per_sec_per_gpu": 12.92, + "tokens/trainable": 228958 + }, + { + "epoch": 0.23877551020408164, + "grad_norm": 0.3378254473209381, + "learning_rate": 9.794918827923458e-05, + "loss": 0.27569928765296936, + "memory/device_reserved (GiB)": 77.14, + "memory/max_active (GiB)": 72.04, + "memory/max_allocated (GiB)": 72.04, + "ppl": 1.31745, + "step": 117, + "tokens/total": 28820800, + "tokens/train_per_sec_per_gpu": 11.08, + "tokens/trainable": 230817 + }, + { + "epoch": 0.24081632653061225, + "grad_norm": 0.4628329575061798, + "learning_rate": 9.790210659706089e-05, + "loss": 0.2634648084640503, + "memory/device_reserved (GiB)": 77.14, + "memory/max_active (GiB)": 75.28, + "memory/max_allocated (GiB)": 75.28, + "ppl": 1.30143, + "step": 118, + "tokens/total": 29091200, + "tokens/train_per_sec_per_gpu": 11.53, + "tokens/trainable": 232783 + }, + { + "epoch": 0.24285714285714285, + "grad_norm": 0.25572919845581055, + "learning_rate": 9.785450216622579e-05, + "loss": 0.2887064814567566, + "memory/device_reserved (GiB)": 77.14, + "memory/max_active (GiB)": 75.69, + "memory/max_allocated (GiB)": 75.69, + "ppl": 1.3347, + "step": 119, + "tokens/total": 29355520, + "tokens/train_per_sec_per_gpu": 10.76, + "tokens/trainable": 234569 + }, + { + "epoch": 0.24489795918367346, + "grad_norm": 0.3095442056655884, + "learning_rate": 9.780637550622941e-05, + "loss": 0.31611189246177673, + "memory/device_reserved (GiB)": 77.14, + "memory/max_active (GiB)": 73.9, + "memory/max_allocated (GiB)": 73.9, + "ppl": 1.37178, + "step": 120, + "tokens/total": 29599296, + "tokens/train_per_sec_per_gpu": 13.69, + "tokens/trainable": 236559 + }, + { + "epoch": 0.24693877551020407, + "grad_norm": 0.42760133743286133, + "learning_rate": 9.775772714227096e-05, + "loss": 0.27209585905075073, + "memory/device_reserved (GiB)": 77.14, + "memory/max_active (GiB)": 74.88, + "memory/max_allocated (GiB)": 74.88, + "ppl": 1.31271, + "step": 121, + "tokens/total": 29828288, + "tokens/train_per_sec_per_gpu": 14.38, + "tokens/trainable": 238519 + }, + { + "epoch": 0.24897959183673468, + "grad_norm": 0.8849247097969055, + "learning_rate": 9.770855760524288e-05, + "loss": 0.22954438626766205, + "memory/device_reserved (GiB)": 77.14, + "memory/max_active (GiB)": 75.83, + "memory/max_allocated (GiB)": 75.83, + "ppl": 1.25803, + "step": 122, + "tokens/total": 30097920, + "tokens/train_per_sec_per_gpu": 9.72, + "tokens/trainable": 240260 + }, + { + "epoch": 0.2510204081632653, + "grad_norm": 0.29249808192253113, + "learning_rate": 9.765886743172511e-05, + "loss": 0.3366485834121704, + "memory/device_reserved (GiB)": 77.14, + "memory/max_active (GiB)": 76.21, + "memory/max_allocated (GiB)": 76.21, + "ppl": 1.40025, + "step": 123, + "tokens/total": 30341760, + "tokens/train_per_sec_per_gpu": 14.18, + "tokens/trainable": 242301 + }, + { + "epoch": 0.2530612244897959, + "grad_norm": 0.2477550059556961, + "learning_rate": 9.760865716397924e-05, + "loss": 0.26008856296539307, + "memory/device_reserved (GiB)": 77.14, + "memory/max_active (GiB)": 76.32, + "memory/max_allocated (GiB)": 76.32, + "ppl": 1.29704, + "step": 124, + "tokens/total": 30606272, + "tokens/train_per_sec_per_gpu": 11.8, + "tokens/trainable": 244133 + }, + { + "epoch": 0.25510204081632654, + "grad_norm": 0.2991919219493866, + "learning_rate": 9.755792734994255e-05, + "loss": 0.2831898033618927, + "memory/device_reserved (GiB)": 77.14, + "memory/max_active (GiB)": 73.51, + "memory/max_allocated (GiB)": 73.51, + "ppl": 1.32736, + "step": 125, + "tokens/total": 30841408, + "tokens/train_per_sec_per_gpu": 11.69, + "tokens/trainable": 245844 + }, + { + "epoch": 0.2571428571428571, + "grad_norm": 0.2869957983493805, + "learning_rate": 9.750667854322207e-05, + "loss": 0.29190540313720703, + "memory/device_reserved (GiB)": 77.29, + "memory/max_active (GiB)": 76.18, + "memory/max_allocated (GiB)": 76.18, + "ppl": 1.33898, + "step": 126, + "tokens/total": 31081472, + "tokens/train_per_sec_per_gpu": 8.64, + "tokens/trainable": 247730 + }, + { + "epoch": 0.25918367346938775, + "grad_norm": 0.2700897455215454, + "learning_rate": 9.745491130308848e-05, + "loss": 0.2809198796749115, + "memory/device_reserved (GiB)": 77.29, + "memory/max_active (GiB)": 75.04, + "memory/max_allocated (GiB)": 75.04, + "ppl": 1.32435, + "step": 127, + "tokens/total": 31318592, + "tokens/train_per_sec_per_gpu": 12.27, + "tokens/trainable": 249627 + }, + { + "epoch": 0.2612244897959184, + "grad_norm": 0.3195943534374237, + "learning_rate": 9.74026261944701e-05, + "loss": 0.2954328656196594, + "memory/device_reserved (GiB)": 77.29, + "memory/max_active (GiB)": 72.15, + "memory/max_allocated (GiB)": 72.15, + "ppl": 1.34371, + "step": 128, + "tokens/total": 31555904, + "tokens/train_per_sec_per_gpu": 11.63, + "tokens/trainable": 251506 + }, + { + "epoch": 0.26326530612244897, + "grad_norm": 0.25746840238571167, + "learning_rate": 9.734982378794661e-05, + "loss": 0.2928959131240845, + "memory/device_reserved (GiB)": 77.29, + "memory/max_active (GiB)": 76.29, + "memory/max_allocated (GiB)": 76.29, + "ppl": 1.3403, + "step": 129, + "tokens/total": 31787776, + "tokens/train_per_sec_per_gpu": 12.94, + "tokens/trainable": 253415 + }, + { + "epoch": 0.2653061224489796, + "grad_norm": 0.27648335695266724, + "learning_rate": 9.729650465974299e-05, + "loss": 0.27435725927352905, + "memory/device_reserved (GiB)": 77.29, + "memory/max_active (GiB)": 74.06, + "memory/max_allocated (GiB)": 74.06, + "ppl": 1.31568, + "step": 130, + "tokens/total": 32039488, + "tokens/train_per_sec_per_gpu": 12.14, + "tokens/trainable": 255297 + }, + { + "epoch": 0.2673469387755102, + "grad_norm": 0.3401206135749817, + "learning_rate": 9.724266939172302e-05, + "loss": 0.3491106629371643, + "memory/device_reserved (GiB)": 77.29, + "memory/max_active (GiB)": 73.11, + "memory/max_allocated (GiB)": 73.11, + "ppl": 1.41781, + "step": 131, + "tokens/total": 32282560, + "tokens/train_per_sec_per_gpu": 13.1, + "tokens/trainable": 257159 + }, + { + "epoch": 0.2693877551020408, + "grad_norm": 0.27172669768333435, + "learning_rate": 9.718831857138309e-05, + "loss": 0.28538137674331665, + "memory/device_reserved (GiB)": 77.29, + "memory/max_active (GiB)": 75.94, + "memory/max_allocated (GiB)": 75.94, + "ppl": 1.33027, + "step": 132, + "tokens/total": 32520768, + "tokens/train_per_sec_per_gpu": 14.36, + "tokens/trainable": 259207 + }, + { + "epoch": 0.2714285714285714, + "grad_norm": 0.2485136240720749, + "learning_rate": 9.713345279184573e-05, + "loss": 0.2645806670188904, + "memory/device_reserved (GiB)": 77.29, + "memory/max_active (GiB)": 73.36, + "memory/max_allocated (GiB)": 73.36, + "ppl": 1.30288, + "step": 133, + "tokens/total": 32780928, + "tokens/train_per_sec_per_gpu": 12.27, + "tokens/trainable": 261322 + }, + { + "epoch": 0.27346938775510204, + "grad_norm": 0.3484657108783722, + "learning_rate": 9.707807265185314e-05, + "loss": 0.35569071769714355, + "memory/device_reserved (GiB)": 77.29, + "memory/max_active (GiB)": 73.24, + "memory/max_allocated (GiB)": 73.24, + "ppl": 1.42717, + "step": 134, + "tokens/total": 33009600, + "tokens/train_per_sec_per_gpu": 12.76, + "tokens/trainable": 263351 + }, + { + "epoch": 0.2755102040816326, + "grad_norm": 0.3266739547252655, + "learning_rate": 9.702217875576068e-05, + "loss": 0.3133898377418518, + "memory/device_reserved (GiB)": 77.29, + "memory/max_active (GiB)": 75.31, + "memory/max_allocated (GiB)": 75.31, + "ppl": 1.36805, + "step": 135, + "tokens/total": 33260288, + "tokens/train_per_sec_per_gpu": 12.93, + "tokens/trainable": 265322 + }, + { + "epoch": 0.27755102040816326, + "grad_norm": 0.30616116523742676, + "learning_rate": 9.69657717135302e-05, + "loss": 0.35253050923347473, + "memory/device_reserved (GiB)": 77.29, + "memory/max_active (GiB)": 74.88, + "memory/max_allocated (GiB)": 74.88, + "ppl": 1.42266, + "step": 136, + "tokens/total": 33525632, + "tokens/train_per_sec_per_gpu": 11.93, + "tokens/trainable": 267374 + }, + { + "epoch": 0.2795918367346939, + "grad_norm": 0.24909131228923798, + "learning_rate": 9.690885214072352e-05, + "loss": 0.2477751225233078, + "memory/device_reserved (GiB)": 77.29, + "memory/max_active (GiB)": 76.05, + "memory/max_allocated (GiB)": 76.05, + "ppl": 1.28117, + "step": 137, + "tokens/total": 33775872, + "tokens/train_per_sec_per_gpu": 12.36, + "tokens/trainable": 269403 + }, + { + "epoch": 0.2816326530612245, + "grad_norm": 0.27541583776474, + "learning_rate": 9.685142065849556e-05, + "loss": 0.26661932468414307, + "memory/device_reserved (GiB)": 77.29, + "memory/max_active (GiB)": 72.83, + "memory/max_allocated (GiB)": 72.83, + "ppl": 1.30554, + "step": 138, + "tokens/total": 34006400, + "tokens/train_per_sec_per_gpu": 13.13, + "tokens/trainable": 271301 + }, + { + "epoch": 0.2836734693877551, + "grad_norm": 0.25799280405044556, + "learning_rate": 9.679347789358766e-05, + "loss": 0.25472038984298706, + "memory/device_reserved (GiB)": 77.29, + "memory/max_active (GiB)": 76.32, + "memory/max_allocated (GiB)": 76.32, + "ppl": 1.2901, + "step": 139, + "tokens/total": 34284992, + "tokens/train_per_sec_per_gpu": 10.12, + "tokens/trainable": 273314 + }, + { + "epoch": 0.2857142857142857, + "grad_norm": 0.30096814036369324, + "learning_rate": 9.673502447832071e-05, + "loss": 0.3603587746620178, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 76.64, + "memory/max_allocated (GiB)": 76.64, + "ppl": 1.43384, + "step": 140, + "tokens/total": 34536000, + "tokens/train_per_sec_per_gpu": 12.71, + "tokens/trainable": 275380 + }, + { + "epoch": 0.28775510204081634, + "grad_norm": 0.3104712665081024, + "learning_rate": 9.667606105058828e-05, + "loss": 0.29542291164398193, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 72.75, + "memory/max_allocated (GiB)": 72.75, + "ppl": 1.34369, + "step": 141, + "tokens/total": 34783360, + "tokens/train_per_sec_per_gpu": 10.77, + "tokens/trainable": 277127 + }, + { + "epoch": 0.2897959183673469, + "grad_norm": 0.30640509724617004, + "learning_rate": 9.661658825384959e-05, + "loss": 0.3147083520889282, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 72.26, + "memory/max_allocated (GiB)": 72.26, + "ppl": 1.36986, + "step": 142, + "tokens/total": 35014912, + "tokens/train_per_sec_per_gpu": 12.62, + "tokens/trainable": 278957 + }, + { + "epoch": 0.29183673469387755, + "grad_norm": 0.2548229992389679, + "learning_rate": 9.655660673712254e-05, + "loss": 0.31100380420684814, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 75.99, + "memory/max_allocated (GiB)": 75.99, + "ppl": 1.36479, + "step": 143, + "tokens/total": 35273536, + "tokens/train_per_sec_per_gpu": 11.7, + "tokens/trainable": 280951 + }, + { + "epoch": 0.2938775510204082, + "grad_norm": 0.29141655564308167, + "learning_rate": 9.649611715497662e-05, + "loss": 0.27038058638572693, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 75.18, + "memory/max_allocated (GiB)": 75.18, + "ppl": 1.31046, + "step": 144, + "tokens/total": 35527168, + "tokens/train_per_sec_per_gpu": 11.5, + "tokens/trainable": 282752 + }, + { + "epoch": 0.29591836734693877, + "grad_norm": 0.43348321318626404, + "learning_rate": 9.643512016752579e-05, + "loss": 0.4798806607723236, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 75.23, + "memory/max_allocated (GiB)": 75.23, + "ppl": 1.61588, + "step": 145, + "tokens/total": 35796480, + "tokens/train_per_sec_per_gpu": 11.25, + "tokens/trainable": 284653 + }, + { + "epoch": 0.2979591836734694, + "grad_norm": 0.2728241980075836, + "learning_rate": 9.637361644042118e-05, + "loss": 0.286792516708374, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 76.56, + "memory/max_allocated (GiB)": 76.56, + "ppl": 1.33215, + "step": 146, + "tokens/total": 36064384, + "tokens/train_per_sec_per_gpu": 10.29, + "tokens/trainable": 286479 + }, + { + "epoch": 0.3, + "grad_norm": 0.30759933590888977, + "learning_rate": 9.631160664484398e-05, + "loss": 0.32493382692337036, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 73.81, + "memory/max_allocated (GiB)": 73.81, + "ppl": 1.38394, + "step": 147, + "tokens/total": 36328320, + "tokens/train_per_sec_per_gpu": 13.16, + "tokens/trainable": 288740 + }, + { + "epoch": 0.3020408163265306, + "grad_norm": 0.2914741635322571, + "learning_rate": 9.624909145749798e-05, + "loss": 0.33574390411376953, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 73.57, + "memory/max_allocated (GiB)": 73.57, + "ppl": 1.39898, + "step": 148, + "tokens/total": 36571648, + "tokens/train_per_sec_per_gpu": 13.25, + "tokens/trainable": 290562 + }, + { + "epoch": 0.3040816326530612, + "grad_norm": 0.29609742760658264, + "learning_rate": 9.618607156060227e-05, + "loss": 0.2782217264175415, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 75.15, + "memory/max_allocated (GiB)": 75.15, + "ppl": 1.32078, + "step": 149, + "tokens/total": 36835264, + "tokens/train_per_sec_per_gpu": 12.51, + "tokens/trainable": 292578 + }, + { + "epoch": 0.30612244897959184, + "grad_norm": 0.28771230578422546, + "learning_rate": 9.612254764188369e-05, + "loss": 0.3707728981971741, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 73.84, + "memory/max_allocated (GiB)": 73.84, + "ppl": 1.44885, + "step": 150, + "tokens/total": 37060352, + "tokens/train_per_sec_per_gpu": 12.85, + "tokens/trainable": 294477 + }, + { + "epoch": 0.3081632653061224, + "grad_norm": 0.3155246675014496, + "learning_rate": 9.605852039456947e-05, + "loss": 0.34479111433029175, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 73.62, + "memory/max_allocated (GiB)": 73.62, + "ppl": 1.4117, + "step": 151, + "tokens/total": 37302080, + "tokens/train_per_sec_per_gpu": 11.81, + "tokens/trainable": 296799 + }, + { + "epoch": 0.31020408163265306, + "grad_norm": 0.24003657698631287, + "learning_rate": 9.599399051737958e-05, + "loss": 0.26617324352264404, + "memory/device_reserved (GiB)": 76.61, + "memory/max_active (GiB)": 75.75, + "memory/max_allocated (GiB)": 75.75, + "ppl": 1.30496, + "step": 152, + "tokens/total": 37547648, + "tokens/train_per_sec_per_gpu": 11.95, + "tokens/trainable": 298641 + }, + { + "epoch": 0.3122448979591837, + "grad_norm": 0.2692473530769348, + "learning_rate": 9.592895871451908e-05, + "loss": 0.26289498805999756, + "memory/device_reserved (GiB)": 76.61, + "memory/max_active (GiB)": 75.31, + "memory/max_allocated (GiB)": 75.31, + "ppl": 1.30069, + "step": 153, + "tokens/total": 37777408, + "tokens/train_per_sec_per_gpu": 10.39, + "tokens/trainable": 300321 + }, + { + "epoch": 0.3142857142857143, + "grad_norm": 0.26663610339164734, + "learning_rate": 9.58634256956705e-05, + "loss": 0.3169020414352417, + "memory/device_reserved (GiB)": 76.61, + "memory/max_active (GiB)": 73.3, + "memory/max_allocated (GiB)": 73.3, + "ppl": 1.37287, + "step": 154, + "tokens/total": 38015744, + "tokens/train_per_sec_per_gpu": 17.09, + "tokens/trainable": 302682 + }, + { + "epoch": 0.3163265306122449, + "grad_norm": 0.276568204164505, + "learning_rate": 9.579739217598608e-05, + "loss": 0.23821114003658295, + "memory/device_reserved (GiB)": 76.61, + "memory/max_active (GiB)": 75.07, + "memory/max_allocated (GiB)": 75.07, + "ppl": 1.26898, + "step": 155, + "tokens/total": 38262912, + "tokens/train_per_sec_per_gpu": 14.11, + "tokens/trainable": 304982 + }, + { + "epoch": 0.3183673469387755, + "grad_norm": 0.2808186411857605, + "learning_rate": 9.573085887607991e-05, + "loss": 0.29930347204208374, + "memory/device_reserved (GiB)": 76.61, + "memory/max_active (GiB)": 72.21, + "memory/max_allocated (GiB)": 72.21, + "ppl": 1.34892, + "step": 156, + "tokens/total": 38483264, + "tokens/train_per_sec_per_gpu": 16.27, + "tokens/trainable": 307115 + }, + { + "epoch": 0.32040816326530613, + "grad_norm": 0.25718843936920166, + "learning_rate": 9.566382652202013e-05, + "loss": 0.22872576117515564, + "memory/device_reserved (GiB)": 77.45, + "memory/max_active (GiB)": 76.56, + "memory/max_allocated (GiB)": 76.56, + "ppl": 1.257, + "step": 157, + "tokens/total": 38730240, + "tokens/train_per_sec_per_gpu": 11.61, + "tokens/trainable": 309080 + }, + { + "epoch": 0.3224489795918367, + "grad_norm": 0.3126402497291565, + "learning_rate": 9.559629584532099e-05, + "loss": 0.38535940647125244, + "memory/device_reserved (GiB)": 77.45, + "memory/max_active (GiB)": 72.48, + "memory/max_allocated (GiB)": 72.48, + "ppl": 1.47014, + "step": 158, + "tokens/total": 38973440, + "tokens/train_per_sec_per_gpu": 12.48, + "tokens/trainable": 311082 + }, + { + "epoch": 0.32448979591836735, + "grad_norm": 0.2776803970336914, + "learning_rate": 9.552826758293487e-05, + "loss": 0.27426379919052124, + "memory/device_reserved (GiB)": 77.45, + "memory/max_active (GiB)": 74.28, + "memory/max_allocated (GiB)": 74.28, + "ppl": 1.31556, + "step": 159, + "tokens/total": 39198784, + "tokens/train_per_sec_per_gpu": 11.78, + "tokens/trainable": 312908 + }, + { + "epoch": 0.32653061224489793, + "grad_norm": 0.2852776050567627, + "learning_rate": 9.54597424772442e-05, + "loss": 0.3446773588657379, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 76.64, + "memory/max_allocated (GiB)": 76.64, + "ppl": 1.41153, + "step": 160, + "tokens/total": 39459840, + "tokens/train_per_sec_per_gpu": 11.07, + "tokens/trainable": 314839 + }, + { + "epoch": 0.32857142857142857, + "grad_norm": 0.28382131457328796, + "learning_rate": 9.539072127605342e-05, + "loss": 0.3068111538887024, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 76.07, + "memory/max_allocated (GiB)": 76.07, + "ppl": 1.35908, + "step": 161, + "tokens/total": 39709184, + "tokens/train_per_sec_per_gpu": 11.66, + "tokens/trainable": 316710 + }, + { + "epoch": 0.3306122448979592, + "grad_norm": 0.3279019892215729, + "learning_rate": 9.532120473258075e-05, + "loss": 0.4070303440093994, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 72.92, + "memory/max_allocated (GiB)": 72.92, + "ppl": 1.50235, + "step": 162, + "tokens/total": 39959232, + "tokens/train_per_sec_per_gpu": 11.75, + "tokens/trainable": 318693 + }, + { + "epoch": 0.3326530612244898, + "grad_norm": 0.2982117235660553, + "learning_rate": 9.525119360545004e-05, + "loss": 0.34509584307670593, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 75.09, + "memory/max_allocated (GiB)": 75.09, + "ppl": 1.41213, + "step": 163, + "tokens/total": 40214080, + "tokens/train_per_sec_per_gpu": 11.81, + "tokens/trainable": 320608 + }, + { + "epoch": 0.3346938775510204, + "grad_norm": 0.25095808506011963, + "learning_rate": 9.518068865868245e-05, + "loss": 0.2849389314651489, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 75.67, + "memory/max_allocated (GiB)": 75.67, + "ppl": 1.32968, + "step": 164, + "tokens/total": 40433280, + "tokens/train_per_sec_per_gpu": 12.92, + "tokens/trainable": 322508 + }, + { + "epoch": 0.336734693877551, + "grad_norm": 0.24955125153064728, + "learning_rate": 9.510969066168813e-05, + "loss": 0.26943549513816833, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 76.75, + "memory/max_allocated (GiB)": 76.75, + "ppl": 1.30923, + "step": 165, + "tokens/total": 40674624, + "tokens/train_per_sec_per_gpu": 12.35, + "tokens/trainable": 324615 + }, + { + "epoch": 0.33877551020408164, + "grad_norm": 0.311704158782959, + "learning_rate": 9.503820038925782e-05, + "loss": 0.3359563946723938, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 76.81, + "memory/max_allocated (GiB)": 76.81, + "ppl": 1.39928, + "step": 166, + "tokens/total": 40929600, + "tokens/train_per_sec_per_gpu": 12.27, + "tokens/trainable": 326703 + }, + { + "epoch": 0.3408163265306122, + "grad_norm": 0.26562872529029846, + "learning_rate": 9.496621862155433e-05, + "loss": 0.33208560943603516, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 74.6, + "memory/max_allocated (GiB)": 74.6, + "ppl": 1.39387, + "step": 167, + "tokens/total": 41178112, + "tokens/train_per_sec_per_gpu": 12.48, + "tokens/trainable": 328707 + }, + { + "epoch": 0.34285714285714286, + "grad_norm": 0.3553466498851776, + "learning_rate": 9.489374614410413e-05, + "loss": 0.3184070587158203, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 73.08, + "memory/max_allocated (GiB)": 73.08, + "ppl": 1.37494, + "step": 168, + "tokens/total": 41413248, + "tokens/train_per_sec_per_gpu": 9.75, + "tokens/trainable": 330454 + }, + { + "epoch": 0.3448979591836735, + "grad_norm": 0.24358566105365753, + "learning_rate": 9.482078374778877e-05, + "loss": 0.2871302366256714, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 75.07, + "memory/max_allocated (GiB)": 75.07, + "ppl": 1.3326, + "step": 169, + "tokens/total": 41664768, + "tokens/train_per_sec_per_gpu": 11.83, + "tokens/trainable": 332291 + }, + { + "epoch": 0.3469387755102041, + "grad_norm": 0.24818101525306702, + "learning_rate": 9.474733222883618e-05, + "loss": 0.30461373925209045, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 74.58, + "memory/max_allocated (GiB)": 74.58, + "ppl": 1.3561, + "step": 170, + "tokens/total": 41912320, + "tokens/train_per_sec_per_gpu": 13.83, + "tokens/trainable": 334598 + }, + { + "epoch": 0.3489795918367347, + "grad_norm": 0.3133099377155304, + "learning_rate": 9.467339238881199e-05, + "loss": 0.27984052896499634, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 72.1, + "memory/max_allocated (GiB)": 72.1, + "ppl": 1.32292, + "step": 171, + "tokens/total": 42158912, + "tokens/train_per_sec_per_gpu": 10.95, + "tokens/trainable": 336300 + }, + { + "epoch": 0.3510204081632653, + "grad_norm": 0.2956582307815552, + "learning_rate": 9.459896503461081e-05, + "loss": 0.3749943971633911, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 76.15, + "memory/max_allocated (GiB)": 76.15, + "ppl": 1.45498, + "step": 172, + "tokens/total": 42403968, + "tokens/train_per_sec_per_gpu": 13.75, + "tokens/trainable": 338598 + }, + { + "epoch": 0.35306122448979593, + "grad_norm": 0.25042635202407837, + "learning_rate": 9.452405097844745e-05, + "loss": 0.32250380516052246, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 76.43, + "memory/max_allocated (GiB)": 76.43, + "ppl": 1.38058, + "step": 173, + "tokens/total": 42656192, + "tokens/train_per_sec_per_gpu": 12.65, + "tokens/trainable": 340730 + }, + { + "epoch": 0.3551020408163265, + "grad_norm": 0.3163611888885498, + "learning_rate": 9.444865103784804e-05, + "loss": 0.358521431684494, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 74.49, + "memory/max_allocated (GiB)": 74.49, + "ppl": 1.43121, + "step": 174, + "tokens/total": 42891776, + "tokens/train_per_sec_per_gpu": 13.19, + "tokens/trainable": 342662 + }, + { + "epoch": 0.35714285714285715, + "grad_norm": 0.2791230380535126, + "learning_rate": 9.437276603564103e-05, + "loss": 0.2755742073059082, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 75.17, + "memory/max_allocated (GiB)": 75.17, + "ppl": 1.31729, + "step": 175, + "tokens/total": 43108672, + "tokens/train_per_sec_per_gpu": 15.83, + "tokens/trainable": 344711 + }, + { + "epoch": 0.35918367346938773, + "grad_norm": 0.2673008441925049, + "learning_rate": 9.429639679994835e-05, + "loss": 0.2903071343898773, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 76.78, + "memory/max_allocated (GiB)": 76.78, + "ppl": 1.33684, + "step": 176, + "tokens/total": 43363648, + "tokens/train_per_sec_per_gpu": 7.56, + "tokens/trainable": 346404 + }, + { + "epoch": 0.36122448979591837, + "grad_norm": 0.255515456199646, + "learning_rate": 9.421954416417624e-05, + "loss": 0.29625675082206726, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 73.43, + "memory/max_allocated (GiB)": 73.43, + "ppl": 1.34482, + "step": 177, + "tokens/total": 43619328, + "tokens/train_per_sec_per_gpu": 10.84, + "tokens/trainable": 348353 + }, + { + "epoch": 0.363265306122449, + "grad_norm": 0.23868580162525177, + "learning_rate": 9.414220896700625e-05, + "loss": 0.243681401014328, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 75.96, + "memory/max_allocated (GiB)": 75.96, + "ppl": 1.27594, + "step": 178, + "tokens/total": 43864576, + "tokens/train_per_sec_per_gpu": 12.37, + "tokens/trainable": 350306 + }, + { + "epoch": 0.3653061224489796, + "grad_norm": 0.2819514870643616, + "learning_rate": 9.406439205238604e-05, + "loss": 0.2865571975708008, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 73.08, + "memory/max_allocated (GiB)": 73.08, + "ppl": 1.33183, + "step": 179, + "tokens/total": 44110464, + "tokens/train_per_sec_per_gpu": 14.06, + "tokens/trainable": 352775 + }, + { + "epoch": 0.3673469387755102, + "grad_norm": 0.379252165555954, + "learning_rate": 9.398609426952019e-05, + "loss": 0.2871363162994385, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 76.07, + "memory/max_allocated (GiB)": 76.07, + "ppl": 1.33261, + "step": 180, + "tokens/total": 44379840, + "tokens/train_per_sec_per_gpu": 10.57, + "tokens/trainable": 354683 + }, + { + "epoch": 0.3693877551020408, + "grad_norm": 0.28169211745262146, + "learning_rate": 9.390731647286089e-05, + "loss": 0.26943761110305786, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 73.32, + "memory/max_allocated (GiB)": 73.32, + "ppl": 1.30923, + "step": 181, + "tokens/total": 44580160, + "tokens/train_per_sec_per_gpu": 11.75, + "tokens/trainable": 356470 + }, + { + "epoch": 0.37142857142857144, + "grad_norm": 0.3467806279659271, + "learning_rate": 9.382805952209871e-05, + "loss": 0.28104543685913086, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 76.29, + "memory/max_allocated (GiB)": 76.29, + "ppl": 1.32451, + "step": 182, + "tokens/total": 44835008, + "tokens/train_per_sec_per_gpu": 12.22, + "tokens/trainable": 358392 + }, + { + "epoch": 0.373469387755102, + "grad_norm": 0.2689661383628845, + "learning_rate": 9.37483242821531e-05, + "loss": 0.27438846230506897, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 72.37, + "memory/max_allocated (GiB)": 72.37, + "ppl": 1.31573, + "step": 183, + "tokens/total": 45072064, + "tokens/train_per_sec_per_gpu": 12.5, + "tokens/trainable": 360336 + }, + { + "epoch": 0.37551020408163266, + "grad_norm": 0.30542707443237305, + "learning_rate": 9.366811162316306e-05, + "loss": 0.32957538962364197, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 73.02, + "memory/max_allocated (GiB)": 73.02, + "ppl": 1.39038, + "step": 184, + "tokens/total": 45318912, + "tokens/train_per_sec_per_gpu": 11.61, + "tokens/trainable": 362210 + }, + { + "epoch": 0.37755102040816324, + "grad_norm": 0.3197149336338043, + "learning_rate": 9.358742242047754e-05, + "loss": 0.31359681487083435, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 74.79, + "memory/max_allocated (GiB)": 74.79, + "ppl": 1.36834, + "step": 185, + "tokens/total": 45543488, + "tokens/train_per_sec_per_gpu": 13.12, + "tokens/trainable": 364203 + }, + { + "epoch": 0.3795918367346939, + "grad_norm": 0.28580036759376526, + "learning_rate": 9.3506257554646e-05, + "loss": 0.28920769691467285, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 71.94, + "memory/max_allocated (GiB)": 71.94, + "ppl": 1.33537, + "step": 186, + "tokens/total": 45772736, + "tokens/train_per_sec_per_gpu": 12.88, + "tokens/trainable": 366163 + }, + { + "epoch": 0.3816326530612245, + "grad_norm": 0.31438928842544556, + "learning_rate": 9.342461791140869e-05, + "loss": 0.2657240331172943, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 75.91, + "memory/max_allocated (GiB)": 75.91, + "ppl": 1.30438, + "step": 187, + "tokens/total": 46011712, + "tokens/train_per_sec_per_gpu": 10.7, + "tokens/trainable": 367799 + }, + { + "epoch": 0.3836734693877551, + "grad_norm": 0.30836033821105957, + "learning_rate": 9.334250438168708e-05, + "loss": 0.3199600577354431, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 75.23, + "memory/max_allocated (GiB)": 75.23, + "ppl": 1.37707, + "step": 188, + "tokens/total": 46228096, + "tokens/train_per_sec_per_gpu": 13.53, + "tokens/trainable": 369776 + }, + { + "epoch": 0.38571428571428573, + "grad_norm": 0.35285985469818115, + "learning_rate": 9.325991786157404e-05, + "loss": 0.31909671425819397, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 76.62, + "memory/max_allocated (GiB)": 76.62, + "ppl": 1.37588, + "step": 189, + "tokens/total": 46467904, + "tokens/train_per_sec_per_gpu": 12.24, + "tokens/trainable": 371863 + }, + { + "epoch": 0.3877551020408163, + "grad_norm": 0.241706982254982, + "learning_rate": 9.317685925232419e-05, + "loss": 0.2664746940135956, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 74.11, + "memory/max_allocated (GiB)": 74.11, + "ppl": 1.30535, + "step": 190, + "tokens/total": 46700224, + "tokens/train_per_sec_per_gpu": 15.81, + "tokens/trainable": 374102 + }, + { + "epoch": 0.38979591836734695, + "grad_norm": 0.2703701853752136, + "learning_rate": 9.309332946034392e-05, + "loss": 0.2703375220298767, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 74.98, + "memory/max_allocated (GiB)": 74.98, + "ppl": 1.31041, + "step": 191, + "tokens/total": 46950848, + "tokens/train_per_sec_per_gpu": 11.92, + "tokens/trainable": 375980 + }, + { + "epoch": 0.39183673469387753, + "grad_norm": 0.2964496612548828, + "learning_rate": 9.300932939718159e-05, + "loss": 0.31706342101097107, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 72.37, + "memory/max_allocated (GiB)": 72.37, + "ppl": 1.37309, + "step": 192, + "tokens/total": 47188160, + "tokens/train_per_sec_per_gpu": 12.62, + "tokens/trainable": 377889 + }, + { + "epoch": 0.39387755102040817, + "grad_norm": 0.29765358567237854, + "learning_rate": 9.292485997951759e-05, + "loss": 0.29366958141326904, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 76.59, + "memory/max_allocated (GiB)": 76.59, + "ppl": 1.34134, + "step": 193, + "tokens/total": 47436288, + "tokens/train_per_sec_per_gpu": 10.37, + "tokens/trainable": 379651 + }, + { + "epoch": 0.39591836734693875, + "grad_norm": 0.26283350586891174, + "learning_rate": 9.283992212915427e-05, + "loss": 0.3029991090297699, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 72.75, + "memory/max_allocated (GiB)": 72.75, + "ppl": 1.35391, + "step": 194, + "tokens/total": 47675776, + "tokens/train_per_sec_per_gpu": 12.82, + "tokens/trainable": 381553 + }, + { + "epoch": 0.3979591836734694, + "grad_norm": 0.25369948148727417, + "learning_rate": 9.275451677300591e-05, + "loss": 0.30793553590774536, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 75.09, + "memory/max_allocated (GiB)": 75.09, + "ppl": 1.36061, + "step": 195, + "tokens/total": 47912704, + "tokens/train_per_sec_per_gpu": 13.38, + "tokens/trainable": 383432 + }, + { + "epoch": 0.4, + "grad_norm": 0.2838038504123688, + "learning_rate": 9.266864484308866e-05, + "loss": 0.2943202257156372, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 76.78, + "memory/max_allocated (GiB)": 76.78, + "ppl": 1.34221, + "step": 196, + "tokens/total": 48137088, + "tokens/train_per_sec_per_gpu": 14.07, + "tokens/trainable": 385535 + }, + { + "epoch": 0.4020408163265306, + "grad_norm": 0.2468489110469818, + "learning_rate": 9.258230727651028e-05, + "loss": 0.25899794697761536, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 73.79, + "memory/max_allocated (GiB)": 73.79, + "ppl": 1.29563, + "step": 197, + "tokens/total": 48394240, + "tokens/train_per_sec_per_gpu": 15.65, + "tokens/trainable": 388061 + }, + { + "epoch": 0.40408163265306124, + "grad_norm": 0.24155408143997192, + "learning_rate": 9.249550501545997e-05, + "loss": 0.30748817324638367, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 75.5, + "memory/max_allocated (GiB)": 75.5, + "ppl": 1.36, + "step": 198, + "tokens/total": 48645184, + "tokens/train_per_sec_per_gpu": 13.67, + "tokens/trainable": 390121 + }, + { + "epoch": 0.4061224489795918, + "grad_norm": 0.2673620879650116, + "learning_rate": 9.240823900719807e-05, + "loss": 0.3003724217414856, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 72.81, + "memory/max_allocated (GiB)": 72.81, + "ppl": 1.35036, + "step": 199, + "tokens/total": 48895616, + "tokens/train_per_sec_per_gpu": 12.07, + "tokens/trainable": 391941 + }, + { + "epoch": 0.40816326530612246, + "grad_norm": 0.2946039140224457, + "learning_rate": 9.232051020404573e-05, + "loss": 0.2821658253669739, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 72.73, + "memory/max_allocated (GiB)": 72.73, + "ppl": 1.326, + "step": 200, + "tokens/total": 49133184, + "tokens/train_per_sec_per_gpu": 10.98, + "tokens/trainable": 393669 + }, + { + "epoch": 0.41020408163265304, + "grad_norm": 0.2829929292201996, + "learning_rate": 9.22323195633745e-05, + "loss": 0.3213472366333008, + "memory/device_reserved (GiB)": 77.88, + "memory/max_active (GiB)": 76.4, + "memory/max_allocated (GiB)": 76.4, + "ppl": 1.37898, + "step": 201, + "tokens/total": 49376000, + "tokens/train_per_sec_per_gpu": 10.63, + "tokens/trainable": 395917 + }, + { + "epoch": 0.4122448979591837, + "grad_norm": 0.37282848358154297, + "learning_rate": 9.214366804759588e-05, + "loss": 0.2857997417449951, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 75.91, + "memory/max_allocated (GiB)": 75.91, + "ppl": 1.33083, + "step": 202, + "tokens/total": 49614720, + "tokens/train_per_sec_per_gpu": 13.19, + "tokens/trainable": 398069 + }, + { + "epoch": 0.4142857142857143, + "grad_norm": 0.4424976706504822, + "learning_rate": 9.205455662415093e-05, + "loss": 0.3311164379119873, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 72.21, + "memory/max_allocated (GiB)": 72.21, + "ppl": 1.39252, + "step": 203, + "tokens/total": 49836480, + "tokens/train_per_sec_per_gpu": 17.48, + "tokens/trainable": 400171 + }, + { + "epoch": 0.4163265306122449, + "grad_norm": 0.2771094739437103, + "learning_rate": 9.196498626549945e-05, + "loss": 0.30752241611480713, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 74.47, + "memory/max_allocated (GiB)": 74.47, + "ppl": 1.36005, + "step": 204, + "tokens/total": 50109120, + "tokens/train_per_sec_per_gpu": 9.94, + "tokens/trainable": 401913 + }, + { + "epoch": 0.41836734693877553, + "grad_norm": 0.21018214523792267, + "learning_rate": 9.187495794910968e-05, + "loss": 0.24815361201763153, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 76.21, + "memory/max_allocated (GiB)": 76.21, + "ppl": 1.28166, + "step": 205, + "tokens/total": 50386240, + "tokens/train_per_sec_per_gpu": 11.55, + "tokens/trainable": 403860 + }, + { + "epoch": 0.4204081632653061, + "grad_norm": 0.31274956464767456, + "learning_rate": 9.17844726574474e-05, + "loss": 0.29557710886001587, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 72.32, + "memory/max_allocated (GiB)": 72.32, + "ppl": 1.3439, + "step": 206, + "tokens/total": 50616768, + "tokens/train_per_sec_per_gpu": 16.45, + "tokens/trainable": 406101 + }, + { + "epoch": 0.42244897959183675, + "grad_norm": 0.2219875305891037, + "learning_rate": 9.169353137796534e-05, + "loss": 0.27335110306739807, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 72.48, + "memory/max_allocated (GiB)": 72.48, + "ppl": 1.31436, + "step": 207, + "tokens/total": 50854272, + "tokens/train_per_sec_per_gpu": 12.29, + "tokens/trainable": 408195 + }, + { + "epoch": 0.42448979591836733, + "grad_norm": 0.275010347366333, + "learning_rate": 9.160213510309234e-05, + "loss": 0.26474326848983765, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 72.62, + "memory/max_allocated (GiB)": 72.62, + "ppl": 1.3031, + "step": 208, + "tokens/total": 51083392, + "tokens/train_per_sec_per_gpu": 12.87, + "tokens/trainable": 410139 + }, + { + "epoch": 0.42653061224489797, + "grad_norm": 0.2345055192708969, + "learning_rate": 9.151028483022253e-05, + "loss": 0.18677058815956116, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 71.96, + "memory/max_allocated (GiB)": 71.96, + "ppl": 1.20535, + "step": 209, + "tokens/total": 51315584, + "tokens/train_per_sec_per_gpu": 10.31, + "tokens/trainable": 411829 + }, + { + "epoch": 0.42857142857142855, + "grad_norm": 0.24785961210727692, + "learning_rate": 9.141798156170446e-05, + "loss": 0.27231812477111816, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 73.68, + "memory/max_allocated (GiB)": 73.68, + "ppl": 1.313, + "step": 210, + "tokens/total": 51565632, + "tokens/train_per_sec_per_gpu": 10.51, + "tokens/trainable": 413776 + }, + { + "epoch": 0.4306122448979592, + "grad_norm": 0.3085176646709442, + "learning_rate": 9.132522630483017e-05, + "loss": 0.33028310537338257, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 76.65, + "memory/max_allocated (GiB)": 76.65, + "ppl": 1.39136, + "step": 211, + "tokens/total": 51805120, + "tokens/train_per_sec_per_gpu": 10.32, + "tokens/trainable": 415533 + }, + { + "epoch": 0.4326530612244898, + "grad_norm": 0.24363301694393158, + "learning_rate": 9.123202007182416e-05, + "loss": 0.287580668926239, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 73.24, + "memory/max_allocated (GiB)": 73.24, + "ppl": 1.3332, + "step": 212, + "tokens/total": 52059776, + "tokens/train_per_sec_per_gpu": 12.47, + "tokens/trainable": 417532 + }, + { + "epoch": 0.4346938775510204, + "grad_norm": 0.3014145493507385, + "learning_rate": 9.113836387983239e-05, + "loss": 0.3328613340854645, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 73.84, + "memory/max_allocated (GiB)": 73.84, + "ppl": 1.39495, + "step": 213, + "tokens/total": 52304832, + "tokens/train_per_sec_per_gpu": 12.72, + "tokens/trainable": 419445 + }, + { + "epoch": 0.43673469387755104, + "grad_norm": 0.28564947843551636, + "learning_rate": 9.104425875091111e-05, + "loss": 0.3582199513912201, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 76.26, + "memory/max_allocated (GiB)": 76.26, + "ppl": 1.43078, + "step": 214, + "tokens/total": 52573184, + "tokens/train_per_sec_per_gpu": 13.0, + "tokens/trainable": 421735 + }, + { + "epoch": 0.4387755102040816, + "grad_norm": 0.24365663528442383, + "learning_rate": 9.094970571201582e-05, + "loss": 0.29590386152267456, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 73.92, + "memory/max_allocated (GiB)": 73.92, + "ppl": 1.34434, + "step": 215, + "tokens/total": 52812416, + "tokens/train_per_sec_per_gpu": 12.89, + "tokens/trainable": 423750 + }, + { + "epoch": 0.44081632653061226, + "grad_norm": 0.2833119034767151, + "learning_rate": 9.085470579498995e-05, + "loss": 0.35366660356521606, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 74.36, + "memory/max_allocated (GiB)": 74.36, + "ppl": 1.42428, + "step": 216, + "tokens/total": 53062336, + "tokens/train_per_sec_per_gpu": 14.44, + "tokens/trainable": 426047 + }, + { + "epoch": 0.44285714285714284, + "grad_norm": 0.25222450494766235, + "learning_rate": 9.075926003655367e-05, + "loss": 0.261421799659729, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 75.28, + "memory/max_allocated (GiB)": 75.28, + "ppl": 1.29878, + "step": 217, + "tokens/total": 53292160, + "tokens/train_per_sec_per_gpu": 14.18, + "tokens/trainable": 428071 + }, + { + "epoch": 0.4448979591836735, + "grad_norm": 0.22040905058383942, + "learning_rate": 9.066336947829251e-05, + "loss": 0.2836398482322693, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 76.05, + "memory/max_allocated (GiB)": 76.05, + "ppl": 1.32795, + "step": 218, + "tokens/total": 53548544, + "tokens/train_per_sec_per_gpu": 12.84, + "tokens/trainable": 430076 + }, + { + "epoch": 0.44693877551020406, + "grad_norm": 0.22903573513031006, + "learning_rate": 9.056703516664606e-05, + "loss": 0.2966764569282532, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 73.54, + "memory/max_allocated (GiB)": 73.54, + "ppl": 1.34538, + "step": 219, + "tokens/total": 53806848, + "tokens/train_per_sec_per_gpu": 15.43, + "tokens/trainable": 432426 + }, + { + "epoch": 0.4489795918367347, + "grad_norm": 0.277042955160141, + "learning_rate": 9.047025815289655e-05, + "loss": 0.2946584224700928, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 72.92, + "memory/max_allocated (GiB)": 72.92, + "ppl": 1.34267, + "step": 220, + "tokens/total": 54049600, + "tokens/train_per_sec_per_gpu": 13.97, + "tokens/trainable": 434720 + }, + { + "epoch": 0.45102040816326533, + "grad_norm": 0.23950393497943878, + "learning_rate": 9.037303949315736e-05, + "loss": 0.26722252368927, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 75.07, + "memory/max_allocated (GiB)": 75.07, + "ppl": 1.30633, + "step": 221, + "tokens/total": 54298176, + "tokens/train_per_sec_per_gpu": 13.58, + "tokens/trainable": 436747 + }, + { + "epoch": 0.4530612244897959, + "grad_norm": 0.27701491117477417, + "learning_rate": 9.027538024836142e-05, + "loss": 0.3390708863735199, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 75.5, + "memory/max_allocated (GiB)": 75.5, + "ppl": 1.40364, + "step": 222, + "tokens/total": 54545792, + "tokens/train_per_sec_per_gpu": 13.7, + "tokens/trainable": 438709 + }, + { + "epoch": 0.45510204081632655, + "grad_norm": 0.27305006980895996, + "learning_rate": 9.017728148424974e-05, + "loss": 0.2793228030204773, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 71.34, + "memory/max_allocated (GiB)": 71.34, + "ppl": 1.32223, + "step": 223, + "tokens/total": 54757952, + "tokens/train_per_sec_per_gpu": 12.92, + "tokens/trainable": 440507 + }, + { + "epoch": 0.45714285714285713, + "grad_norm": 0.3301241993904114, + "learning_rate": 9.007874427135977e-05, + "loss": 0.2435423582792282, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 75.86, + "memory/max_allocated (GiB)": 75.86, + "ppl": 1.27576, + "step": 224, + "tokens/total": 55005696, + "tokens/train_per_sec_per_gpu": 11.19, + "tokens/trainable": 442449 + }, + { + "epoch": 0.45918367346938777, + "grad_norm": 0.26555532217025757, + "learning_rate": 8.997976968501361e-05, + "loss": 0.28337737917900085, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 72.83, + "memory/max_allocated (GiB)": 72.83, + "ppl": 1.32761, + "step": 225, + "tokens/total": 55256896, + "tokens/train_per_sec_per_gpu": 13.91, + "tokens/trainable": 444834 + }, + { + "epoch": 0.46122448979591835, + "grad_norm": 0.24408915638923645, + "learning_rate": 8.988035880530646e-05, + "loss": 0.26556074619293213, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 75.53, + "memory/max_allocated (GiB)": 75.53, + "ppl": 1.30416, + "step": 226, + "tokens/total": 55483776, + "tokens/train_per_sec_per_gpu": 9.15, + "tokens/trainable": 446848 + }, + { + "epoch": 0.463265306122449, + "grad_norm": 0.29348307847976685, + "learning_rate": 8.978051271709461e-05, + "loss": 0.26864713430404663, + "memory/device_reserved (GiB)": 77.08, + "memory/max_active (GiB)": 76.35, + "memory/max_allocated (GiB)": 76.35, + "ppl": 1.30819, + "step": 227, + "tokens/total": 55743616, + "tokens/train_per_sec_per_gpu": 11.26, + "tokens/trainable": 448799 + }, + { + "epoch": 0.46530612244897956, + "grad_norm": 0.2565111219882965, + "learning_rate": 8.968023250998381e-05, + "loss": 0.2575717270374298, + "memory/device_reserved (GiB)": 77.08, + "memory/max_active (GiB)": 72.94, + "memory/max_allocated (GiB)": 72.94, + "ppl": 1.29378, + "step": 228, + "tokens/total": 55997760, + "tokens/train_per_sec_per_gpu": 10.46, + "tokens/trainable": 450576 + }, + { + "epoch": 0.4673469387755102, + "grad_norm": 0.2683160901069641, + "learning_rate": 8.957951927831722e-05, + "loss": 0.3322480320930481, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 76.78, + "memory/max_allocated (GiB)": 76.78, + "ppl": 1.3941, + "step": 229, + "tokens/total": 56248960, + "tokens/train_per_sec_per_gpu": 12.09, + "tokens/trainable": 452647 + }, + { + "epoch": 0.46938775510204084, + "grad_norm": 0.24362893402576447, + "learning_rate": 8.94783741211636e-05, + "loss": 0.2575136721134186, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 75.91, + "memory/max_allocated (GiB)": 75.91, + "ppl": 1.29371, + "step": 230, + "tokens/total": 56543872, + "tokens/train_per_sec_per_gpu": 10.94, + "tokens/trainable": 454763 + }, + { + "epoch": 0.4714285714285714, + "grad_norm": 0.2578815519809723, + "learning_rate": 8.937679814230517e-05, + "loss": 0.22576625645160675, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 73.22, + "memory/max_allocated (GiB)": 73.22, + "ppl": 1.25328, + "step": 231, + "tokens/total": 56784320, + "tokens/train_per_sec_per_gpu": 11.18, + "tokens/trainable": 456672 + }, + { + "epoch": 0.47346938775510206, + "grad_norm": 0.2743871212005615, + "learning_rate": 8.927479245022569e-05, + "loss": 0.2976492941379547, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 74.36, + "memory/max_allocated (GiB)": 74.36, + "ppl": 1.34669, + "step": 232, + "tokens/total": 57039808, + "tokens/train_per_sec_per_gpu": 11.2, + "tokens/trainable": 458524 + }, + { + "epoch": 0.47551020408163264, + "grad_norm": 0.26580774784088135, + "learning_rate": 8.917235815809835e-05, + "loss": 0.29766151309013367, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 74.25, + "memory/max_allocated (GiB)": 74.25, + "ppl": 1.34671, + "step": 233, + "tokens/total": 57304832, + "tokens/train_per_sec_per_gpu": 11.22, + "tokens/trainable": 460558 + }, + { + "epoch": 0.4775510204081633, + "grad_norm": 0.25239887833595276, + "learning_rate": 8.90694963837735e-05, + "loss": 0.29248756170272827, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 73.73, + "memory/max_allocated (GiB)": 73.73, + "ppl": 1.33976, + "step": 234, + "tokens/total": 57573312, + "tokens/train_per_sec_per_gpu": 12.46, + "tokens/trainable": 462527 + }, + { + "epoch": 0.47959183673469385, + "grad_norm": 0.2836940884590149, + "learning_rate": 8.896620824976665e-05, + "loss": 0.28166377544403076, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 72.4, + "memory/max_allocated (GiB)": 72.4, + "ppl": 1.32533, + "step": 235, + "tokens/total": 57816064, + "tokens/train_per_sec_per_gpu": 13.04, + "tokens/trainable": 464488 + }, + { + "epoch": 0.4816326530612245, + "grad_norm": 0.2632487118244171, + "learning_rate": 8.886249488324599e-05, + "loss": 0.2709941267967224, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 75.28, + "memory/max_allocated (GiB)": 75.28, + "ppl": 1.31127, + "step": 236, + "tokens/total": 58080256, + "tokens/train_per_sec_per_gpu": 10.53, + "tokens/trainable": 466331 + }, + { + "epoch": 0.48367346938775513, + "grad_norm": 0.22828401625156403, + "learning_rate": 8.87583574160203e-05, + "loss": 0.30610761046409607, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 76.02, + "memory/max_allocated (GiB)": 76.02, + "ppl": 1.35813, + "step": 237, + "tokens/total": 58346176, + "tokens/train_per_sec_per_gpu": 13.59, + "tokens/trainable": 468759 + }, + { + "epoch": 0.4857142857142857, + "grad_norm": 0.24178864061832428, + "learning_rate": 8.865379698452649e-05, + "loss": 0.2549661695957184, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 74.79, + "memory/max_allocated (GiB)": 74.79, + "ppl": 1.29042, + "step": 238, + "tokens/total": 58588672, + "tokens/train_per_sec_per_gpu": 12.62, + "tokens/trainable": 470776 + }, + { + "epoch": 0.48775510204081635, + "grad_norm": 0.4264017343521118, + "learning_rate": 8.854881472981721e-05, + "loss": 0.25378647446632385, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 74.66, + "memory/max_allocated (GiB)": 74.66, + "ppl": 1.2889, + "step": 239, + "tokens/total": 58832320, + "tokens/train_per_sec_per_gpu": 12.72, + "tokens/trainable": 472688 + }, + { + "epoch": 0.4897959183673469, + "grad_norm": 0.25247877836227417, + "learning_rate": 8.844341179754839e-05, + "loss": 0.27283287048339844, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 73.13, + "memory/max_allocated (GiB)": 73.13, + "ppl": 1.31368, + "step": 240, + "tokens/total": 59062400, + "tokens/train_per_sec_per_gpu": 15.46, + "tokens/trainable": 474748 + }, + { + "epoch": 0.49183673469387756, + "grad_norm": 0.2840617895126343, + "learning_rate": 8.833758933796678e-05, + "loss": 0.3166944682598114, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 73.92, + "memory/max_allocated (GiB)": 73.92, + "ppl": 1.37258, + "step": 241, + "tokens/total": 59304768, + "tokens/train_per_sec_per_gpu": 11.44, + "tokens/trainable": 476700 + }, + { + "epoch": 0.49387755102040815, + "grad_norm": 0.2712841033935547, + "learning_rate": 8.823134850589733e-05, + "loss": 0.2397720366716385, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 74.14, + "memory/max_allocated (GiB)": 74.14, + "ppl": 1.27096, + "step": 242, + "tokens/total": 59518720, + "tokens/train_per_sec_per_gpu": 14.09, + "tokens/trainable": 478590 + }, + { + "epoch": 0.4959183673469388, + "grad_norm": 0.2553960680961609, + "learning_rate": 8.812469046073069e-05, + "loss": 0.3182368278503418, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 72.04, + "memory/max_allocated (GiB)": 72.04, + "ppl": 1.3747, + "step": 243, + "tokens/total": 59745280, + "tokens/train_per_sec_per_gpu": 14.05, + "tokens/trainable": 480586 + }, + { + "epoch": 0.49795918367346936, + "grad_norm": 0.3267519176006317, + "learning_rate": 8.801761636641042e-05, + "loss": 0.30194833874702454, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 72.29, + "memory/max_allocated (GiB)": 72.29, + "ppl": 1.35249, + "step": 244, + "tokens/total": 59979520, + "tokens/train_per_sec_per_gpu": 11.58, + "tokens/trainable": 482413 + }, + { + "epoch": 0.5, + "grad_norm": 0.2530391216278076, + "learning_rate": 8.791012739142044e-05, + "loss": 0.2214522361755371, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 72.48, + "memory/max_allocated (GiB)": 72.48, + "ppl": 1.24789, + "step": 245, + "tokens/total": 60207040, + "tokens/train_per_sec_per_gpu": 14.89, + "tokens/trainable": 484421 + }, + { + "epoch": 0.5020408163265306, + "grad_norm": 0.2686348855495453, + "learning_rate": 8.780222470877214e-05, + "loss": 0.2083885818719864, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 73.05, + "memory/max_allocated (GiB)": 73.05, + "ppl": 1.23169, + "step": 246, + "tokens/total": 60420736, + "tokens/train_per_sec_per_gpu": 9.81, + "tokens/trainable": 486090 + }, + { + "epoch": 0.5040816326530613, + "grad_norm": 0.3877030313014984, + "learning_rate": 8.769390949599165e-05, + "loss": 0.2708570957183838, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 74.14, + "memory/max_allocated (GiB)": 74.14, + "ppl": 1.31109, + "step": 247, + "tokens/total": 60655424, + "tokens/train_per_sec_per_gpu": 13.27, + "tokens/trainable": 488011 + }, + { + "epoch": 0.5061224489795918, + "grad_norm": 0.2609875500202179, + "learning_rate": 8.758518293510704e-05, + "loss": 0.2860894501209259, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 74.22, + "memory/max_allocated (GiB)": 74.22, + "ppl": 1.33121, + "step": 248, + "tokens/total": 60923136, + "tokens/train_per_sec_per_gpu": 12.54, + "tokens/trainable": 490304 + }, + { + "epoch": 0.5081632653061224, + "grad_norm": 0.31066346168518066, + "learning_rate": 8.74760462126353e-05, + "loss": 0.26097726821899414, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 73.3, + "memory/max_allocated (GiB)": 73.3, + "ppl": 1.2982, + "step": 249, + "tokens/total": 61157312, + "tokens/train_per_sec_per_gpu": 12.28, + "tokens/trainable": 492072 + }, + { + "epoch": 0.5102040816326531, + "grad_norm": 0.25900799036026, + "learning_rate": 8.736650051956944e-05, + "loss": 0.27631238102912903, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 76.48, + "memory/max_allocated (GiB)": 76.48, + "ppl": 1.31826, + "step": 250, + "tokens/total": 61406464, + "tokens/train_per_sec_per_gpu": 12.73, + "tokens/trainable": 494145 + }, + { + "epoch": 0.5122448979591837, + "grad_norm": 0.29282519221305847, + "learning_rate": 8.725654705136558e-05, + "loss": 0.2838340699672699, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 74.52, + "memory/max_allocated (GiB)": 74.52, + "ppl": 1.32821, + "step": 251, + "tokens/total": 61671360, + "tokens/train_per_sec_per_gpu": 10.15, + "tokens/trainable": 496207 + }, + { + "epoch": 0.5142857142857142, + "grad_norm": 0.3111077845096588, + "learning_rate": 8.714618700792975e-05, + "loss": 0.2740340530872345, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 76.45, + "memory/max_allocated (GiB)": 76.45, + "ppl": 1.31526, + "step": 252, + "tokens/total": 61930368, + "tokens/train_per_sec_per_gpu": 12.53, + "tokens/trainable": 498397 + }, + { + "epoch": 0.5163265306122449, + "grad_norm": 0.32716354727745056, + "learning_rate": 8.703542159360495e-05, + "loss": 0.34461355209350586, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 75.09, + "memory/max_allocated (GiB)": 75.09, + "ppl": 1.41144, + "step": 253, + "tokens/total": 62178176, + "tokens/train_per_sec_per_gpu": 11.8, + "tokens/trainable": 500274 + }, + { + "epoch": 0.5183673469387755, + "grad_norm": 0.20006389915943146, + "learning_rate": 8.692425201715787e-05, + "loss": 0.21290145814418793, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 76.54, + "memory/max_allocated (GiB)": 76.54, + "ppl": 1.23726, + "step": 254, + "tokens/total": 62456000, + "tokens/train_per_sec_per_gpu": 9.72, + "tokens/trainable": 501999 + }, + { + "epoch": 0.5204081632653061, + "grad_norm": 0.23467357456684113, + "learning_rate": 8.681267949176579e-05, + "loss": 0.26032721996307373, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 75.67, + "memory/max_allocated (GiB)": 75.67, + "ppl": 1.29735, + "step": 255, + "tokens/total": 62711168, + "tokens/train_per_sec_per_gpu": 10.91, + "tokens/trainable": 503899 + }, + { + "epoch": 0.5224489795918368, + "grad_norm": 0.24094875156879425, + "learning_rate": 8.670070523500331e-05, + "loss": 0.2916109561920166, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 73.79, + "memory/max_allocated (GiB)": 73.79, + "ppl": 1.33858, + "step": 256, + "tokens/total": 62945856, + "tokens/train_per_sec_per_gpu": 13.03, + "tokens/trainable": 505937 + }, + { + "epoch": 0.5244897959183673, + "grad_norm": 0.23938961327075958, + "learning_rate": 8.658833046882908e-05, + "loss": 0.28174662590026855, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 74.03, + "memory/max_allocated (GiB)": 74.03, + "ppl": 1.32544, + "step": 257, + "tokens/total": 63178240, + "tokens/train_per_sec_per_gpu": 12.91, + "tokens/trainable": 507886 + }, + { + "epoch": 0.5265306122448979, + "grad_norm": 0.2992979884147644, + "learning_rate": 8.647555641957243e-05, + "loss": 0.3498281240463257, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 75.99, + "memory/max_allocated (GiB)": 75.99, + "ppl": 1.41882, + "step": 258, + "tokens/total": 63446144, + "tokens/train_per_sec_per_gpu": 12.78, + "tokens/trainable": 510118 + }, + { + "epoch": 0.5285714285714286, + "grad_norm": 0.2319977581501007, + "learning_rate": 8.636238431792001e-05, + "loss": 0.26939520239830017, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 71.74, + "memory/max_allocated (GiB)": 71.74, + "ppl": 1.30917, + "step": 259, + "tokens/total": 63660352, + "tokens/train_per_sec_per_gpu": 14.66, + "tokens/trainable": 511995 + }, + { + "epoch": 0.5306122448979592, + "grad_norm": 0.29739317297935486, + "learning_rate": 8.624881539890235e-05, + "loss": 0.26076313853263855, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 72.56, + "memory/max_allocated (GiB)": 72.56, + "ppl": 1.29792, + "step": 260, + "tokens/total": 63898944, + "tokens/train_per_sec_per_gpu": 13.62, + "tokens/trainable": 513979 + }, + { + "epoch": 0.5326530612244897, + "grad_norm": 0.3570258319377899, + "learning_rate": 8.613485090188042e-05, + "loss": 0.335449755191803, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 76.7, + "memory/max_allocated (GiB)": 76.7, + "ppl": 1.39857, + "step": 261, + "tokens/total": 64172160, + "tokens/train_per_sec_per_gpu": 11.16, + "tokens/trainable": 515977 + }, + { + "epoch": 0.5346938775510204, + "grad_norm": 0.227840393781662, + "learning_rate": 8.602049207053206e-05, + "loss": 0.27546989917755127, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 73.9, + "memory/max_allocated (GiB)": 73.9, + "ppl": 1.31715, + "step": 262, + "tokens/total": 64437632, + "tokens/train_per_sec_per_gpu": 11.8, + "tokens/trainable": 518051 + }, + { + "epoch": 0.536734693877551, + "grad_norm": 0.2699694335460663, + "learning_rate": 8.59057401528384e-05, + "loss": 0.3363790512084961, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 76.37, + "memory/max_allocated (GiB)": 76.37, + "ppl": 1.39987, + "step": 263, + "tokens/total": 64686784, + "tokens/train_per_sec_per_gpu": 10.55, + "tokens/trainable": 520106 + }, + { + "epoch": 0.5387755102040817, + "grad_norm": 0.26675179600715637, + "learning_rate": 8.57905964010703e-05, + "loss": 0.23343563079833984, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 73.87, + "memory/max_allocated (GiB)": 73.87, + "ppl": 1.26293, + "step": 264, + "tokens/total": 64922304, + "tokens/train_per_sec_per_gpu": 10.38, + "tokens/trainable": 521907 + }, + { + "epoch": 0.5408163265306123, + "grad_norm": 0.26130393147468567, + "learning_rate": 8.567506207177461e-05, + "loss": 0.30665674805641174, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 74.66, + "memory/max_allocated (GiB)": 74.66, + "ppl": 1.35887, + "step": 265, + "tokens/total": 65173888, + "tokens/train_per_sec_per_gpu": 12.68, + "tokens/trainable": 523953 + }, + { + "epoch": 0.5428571428571428, + "grad_norm": 0.27383264899253845, + "learning_rate": 8.555913842576053e-05, + "loss": 0.29130589962005615, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 75.28, + "memory/max_allocated (GiB)": 75.28, + "ppl": 1.33817, + "step": 266, + "tokens/total": 65411456, + "tokens/train_per_sec_per_gpu": 12.48, + "tokens/trainable": 525828 + }, + { + "epoch": 0.5448979591836735, + "grad_norm": 0.33510392904281616, + "learning_rate": 8.544282672808578e-05, + "loss": 0.29016339778900146, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 73.76, + "memory/max_allocated (GiB)": 73.76, + "ppl": 1.33665, + "step": 267, + "tokens/total": 65632384, + "tokens/train_per_sec_per_gpu": 14.4, + "tokens/trainable": 527801 + }, + { + "epoch": 0.5469387755102041, + "grad_norm": 0.2640371322631836, + "learning_rate": 8.53261282480429e-05, + "loss": 0.28861451148986816, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 75.34, + "memory/max_allocated (GiB)": 75.34, + "ppl": 1.33458, + "step": 268, + "tokens/total": 65862912, + "tokens/train_per_sec_per_gpu": 12.48, + "tokens/trainable": 529816 + }, + { + "epoch": 0.5489795918367347, + "grad_norm": 0.26248452067375183, + "learning_rate": 8.520904425914525e-05, + "loss": 0.3020923137664795, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 75.28, + "memory/max_allocated (GiB)": 75.28, + "ppl": 1.35269, + "step": 269, + "tokens/total": 66114816, + "tokens/train_per_sec_per_gpu": 11.84, + "tokens/trainable": 531847 + }, + { + "epoch": 0.5510204081632653, + "grad_norm": 0.28532981872558594, + "learning_rate": 8.50915760391132e-05, + "loss": 0.24453142285346985, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 71.96, + "memory/max_allocated (GiB)": 71.96, + "ppl": 1.27702, + "step": 270, + "tokens/total": 66343168, + "tokens/train_per_sec_per_gpu": 14.43, + "tokens/trainable": 533872 + }, + { + "epoch": 0.5530612244897959, + "grad_norm": 0.31190329790115356, + "learning_rate": 8.497372486986024e-05, + "loss": 0.42315733432769775, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 72.18, + "memory/max_allocated (GiB)": 72.18, + "ppl": 1.52677, + "step": 271, + "tokens/total": 66600576, + "tokens/train_per_sec_per_gpu": 13.46, + "tokens/trainable": 536013 + }, + { + "epoch": 0.5551020408163265, + "grad_norm": 0.26890113949775696, + "learning_rate": 8.485549203747886e-05, + "loss": 0.28477826714515686, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 75.07, + "memory/max_allocated (GiB)": 75.07, + "ppl": 1.32947, + "step": 272, + "tokens/total": 66844288, + "tokens/train_per_sec_per_gpu": 12.87, + "tokens/trainable": 538038 + }, + { + "epoch": 0.5571428571428572, + "grad_norm": 0.25644317269325256, + "learning_rate": 8.473687883222665e-05, + "loss": 0.2856728434562683, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 75.34, + "memory/max_allocated (GiB)": 75.34, + "ppl": 1.33066, + "step": 273, + "tokens/total": 67100736, + "tokens/train_per_sec_per_gpu": 12.41, + "tokens/trainable": 540087 + }, + { + "epoch": 0.5591836734693878, + "grad_norm": 0.22711892426013947, + "learning_rate": 8.461788654851209e-05, + "loss": 0.24410194158554077, + "memory/device_reserved (GiB)": 77.71, + "memory/max_active (GiB)": 76.67, + "memory/max_allocated (GiB)": 76.67, + "ppl": 1.27647, + "step": 274, + "tokens/total": 67362816, + "tokens/train_per_sec_per_gpu": 11.58, + "tokens/trainable": 542133 + }, + { + "epoch": 0.5612244897959183, + "grad_norm": 0.2506448030471802, + "learning_rate": 8.44985164848805e-05, + "loss": 0.25618839263916016, + "memory/device_reserved (GiB)": 77.71, + "memory/max_active (GiB)": 75.72, + "memory/max_allocated (GiB)": 75.72, + "ppl": 1.292, + "step": 275, + "tokens/total": 67593216, + "tokens/train_per_sec_per_gpu": 17.96, + "tokens/trainable": 544661 + }, + { + "epoch": 0.563265306122449, + "grad_norm": 0.24065186083316803, + "learning_rate": 8.437876994399991e-05, + "loss": 0.2728542685508728, + "memory/device_reserved (GiB)": 77.71, + "memory/max_active (GiB)": 75.8, + "memory/max_allocated (GiB)": 75.8, + "ppl": 1.31371, + "step": 276, + "tokens/total": 67841792, + "tokens/train_per_sec_per_gpu": 9.02, + "tokens/trainable": 546596 + }, + { + "epoch": 0.5653061224489796, + "grad_norm": 0.24970397353172302, + "learning_rate": 8.425864823264672e-05, + "loss": 0.24952834844589233, + "memory/device_reserved (GiB)": 76.63, + "memory/max_active (GiB)": 75.66, + "memory/max_allocated (GiB)": 75.66, + "ppl": 1.28342, + "step": 277, + "tokens/total": 68092864, + "tokens/train_per_sec_per_gpu": 11.34, + "tokens/trainable": 548572 + }, + { + "epoch": 0.5673469387755102, + "grad_norm": 0.25488054752349854, + "learning_rate": 8.413815266169158e-05, + "loss": 0.2789323925971985, + "memory/device_reserved (GiB)": 76.63, + "memory/max_active (GiB)": 72.72, + "memory/max_allocated (GiB)": 72.72, + "ppl": 1.32172, + "step": 278, + "tokens/total": 68328448, + "tokens/train_per_sec_per_gpu": 12.76, + "tokens/trainable": 550530 + }, + { + "epoch": 0.5693877551020409, + "grad_norm": 0.25996133685112, + "learning_rate": 8.401728454608494e-05, + "loss": 0.26806801557540894, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 76.73, + "memory/max_allocated (GiB)": 76.73, + "ppl": 1.30744, + "step": 279, + "tokens/total": 68577856, + "tokens/train_per_sec_per_gpu": 10.53, + "tokens/trainable": 552339 + }, + { + "epoch": 0.5714285714285714, + "grad_norm": 0.2861132323741913, + "learning_rate": 8.389604520484282e-05, + "loss": 0.2793717086315155, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 72.15, + "memory/max_allocated (GiB)": 72.15, + "ppl": 1.3223, + "step": 280, + "tokens/total": 68819072, + "tokens/train_per_sec_per_gpu": 14.12, + "tokens/trainable": 554395 + }, + { + "epoch": 0.573469387755102, + "grad_norm": 0.27016931772232056, + "learning_rate": 8.377443596103239e-05, + "loss": 0.3062211871147156, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 76.84, + "memory/max_allocated (GiB)": 76.84, + "ppl": 1.35828, + "step": 281, + "tokens/total": 69092800, + "tokens/train_per_sec_per_gpu": 11.69, + "tokens/trainable": 556475 + }, + { + "epoch": 0.5755102040816327, + "grad_norm": 0.3452536463737488, + "learning_rate": 8.365245814175743e-05, + "loss": 0.34729278087615967, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 75.64, + "memory/max_allocated (GiB)": 75.64, + "ppl": 1.41523, + "step": 282, + "tokens/total": 69357184, + "tokens/train_per_sec_per_gpu": 12.02, + "tokens/trainable": 558522 + }, + { + "epoch": 0.5775510204081633, + "grad_norm": 0.4534103274345398, + "learning_rate": 8.353011307814405e-05, + "loss": 0.2164735496044159, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 72.07, + "memory/max_allocated (GiB)": 72.07, + "ppl": 1.24169, + "step": 283, + "tokens/total": 69610112, + "tokens/train_per_sec_per_gpu": 12.31, + "tokens/trainable": 560719 + }, + { + "epoch": 0.5795918367346938, + "grad_norm": 0.2472052276134491, + "learning_rate": 8.340740210532595e-05, + "loss": 0.3178255259990692, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 73.9, + "memory/max_allocated (GiB)": 73.9, + "ppl": 1.37414, + "step": 284, + "tokens/total": 69852864, + "tokens/train_per_sec_per_gpu": 15.36, + "tokens/trainable": 562953 + }, + { + "epoch": 0.5816326530612245, + "grad_norm": 0.23412083089351654, + "learning_rate": 8.328432656242997e-05, + "loss": 0.27370530366897583, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 74.88, + "memory/max_allocated (GiB)": 74.88, + "ppl": 1.31483, + "step": 285, + "tokens/total": 70100224, + "tokens/train_per_sec_per_gpu": 13.31, + "tokens/trainable": 565039 + }, + { + "epoch": 0.5836734693877551, + "grad_norm": 0.28101804852485657, + "learning_rate": 8.316088779256145e-05, + "loss": 0.32362085580825806, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 76.32, + "memory/max_allocated (GiB)": 76.32, + "ppl": 1.38212, + "step": 286, + "tokens/total": 70355264, + "tokens/train_per_sec_per_gpu": 11.76, + "tokens/trainable": 567145 + }, + { + "epoch": 0.5857142857142857, + "grad_norm": 0.24828389286994934, + "learning_rate": 8.303708714278959e-05, + "loss": 0.2671997845172882, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 72.62, + "memory/max_allocated (GiB)": 72.62, + "ppl": 1.3063, + "step": 287, + "tokens/total": 70590976, + "tokens/train_per_sec_per_gpu": 13.82, + "tokens/trainable": 569069 + }, + { + "epoch": 0.5877551020408164, + "grad_norm": 0.2571806311607361, + "learning_rate": 8.291292596413272e-05, + "loss": 0.34625381231307983, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 73.32, + "memory/max_allocated (GiB)": 73.32, + "ppl": 1.41376, + "step": 288, + "tokens/total": 70839104, + "tokens/train_per_sec_per_gpu": 12.37, + "tokens/trainable": 571057 + }, + { + "epoch": 0.5897959183673469, + "grad_norm": 0.27506640553474426, + "learning_rate": 8.278840561154355e-05, + "loss": 0.32102882862091064, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 73.38, + "memory/max_allocated (GiB)": 73.38, + "ppl": 1.37855, + "step": 289, + "tokens/total": 71089472, + "tokens/train_per_sec_per_gpu": 12.42, + "tokens/trainable": 573199 + }, + { + "epoch": 0.5918367346938775, + "grad_norm": 0.22203350067138672, + "learning_rate": 8.26635274438944e-05, + "loss": 0.22201672196388245, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 75.07, + "memory/max_allocated (GiB)": 75.07, + "ppl": 1.24859, + "step": 290, + "tokens/total": 71380544, + "tokens/train_per_sec_per_gpu": 10.78, + "tokens/trainable": 575217 + }, + { + "epoch": 0.5938775510204082, + "grad_norm": 0.24041269719600677, + "learning_rate": 8.253829282396245e-05, + "loss": 0.310225248336792, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 72.94, + "memory/max_allocated (GiB)": 72.94, + "ppl": 1.36373, + "step": 291, + "tokens/total": 71612352, + "tokens/train_per_sec_per_gpu": 14.92, + "tokens/trainable": 577380 + }, + { + "epoch": 0.5959183673469388, + "grad_norm": 0.25376105308532715, + "learning_rate": 8.24127031184147e-05, + "loss": 0.28667837381362915, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 73.08, + "memory/max_allocated (GiB)": 73.08, + "ppl": 1.332, + "step": 292, + "tokens/total": 71846912, + "tokens/train_per_sec_per_gpu": 13.43, + "tokens/trainable": 579320 + }, + { + "epoch": 0.5979591836734693, + "grad_norm": 0.2470804899930954, + "learning_rate": 8.228675969779318e-05, + "loss": 0.2973708510398865, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 75.39, + "memory/max_allocated (GiB)": 75.39, + "ppl": 1.34631, + "step": 293, + "tokens/total": 72104512, + "tokens/train_per_sec_per_gpu": 10.7, + "tokens/trainable": 581313 + }, + { + "epoch": 0.6, + "grad_norm": 0.29593175649642944, + "learning_rate": 8.216046393649996e-05, + "loss": 0.34688469767570496, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 74.12, + "memory/max_allocated (GiB)": 74.12, + "ppl": 1.41465, + "step": 294, + "tokens/total": 72334464, + "tokens/train_per_sec_per_gpu": 11.91, + "tokens/trainable": 583198 + }, + { + "epoch": 0.6020408163265306, + "grad_norm": 0.24209007620811462, + "learning_rate": 8.203381721278216e-05, + "loss": 0.255124032497406, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 74.74, + "memory/max_allocated (GiB)": 74.74, + "ppl": 1.29062, + "step": 295, + "tokens/total": 72570240, + "tokens/train_per_sec_per_gpu": 11.22, + "tokens/trainable": 585096 + }, + { + "epoch": 0.6040816326530613, + "grad_norm": 0.251467764377594, + "learning_rate": 8.190682090871691e-05, + "loss": 0.24599887430667877, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 76.73, + "memory/max_allocated (GiB)": 76.73, + "ppl": 1.2789, + "step": 296, + "tokens/total": 72804864, + "tokens/train_per_sec_per_gpu": 12.38, + "tokens/trainable": 587097 + }, + { + "epoch": 0.6061224489795919, + "grad_norm": 0.23787440359592438, + "learning_rate": 8.177947641019621e-05, + "loss": 0.2537440061569214, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 74.11, + "memory/max_allocated (GiB)": 74.11, + "ppl": 1.28884, + "step": 297, + "tokens/total": 73031104, + "tokens/train_per_sec_per_gpu": 13.25, + "tokens/trainable": 589001 + }, + { + "epoch": 0.6081632653061224, + "grad_norm": 0.30218034982681274, + "learning_rate": 8.16517851069119e-05, + "loss": 0.3335407078266144, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 76.75, + "memory/max_allocated (GiB)": 76.75, + "ppl": 1.3959, + "step": 298, + "tokens/total": 73290304, + "tokens/train_per_sec_per_gpu": 10.66, + "tokens/trainable": 590853 + }, + { + "epoch": 0.610204081632653, + "grad_norm": 0.29110586643218994, + "learning_rate": 8.152374839234048e-05, + "loss": 0.2768510580062866, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 76.24, + "memory/max_allocated (GiB)": 76.24, + "ppl": 1.31897, + "step": 299, + "tokens/total": 73538624, + "tokens/train_per_sec_per_gpu": 11.47, + "tokens/trainable": 592628 + }, + { + "epoch": 0.6122448979591837, + "grad_norm": 0.2706233859062195, + "learning_rate": 8.139536766372775e-05, + "loss": 0.24794849753379822, + "memory/device_reserved (GiB)": 77.63, + "memory/max_active (GiB)": 74.25, + "memory/max_allocated (GiB)": 74.25, + "ppl": 1.28139, + "step": 300, + "tokens/total": 73763520, + "tokens/train_per_sec_per_gpu": 11.47, + "tokens/trainable": 594362 + }, + { + "epoch": 0.6142857142857143, + "grad_norm": 0.2760632634162903, + "learning_rate": 8.126664432207385e-05, + "loss": 0.344487726688385, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 76.83, + "memory/max_allocated (GiB)": 76.83, + "ppl": 1.41127, + "step": 301, + "tokens/total": 74021568, + "tokens/train_per_sec_per_gpu": 9.99, + "tokens/trainable": 596581 + }, + { + "epoch": 0.6163265306122448, + "grad_norm": 0.3173147141933441, + "learning_rate": 8.113757977211768e-05, + "loss": 0.32349497079849243, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 75.37, + "memory/max_allocated (GiB)": 75.37, + "ppl": 1.38195, + "step": 302, + "tokens/total": 74246528, + "tokens/train_per_sec_per_gpu": 10.36, + "tokens/trainable": 598390 + }, + { + "epoch": 0.6183673469387755, + "grad_norm": 0.3143511712551117, + "learning_rate": 8.100817542232174e-05, + "loss": 0.337619423866272, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 76.62, + "memory/max_allocated (GiB)": 76.62, + "ppl": 1.40161, + "step": 303, + "tokens/total": 74506048, + "tokens/train_per_sec_per_gpu": 9.99, + "tokens/trainable": 600327 + }, + { + "epoch": 0.6204081632653061, + "grad_norm": 0.3263002932071686, + "learning_rate": 8.087843268485671e-05, + "loss": 0.37967658042907715, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 72.45, + "memory/max_allocated (GiB)": 72.45, + "ppl": 1.46181, + "step": 304, + "tokens/total": 74714240, + "tokens/train_per_sec_per_gpu": 16.04, + "tokens/trainable": 602364 + }, + { + "epoch": 0.6224489795918368, + "grad_norm": 0.28157365322113037, + "learning_rate": 8.074835297558608e-05, + "loss": 0.2802862226963043, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 75.23, + "memory/max_allocated (GiB)": 75.23, + "ppl": 1.32351, + "step": 305, + "tokens/total": 74951936, + "tokens/train_per_sec_per_gpu": 10.7, + "tokens/trainable": 604134 + }, + { + "epoch": 0.6244897959183674, + "grad_norm": 0.3234773576259613, + "learning_rate": 8.061793771405059e-05, + "loss": 0.30056077241897583, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 76.26, + "memory/max_allocated (GiB)": 76.26, + "ppl": 1.35062, + "step": 306, + "tokens/total": 75190400, + "tokens/train_per_sec_per_gpu": 12.44, + "tokens/trainable": 606064 + }, + { + "epoch": 0.6265306122448979, + "grad_norm": 0.2672911584377289, + "learning_rate": 8.048718832345288e-05, + "loss": 0.3324597179889679, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 76.4, + "memory/max_allocated (GiB)": 76.4, + "ppl": 1.39439, + "step": 307, + "tokens/total": 75441728, + "tokens/train_per_sec_per_gpu": 11.32, + "tokens/trainable": 608000 + }, + { + "epoch": 0.6285714285714286, + "grad_norm": 0.23055435717105865, + "learning_rate": 8.035610623064187e-05, + "loss": 0.22188040614128113, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 73.6, + "memory/max_allocated (GiB)": 73.6, + "ppl": 1.24842, + "step": 308, + "tokens/total": 75671744, + "tokens/train_per_sec_per_gpu": 12.54, + "tokens/trainable": 609854 + }, + { + "epoch": 0.6306122448979592, + "grad_norm": 0.25100985169410706, + "learning_rate": 8.02246928660972e-05, + "loss": 0.25848251581192017, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 76.73, + "memory/max_allocated (GiB)": 76.73, + "ppl": 1.29496, + "step": 309, + "tokens/total": 75933632, + "tokens/train_per_sec_per_gpu": 13.0, + "tokens/trainable": 612055 + }, + { + "epoch": 0.6326530612244898, + "grad_norm": 0.28883853554725647, + "learning_rate": 8.009294966391363e-05, + "loss": 0.3101077079772949, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 74.06, + "memory/max_allocated (GiB)": 74.06, + "ppl": 1.36357, + "step": 310, + "tokens/total": 76196416, + "tokens/train_per_sec_per_gpu": 11.23, + "tokens/trainable": 614127 + }, + { + "epoch": 0.6346938775510204, + "grad_norm": 0.2648436427116394, + "learning_rate": 7.99608780617854e-05, + "loss": 0.32507240772247314, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 73.81, + "memory/max_allocated (GiB)": 73.81, + "ppl": 1.38413, + "step": 311, + "tokens/total": 76438400, + "tokens/train_per_sec_per_gpu": 14.56, + "tokens/trainable": 616331 + }, + { + "epoch": 0.636734693877551, + "grad_norm": 0.2828190326690674, + "learning_rate": 7.982847950099054e-05, + "loss": 0.29658567905426025, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 71.96, + "memory/max_allocated (GiB)": 71.96, + "ppl": 1.34526, + "step": 312, + "tokens/total": 76672064, + "tokens/train_per_sec_per_gpu": 14.47, + "tokens/trainable": 618435 + }, + { + "epoch": 0.6387755102040816, + "grad_norm": 0.2325240820646286, + "learning_rate": 7.969575542637513e-05, + "loss": 0.27081894874572754, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 75.26, + "memory/max_allocated (GiB)": 75.26, + "ppl": 1.31104, + "step": 313, + "tokens/total": 76922752, + "tokens/train_per_sec_per_gpu": 13.21, + "tokens/trainable": 620498 + }, + { + "epoch": 0.6408163265306123, + "grad_norm": 0.9377326369285583, + "learning_rate": 7.956270728633748e-05, + "loss": 0.2555592656135559, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 72.07, + "memory/max_allocated (GiB)": 72.07, + "ppl": 1.29118, + "step": 314, + "tokens/total": 77174912, + "tokens/train_per_sec_per_gpu": 12.63, + "tokens/trainable": 622434 + }, + { + "epoch": 0.6428571428571429, + "grad_norm": 0.2372015416622162, + "learning_rate": 7.942933653281244e-05, + "loss": 0.23461508750915527, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 72.75, + "memory/max_allocated (GiB)": 72.75, + "ppl": 1.26442, + "step": 315, + "tokens/total": 77409152, + "tokens/train_per_sec_per_gpu": 11.88, + "tokens/trainable": 624143 + }, + { + "epoch": 0.6448979591836734, + "grad_norm": 0.27768975496292114, + "learning_rate": 7.929564462125551e-05, + "loss": 0.36427009105682373, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 73.68, + "memory/max_allocated (GiB)": 73.68, + "ppl": 1.43946, + "step": 316, + "tokens/total": 77644992, + "tokens/train_per_sec_per_gpu": 12.13, + "tokens/trainable": 626058 + }, + { + "epoch": 0.6469387755102041, + "grad_norm": 0.25564926862716675, + "learning_rate": 7.916163301062686e-05, + "loss": 0.3089619576931, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 76.24, + "memory/max_allocated (GiB)": 76.24, + "ppl": 1.36201, + "step": 317, + "tokens/total": 77922304, + "tokens/train_per_sec_per_gpu": 13.11, + "tokens/trainable": 628212 + }, + { + "epoch": 0.6489795918367347, + "grad_norm": 0.2081686109304428, + "learning_rate": 7.902730316337556e-05, + "loss": 0.2701936960220337, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 74.25, + "memory/max_allocated (GiB)": 74.25, + "ppl": 1.31022, + "step": 318, + "tokens/total": 78199680, + "tokens/train_per_sec_per_gpu": 13.17, + "tokens/trainable": 630423 + }, + { + "epoch": 0.6510204081632653, + "grad_norm": 0.2430022954940796, + "learning_rate": 7.889265654542353e-05, + "loss": 0.23716434836387634, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 76.48, + "memory/max_allocated (GiB)": 76.48, + "ppl": 1.26765, + "step": 319, + "tokens/total": 78428992, + "tokens/train_per_sec_per_gpu": 14.25, + "tokens/trainable": 632427 + }, + { + "epoch": 0.6530612244897959, + "grad_norm": 0.22756606340408325, + "learning_rate": 7.875769462614955e-05, + "loss": 0.23425257205963135, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 75.26, + "memory/max_allocated (GiB)": 75.26, + "ppl": 1.26396, + "step": 320, + "tokens/total": 78681920, + "tokens/train_per_sec_per_gpu": 12.84, + "tokens/trainable": 634523 + }, + { + "epoch": 0.6551020408163265, + "grad_norm": 0.23997710645198822, + "learning_rate": 7.862241887837322e-05, + "loss": 0.22164437174797058, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 73.0, + "memory/max_allocated (GiB)": 73.0, + "ppl": 1.24813, + "step": 321, + "tokens/total": 78936576, + "tokens/train_per_sec_per_gpu": 11.23, + "tokens/trainable": 636593 + }, + { + "epoch": 0.6571428571428571, + "grad_norm": 0.34027695655822754, + "learning_rate": 7.848683077833897e-05, + "loss": 0.2563852071762085, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 76.86, + "memory/max_allocated (GiB)": 76.86, + "ppl": 1.29225, + "step": 322, + "tokens/total": 79219264, + "tokens/train_per_sec_per_gpu": 11.88, + "tokens/trainable": 638612 + }, + { + "epoch": 0.6591836734693878, + "grad_norm": 0.23914527893066406, + "learning_rate": 7.835093180569981e-05, + "loss": 0.2441192865371704, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 72.75, + "memory/max_allocated (GiB)": 72.75, + "ppl": 1.2765, + "step": 323, + "tokens/total": 79445632, + "tokens/train_per_sec_per_gpu": 11.93, + "tokens/trainable": 640327 + }, + { + "epoch": 0.6612244897959184, + "grad_norm": 0.32155025005340576, + "learning_rate": 7.821472344350132e-05, + "loss": 0.2721825838088989, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 73.49, + "memory/max_allocated (GiB)": 73.49, + "ppl": 1.31283, + "step": 324, + "tokens/total": 79679616, + "tokens/train_per_sec_per_gpu": 10.95, + "tokens/trainable": 642043 + }, + { + "epoch": 0.6632653061224489, + "grad_norm": 0.26682719588279724, + "learning_rate": 7.807820717816534e-05, + "loss": 0.31224024295806885, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 76.1, + "memory/max_allocated (GiB)": 76.1, + "ppl": 1.36648, + "step": 325, + "tokens/total": 79929792, + "tokens/train_per_sec_per_gpu": 12.86, + "tokens/trainable": 644073 + }, + { + "epoch": 0.6653061224489796, + "grad_norm": 0.2771010994911194, + "learning_rate": 7.794138449947388e-05, + "loss": 0.3241228759288788, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 72.1, + "memory/max_allocated (GiB)": 72.1, + "ppl": 1.38282, + "step": 326, + "tokens/total": 80175552, + "tokens/train_per_sec_per_gpu": 10.08, + "tokens/trainable": 646254 + }, + { + "epoch": 0.6673469387755102, + "grad_norm": 0.28892582654953003, + "learning_rate": 7.780425690055274e-05, + "loss": 0.29025834798812866, + "memory/device_reserved (GiB)": 74.58, + "memory/max_active (GiB)": 73.81, + "memory/max_allocated (GiB)": 73.81, + "ppl": 1.33677, + "step": 327, + "tokens/total": 80410432, + "tokens/train_per_sec_per_gpu": 12.1, + "tokens/trainable": 648069 + }, + { + "epoch": 0.6693877551020408, + "grad_norm": 0.21936146914958954, + "learning_rate": 7.766682587785529e-05, + "loss": 0.2003520280122757, + "memory/device_reserved (GiB)": 77.26, + "memory/max_active (GiB)": 76.54, + "memory/max_allocated (GiB)": 76.54, + "ppl": 1.22183, + "step": 328, + "tokens/total": 80694144, + "tokens/train_per_sec_per_gpu": 11.37, + "tokens/trainable": 650043 + }, + { + "epoch": 0.6714285714285714, + "grad_norm": 0.2513374388217926, + "learning_rate": 7.75290929311461e-05, + "loss": 0.3193785846233368, + "memory/device_reserved (GiB)": 77.26, + "memory/max_active (GiB)": 75.37, + "memory/max_allocated (GiB)": 75.37, + "ppl": 1.37627, + "step": 329, + "tokens/total": 80952192, + "tokens/train_per_sec_per_gpu": 13.07, + "tokens/trainable": 652165 + }, + { + "epoch": 0.673469387755102, + "grad_norm": 0.3239321708679199, + "learning_rate": 7.739105956348464e-05, + "loss": 0.2956297993659973, + "memory/device_reserved (GiB)": 77.26, + "memory/max_active (GiB)": 75.34, + "memory/max_allocated (GiB)": 75.34, + "ppl": 1.34397, + "step": 330, + "tokens/total": 81243200, + "tokens/train_per_sec_per_gpu": 10.37, + "tokens/trainable": 654102 + }, + { + "epoch": 0.6755102040816326, + "grad_norm": 0.2167237251996994, + "learning_rate": 7.72527272812088e-05, + "loss": 0.23567694425582886, + "memory/device_reserved (GiB)": 77.26, + "memory/max_active (GiB)": 73.99, + "memory/max_allocated (GiB)": 73.99, + "ppl": 1.26577, + "step": 331, + "tokens/total": 81468480, + "tokens/train_per_sec_per_gpu": 11.89, + "tokens/trainable": 655910 + }, + { + "epoch": 0.6775510204081633, + "grad_norm": 0.2764493227005005, + "learning_rate": 7.711409759391845e-05, + "loss": 0.27560240030288696, + "memory/device_reserved (GiB)": 77.26, + "memory/max_active (GiB)": 76.32, + "memory/max_allocated (GiB)": 76.32, + "ppl": 1.31732, + "step": 332, + "tokens/total": 81699968, + "tokens/train_per_sec_per_gpu": 14.39, + "tokens/trainable": 658045 + }, + { + "epoch": 0.6795918367346939, + "grad_norm": 0.23673021793365479, + "learning_rate": 7.697517201445905e-05, + "loss": 0.24529802799224854, + "memory/device_reserved (GiB)": 77.26, + "memory/max_active (GiB)": 72.67, + "memory/max_allocated (GiB)": 72.67, + "ppl": 1.278, + "step": 333, + "tokens/total": 81918400, + "tokens/train_per_sec_per_gpu": 14.81, + "tokens/trainable": 660055 + }, + { + "epoch": 0.6816326530612244, + "grad_norm": 0.2029729187488556, + "learning_rate": 7.683595205890508e-05, + "loss": 0.23150934278964996, + "memory/device_reserved (GiB)": 77.26, + "memory/max_active (GiB)": 73.16, + "memory/max_allocated (GiB)": 73.16, + "ppl": 1.2605, + "step": 334, + "tokens/total": 82173568, + "tokens/train_per_sec_per_gpu": 10.39, + "tokens/trainable": 661795 + }, + { + "epoch": 0.6836734693877551, + "grad_norm": 0.2765744626522064, + "learning_rate": 7.669643924654346e-05, + "loss": 0.31045764684677124, + "memory/device_reserved (GiB)": 77.26, + "memory/max_active (GiB)": 71.91, + "memory/max_allocated (GiB)": 71.91, + "ppl": 1.36405, + "step": 335, + "tokens/total": 82399936, + "tokens/train_per_sec_per_gpu": 13.76, + "tokens/trainable": 663787 + }, + { + "epoch": 0.6857142857142857, + "grad_norm": 0.2437237948179245, + "learning_rate": 7.655663509985707e-05, + "loss": 0.2277202606201172, + "memory/device_reserved (GiB)": 77.26, + "memory/max_active (GiB)": 76.07, + "memory/max_allocated (GiB)": 76.07, + "ppl": 1.25573, + "step": 336, + "tokens/total": 82645312, + "tokens/train_per_sec_per_gpu": 10.32, + "tokens/trainable": 665543 + }, + { + "epoch": 0.6877551020408164, + "grad_norm": 0.27684658765792847, + "learning_rate": 7.641654114450802e-05, + "loss": 0.3041114807128906, + "memory/device_reserved (GiB)": 77.26, + "memory/max_active (GiB)": 76.45, + "memory/max_allocated (GiB)": 76.45, + "ppl": 1.35542, + "step": 337, + "tokens/total": 82916736, + "tokens/train_per_sec_per_gpu": 11.5, + "tokens/trainable": 667685 + }, + { + "epoch": 0.689795918367347, + "grad_norm": 0.24407772719860077, + "learning_rate": 7.627615890932111e-05, + "loss": 0.33357661962509155, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 76.59, + "memory/max_allocated (GiB)": 76.59, + "ppl": 1.39595, + "step": 338, + "tokens/total": 83175104, + "tokens/train_per_sec_per_gpu": 12.66, + "tokens/trainable": 669843 + }, + { + "epoch": 0.6918367346938775, + "grad_norm": 0.20234188437461853, + "learning_rate": 7.61354899262671e-05, + "loss": 0.17819929122924805, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 76.07, + "memory/max_allocated (GiB)": 76.07, + "ppl": 1.19506, + "step": 339, + "tokens/total": 83405824, + "tokens/train_per_sec_per_gpu": 11.3, + "tokens/trainable": 671647 + }, + { + "epoch": 0.6938775510204082, + "grad_norm": 0.21054276823997498, + "learning_rate": 7.599453573044593e-05, + "loss": 0.1758253425359726, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 75.09, + "memory/max_allocated (GiB)": 75.09, + "ppl": 1.19223, + "step": 340, + "tokens/total": 83651008, + "tokens/train_per_sec_per_gpu": 10.96, + "tokens/trainable": 673387 + }, + { + "epoch": 0.6959183673469388, + "grad_norm": 0.2889537215232849, + "learning_rate": 7.585329786007006e-05, + "loss": 0.22698107361793518, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 73.46, + "memory/max_allocated (GiB)": 73.46, + "ppl": 1.25481, + "step": 341, + "tokens/total": 83890624, + "tokens/train_per_sec_per_gpu": 10.8, + "tokens/trainable": 675240 + }, + { + "epoch": 0.6979591836734694, + "grad_norm": 0.2577192485332489, + "learning_rate": 7.571177785644766e-05, + "loss": 0.2953447997570038, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 76.75, + "memory/max_allocated (GiB)": 76.75, + "ppl": 1.34359, + "step": 342, + "tokens/total": 84128384, + "tokens/train_per_sec_per_gpu": 10.72, + "tokens/trainable": 676963 + }, + { + "epoch": 0.7, + "grad_norm": 0.2311595231294632, + "learning_rate": 7.556997726396578e-05, + "loss": 0.20930227637290955, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 76.32, + "memory/max_allocated (GiB)": 76.32, + "ppl": 1.23282, + "step": 343, + "tokens/total": 84374400, + "tokens/train_per_sec_per_gpu": 14.01, + "tokens/trainable": 679155 + }, + { + "epoch": 0.7020408163265306, + "grad_norm": 0.22326231002807617, + "learning_rate": 7.542789763007346e-05, + "loss": 0.27322065830230713, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 76.15, + "memory/max_allocated (GiB)": 76.15, + "ppl": 1.31419, + "step": 344, + "tokens/total": 84625152, + "tokens/train_per_sec_per_gpu": 14.36, + "tokens/trainable": 681208 + }, + { + "epoch": 0.7040816326530612, + "grad_norm": 0.22791293263435364, + "learning_rate": 7.528554050526489e-05, + "loss": 0.234121635556221, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 73.9, + "memory/max_allocated (GiB)": 73.9, + "ppl": 1.2638, + "step": 345, + "tokens/total": 84869248, + "tokens/train_per_sec_per_gpu": 13.52, + "tokens/trainable": 683237 + }, + { + "epoch": 0.7061224489795919, + "grad_norm": 0.28647738695144653, + "learning_rate": 7.514290744306251e-05, + "loss": 0.36105626821517944, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 74.28, + "memory/max_allocated (GiB)": 74.28, + "ppl": 1.43484, + "step": 346, + "tokens/total": 85130112, + "tokens/train_per_sec_per_gpu": 12.65, + "tokens/trainable": 685215 + }, + { + "epoch": 0.7081632653061225, + "grad_norm": 0.2750461995601654, + "learning_rate": 7.500000000000001e-05, + "loss": 0.2888035476207733, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 76.02, + "memory/max_allocated (GiB)": 76.02, + "ppl": 1.33483, + "step": 347, + "tokens/total": 85351552, + "tokens/train_per_sec_per_gpu": 12.84, + "tokens/trainable": 687017 + }, + { + "epoch": 0.710204081632653, + "grad_norm": 0.3078104555606842, + "learning_rate": 7.485681973560533e-05, + "loss": 0.2835176885128021, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 76.54, + "memory/max_allocated (GiB)": 76.54, + "ppl": 1.32779, + "step": 348, + "tokens/total": 85602176, + "tokens/train_per_sec_per_gpu": 13.78, + "tokens/trainable": 689235 + }, + { + "epoch": 0.7122448979591837, + "grad_norm": 0.270408570766449, + "learning_rate": 7.471336821238367e-05, + "loss": 0.2432028353214264, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 74.03, + "memory/max_allocated (GiB)": 74.03, + "ppl": 1.27533, + "step": 349, + "tokens/total": 85864256, + "tokens/train_per_sec_per_gpu": 10.9, + "tokens/trainable": 691142 + }, + { + "epoch": 0.7142857142857143, + "grad_norm": 0.2373763918876648, + "learning_rate": 7.45696469958005e-05, + "loss": 0.27240806818008423, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 74.98, + "memory/max_allocated (GiB)": 74.98, + "ppl": 1.31312, + "step": 350, + "tokens/total": 86087552, + "tokens/train_per_sec_per_gpu": 13.51, + "tokens/trainable": 693021 + }, + { + "epoch": 0.7163265306122449, + "grad_norm": 0.23152503371238708, + "learning_rate": 7.442565765426436e-05, + "loss": 0.3414421081542969, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 76.16, + "memory/max_allocated (GiB)": 76.16, + "ppl": 1.40698, + "step": 351, + "tokens/total": 86346048, + "tokens/train_per_sec_per_gpu": 11.25, + "tokens/trainable": 695349 + }, + { + "epoch": 0.7183673469387755, + "grad_norm": 0.2574107348918915, + "learning_rate": 7.428140175910977e-05, + "loss": 0.3076707720756531, + "memory/device_reserved (GiB)": 76.94, + "memory/max_active (GiB)": 75.09, + "memory/max_allocated (GiB)": 75.09, + "ppl": 1.36025, + "step": 352, + "tokens/total": 86614336, + "tokens/train_per_sec_per_gpu": 13.1, + "tokens/trainable": 697531 + }, + { + "epoch": 0.7204081632653061, + "grad_norm": 0.22954194247722626, + "learning_rate": 7.413688088458021e-05, + "loss": 0.2560620903968811, + "memory/device_reserved (GiB)": 76.95, + "memory/max_active (GiB)": 74.66, + "memory/max_allocated (GiB)": 74.66, + "ppl": 1.29183, + "step": 353, + "tokens/total": 86862144, + "tokens/train_per_sec_per_gpu": 11.35, + "tokens/trainable": 699411 + }, + { + "epoch": 0.7224489795918367, + "grad_norm": 0.24674659967422485, + "learning_rate": 7.399209660781075e-05, + "loss": 0.2642272412776947, + "memory/device_reserved (GiB)": 77.43, + "memory/max_active (GiB)": 76.7, + "memory/max_allocated (GiB)": 76.7, + "ppl": 1.30242, + "step": 354, + "tokens/total": 87117248, + "tokens/train_per_sec_per_gpu": 14.62, + "tokens/trainable": 701912 + }, + { + "epoch": 0.7244897959183674, + "grad_norm": 0.329628586769104, + "learning_rate": 7.384705050881096e-05, + "loss": 0.27190959453582764, + "memory/device_reserved (GiB)": 77.43, + "memory/max_active (GiB)": 75.04, + "memory/max_allocated (GiB)": 75.04, + "ppl": 1.31247, + "step": 355, + "tokens/total": 87360320, + "tokens/train_per_sec_per_gpu": 12.52, + "tokens/trainable": 703850 + }, + { + "epoch": 0.726530612244898, + "grad_norm": 0.25129181146621704, + "learning_rate": 7.370174417044763e-05, + "loss": 0.2955136001110077, + "memory/device_reserved (GiB)": 77.43, + "memory/max_active (GiB)": 73.21, + "memory/max_allocated (GiB)": 73.21, + "ppl": 1.34382, + "step": 356, + "tokens/total": 87583680, + "tokens/train_per_sec_per_gpu": 12.57, + "tokens/trainable": 705651 + }, + { + "epoch": 0.7285714285714285, + "grad_norm": 0.360731840133667, + "learning_rate": 7.35561791784275e-05, + "loss": 0.3805921971797943, + "memory/device_reserved (GiB)": 77.43, + "memory/max_active (GiB)": 72.78, + "memory/max_allocated (GiB)": 72.78, + "ppl": 1.46315, + "step": 357, + "tokens/total": 87832832, + "tokens/train_per_sec_per_gpu": 14.27, + "tokens/trainable": 707880 + }, + { + "epoch": 0.7306122448979592, + "grad_norm": 0.22540250420570374, + "learning_rate": 7.341035712128002e-05, + "loss": 0.2772623300552368, + "memory/device_reserved (GiB)": 77.43, + "memory/max_active (GiB)": 73.35, + "memory/max_allocated (GiB)": 73.35, + "ppl": 1.31951, + "step": 358, + "tokens/total": 88082432, + "tokens/train_per_sec_per_gpu": 13.11, + "tokens/trainable": 709933 + }, + { + "epoch": 0.7326530612244898, + "grad_norm": 0.23541003465652466, + "learning_rate": 7.326427959033988e-05, + "loss": 0.29401713609695435, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 76.78, + "memory/max_allocated (GiB)": 76.78, + "ppl": 1.34181, + "step": 359, + "tokens/total": 88352000, + "tokens/train_per_sec_per_gpu": 12.54, + "tokens/trainable": 712011 + }, + { + "epoch": 0.7346938775510204, + "grad_norm": 0.24131914973258972, + "learning_rate": 7.311794817972974e-05, + "loss": 0.23980391025543213, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 75.77, + "memory/max_allocated (GiB)": 75.77, + "ppl": 1.271, + "step": 360, + "tokens/total": 88601984, + "tokens/train_per_sec_per_gpu": 11.34, + "tokens/trainable": 713965 + }, + { + "epoch": 0.736734693877551, + "grad_norm": 0.24607552587985992, + "learning_rate": 7.297136448634282e-05, + "loss": 0.24811792373657227, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 75.53, + "memory/max_allocated (GiB)": 75.53, + "ppl": 1.28161, + "step": 361, + "tokens/total": 88855104, + "tokens/train_per_sec_per_gpu": 11.24, + "tokens/trainable": 715899 + }, + { + "epoch": 0.7387755102040816, + "grad_norm": 0.23426595330238342, + "learning_rate": 7.282453010982545e-05, + "loss": 0.27236995100975037, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 74.49, + "memory/max_allocated (GiB)": 74.49, + "ppl": 1.31307, + "step": 362, + "tokens/total": 89091520, + "tokens/train_per_sec_per_gpu": 11.78, + "tokens/trainable": 717756 + }, + { + "epoch": 0.7408163265306122, + "grad_norm": 0.25521305203437805, + "learning_rate": 7.267744665255966e-05, + "loss": 0.32716047763824463, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 74.55, + "memory/max_allocated (GiB)": 74.55, + "ppl": 1.38702, + "step": 363, + "tokens/total": 89352128, + "tokens/train_per_sec_per_gpu": 13.56, + "tokens/trainable": 719848 + }, + { + "epoch": 0.7428571428571429, + "grad_norm": 0.33170413970947266, + "learning_rate": 7.253011571964562e-05, + "loss": 0.3414410948753357, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 75.37, + "memory/max_allocated (GiB)": 75.37, + "ppl": 1.40697, + "step": 364, + "tokens/total": 89605440, + "tokens/train_per_sec_per_gpu": 10.37, + "tokens/trainable": 721680 + }, + { + "epoch": 0.7448979591836735, + "grad_norm": 0.23171629011631012, + "learning_rate": 7.238253891888418e-05, + "loss": 0.2564217150211334, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 75.99, + "memory/max_allocated (GiB)": 75.99, + "ppl": 1.2923, + "step": 365, + "tokens/total": 89860416, + "tokens/train_per_sec_per_gpu": 11.98, + "tokens/trainable": 723603 + }, + { + "epoch": 0.746938775510204, + "grad_norm": 0.1833423674106598, + "learning_rate": 7.223471786075933e-05, + "loss": 0.19894462823867798, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 75.58, + "memory/max_allocated (GiB)": 75.58, + "ppl": 1.22011, + "step": 366, + "tokens/total": 90116416, + "tokens/train_per_sec_per_gpu": 11.57, + "tokens/trainable": 725611 + }, + { + "epoch": 0.7489795918367347, + "grad_norm": 0.25275909900665283, + "learning_rate": 7.20866541584206e-05, + "loss": 0.23714226484298706, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 75.86, + "memory/max_allocated (GiB)": 75.86, + "ppl": 1.26762, + "step": 367, + "tokens/total": 90371840, + "tokens/train_per_sec_per_gpu": 12.89, + "tokens/trainable": 727663 + }, + { + "epoch": 0.7510204081632653, + "grad_norm": 0.25915607810020447, + "learning_rate": 7.193834942766543e-05, + "loss": 0.2875545620918274, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 73.79, + "memory/max_allocated (GiB)": 73.79, + "ppl": 1.33316, + "step": 368, + "tokens/total": 90609344, + "tokens/train_per_sec_per_gpu": 13.83, + "tokens/trainable": 729759 + }, + { + "epoch": 0.753061224489796, + "grad_norm": 0.4674554765224457, + "learning_rate": 7.17898052869216e-05, + "loss": 0.2860286235809326, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 75.83, + "memory/max_allocated (GiB)": 75.83, + "ppl": 1.33113, + "step": 369, + "tokens/total": 90881984, + "tokens/train_per_sec_per_gpu": 13.99, + "tokens/trainable": 732003 + }, + { + "epoch": 0.7551020408163265, + "grad_norm": 0.34587299823760986, + "learning_rate": 7.164102335722954e-05, + "loss": 0.3153100907802582, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 72.86, + "memory/max_allocated (GiB)": 72.86, + "ppl": 1.37068, + "step": 370, + "tokens/total": 91105344, + "tokens/train_per_sec_per_gpu": 16.73, + "tokens/trainable": 733967 + }, + { + "epoch": 0.7571428571428571, + "grad_norm": 0.23913773894309998, + "learning_rate": 7.149200526222459e-05, + "loss": 0.3215009570121765, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 76.78, + "memory/max_allocated (GiB)": 76.78, + "ppl": 1.3792, + "step": 371, + "tokens/total": 91375040, + "tokens/train_per_sec_per_gpu": 11.87, + "tokens/trainable": 736037 + }, + { + "epoch": 0.7591836734693878, + "grad_norm": 0.23976799845695496, + "learning_rate": 7.134275262811934e-05, + "loss": 0.27008774876594543, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 73.41, + "memory/max_allocated (GiB)": 73.41, + "ppl": 1.31008, + "step": 372, + "tokens/total": 91616640, + "tokens/train_per_sec_per_gpu": 12.23, + "tokens/trainable": 737784 + }, + { + "epoch": 0.7612244897959184, + "grad_norm": 0.21629121899604797, + "learning_rate": 7.119326708368595e-05, + "loss": 0.2583259642124176, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 73.76, + "memory/max_allocated (GiB)": 73.76, + "ppl": 1.29476, + "step": 373, + "tokens/total": 91863488, + "tokens/train_per_sec_per_gpu": 11.93, + "tokens/trainable": 739736 + }, + { + "epoch": 0.763265306122449, + "grad_norm": 0.2703351378440857, + "learning_rate": 7.104355026023819e-05, + "loss": 0.32730811834335327, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 75.42, + "memory/max_allocated (GiB)": 75.42, + "ppl": 1.38723, + "step": 374, + "tokens/total": 92112128, + "tokens/train_per_sec_per_gpu": 13.04, + "tokens/trainable": 741745 + }, + { + "epoch": 0.7653061224489796, + "grad_norm": 0.25079062581062317, + "learning_rate": 7.089360379161381e-05, + "loss": 0.3121660351753235, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 75.94, + "memory/max_allocated (GiB)": 75.94, + "ppl": 1.36638, + "step": 375, + "tokens/total": 92368128, + "tokens/train_per_sec_per_gpu": 12.99, + "tokens/trainable": 744026 + }, + { + "epoch": 0.7673469387755102, + "grad_norm": 0.23077702522277832, + "learning_rate": 7.074342931415661e-05, + "loss": 0.30287420749664307, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 72.97, + "memory/max_allocated (GiB)": 72.97, + "ppl": 1.35374, + "step": 376, + "tokens/total": 92619264, + "tokens/train_per_sec_per_gpu": 10.19, + "tokens/trainable": 746081 + }, + { + "epoch": 0.7693877551020408, + "grad_norm": 0.22504407167434692, + "learning_rate": 7.059302846669864e-05, + "loss": 0.2365780472755432, + "memory/device_reserved (GiB)": 73.9, + "memory/max_active (GiB)": 72.92, + "memory/max_allocated (GiB)": 72.92, + "ppl": 1.26691, + "step": 377, + "tokens/total": 92870592, + "tokens/train_per_sec_per_gpu": 15.79, + "tokens/trainable": 748400 + }, + { + "epoch": 0.7714285714285715, + "grad_norm": 0.23904208838939667, + "learning_rate": 7.044240289054227e-05, + "loss": 0.255153089761734, + "memory/device_reserved (GiB)": 76.83, + "memory/max_active (GiB)": 75.88, + "memory/max_allocated (GiB)": 75.88, + "ppl": 1.29066, + "step": 378, + "tokens/total": 93114624, + "tokens/train_per_sec_per_gpu": 13.6, + "tokens/trainable": 750598 + }, + { + "epoch": 0.773469387755102, + "grad_norm": 0.2344440072774887, + "learning_rate": 7.029155422944234e-05, + "loss": 0.2139434516429901, + "memory/device_reserved (GiB)": 76.83, + "memory/max_active (GiB)": 74.06, + "memory/max_allocated (GiB)": 74.06, + "ppl": 1.23855, + "step": 379, + "tokens/total": 93329216, + "tokens/train_per_sec_per_gpu": 11.69, + "tokens/trainable": 752458 + }, + { + "epoch": 0.7755102040816326, + "grad_norm": 0.24536116421222687, + "learning_rate": 7.014048412958812e-05, + "loss": 0.32632678747177124, + "memory/device_reserved (GiB)": 76.83, + "memory/max_active (GiB)": 73.9, + "memory/max_allocated (GiB)": 73.9, + "ppl": 1.38587, + "step": 380, + "tokens/total": 93589504, + "tokens/train_per_sec_per_gpu": 11.37, + "tokens/trainable": 754662 + }, + { + "epoch": 0.7775510204081633, + "grad_norm": 0.2688855826854706, + "learning_rate": 6.998919423958547e-05, + "loss": 0.2734697461128235, + "memory/device_reserved (GiB)": 76.83, + "memory/max_active (GiB)": 75.34, + "memory/max_allocated (GiB)": 75.34, + "ppl": 1.31452, + "step": 381, + "tokens/total": 93851968, + "tokens/train_per_sec_per_gpu": 11.87, + "tokens/trainable": 756789 + }, + { + "epoch": 0.7795918367346939, + "grad_norm": 0.26867541670799255, + "learning_rate": 6.983768621043876e-05, + "loss": 0.26821669936180115, + "memory/device_reserved (GiB)": 76.83, + "memory/max_active (GiB)": 74.52, + "memory/max_allocated (GiB)": 74.52, + "ppl": 1.30763, + "step": 382, + "tokens/total": 94114048, + "tokens/train_per_sec_per_gpu": 9.4, + "tokens/trainable": 758492 + }, + { + "epoch": 0.7816326530612245, + "grad_norm": 0.23846036195755005, + "learning_rate": 6.968596169553288e-05, + "loss": 0.2921111583709717, + "memory/device_reserved (GiB)": 76.83, + "memory/max_active (GiB)": 72.89, + "memory/max_allocated (GiB)": 72.89, + "ppl": 1.33925, + "step": 383, + "tokens/total": 94382016, + "tokens/train_per_sec_per_gpu": 12.59, + "tokens/trainable": 760587 + }, + { + "epoch": 0.7836734693877551, + "grad_norm": 0.23699703812599182, + "learning_rate": 6.953402235061519e-05, + "loss": 0.2609871029853821, + "memory/device_reserved (GiB)": 76.83, + "memory/max_active (GiB)": 74.17, + "memory/max_allocated (GiB)": 74.17, + "ppl": 1.29821, + "step": 384, + "tokens/total": 94625792, + "tokens/train_per_sec_per_gpu": 14.53, + "tokens/trainable": 762765 + }, + { + "epoch": 0.7857142857142857, + "grad_norm": 0.20090660452842712, + "learning_rate": 6.938186983377751e-05, + "loss": 0.2041926383972168, + "memory/device_reserved (GiB)": 76.83, + "memory/max_active (GiB)": 75.07, + "memory/max_allocated (GiB)": 75.07, + "ppl": 1.22653, + "step": 385, + "tokens/total": 94897408, + "tokens/train_per_sec_per_gpu": 10.61, + "tokens/trainable": 764513 + }, + { + "epoch": 0.7877551020408163, + "grad_norm": 0.2212802767753601, + "learning_rate": 6.922950580543788e-05, + "loss": 0.25139063596725464, + "memory/device_reserved (GiB)": 76.83, + "memory/max_active (GiB)": 73.13, + "memory/max_allocated (GiB)": 73.13, + "ppl": 1.28581, + "step": 386, + "tokens/total": 95117952, + "tokens/train_per_sec_per_gpu": 14.28, + "tokens/trainable": 766648 + }, + { + "epoch": 0.789795918367347, + "grad_norm": 0.23277297616004944, + "learning_rate": 6.907693192832263e-05, + "loss": 0.19786489009857178, + "memory/device_reserved (GiB)": 76.83, + "memory/max_active (GiB)": 74.2, + "memory/max_allocated (GiB)": 74.2, + "ppl": 1.2188, + "step": 387, + "tokens/total": 95365248, + "tokens/train_per_sec_per_gpu": 10.45, + "tokens/trainable": 768352 + }, + { + "epoch": 0.7918367346938775, + "grad_norm": 0.33522191643714905, + "learning_rate": 6.892414986744807e-05, + "loss": 0.3165251612663269, + "memory/device_reserved (GiB)": 76.83, + "memory/max_active (GiB)": 75.2, + "memory/max_allocated (GiB)": 75.2, + "ppl": 1.37235, + "step": 388, + "tokens/total": 95611520, + "tokens/train_per_sec_per_gpu": 15.0, + "tokens/trainable": 770596 + }, + { + "epoch": 0.7938775510204081, + "grad_norm": 0.25158795714378357, + "learning_rate": 6.877116129010243e-05, + "loss": 0.2680057883262634, + "memory/device_reserved (GiB)": 76.83, + "memory/max_active (GiB)": 73.3, + "memory/max_allocated (GiB)": 73.3, + "ppl": 1.30735, + "step": 389, + "tokens/total": 95874944, + "tokens/train_per_sec_per_gpu": 12.92, + "tokens/trainable": 772629 + }, + { + "epoch": 0.7959183673469388, + "grad_norm": 0.2529192566871643, + "learning_rate": 6.861796786582761e-05, + "loss": 0.28054893016815186, + "memory/device_reserved (GiB)": 76.83, + "memory/max_active (GiB)": 74.2, + "memory/max_allocated (GiB)": 74.2, + "ppl": 1.32386, + "step": 390, + "tokens/total": 96126528, + "tokens/train_per_sec_per_gpu": 12.81, + "tokens/trainable": 774667 + }, + { + "epoch": 0.7979591836734694, + "grad_norm": 0.24691379070281982, + "learning_rate": 6.846457126640096e-05, + "loss": 0.29811322689056396, + "memory/device_reserved (GiB)": 76.83, + "memory/max_active (GiB)": 74.17, + "memory/max_allocated (GiB)": 74.17, + "ppl": 1.34731, + "step": 391, + "tokens/total": 96382912, + "tokens/train_per_sec_per_gpu": 13.04, + "tokens/trainable": 776828 + }, + { + "epoch": 0.8, + "grad_norm": 0.25443676114082336, + "learning_rate": 6.831097316581707e-05, + "loss": 0.2967149615287781, + "memory/device_reserved (GiB)": 76.83, + "memory/max_active (GiB)": 75.47, + "memory/max_allocated (GiB)": 75.47, + "ppl": 1.34543, + "step": 392, + "tokens/total": 96638336, + "tokens/train_per_sec_per_gpu": 13.25, + "tokens/trainable": 779052 + }, + { + "epoch": 0.8020408163265306, + "grad_norm": 0.22823578119277954, + "learning_rate": 6.815717524026948e-05, + "loss": 0.2128680795431137, + "memory/device_reserved (GiB)": 76.83, + "memory/max_active (GiB)": 74.88, + "memory/max_allocated (GiB)": 74.88, + "ppl": 1.23722, + "step": 393, + "tokens/total": 96879104, + "tokens/train_per_sec_per_gpu": 12.58, + "tokens/trainable": 780951 + }, + { + "epoch": 0.8040816326530612, + "grad_norm": 0.23700584471225739, + "learning_rate": 6.80031791681324e-05, + "loss": 0.25681912899017334, + "memory/device_reserved (GiB)": 77.31, + "memory/max_active (GiB)": 76.4, + "memory/max_allocated (GiB)": 76.4, + "ppl": 1.29281, + "step": 394, + "tokens/total": 97160256, + "tokens/train_per_sec_per_gpu": 9.91, + "tokens/trainable": 782833 + }, + { + "epoch": 0.8061224489795918, + "grad_norm": 0.3007517158985138, + "learning_rate": 6.784898662994234e-05, + "loss": 0.23515839874744415, + "memory/device_reserved (GiB)": 77.31, + "memory/max_active (GiB)": 75.39, + "memory/max_allocated (GiB)": 75.39, + "ppl": 1.26511, + "step": 395, + "tokens/total": 97384512, + "tokens/train_per_sec_per_gpu": 11.88, + "tokens/trainable": 784600 + }, + { + "epoch": 0.8081632653061225, + "grad_norm": 0.2689831554889679, + "learning_rate": 6.769459930837989e-05, + "loss": 0.24911612272262573, + "memory/device_reserved (GiB)": 77.31, + "memory/max_active (GiB)": 75.28, + "memory/max_allocated (GiB)": 75.28, + "ppl": 1.28289, + "step": 396, + "tokens/total": 97632064, + "tokens/train_per_sec_per_gpu": 12.46, + "tokens/trainable": 786600 + }, + { + "epoch": 0.810204081632653, + "grad_norm": 0.30507197976112366, + "learning_rate": 6.75400188882512e-05, + "loss": 0.33416375517845154, + "memory/device_reserved (GiB)": 77.31, + "memory/max_active (GiB)": 72.62, + "memory/max_allocated (GiB)": 72.62, + "ppl": 1.39677, + "step": 397, + "tokens/total": 97885824, + "tokens/train_per_sec_per_gpu": 14.24, + "tokens/trainable": 789091 + }, + { + "epoch": 0.8122448979591836, + "grad_norm": 0.25920626521110535, + "learning_rate": 6.738524705646973e-05, + "loss": 0.3228701949119568, + "memory/device_reserved (GiB)": 77.31, + "memory/max_active (GiB)": 75.91, + "memory/max_allocated (GiB)": 75.91, + "ppl": 1.38109, + "step": 398, + "tokens/total": 98126592, + "tokens/train_per_sec_per_gpu": 12.64, + "tokens/trainable": 791039 + }, + { + "epoch": 0.8142857142857143, + "grad_norm": 0.21600422263145447, + "learning_rate": 6.723028550203779e-05, + "loss": 0.2245880663394928, + "memory/device_reserved (GiB)": 77.31, + "memory/max_active (GiB)": 73.16, + "memory/max_allocated (GiB)": 73.16, + "ppl": 1.25181, + "step": 399, + "tokens/total": 98339136, + "tokens/train_per_sec_per_gpu": 14.12, + "tokens/trainable": 792943 + }, + { + "epoch": 0.8163265306122449, + "grad_norm": 0.21295800805091858, + "learning_rate": 6.707513591602804e-05, + "loss": 0.23633459210395813, + "memory/device_reserved (GiB)": 77.31, + "memory/max_active (GiB)": 75.34, + "memory/max_allocated (GiB)": 75.34, + "ppl": 1.2666, + "step": 400, + "tokens/total": 98615616, + "tokens/train_per_sec_per_gpu": 9.75, + "tokens/trainable": 794687 + }, + { + "epoch": 0.8183673469387756, + "grad_norm": 0.2668573558330536, + "learning_rate": 6.691979999156521e-05, + "loss": 0.3033875524997711, + "memory/device_reserved (GiB)": 77.72, + "memory/max_active (GiB)": 76.59, + "memory/max_allocated (GiB)": 76.59, + "ppl": 1.35444, + "step": 401, + "tokens/total": 98879424, + "tokens/train_per_sec_per_gpu": 9.23, + "tokens/trainable": 796784 + }, + { + "epoch": 0.8204081632653061, + "grad_norm": 0.2711150646209717, + "learning_rate": 6.67642794238074e-05, + "loss": 0.2761994004249573, + "memory/device_reserved (GiB)": 77.72, + "memory/max_active (GiB)": 73.0, + "memory/max_allocated (GiB)": 73.0, + "ppl": 1.31811, + "step": 402, + "tokens/total": 99115776, + "tokens/train_per_sec_per_gpu": 12.37, + "tokens/trainable": 798648 + }, + { + "epoch": 0.8224489795918367, + "grad_norm": 0.20724070072174072, + "learning_rate": 6.66085759099278e-05, + "loss": 0.21089161932468414, + "memory/device_reserved (GiB)": 77.72, + "memory/max_active (GiB)": 74.39, + "memory/max_allocated (GiB)": 74.39, + "ppl": 1.23478, + "step": 403, + "tokens/total": 99375104, + "tokens/train_per_sec_per_gpu": 11.54, + "tokens/trainable": 800651 + }, + { + "epoch": 0.8244897959183674, + "grad_norm": 0.277954638004303, + "learning_rate": 6.6452691149096e-05, + "loss": 0.2901919484138489, + "memory/device_reserved (GiB)": 77.72, + "memory/max_active (GiB)": 74.06, + "memory/max_allocated (GiB)": 74.06, + "ppl": 1.33668, + "step": 404, + "tokens/total": 99622784, + "tokens/train_per_sec_per_gpu": 12.07, + "tokens/trainable": 802578 + }, + { + "epoch": 0.826530612244898, + "grad_norm": 0.24895675480365753, + "learning_rate": 6.629662684245948e-05, + "loss": 0.2913890779018402, + "memory/device_reserved (GiB)": 77.72, + "memory/max_active (GiB)": 72.34, + "memory/max_allocated (GiB)": 72.34, + "ppl": 1.33829, + "step": 405, + "tokens/total": 99868480, + "tokens/train_per_sec_per_gpu": 12.05, + "tokens/trainable": 804481 + }, + { + "epoch": 0.8285714285714286, + "grad_norm": 0.23595218360424042, + "learning_rate": 6.614038469312517e-05, + "loss": 0.24025043845176697, + "memory/device_reserved (GiB)": 77.72, + "memory/max_active (GiB)": 74.0, + "memory/max_allocated (GiB)": 74.0, + "ppl": 1.27157, + "step": 406, + "tokens/total": 100088768, + "tokens/train_per_sec_per_gpu": 14.89, + "tokens/trainable": 806348 + }, + { + "epoch": 0.8306122448979592, + "grad_norm": 0.24337661266326904, + "learning_rate": 6.598396640614072e-05, + "loss": 0.2879195213317871, + "memory/device_reserved (GiB)": 77.72, + "memory/max_active (GiB)": 75.23, + "memory/max_allocated (GiB)": 75.23, + "ppl": 1.33365, + "step": 407, + "tokens/total": 100327296, + "tokens/train_per_sec_per_gpu": 15.09, + "tokens/trainable": 808502 + }, + { + "epoch": 0.8326530612244898, + "grad_norm": 0.2543904185295105, + "learning_rate": 6.582737368847592e-05, + "loss": 0.30715101957321167, + "memory/device_reserved (GiB)": 77.76, + "memory/max_active (GiB)": 76.67, + "memory/max_allocated (GiB)": 76.67, + "ppl": 1.35955, + "step": 408, + "tokens/total": 100589696, + "tokens/train_per_sec_per_gpu": 12.37, + "tokens/trainable": 810671 + }, + { + "epoch": 0.8346938775510204, + "grad_norm": 0.22101710736751556, + "learning_rate": 6.567060824900414e-05, + "loss": 0.2784162163734436, + "memory/device_reserved (GiB)": 77.76, + "memory/max_active (GiB)": 76.18, + "memory/max_allocated (GiB)": 76.18, + "ppl": 1.32104, + "step": 409, + "tokens/total": 100849600, + "tokens/train_per_sec_per_gpu": 12.66, + "tokens/trainable": 812785 + }, + { + "epoch": 0.8367346938775511, + "grad_norm": 0.2810465693473816, + "learning_rate": 6.551367179848362e-05, + "loss": 0.270579993724823, + "memory/device_reserved (GiB)": 77.76, + "memory/max_active (GiB)": 76.4, + "memory/max_allocated (GiB)": 76.4, + "ppl": 1.31072, + "step": 410, + "tokens/total": 101087360, + "tokens/train_per_sec_per_gpu": 11.83, + "tokens/trainable": 814717 + }, + { + "epoch": 0.8387755102040816, + "grad_norm": 0.21870507299900055, + "learning_rate": 6.535656604953884e-05, + "loss": 0.21983875334262848, + "memory/device_reserved (GiB)": 77.76, + "memory/max_active (GiB)": 73.65, + "memory/max_allocated (GiB)": 73.65, + "ppl": 1.24588, + "step": 411, + "tokens/total": 101313792, + "tokens/train_per_sec_per_gpu": 13.29, + "tokens/trainable": 816693 + }, + { + "epoch": 0.8408163265306122, + "grad_norm": 0.2483329474925995, + "learning_rate": 6.519929271664175e-05, + "loss": 0.28504157066345215, + "memory/device_reserved (GiB)": 77.76, + "memory/max_active (GiB)": 73.3, + "memory/max_allocated (GiB)": 73.3, + "ppl": 1.32982, + "step": 412, + "tokens/total": 101544064, + "tokens/train_per_sec_per_gpu": 12.46, + "tokens/trainable": 818684 + }, + { + "epoch": 0.8428571428571429, + "grad_norm": 0.2841825783252716, + "learning_rate": 6.504185351609317e-05, + "loss": 0.2541833817958832, + "memory/device_reserved (GiB)": 77.76, + "memory/max_active (GiB)": 76.45, + "memory/max_allocated (GiB)": 76.45, + "ppl": 1.28941, + "step": 413, + "tokens/total": 101759040, + "tokens/train_per_sec_per_gpu": 14.52, + "tokens/trainable": 820751 + }, + { + "epoch": 0.8448979591836735, + "grad_norm": 0.2572484612464905, + "learning_rate": 6.488425016600403e-05, + "loss": 0.21007731556892395, + "memory/device_reserved (GiB)": 77.76, + "memory/max_active (GiB)": 73.81, + "memory/max_allocated (GiB)": 73.81, + "ppl": 1.23377, + "step": 414, + "tokens/total": 101995840, + "tokens/train_per_sec_per_gpu": 12.4, + "tokens/trainable": 822616 + }, + { + "epoch": 0.8469387755102041, + "grad_norm": 0.24220719933509827, + "learning_rate": 6.472648438627652e-05, + "loss": 0.28219708800315857, + "memory/device_reserved (GiB)": 77.76, + "memory/max_active (GiB)": 74.88, + "memory/max_allocated (GiB)": 74.88, + "ppl": 1.32604, + "step": 415, + "tokens/total": 102253888, + "tokens/train_per_sec_per_gpu": 15.21, + "tokens/trainable": 825258 + }, + { + "epoch": 0.8489795918367347, + "grad_norm": 0.28009212017059326, + "learning_rate": 6.456855789858548e-05, + "loss": 0.3138611316680908, + "memory/device_reserved (GiB)": 77.76, + "memory/max_active (GiB)": 74.96, + "memory/max_allocated (GiB)": 74.96, + "ppl": 1.3687, + "step": 416, + "tokens/total": 102513728, + "tokens/train_per_sec_per_gpu": 13.18, + "tokens/trainable": 827403 + }, + { + "epoch": 0.8510204081632653, + "grad_norm": 0.2708995044231415, + "learning_rate": 6.441047242635947e-05, + "loss": 0.33378952741622925, + "memory/device_reserved (GiB)": 77.76, + "memory/max_active (GiB)": 74.25, + "memory/max_allocated (GiB)": 74.25, + "ppl": 1.39625, + "step": 417, + "tokens/total": 102750528, + "tokens/train_per_sec_per_gpu": 14.69, + "tokens/trainable": 829464 + }, + { + "epoch": 0.8530612244897959, + "grad_norm": 0.31186723709106445, + "learning_rate": 6.42522296947621e-05, + "loss": 0.31511160731315613, + "memory/device_reserved (GiB)": 77.76, + "memory/max_active (GiB)": 76.7, + "memory/max_allocated (GiB)": 76.7, + "ppl": 1.37041, + "step": 418, + "tokens/total": 102980736, + "tokens/train_per_sec_per_gpu": 10.03, + "tokens/trainable": 831148 + }, + { + "epoch": 0.8551020408163266, + "grad_norm": 0.2542271912097931, + "learning_rate": 6.409383143067305e-05, + "loss": 0.27176880836486816, + "memory/device_reserved (GiB)": 77.76, + "memory/max_active (GiB)": 75.58, + "memory/max_allocated (GiB)": 75.58, + "ppl": 1.31228, + "step": 419, + "tokens/total": 103202304, + "tokens/train_per_sec_per_gpu": 12.37, + "tokens/trainable": 832859 + }, + { + "epoch": 0.8571428571428571, + "grad_norm": 0.29588770866394043, + "learning_rate": 6.393527936266933e-05, + "loss": 0.24880217015743256, + "memory/device_reserved (GiB)": 77.76, + "memory/max_active (GiB)": 73.08, + "memory/max_allocated (GiB)": 73.08, + "ppl": 1.28249, + "step": 420, + "tokens/total": 103432896, + "tokens/train_per_sec_per_gpu": 15.78, + "tokens/trainable": 835125 + }, + { + "epoch": 0.8591836734693877, + "grad_norm": 0.2471047341823578, + "learning_rate": 6.377657522100644e-05, + "loss": 0.2640013098716736, + "memory/device_reserved (GiB)": 77.76, + "memory/max_active (GiB)": 72.67, + "memory/max_allocated (GiB)": 72.67, + "ppl": 1.30213, + "step": 421, + "tokens/total": 103687424, + "tokens/train_per_sec_per_gpu": 12.8, + "tokens/trainable": 837239 + }, + { + "epoch": 0.8612244897959184, + "grad_norm": 0.21796868741512299, + "learning_rate": 6.361772073759935e-05, + "loss": 0.247511088848114, + "memory/device_reserved (GiB)": 77.76, + "memory/max_active (GiB)": 74.17, + "memory/max_allocated (GiB)": 74.17, + "ppl": 1.28083, + "step": 422, + "tokens/total": 103933248, + "tokens/train_per_sec_per_gpu": 13.07, + "tokens/trainable": 839147 + }, + { + "epoch": 0.863265306122449, + "grad_norm": 0.23844701051712036, + "learning_rate": 6.345871764600374e-05, + "loss": 0.23742735385894775, + "memory/device_reserved (GiB)": 77.76, + "memory/max_active (GiB)": 74.68, + "memory/max_allocated (GiB)": 74.68, + "ppl": 1.26798, + "step": 423, + "tokens/total": 104186688, + "tokens/train_per_sec_per_gpu": 11.85, + "tokens/trainable": 841045 + }, + { + "epoch": 0.8653061224489796, + "grad_norm": 0.23335368931293488, + "learning_rate": 6.329956768139704e-05, + "loss": 0.2869156301021576, + "memory/device_reserved (GiB)": 77.76, + "memory/max_active (GiB)": 72.51, + "memory/max_allocated (GiB)": 72.51, + "ppl": 1.33231, + "step": 424, + "tokens/total": 104447872, + "tokens/train_per_sec_per_gpu": 12.35, + "tokens/trainable": 843215 + }, + { + "epoch": 0.8673469387755102, + "grad_norm": 0.2918064296245575, + "learning_rate": 6.314027258055944e-05, + "loss": 0.2789589762687683, + "memory/device_reserved (GiB)": 77.76, + "memory/max_active (GiB)": 72.21, + "memory/max_allocated (GiB)": 72.21, + "ppl": 1.32175, + "step": 425, + "tokens/total": 104664384, + "tokens/train_per_sec_per_gpu": 11.08, + "tokens/trainable": 845002 + }, + { + "epoch": 0.8693877551020408, + "grad_norm": 0.2757396399974823, + "learning_rate": 6.298083408185503e-05, + "loss": 0.3008022904396057, + "memory/device_reserved (GiB)": 77.76, + "memory/max_active (GiB)": 76.07, + "memory/max_allocated (GiB)": 76.07, + "ppl": 1.35094, + "step": 426, + "tokens/total": 104935104, + "tokens/train_per_sec_per_gpu": 9.66, + "tokens/trainable": 847171 + }, + { + "epoch": 0.8714285714285714, + "grad_norm": 0.28692859411239624, + "learning_rate": 6.282125392521273e-05, + "loss": 0.22909744083881378, + "memory/device_reserved (GiB)": 77.39, + "memory/max_active (GiB)": 76.67, + "memory/max_allocated (GiB)": 76.67, + "ppl": 1.25746, + "step": 427, + "tokens/total": 105177792, + "tokens/train_per_sec_per_gpu": 10.97, + "tokens/trainable": 849003 + }, + { + "epoch": 0.8734693877551021, + "grad_norm": 0.24682961404323578, + "learning_rate": 6.266153385210739e-05, + "loss": 0.24241462349891663, + "memory/device_reserved (GiB)": 77.39, + "memory/max_active (GiB)": 74.6, + "memory/max_allocated (GiB)": 74.6, + "ppl": 1.27432, + "step": 428, + "tokens/total": 105405248, + "tokens/train_per_sec_per_gpu": 11.32, + "tokens/trainable": 850792 + }, + { + "epoch": 0.8755102040816326, + "grad_norm": 0.3659954071044922, + "learning_rate": 6.250167560554075e-05, + "loss": 0.33554619550704956, + "memory/device_reserved (GiB)": 77.39, + "memory/max_active (GiB)": 72.45, + "memory/max_allocated (GiB)": 72.45, + "ppl": 1.3987, + "step": 429, + "tokens/total": 105636864, + "tokens/train_per_sec_per_gpu": 10.26, + "tokens/trainable": 852551 + }, + { + "epoch": 0.8775510204081632, + "grad_norm": 0.26704132556915283, + "learning_rate": 6.234168093002237e-05, + "loss": 0.31336140632629395, + "memory/device_reserved (GiB)": 77.39, + "memory/max_active (GiB)": 76.51, + "memory/max_allocated (GiB)": 76.51, + "ppl": 1.36802, + "step": 430, + "tokens/total": 105907776, + "tokens/train_per_sec_per_gpu": 12.05, + "tokens/trainable": 854721 + }, + { + "epoch": 0.8795918367346939, + "grad_norm": 0.2755028307437897, + "learning_rate": 6.21815515715507e-05, + "loss": 0.2989242672920227, + "memory/device_reserved (GiB)": 77.39, + "memory/max_active (GiB)": 74.96, + "memory/max_allocated (GiB)": 74.96, + "ppl": 1.34841, + "step": 431, + "tokens/total": 106149824, + "tokens/train_per_sec_per_gpu": 12.76, + "tokens/trainable": 856854 + }, + { + "epoch": 0.8816326530612245, + "grad_norm": 0.2054336965084076, + "learning_rate": 6.202128927759391e-05, + "loss": 0.24352321028709412, + "memory/device_reserved (GiB)": 77.39, + "memory/max_active (GiB)": 75.8, + "memory/max_allocated (GiB)": 75.8, + "ppl": 1.27574, + "step": 432, + "tokens/total": 106407488, + "tokens/train_per_sec_per_gpu": 9.75, + "tokens/trainable": 858527 + }, + { + "epoch": 0.8836734693877552, + "grad_norm": 0.24481087923049927, + "learning_rate": 6.18608957970709e-05, + "loss": 0.3159305453300476, + "memory/device_reserved (GiB)": 77.39, + "memory/max_active (GiB)": 71.94, + "memory/max_allocated (GiB)": 71.94, + "ppl": 1.37153, + "step": 433, + "tokens/total": 106646720, + "tokens/train_per_sec_per_gpu": 12.72, + "tokens/trainable": 860570 + }, + { + "epoch": 0.8857142857142857, + "grad_norm": 0.2475660741329193, + "learning_rate": 6.170037288033222e-05, + "loss": 0.28183165192604065, + "memory/device_reserved (GiB)": 77.39, + "memory/max_active (GiB)": 73.54, + "memory/max_allocated (GiB)": 73.54, + "ppl": 1.32556, + "step": 434, + "tokens/total": 106896960, + "tokens/train_per_sec_per_gpu": 11.7, + "tokens/trainable": 862447 + }, + { + "epoch": 0.8877551020408163, + "grad_norm": 0.26032233238220215, + "learning_rate": 6.153972227914088e-05, + "loss": 0.2791120409965515, + "memory/device_reserved (GiB)": 77.39, + "memory/max_active (GiB)": 73.79, + "memory/max_allocated (GiB)": 73.79, + "ppl": 1.32196, + "step": 435, + "tokens/total": 107168704, + "tokens/train_per_sec_per_gpu": 11.48, + "tokens/trainable": 864540 + }, + { + "epoch": 0.889795918367347, + "grad_norm": 0.26088958978652954, + "learning_rate": 6.137894574665336e-05, + "loss": 0.25506502389907837, + "memory/device_reserved (GiB)": 77.39, + "memory/max_active (GiB)": 72.91, + "memory/max_allocated (GiB)": 72.91, + "ppl": 1.29055, + "step": 436, + "tokens/total": 107392960, + "tokens/train_per_sec_per_gpu": 13.84, + "tokens/trainable": 866462 + }, + { + "epoch": 0.8918367346938776, + "grad_norm": 0.20378683507442474, + "learning_rate": 6.121804503740035e-05, + "loss": 0.20120058953762054, + "memory/device_reserved (GiB)": 77.39, + "memory/max_active (GiB)": 73.43, + "memory/max_allocated (GiB)": 73.43, + "ppl": 1.22287, + "step": 437, + "tokens/total": 107613312, + "tokens/train_per_sec_per_gpu": 12.03, + "tokens/trainable": 868200 + }, + { + "epoch": 0.8938775510204081, + "grad_norm": 1.791853427886963, + "learning_rate": 6.105702190726765e-05, + "loss": 0.297176718711853, + "memory/device_reserved (GiB)": 77.39, + "memory/max_active (GiB)": 74.66, + "memory/max_allocated (GiB)": 74.66, + "ppl": 1.34605, + "step": 438, + "tokens/total": 107881664, + "tokens/train_per_sec_per_gpu": 11.4, + "tokens/trainable": 870401 + }, + { + "epoch": 0.8959183673469387, + "grad_norm": 0.2114609181880951, + "learning_rate": 6.08958781134771e-05, + "loss": 0.217289537191391, + "memory/device_reserved (GiB)": 77.39, + "memory/max_active (GiB)": 76.4, + "memory/max_allocated (GiB)": 76.4, + "ppl": 1.2427, + "step": 439, + "tokens/total": 108135488, + "tokens/train_per_sec_per_gpu": 11.16, + "tokens/trainable": 872309 + }, + { + "epoch": 0.8979591836734694, + "grad_norm": 0.2895108461380005, + "learning_rate": 6.073461541456725e-05, + "loss": 0.2577347755432129, + "memory/device_reserved (GiB)": 77.39, + "memory/max_active (GiB)": 75.72, + "memory/max_allocated (GiB)": 75.72, + "ppl": 1.294, + "step": 440, + "tokens/total": 108375296, + "tokens/train_per_sec_per_gpu": 11.21, + "tokens/trainable": 874206 + }, + { + "epoch": 0.9, + "grad_norm": 0.23809455335140228, + "learning_rate": 6.057323557037431e-05, + "loss": 0.2702704071998596, + "memory/device_reserved (GiB)": 77.39, + "memory/max_active (GiB)": 71.88, + "memory/max_allocated (GiB)": 71.88, + "ppl": 1.31032, + "step": 441, + "tokens/total": 108621888, + "tokens/train_per_sec_per_gpu": 11.1, + "tokens/trainable": 876148 + }, + { + "epoch": 0.9020408163265307, + "grad_norm": 0.23602895438671112, + "learning_rate": 6.04117403420128e-05, + "loss": 0.32741737365722656, + "memory/device_reserved (GiB)": 77.39, + "memory/max_active (GiB)": 75.61, + "memory/max_allocated (GiB)": 75.61, + "ppl": 1.38738, + "step": 442, + "tokens/total": 108898944, + "tokens/train_per_sec_per_gpu": 12.19, + "tokens/trainable": 878248 + }, + { + "epoch": 0.9040816326530612, + "grad_norm": 0.2687036693096161, + "learning_rate": 6.02501314918565e-05, + "loss": 0.29402047395706177, + "memory/device_reserved (GiB)": 77.39, + "memory/max_active (GiB)": 73.41, + "memory/max_allocated (GiB)": 73.41, + "ppl": 1.34181, + "step": 443, + "tokens/total": 109123072, + "tokens/train_per_sec_per_gpu": 13.93, + "tokens/trainable": 880136 + }, + { + "epoch": 0.9061224489795918, + "grad_norm": 0.3213645815849304, + "learning_rate": 6.0088410783519035e-05, + "loss": 0.2375052273273468, + "memory/device_reserved (GiB)": 77.39, + "memory/max_active (GiB)": 72.97, + "memory/max_allocated (GiB)": 72.97, + "ppl": 1.26808, + "step": 444, + "tokens/total": 109347392, + "tokens/train_per_sec_per_gpu": 9.85, + "tokens/trainable": 881758 + }, + { + "epoch": 0.9081632653061225, + "grad_norm": 0.22513405978679657, + "learning_rate": 5.99265799818348e-05, + "loss": 0.19214271008968353, + "memory/device_reserved (GiB)": 77.39, + "memory/max_active (GiB)": 72.59, + "memory/max_allocated (GiB)": 72.59, + "ppl": 1.21184, + "step": 445, + "tokens/total": 109593152, + "tokens/train_per_sec_per_gpu": 9.9, + "tokens/trainable": 883475 + }, + { + "epoch": 0.9102040816326531, + "grad_norm": 0.2382912039756775, + "learning_rate": 5.9764640852839585e-05, + "loss": 0.28363141417503357, + "memory/device_reserved (GiB)": 77.39, + "memory/max_active (GiB)": 75.83, + "memory/max_allocated (GiB)": 75.83, + "ppl": 1.32794, + "step": 446, + "tokens/total": 109832896, + "tokens/train_per_sec_per_gpu": 12.73, + "tokens/trainable": 885442 + }, + { + "epoch": 0.9122448979591836, + "grad_norm": 0.23997141420841217, + "learning_rate": 5.960259516375134e-05, + "loss": 0.26503628492355347, + "memory/device_reserved (GiB)": 77.39, + "memory/max_active (GiB)": 72.37, + "memory/max_allocated (GiB)": 72.37, + "ppl": 1.30348, + "step": 447, + "tokens/total": 110051328, + "tokens/train_per_sec_per_gpu": 15.14, + "tokens/trainable": 887497 + }, + { + "epoch": 0.9142857142857143, + "grad_norm": 0.28751108050346375, + "learning_rate": 5.944044468295089e-05, + "loss": 0.31920450925827026, + "memory/device_reserved (GiB)": 77.39, + "memory/max_active (GiB)": 75.53, + "memory/max_allocated (GiB)": 75.53, + "ppl": 1.37603, + "step": 448, + "tokens/total": 110282944, + "tokens/train_per_sec_per_gpu": 11.59, + "tokens/trainable": 889326 + }, + { + "epoch": 0.9163265306122449, + "grad_norm": 0.21982364356517792, + "learning_rate": 5.927819117996266e-05, + "loss": 0.22485430538654327, + "memory/device_reserved (GiB)": 77.43, + "memory/max_active (GiB)": 76.51, + "memory/max_allocated (GiB)": 76.51, + "ppl": 1.25214, + "step": 449, + "tokens/total": 110584512, + "tokens/train_per_sec_per_gpu": 9.88, + "tokens/trainable": 891187 + }, + { + "epoch": 0.9183673469387755, + "grad_norm": 0.2274937629699707, + "learning_rate": 5.9115836425435314e-05, + "loss": 0.2748928666114807, + "memory/device_reserved (GiB)": 77.43, + "memory/max_active (GiB)": 75.69, + "memory/max_allocated (GiB)": 75.69, + "ppl": 1.31639, + "step": 450, + "tokens/total": 110839808, + "tokens/train_per_sec_per_gpu": 12.03, + "tokens/trainable": 893073 + }, + { + "epoch": 0.9204081632653062, + "grad_norm": 0.2462545782327652, + "learning_rate": 5.895338219112246e-05, + "loss": 0.3169158399105072, + "memory/device_reserved (GiB)": 77.43, + "memory/max_active (GiB)": 73.79, + "memory/max_allocated (GiB)": 73.79, + "ppl": 1.37289, + "step": 451, + "tokens/total": 111086656, + "tokens/train_per_sec_per_gpu": 9.16, + "tokens/trainable": 895037 + }, + { + "epoch": 0.9224489795918367, + "grad_norm": 0.25683456659317017, + "learning_rate": 5.879083024986334e-05, + "loss": 0.3225615620613098, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 76.7, + "memory/max_allocated (GiB)": 76.7, + "ppl": 1.38066, + "step": 452, + "tokens/total": 111324800, + "tokens/train_per_sec_per_gpu": 13.19, + "tokens/trainable": 896949 + }, + { + "epoch": 0.9244897959183673, + "grad_norm": 0.25344595313072205, + "learning_rate": 5.862818237556343e-05, + "loss": 0.20440013706684113, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 76.29, + "memory/max_allocated (GiB)": 76.29, + "ppl": 1.22679, + "step": 453, + "tokens/total": 111596480, + "tokens/train_per_sec_per_gpu": 10.3, + "tokens/trainable": 898857 + }, + { + "epoch": 0.926530612244898, + "grad_norm": 0.23042747378349304, + "learning_rate": 5.846544034317513e-05, + "loss": 0.24026471376419067, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 75.47, + "memory/max_allocated (GiB)": 75.47, + "ppl": 1.27159, + "step": 454, + "tokens/total": 111849728, + "tokens/train_per_sec_per_gpu": 10.78, + "tokens/trainable": 900650 + }, + { + "epoch": 0.9285714285714286, + "grad_norm": 0.275783509016037, + "learning_rate": 5.8302605928678345e-05, + "loss": 0.23303017020225525, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 76.78, + "memory/max_allocated (GiB)": 76.78, + "ppl": 1.26242, + "step": 455, + "tokens/total": 112116160, + "tokens/train_per_sec_per_gpu": 11.37, + "tokens/trainable": 902532 + }, + { + "epoch": 0.9306122448979591, + "grad_norm": 0.27579089999198914, + "learning_rate": 5.813968090906117e-05, + "loss": 0.289935439825058, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 74.01, + "memory/max_allocated (GiB)": 74.01, + "ppl": 1.33634, + "step": 456, + "tokens/total": 112378816, + "tokens/train_per_sec_per_gpu": 9.59, + "tokens/trainable": 904371 + }, + { + "epoch": 0.9326530612244898, + "grad_norm": 0.22608570754528046, + "learning_rate": 5.797666706230042e-05, + "loss": 0.23607102036476135, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 73.57, + "memory/max_allocated (GiB)": 73.57, + "ppl": 1.26626, + "step": 457, + "tokens/total": 112648128, + "tokens/train_per_sec_per_gpu": 11.11, + "tokens/trainable": 906372 + }, + { + "epoch": 0.9346938775510204, + "grad_norm": 0.24509814381599426, + "learning_rate": 5.78135661673423e-05, + "loss": 0.2416107952594757, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 75.42, + "memory/max_allocated (GiB)": 75.42, + "ppl": 1.2733, + "step": 458, + "tokens/total": 112869504, + "tokens/train_per_sec_per_gpu": 14.94, + "tokens/trainable": 908373 + }, + { + "epoch": 0.936734693877551, + "grad_norm": 0.2433391511440277, + "learning_rate": 5.765038000408295e-05, + "loss": 0.3300582468509674, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 76.07, + "memory/max_allocated (GiB)": 76.07, + "ppl": 1.39105, + "step": 459, + "tokens/total": 113124864, + "tokens/train_per_sec_per_gpu": 11.36, + "tokens/trainable": 910387 + }, + { + "epoch": 0.9387755102040817, + "grad_norm": 0.2222221940755844, + "learning_rate": 5.748711035334902e-05, + "loss": 0.20686075091362, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 74.22, + "memory/max_allocated (GiB)": 74.22, + "ppl": 1.22981, + "step": 460, + "tokens/total": 113368320, + "tokens/train_per_sec_per_gpu": 12.17, + "tokens/trainable": 912364 + }, + { + "epoch": 0.9408163265306122, + "grad_norm": 0.2881181240081787, + "learning_rate": 5.732375899687827e-05, + "loss": 0.2614569365978241, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 73.11, + "memory/max_allocated (GiB)": 73.11, + "ppl": 1.29882, + "step": 461, + "tokens/total": 113608384, + "tokens/train_per_sec_per_gpu": 12.66, + "tokens/trainable": 914225 + }, + { + "epoch": 0.9428571428571428, + "grad_norm": 0.24964448809623718, + "learning_rate": 5.7160327717300076e-05, + "loss": 0.2843072712421417, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 76.32, + "memory/max_allocated (GiB)": 76.32, + "ppl": 1.32884, + "step": 462, + "tokens/total": 113862912, + "tokens/train_per_sec_per_gpu": 14.57, + "tokens/trainable": 916421 + }, + { + "epoch": 0.9448979591836735, + "grad_norm": 0.21044345200061798, + "learning_rate": 5.6996818298116016e-05, + "loss": 0.19538094103336334, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 75.47, + "memory/max_allocated (GiB)": 75.47, + "ppl": 1.21577, + "step": 463, + "tokens/total": 114108032, + "tokens/train_per_sec_per_gpu": 11.47, + "tokens/trainable": 918176 + }, + { + "epoch": 0.9469387755102041, + "grad_norm": 0.24659273028373718, + "learning_rate": 5.68332325236804e-05, + "loss": 0.32539939880371094, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 75.91, + "memory/max_allocated (GiB)": 75.91, + "ppl": 1.38458, + "step": 464, + "tokens/total": 114349696, + "tokens/train_per_sec_per_gpu": 12.7, + "tokens/trainable": 920119 + }, + { + "epoch": 0.9489795918367347, + "grad_norm": 0.19757148623466492, + "learning_rate": 5.666957217918076e-05, + "loss": 0.20354850590229034, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 76.24, + "memory/max_allocated (GiB)": 76.24, + "ppl": 1.22574, + "step": 465, + "tokens/total": 114611136, + "tokens/train_per_sec_per_gpu": 10.7, + "tokens/trainable": 922045 + }, + { + "epoch": 0.9510204081632653, + "grad_norm": 0.2432941198348999, + "learning_rate": 5.6505839050618455e-05, + "loss": 0.2771606743335724, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 72.89, + "memory/max_allocated (GiB)": 72.89, + "ppl": 1.31938, + "step": 466, + "tokens/total": 114867136, + "tokens/train_per_sec_per_gpu": 11.2, + "tokens/trainable": 923918 + }, + { + "epoch": 0.9530612244897959, + "grad_norm": 0.21747350692749023, + "learning_rate": 5.634203492478909e-05, + "loss": 0.2509821057319641, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 73.68, + "memory/max_allocated (GiB)": 73.68, + "ppl": 1.28529, + "step": 467, + "tokens/total": 115120320, + "tokens/train_per_sec_per_gpu": 13.13, + "tokens/trainable": 925876 + }, + { + "epoch": 0.9551020408163265, + "grad_norm": 0.2742849588394165, + "learning_rate": 5.6178161589263024e-05, + "loss": 0.30438852310180664, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 72.26, + "memory/max_allocated (GiB)": 72.26, + "ppl": 1.3558, + "step": 468, + "tokens/total": 115362304, + "tokens/train_per_sec_per_gpu": 13.08, + "tokens/trainable": 927858 + }, + { + "epoch": 0.9571428571428572, + "grad_norm": 0.2456769347190857, + "learning_rate": 5.601422083236596e-05, + "loss": 0.19210582971572876, + "memory/device_reserved (GiB)": 77.57, + "memory/max_active (GiB)": 75.12, + "memory/max_allocated (GiB)": 75.12, + "ppl": 1.2118, + "step": 469, + "tokens/total": 115602240, + "tokens/train_per_sec_per_gpu": 12.48, + "tokens/trainable": 929668 + }, + { + "epoch": 0.9591836734693877, + "grad_norm": 0.270745187997818, + "learning_rate": 5.585021444315932e-05, + "loss": 0.33105525374412537, + "memory/device_reserved (GiB)": 77.72, + "memory/max_active (GiB)": 76.73, + "memory/max_allocated (GiB)": 76.73, + "ppl": 1.39244, + "step": 470, + "tokens/total": 115863360, + "tokens/train_per_sec_per_gpu": 12.67, + "tokens/trainable": 931965 + }, + { + "epoch": 0.9612244897959183, + "grad_norm": 0.28040361404418945, + "learning_rate": 5.5686144211420775e-05, + "loss": 0.30292609333992004, + "memory/device_reserved (GiB)": 77.72, + "memory/max_active (GiB)": 73.51, + "memory/max_allocated (GiB)": 73.51, + "ppl": 1.35381, + "step": 471, + "tokens/total": 116118848, + "tokens/train_per_sec_per_gpu": 11.0, + "tokens/trainable": 933871 + }, + { + "epoch": 0.963265306122449, + "grad_norm": 0.3054877817630768, + "learning_rate": 5.5522011927624674e-05, + "loss": 0.36382973194122314, + "memory/device_reserved (GiB)": 77.72, + "memory/max_active (GiB)": 74.79, + "memory/max_allocated (GiB)": 74.79, + "ppl": 1.43883, + "step": 472, + "tokens/total": 116376256, + "tokens/train_per_sec_per_gpu": 10.96, + "tokens/trainable": 935828 + }, + { + "epoch": 0.9653061224489796, + "grad_norm": 0.1959354430437088, + "learning_rate": 5.535781938292255e-05, + "loss": 0.1597578376531601, + "memory/device_reserved (GiB)": 77.72, + "memory/max_active (GiB)": 74.14, + "memory/max_allocated (GiB)": 74.14, + "ppl": 1.17323, + "step": 473, + "tokens/total": 116659968, + "tokens/train_per_sec_per_gpu": 8.47, + "tokens/trainable": 937442 + }, + { + "epoch": 0.9673469387755103, + "grad_norm": 0.2228005826473236, + "learning_rate": 5.5193568369123574e-05, + "loss": 0.27887964248657227, + "memory/device_reserved (GiB)": 77.72, + "memory/max_active (GiB)": 71.88, + "memory/max_allocated (GiB)": 71.88, + "ppl": 1.32165, + "step": 474, + "tokens/total": 116900544, + "tokens/train_per_sec_per_gpu": 13.19, + "tokens/trainable": 939514 + }, + { + "epoch": 0.9693877551020408, + "grad_norm": 0.24833163619041443, + "learning_rate": 5.5029260678674946e-05, + "loss": 0.2714032530784607, + "memory/device_reserved (GiB)": 77.72, + "memory/max_active (GiB)": 72.53, + "memory/max_allocated (GiB)": 72.53, + "ppl": 1.3118, + "step": 475, + "tokens/total": 117173952, + "tokens/train_per_sec_per_gpu": 12.5, + "tokens/trainable": 941504 + }, + { + "epoch": 0.9714285714285714, + "grad_norm": 0.23854611814022064, + "learning_rate": 5.486489810464238e-05, + "loss": 0.23803013563156128, + "memory/device_reserved (GiB)": 77.72, + "memory/max_active (GiB)": 75.01, + "memory/max_allocated (GiB)": 75.01, + "ppl": 1.26875, + "step": 476, + "tokens/total": 117403520, + "tokens/train_per_sec_per_gpu": 8.56, + "tokens/trainable": 943229 + }, + { + "epoch": 0.9734693877551021, + "grad_norm": 0.21503248810768127, + "learning_rate": 5.470048244069055e-05, + "loss": 0.2587663531303406, + "memory/device_reserved (GiB)": 75.95, + "memory/max_active (GiB)": 71.96, + "memory/max_allocated (GiB)": 71.96, + "ppl": 1.29533, + "step": 477, + "tokens/total": 117655552, + "tokens/train_per_sec_per_gpu": 10.21, + "tokens/trainable": 945160 + }, + { + "epoch": 0.9755102040816327, + "grad_norm": 0.20460256934165955, + "learning_rate": 5.453601548106346e-05, + "loss": 0.20304179191589355, + "memory/device_reserved (GiB)": 77.33, + "memory/max_active (GiB)": 76.4, + "memory/max_allocated (GiB)": 76.4, + "ppl": 1.22512, + "step": 478, + "tokens/total": 117909760, + "tokens/train_per_sec_per_gpu": 12.96, + "tokens/trainable": 947150 + }, + { + "epoch": 0.9775510204081632, + "grad_norm": 0.2662351131439209, + "learning_rate": 5.437149902056493e-05, + "loss": 0.28601670265197754, + "memory/device_reserved (GiB)": 77.33, + "memory/max_active (GiB)": 72.54, + "memory/max_allocated (GiB)": 72.54, + "ppl": 1.33111, + "step": 479, + "tokens/total": 118140608, + "tokens/train_per_sec_per_gpu": 13.37, + "tokens/trainable": 949054 + }, + { + "epoch": 0.9795918367346939, + "grad_norm": 0.23290075361728668, + "learning_rate": 5.4206934854538924e-05, + "loss": 0.2588629424571991, + "memory/device_reserved (GiB)": 77.33, + "memory/max_active (GiB)": 71.91, + "memory/max_allocated (GiB)": 71.91, + "ppl": 1.29546, + "step": 480, + "tokens/total": 118367040, + "tokens/train_per_sec_per_gpu": 10.45, + "tokens/trainable": 950659 + }, + { + "epoch": 0.9816326530612245, + "grad_norm": 0.2183947116136551, + "learning_rate": 5.40423247788501e-05, + "loss": 0.267020046710968, + "memory/device_reserved (GiB)": 77.33, + "memory/max_active (GiB)": 73.73, + "memory/max_allocated (GiB)": 73.73, + "ppl": 1.30607, + "step": 481, + "tokens/total": 118613504, + "tokens/train_per_sec_per_gpu": 11.01, + "tokens/trainable": 952450 + }, + { + "epoch": 0.9836734693877551, + "grad_norm": 0.21648745238780975, + "learning_rate": 5.387767058986399e-05, + "loss": 0.20945556461811066, + "memory/device_reserved (GiB)": 77.34, + "memory/max_active (GiB)": 75.42, + "memory/max_allocated (GiB)": 75.42, + "ppl": 1.23301, + "step": 482, + "tokens/total": 118871872, + "tokens/train_per_sec_per_gpu": 9.99, + "tokens/trainable": 954253 + }, + { + "epoch": 0.9857142857142858, + "grad_norm": 0.2088627815246582, + "learning_rate": 5.3712974084427644e-05, + "loss": 0.22902044653892517, + "memory/device_reserved (GiB)": 77.34, + "memory/max_active (GiB)": 72.07, + "memory/max_allocated (GiB)": 72.07, + "ppl": 1.25737, + "step": 483, + "tokens/total": 119107776, + "tokens/train_per_sec_per_gpu": 12.27, + "tokens/trainable": 956219 + }, + { + "epoch": 0.9877551020408163, + "grad_norm": 0.21806953847408295, + "learning_rate": 5.354823705984987e-05, + "loss": 0.23744834959506989, + "memory/device_reserved (GiB)": 77.34, + "memory/max_active (GiB)": 75.07, + "memory/max_allocated (GiB)": 75.07, + "ppl": 1.26801, + "step": 484, + "tokens/total": 119375936, + "tokens/train_per_sec_per_gpu": 11.58, + "tokens/trainable": 958184 + }, + { + "epoch": 0.9897959183673469, + "grad_norm": 0.2159849852323532, + "learning_rate": 5.338346131388161e-05, + "loss": 0.28593748807907104, + "memory/device_reserved (GiB)": 77.34, + "memory/max_active (GiB)": 73.62, + "memory/max_allocated (GiB)": 73.62, + "ppl": 1.33101, + "step": 485, + "tokens/total": 119658624, + "tokens/train_per_sec_per_gpu": 12.17, + "tokens/trainable": 960532 + }, + { + "epoch": 0.9918367346938776, + "grad_norm": 0.2770880460739136, + "learning_rate": 5.321864864469646e-05, + "loss": 0.3062543272972107, + "memory/device_reserved (GiB)": 77.34, + "memory/max_active (GiB)": 74.88, + "memory/max_allocated (GiB)": 74.88, + "ppl": 1.35833, + "step": 486, + "tokens/total": 119883072, + "tokens/train_per_sec_per_gpu": 12.53, + "tokens/trainable": 962362 + }, + { + "epoch": 0.9938775510204082, + "grad_norm": 0.29565325379371643, + "learning_rate": 5.305380085087085e-05, + "loss": 0.3190445899963379, + "memory/device_reserved (GiB)": 77.34, + "memory/max_active (GiB)": 75.66, + "memory/max_allocated (GiB)": 75.66, + "ppl": 1.37581, + "step": 487, + "tokens/total": 120155456, + "tokens/train_per_sec_per_gpu": 12.55, + "tokens/trainable": 964473 + }, + { + "epoch": 0.9959183673469387, + "grad_norm": 0.2474423050880432, + "learning_rate": 5.2888919731364594e-05, + "loss": 0.30566027760505676, + "memory/device_reserved (GiB)": 77.34, + "memory/max_active (GiB)": 71.8, + "memory/max_allocated (GiB)": 71.8, + "ppl": 1.35752, + "step": 488, + "tokens/total": 120412416, + "tokens/train_per_sec_per_gpu": 11.49, + "tokens/trainable": 966342 + }, + { + "epoch": 0.9979591836734694, + "grad_norm": 0.26179268956184387, + "learning_rate": 5.2724007085501137e-05, + "loss": 0.32216572761535645, + "memory/device_reserved (GiB)": 77.34, + "memory/max_active (GiB)": 74.47, + "memory/max_allocated (GiB)": 74.47, + "ppl": 1.38011, + "step": 489, + "tokens/total": 120643264, + "tokens/train_per_sec_per_gpu": 13.39, + "tokens/trainable": 968356 + }, + { + "epoch": 1.0, + "grad_norm": 0.222548708319664, + "learning_rate": 5.2559064712948016e-05, + "loss": 0.2377071976661682, + "memory/device_reserved (GiB)": 77.34, + "memory/max_active (GiB)": 75.47, + "memory/max_allocated (GiB)": 75.47, + "ppl": 1.26834, + "step": 490, + "tokens/total": 120899392, + "tokens/train_per_sec_per_gpu": 11.88, + "tokens/trainable": 970260 + }, + { + "epoch": 1.0020408163265306, + "grad_norm": 0.695749044418335, + "learning_rate": 5.2394094413697136e-05, + "loss": 0.19689464569091797, + "memory/device_reserved (GiB)": 77.34, + "memory/max_active (GiB)": 74.77, + "memory/max_allocated (GiB)": 74.77, + "ppl": 1.21762, + "step": 491, + "tokens/total": 121135360, + "tokens/train_per_sec_per_gpu": 10.62, + "tokens/trainable": 972107 + }, + { + "epoch": 1.0040816326530613, + "grad_norm": 0.2337198555469513, + "learning_rate": 5.222909798804515e-05, + "loss": 0.23959970474243164, + "memory/device_reserved (GiB)": 77.34, + "memory/max_active (GiB)": 73.62, + "memory/max_allocated (GiB)": 73.62, + "ppl": 1.27074, + "step": 492, + "tokens/total": 121360576, + "tokens/train_per_sec_per_gpu": 13.04, + "tokens/trainable": 973888 + }, + { + "epoch": 1.006122448979592, + "grad_norm": 0.18215812742710114, + "learning_rate": 5.206407723657386e-05, + "loss": 0.18581895530223846, + "memory/device_reserved (GiB)": 77.34, + "memory/max_active (GiB)": 74.85, + "memory/max_allocated (GiB)": 74.85, + "ppl": 1.2042, + "step": 493, + "tokens/total": 121588096, + "tokens/train_per_sec_per_gpu": 11.99, + "tokens/trainable": 975754 + }, + { + "epoch": 1.0081632653061225, + "grad_norm": 0.21457408368587494, + "learning_rate": 5.189903396013052e-05, + "loss": 0.23084910213947296, + "memory/device_reserved (GiB)": 77.34, + "memory/max_active (GiB)": 75.69, + "memory/max_allocated (GiB)": 75.69, + "ppl": 1.25967, + "step": 494, + "tokens/total": 121857216, + "tokens/train_per_sec_per_gpu": 13.93, + "tokens/trainable": 978156 + }, + { + "epoch": 1.010204081632653, + "grad_norm": 0.21651597321033478, + "learning_rate": 5.173396995980818e-05, + "loss": 0.24254319071769714, + "memory/device_reserved (GiB)": 77.34, + "memory/max_active (GiB)": 74.55, + "memory/max_allocated (GiB)": 74.55, + "ppl": 1.27449, + "step": 495, + "tokens/total": 122128512, + "tokens/train_per_sec_per_gpu": 12.7, + "tokens/trainable": 980324 + }, + { + "epoch": 1.0122448979591836, + "grad_norm": 0.23706747591495514, + "learning_rate": 5.156888703692608e-05, + "loss": 0.26104283332824707, + "memory/device_reserved (GiB)": 77.34, + "memory/max_active (GiB)": 74.68, + "memory/max_allocated (GiB)": 74.68, + "ppl": 1.29828, + "step": 496, + "tokens/total": 122391296, + "tokens/train_per_sec_per_gpu": 11.21, + "tokens/trainable": 982307 + }, + { + "epoch": 1.0142857142857142, + "grad_norm": 0.20237280428409576, + "learning_rate": 5.1403786993009906e-05, + "loss": 0.19621878862380981, + "memory/device_reserved (GiB)": 77.43, + "memory/max_active (GiB)": 76.54, + "memory/max_allocated (GiB)": 76.54, + "ppl": 1.21679, + "step": 497, + "tokens/total": 122656000, + "tokens/train_per_sec_per_gpu": 12.29, + "tokens/trainable": 984177 + }, + { + "epoch": 1.0163265306122449, + "grad_norm": 0.29173213243484497, + "learning_rate": 5.123867162977224e-05, + "loss": 0.27584266662597656, + "memory/device_reserved (GiB)": 77.43, + "memory/max_active (GiB)": 74.06, + "memory/max_allocated (GiB)": 74.06, + "ppl": 1.31764, + "step": 498, + "tokens/total": 122892864, + "tokens/train_per_sec_per_gpu": 13.39, + "tokens/trainable": 986057 + }, + { + "epoch": 1.0183673469387755, + "grad_norm": 0.2208496332168579, + "learning_rate": 5.1073542749092804e-05, + "loss": 0.18701229989528656, + "memory/device_reserved (GiB)": 77.43, + "memory/max_active (GiB)": 71.83, + "memory/max_allocated (GiB)": 71.83, + "ppl": 1.20564, + "step": 499, + "tokens/total": 123103744, + "tokens/train_per_sec_per_gpu": 12.85, + "tokens/trainable": 987822 + }, + { + "epoch": 1.0204081632653061, + "grad_norm": 0.21554070711135864, + "learning_rate": 5.0908402152998844e-05, + "loss": 0.2235528528690338, + "memory/device_reserved (GiB)": 77.43, + "memory/max_active (GiB)": 76.43, + "memory/max_allocated (GiB)": 76.43, + "ppl": 1.25051, + "step": 500, + "tokens/total": 123352704, + "tokens/train_per_sec_per_gpu": 12.3, + "tokens/trainable": 989861 + }, + { + "epoch": 1.0224489795918368, + "grad_norm": 0.3275497853755951, + "learning_rate": 5.074325164364548e-05, + "loss": 0.2499224990606308, + "memory/device_reserved (GiB)": 77.43, + "memory/max_active (GiB)": 75.23, + "memory/max_allocated (GiB)": 75.23, + "ppl": 1.28393, + "step": 501, + "tokens/total": 123575616, + "tokens/train_per_sec_per_gpu": 11.37, + "tokens/trainable": 991966 + }, + { + "epoch": 1.0244897959183674, + "grad_norm": 0.23635686933994293, + "learning_rate": 5.0578093023295944e-05, + "loss": 0.2355385422706604, + "memory/device_reserved (GiB)": 76.84, + "memory/max_active (GiB)": 76.05, + "memory/max_allocated (GiB)": 76.05, + "ppl": 1.26559, + "step": 502, + "tokens/total": 123822464, + "tokens/train_per_sec_per_gpu": 12.44, + "tokens/trainable": 994030 + }, + { + "epoch": 1.026530612244898, + "grad_norm": 0.2575612962245941, + "learning_rate": 5.041292809430209e-05, + "loss": 0.21649597585201263, + "memory/device_reserved (GiB)": 76.85, + "memory/max_active (GiB)": 75.37, + "memory/max_allocated (GiB)": 75.37, + "ppl": 1.24172, + "step": 503, + "tokens/total": 124072576, + "tokens/train_per_sec_per_gpu": 14.33, + "tokens/trainable": 996357 + }, + { + "epoch": 1.0285714285714285, + "grad_norm": 0.3012147843837738, + "learning_rate": 5.024775865908451e-05, + "loss": 0.24610432982444763, + "memory/device_reserved (GiB)": 76.85, + "memory/max_active (GiB)": 73.71, + "memory/max_allocated (GiB)": 73.71, + "ppl": 1.27903, + "step": 504, + "tokens/total": 124323456, + "tokens/train_per_sec_per_gpu": 11.92, + "tokens/trainable": 998330 + }, + { + "epoch": 1.030612244897959, + "grad_norm": 0.2535402774810791, + "learning_rate": 5.008258652011303e-05, + "loss": 0.3021504580974579, + "memory/device_reserved (GiB)": 76.85, + "memory/max_active (GiB)": 72.81, + "memory/max_allocated (GiB)": 72.81, + "ppl": 1.35276, + "step": 505, + "tokens/total": 124590912, + "tokens/train_per_sec_per_gpu": 12.89, + "tokens/trainable": 1000519 + }, + { + "epoch": 1.0326530612244897, + "grad_norm": 0.23790322244167328, + "learning_rate": 4.9917413479886984e-05, + "loss": 0.25011152029037476, + "memory/device_reserved (GiB)": 77.28, + "memory/max_active (GiB)": 76.51, + "memory/max_allocated (GiB)": 76.51, + "ppl": 1.28417, + "step": 506, + "tokens/total": 124844800, + "tokens/train_per_sec_per_gpu": 12.93, + "tokens/trainable": 1002471 + }, + { + "epoch": 1.0346938775510204, + "grad_norm": 0.2475406378507614, + "learning_rate": 4.975224134091551e-05, + "loss": 0.28482577204704285, + "memory/device_reserved (GiB)": 77.28, + "memory/max_active (GiB)": 74.17, + "memory/max_allocated (GiB)": 74.17, + "ppl": 1.32953, + "step": 507, + "tokens/total": 125072576, + "tokens/train_per_sec_per_gpu": 13.65, + "tokens/trainable": 1004612 + }, + { + "epoch": 1.036734693877551, + "grad_norm": 0.22571589052677155, + "learning_rate": 4.9587071905697926e-05, + "loss": 0.18567447364330292, + "memory/device_reserved (GiB)": 77.28, + "memory/max_active (GiB)": 73.92, + "memory/max_allocated (GiB)": 73.92, + "ppl": 1.20403, + "step": 508, + "tokens/total": 125328896, + "tokens/train_per_sec_per_gpu": 12.5, + "tokens/trainable": 1006465 + }, + { + "epoch": 1.0387755102040817, + "grad_norm": 0.23652319610118866, + "learning_rate": 4.942190697670406e-05, + "loss": 0.21158185601234436, + "memory/device_reserved (GiB)": 77.28, + "memory/max_active (GiB)": 72.81, + "memory/max_allocated (GiB)": 72.81, + "ppl": 1.23563, + "step": 509, + "tokens/total": 125588672, + "tokens/train_per_sec_per_gpu": 9.93, + "tokens/trainable": 1008270 + }, + { + "epoch": 1.0408163265306123, + "grad_norm": 0.24855706095695496, + "learning_rate": 4.9256748356354544e-05, + "loss": 0.21951709687709808, + "memory/device_reserved (GiB)": 77.51, + "memory/max_active (GiB)": 76.62, + "memory/max_allocated (GiB)": 76.62, + "ppl": 1.24548, + "step": 510, + "tokens/total": 125892416, + "tokens/train_per_sec_per_gpu": 10.54, + "tokens/trainable": 1010296 + }, + { + "epoch": 1.042857142857143, + "grad_norm": 0.2422824651002884, + "learning_rate": 4.9091597847001175e-05, + "loss": 0.2284081131219864, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 76.81, + "memory/max_allocated (GiB)": 76.81, + "ppl": 1.2566, + "step": 511, + "tokens/total": 126150016, + "tokens/train_per_sec_per_gpu": 12.65, + "tokens/trainable": 1012385 + }, + { + "epoch": 1.0448979591836736, + "grad_norm": 0.28957000374794006, + "learning_rate": 4.8926457250907214e-05, + "loss": 0.21849867701530457, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 76.1, + "memory/max_allocated (GiB)": 76.1, + "ppl": 1.24421, + "step": 512, + "tokens/total": 126396160, + "tokens/train_per_sec_per_gpu": 11.62, + "tokens/trainable": 1014252 + }, + { + "epoch": 1.046938775510204, + "grad_norm": 0.22412888705730438, + "learning_rate": 4.8761328370227774e-05, + "loss": 0.19620712101459503, + "memory/device_reserved (GiB)": 77.74, + "memory/max_active (GiB)": 76.84, + "memory/max_allocated (GiB)": 76.84, + "ppl": 1.21678, + "step": 513, + "tokens/total": 126674176, + "tokens/train_per_sec_per_gpu": 12.0, + "tokens/trainable": 1016389 + }, + { + "epoch": 1.0489795918367346, + "grad_norm": 0.25189709663391113, + "learning_rate": 4.859621300699009e-05, + "loss": 0.22615531086921692, + "memory/device_reserved (GiB)": 77.74, + "memory/max_active (GiB)": 75.5, + "memory/max_allocated (GiB)": 75.5, + "ppl": 1.25377, + "step": 514, + "tokens/total": 126909120, + "tokens/train_per_sec_per_gpu": 11.4, + "tokens/trainable": 1018328 + }, + { + "epoch": 1.0510204081632653, + "grad_norm": 0.23306973278522491, + "learning_rate": 4.8431112963073924e-05, + "loss": 0.24342980980873108, + "memory/device_reserved (GiB)": 77.74, + "memory/max_active (GiB)": 76.07, + "memory/max_allocated (GiB)": 76.07, + "ppl": 1.27562, + "step": 515, + "tokens/total": 127169344, + "tokens/train_per_sec_per_gpu": 11.91, + "tokens/trainable": 1020390 + }, + { + "epoch": 1.0530612244897959, + "grad_norm": 0.24464914202690125, + "learning_rate": 4.826603004019182e-05, + "loss": 0.23267650604248047, + "memory/device_reserved (GiB)": 77.74, + "memory/max_active (GiB)": 76.62, + "memory/max_allocated (GiB)": 76.62, + "ppl": 1.26197, + "step": 516, + "tokens/total": 127414080, + "tokens/train_per_sec_per_gpu": 11.84, + "tokens/trainable": 1022321 + }, + { + "epoch": 1.0551020408163265, + "grad_norm": 0.24754460155963898, + "learning_rate": 4.8100966039869474e-05, + "loss": 0.1910042017698288, + "memory/device_reserved (GiB)": 77.74, + "memory/max_active (GiB)": 74.36, + "memory/max_allocated (GiB)": 74.36, + "ppl": 1.21046, + "step": 517, + "tokens/total": 127663680, + "tokens/train_per_sec_per_gpu": 9.42, + "tokens/trainable": 1024077 + }, + { + "epoch": 1.0571428571428572, + "grad_norm": 0.23382535576820374, + "learning_rate": 4.793592276342614e-05, + "loss": 0.21250295639038086, + "memory/device_reserved (GiB)": 77.74, + "memory/max_active (GiB)": 76.73, + "memory/max_allocated (GiB)": 76.73, + "ppl": 1.23677, + "step": 518, + "tokens/total": 127912576, + "tokens/train_per_sec_per_gpu": 11.26, + "tokens/trainable": 1025992 + }, + { + "epoch": 1.0591836734693878, + "grad_norm": 0.23026596009731293, + "learning_rate": 4.777090201195486e-05, + "loss": 0.16110140085220337, + "memory/device_reserved (GiB)": 77.74, + "memory/max_active (GiB)": 74.06, + "memory/max_allocated (GiB)": 74.06, + "ppl": 1.1748, + "step": 519, + "tokens/total": 128142528, + "tokens/train_per_sec_per_gpu": 13.52, + "tokens/trainable": 1027984 + }, + { + "epoch": 1.0612244897959184, + "grad_norm": 0.2506268322467804, + "learning_rate": 4.7605905586302875e-05, + "loss": 0.22823640704154968, + "memory/device_reserved (GiB)": 77.74, + "memory/max_active (GiB)": 75.56, + "memory/max_allocated (GiB)": 75.56, + "ppl": 1.25638, + "step": 520, + "tokens/total": 128406208, + "tokens/train_per_sec_per_gpu": 10.52, + "tokens/trainable": 1029994 + }, + { + "epoch": 1.063265306122449, + "grad_norm": 0.23237290978431702, + "learning_rate": 4.7440935287051996e-05, + "loss": 0.16627773642539978, + "memory/device_reserved (GiB)": 77.74, + "memory/max_active (GiB)": 72.26, + "memory/max_allocated (GiB)": 72.26, + "ppl": 1.1809, + "step": 521, + "tokens/total": 128637184, + "tokens/train_per_sec_per_gpu": 12.76, + "tokens/trainable": 1031790 + }, + { + "epoch": 1.0653061224489795, + "grad_norm": 0.22715282440185547, + "learning_rate": 4.727599291449887e-05, + "loss": 0.1747395098209381, + "memory/device_reserved (GiB)": 77.74, + "memory/max_active (GiB)": 75.94, + "memory/max_allocated (GiB)": 75.94, + "ppl": 1.19094, + "step": 522, + "tokens/total": 128879616, + "tokens/train_per_sec_per_gpu": 12.25, + "tokens/trainable": 1033602 + }, + { + "epoch": 1.0673469387755101, + "grad_norm": 0.2602156698703766, + "learning_rate": 4.7111080268635424e-05, + "loss": 0.23783069849014282, + "memory/device_reserved (GiB)": 77.74, + "memory/max_active (GiB)": 73.68, + "memory/max_allocated (GiB)": 73.68, + "ppl": 1.26849, + "step": 523, + "tokens/total": 129130496, + "tokens/train_per_sec_per_gpu": 12.81, + "tokens/trainable": 1035596 + }, + { + "epoch": 1.0693877551020408, + "grad_norm": 0.2306244820356369, + "learning_rate": 4.694619914912917e-05, + "loss": 0.21332740783691406, + "memory/device_reserved (GiB)": 77.74, + "memory/max_active (GiB)": 75.8, + "memory/max_allocated (GiB)": 75.8, + "ppl": 1.23779, + "step": 524, + "tokens/total": 129395904, + "tokens/train_per_sec_per_gpu": 13.92, + "tokens/trainable": 1037875 + }, + { + "epoch": 1.0714285714285714, + "grad_norm": 0.25532639026641846, + "learning_rate": 4.6781351355303553e-05, + "loss": 0.24508503079414368, + "memory/device_reserved (GiB)": 77.74, + "memory/max_active (GiB)": 76.65, + "memory/max_allocated (GiB)": 76.65, + "ppl": 1.27773, + "step": 525, + "tokens/total": 129646720, + "tokens/train_per_sec_per_gpu": 12.46, + "tokens/trainable": 1040071 + }, + { + "epoch": 1.073469387755102, + "grad_norm": 0.2608277499675751, + "learning_rate": 4.66165386861184e-05, + "loss": 0.1986381560564041, + "memory/device_reserved (GiB)": 77.74, + "memory/max_active (GiB)": 75.09, + "memory/max_allocated (GiB)": 75.09, + "ppl": 1.21974, + "step": 526, + "tokens/total": 129923776, + "tokens/train_per_sec_per_gpu": 9.08, + "tokens/trainable": 1042089 + }, + { + "epoch": 1.0755102040816327, + "grad_norm": 0.27311381697654724, + "learning_rate": 4.645176294015015e-05, + "loss": 0.19322440028190613, + "memory/device_reserved (GiB)": 76.08, + "memory/max_active (GiB)": 73.16, + "memory/max_allocated (GiB)": 73.16, + "ppl": 1.21315, + "step": 527, + "tokens/total": 130162176, + "tokens/train_per_sec_per_gpu": 15.07, + "tokens/trainable": 1044330 + }, + { + "epoch": 1.0775510204081633, + "grad_norm": 0.27017319202423096, + "learning_rate": 4.628702591557237e-05, + "loss": 0.22304826974868774, + "memory/device_reserved (GiB)": 76.08, + "memory/max_active (GiB)": 75.04, + "memory/max_allocated (GiB)": 75.04, + "ppl": 1.24988, + "step": 528, + "tokens/total": 130402688, + "tokens/train_per_sec_per_gpu": 10.65, + "tokens/trainable": 1046138 + }, + { + "epoch": 1.079591836734694, + "grad_norm": 0.30155065655708313, + "learning_rate": 4.612232941013603e-05, + "loss": 0.2504119873046875, + "memory/device_reserved (GiB)": 76.08, + "memory/max_active (GiB)": 73.35, + "memory/max_allocated (GiB)": 73.35, + "ppl": 1.28455, + "step": 529, + "tokens/total": 130628544, + "tokens/train_per_sec_per_gpu": 13.88, + "tokens/trainable": 1048180 + }, + { + "epoch": 1.0816326530612246, + "grad_norm": 0.26590681076049805, + "learning_rate": 4.595767522114993e-05, + "loss": 0.25388896465301514, + "memory/device_reserved (GiB)": 76.1, + "memory/max_active (GiB)": 75.2, + "memory/max_allocated (GiB)": 75.2, + "ppl": 1.28903, + "step": 530, + "tokens/total": 130854720, + "tokens/train_per_sec_per_gpu": 13.7, + "tokens/trainable": 1050029 + }, + { + "epoch": 1.083673469387755, + "grad_norm": 0.3531028926372528, + "learning_rate": 4.579306514546107e-05, + "loss": 0.3330848217010498, + "memory/device_reserved (GiB)": 76.1, + "memory/max_active (GiB)": 74.11, + "memory/max_allocated (GiB)": 74.11, + "ppl": 1.39527, + "step": 531, + "tokens/total": 131120192, + "tokens/train_per_sec_per_gpu": 12.49, + "tokens/trainable": 1052178 + }, + { + "epoch": 1.0857142857142856, + "grad_norm": 0.23187237977981567, + "learning_rate": 4.562850097943507e-05, + "loss": 0.2286103069782257, + "memory/device_reserved (GiB)": 76.77, + "memory/max_active (GiB)": 75.86, + "memory/max_allocated (GiB)": 75.86, + "ppl": 1.25685, + "step": 532, + "tokens/total": 131386112, + "tokens/train_per_sec_per_gpu": 12.81, + "tokens/trainable": 1054418 + }, + { + "epoch": 1.0877551020408163, + "grad_norm": 0.3015822768211365, + "learning_rate": 4.5463984518936544e-05, + "loss": 0.2324981540441513, + "memory/device_reserved (GiB)": 76.77, + "memory/max_active (GiB)": 73.49, + "memory/max_allocated (GiB)": 73.49, + "ppl": 1.26175, + "step": 533, + "tokens/total": 131611456, + "tokens/train_per_sec_per_gpu": 13.2, + "tokens/trainable": 1056251 + }, + { + "epoch": 1.089795918367347, + "grad_norm": 0.24493113160133362, + "learning_rate": 4.529951755930946e-05, + "loss": 0.22329843044281006, + "memory/device_reserved (GiB)": 77.45, + "memory/max_active (GiB)": 76.67, + "memory/max_allocated (GiB)": 76.67, + "ppl": 1.25019, + "step": 534, + "tokens/total": 131876416, + "tokens/train_per_sec_per_gpu": 10.29, + "tokens/trainable": 1058208 + }, + { + "epoch": 1.0918367346938775, + "grad_norm": 0.38875845074653625, + "learning_rate": 4.5135101895357624e-05, + "loss": 0.25845324993133545, + "memory/device_reserved (GiB)": 77.45, + "memory/max_active (GiB)": 75.8, + "memory/max_allocated (GiB)": 75.8, + "ppl": 1.29493, + "step": 535, + "tokens/total": 132113600, + "tokens/train_per_sec_per_gpu": 13.55, + "tokens/trainable": 1060365 + }, + { + "epoch": 1.0938775510204082, + "grad_norm": 0.25096291303634644, + "learning_rate": 4.497073932132507e-05, + "loss": 0.2262846827507019, + "memory/device_reserved (GiB)": 77.45, + "memory/max_active (GiB)": 76.07, + "memory/max_allocated (GiB)": 76.07, + "ppl": 1.25393, + "step": 536, + "tokens/total": 132360896, + "tokens/train_per_sec_per_gpu": 11.28, + "tokens/trainable": 1062198 + }, + { + "epoch": 1.0959183673469388, + "grad_norm": 0.2903803288936615, + "learning_rate": 4.480643163087644e-05, + "loss": 0.26870405673980713, + "memory/device_reserved (GiB)": 77.45, + "memory/max_active (GiB)": 74.8, + "memory/max_allocated (GiB)": 74.8, + "ppl": 1.30827, + "step": 537, + "tokens/total": 132606336, + "tokens/train_per_sec_per_gpu": 12.05, + "tokens/trainable": 1064234 + }, + { + "epoch": 1.0979591836734695, + "grad_norm": 0.24137850105762482, + "learning_rate": 4.4642180617077464e-05, + "loss": 0.1940939575433731, + "memory/device_reserved (GiB)": 77.45, + "memory/max_active (GiB)": 72.72, + "memory/max_allocated (GiB)": 72.72, + "ppl": 1.21421, + "step": 538, + "tokens/total": 132857920, + "tokens/train_per_sec_per_gpu": 11.57, + "tokens/trainable": 1066099 + }, + { + "epoch": 1.1, + "grad_norm": 0.22201845049858093, + "learning_rate": 4.4477988072375344e-05, + "loss": 0.17957425117492676, + "memory/device_reserved (GiB)": 77.45, + "memory/max_active (GiB)": 75.45, + "memory/max_allocated (GiB)": 75.45, + "ppl": 1.19671, + "step": 539, + "tokens/total": 133101568, + "tokens/train_per_sec_per_gpu": 10.58, + "tokens/trainable": 1067989 + }, + { + "epoch": 1.1020408163265305, + "grad_norm": 0.2741422951221466, + "learning_rate": 4.431385578857924e-05, + "loss": 0.24984097480773926, + "memory/device_reserved (GiB)": 77.45, + "memory/max_active (GiB)": 72.94, + "memory/max_allocated (GiB)": 72.94, + "ppl": 1.28382, + "step": 540, + "tokens/total": 133341696, + "tokens/train_per_sec_per_gpu": 14.17, + "tokens/trainable": 1069962 + }, + { + "epoch": 1.1040816326530611, + "grad_norm": 0.24486468732357025, + "learning_rate": 4.414978555684069e-05, + "loss": 0.22328521311283112, + "memory/device_reserved (GiB)": 77.45, + "memory/max_active (GiB)": 74.66, + "memory/max_allocated (GiB)": 74.66, + "ppl": 1.25018, + "step": 541, + "tokens/total": 133598720, + "tokens/train_per_sec_per_gpu": 12.43, + "tokens/trainable": 1072125 + }, + { + "epoch": 1.1061224489795918, + "grad_norm": 0.24908317625522614, + "learning_rate": 4.3985779167634045e-05, + "loss": 0.195266991853714, + "memory/device_reserved (GiB)": 77.45, + "memory/max_active (GiB)": 75.99, + "memory/max_allocated (GiB)": 75.99, + "ppl": 1.21564, + "step": 542, + "tokens/total": 133818304, + "tokens/train_per_sec_per_gpu": 11.17, + "tokens/trainable": 1074005 + }, + { + "epoch": 1.1081632653061224, + "grad_norm": 0.5780074000358582, + "learning_rate": 4.382183841073698e-05, + "loss": 0.2370929718017578, + "memory/device_reserved (GiB)": 77.45, + "memory/max_active (GiB)": 72.94, + "memory/max_allocated (GiB)": 72.94, + "ppl": 1.26756, + "step": 543, + "tokens/total": 134036928, + "tokens/train_per_sec_per_gpu": 13.16, + "tokens/trainable": 1075982 + }, + { + "epoch": 1.110204081632653, + "grad_norm": 0.24174433946609497, + "learning_rate": 4.365796507521094e-05, + "loss": 0.21796806156635284, + "memory/device_reserved (GiB)": 77.45, + "memory/max_active (GiB)": 75.69, + "memory/max_allocated (GiB)": 75.69, + "ppl": 1.24355, + "step": 544, + "tokens/total": 134307712, + "tokens/train_per_sec_per_gpu": 11.72, + "tokens/trainable": 1078083 + }, + { + "epoch": 1.1122448979591837, + "grad_norm": 0.26919224858283997, + "learning_rate": 4.3494160949381556e-05, + "loss": 0.17404673993587494, + "memory/device_reserved (GiB)": 77.45, + "memory/max_active (GiB)": 72.53, + "memory/max_allocated (GiB)": 72.53, + "ppl": 1.19011, + "step": 545, + "tokens/total": 134565568, + "tokens/train_per_sec_per_gpu": 11.1, + "tokens/trainable": 1079936 + }, + { + "epoch": 1.1142857142857143, + "grad_norm": 0.2523377239704132, + "learning_rate": 4.333042782081926e-05, + "loss": 0.20959612727165222, + "memory/device_reserved (GiB)": 77.45, + "memory/max_active (GiB)": 73.41, + "memory/max_allocated (GiB)": 73.41, + "ppl": 1.23318, + "step": 546, + "tokens/total": 134815424, + "tokens/train_per_sec_per_gpu": 9.77, + "tokens/trainable": 1081763 + }, + { + "epoch": 1.116326530612245, + "grad_norm": 0.23157450556755066, + "learning_rate": 4.3166767476319624e-05, + "loss": 0.20231910049915314, + "memory/device_reserved (GiB)": 77.75, + "memory/max_active (GiB)": 76.73, + "memory/max_allocated (GiB)": 76.73, + "ppl": 1.22424, + "step": 547, + "tokens/total": 135062464, + "tokens/train_per_sec_per_gpu": 11.54, + "tokens/trainable": 1083758 + }, + { + "epoch": 1.1183673469387756, + "grad_norm": 0.3401866853237152, + "learning_rate": 4.300318170188399e-05, + "loss": 0.21560195088386536, + "memory/device_reserved (GiB)": 77.75, + "memory/max_active (GiB)": 73.65, + "memory/max_allocated (GiB)": 73.65, + "ppl": 1.24061, + "step": 548, + "tokens/total": 135307712, + "tokens/train_per_sec_per_gpu": 11.13, + "tokens/trainable": 1085536 + }, + { + "epoch": 1.120408163265306, + "grad_norm": 0.25663185119628906, + "learning_rate": 4.283967228269992e-05, + "loss": 0.2657681703567505, + "memory/device_reserved (GiB)": 77.75, + "memory/max_active (GiB)": 76.48, + "memory/max_allocated (GiB)": 76.48, + "ppl": 1.30443, + "step": 549, + "tokens/total": 135584768, + "tokens/train_per_sec_per_gpu": 11.15, + "tokens/trainable": 1087718 + }, + { + "epoch": 1.1224489795918366, + "grad_norm": 0.256232351064682, + "learning_rate": 4.2676241003121725e-05, + "loss": 0.20277030766010284, + "memory/device_reserved (GiB)": 77.75, + "memory/max_active (GiB)": 75.28, + "memory/max_allocated (GiB)": 75.28, + "ppl": 1.22479, + "step": 550, + "tokens/total": 135811520, + "tokens/train_per_sec_per_gpu": 12.1, + "tokens/trainable": 1089512 + }, + { + "epoch": 1.1244897959183673, + "grad_norm": 0.29823067784309387, + "learning_rate": 4.251288964665098e-05, + "loss": 0.2368815541267395, + "memory/device_reserved (GiB)": 77.75, + "memory/max_active (GiB)": 75.5, + "memory/max_allocated (GiB)": 75.5, + "ppl": 1.26729, + "step": 551, + "tokens/total": 136041088, + "tokens/train_per_sec_per_gpu": 10.23, + "tokens/trainable": 1091423 + }, + { + "epoch": 1.126530612244898, + "grad_norm": 0.28202635049819946, + "learning_rate": 4.234961999591706e-05, + "loss": 0.2961483597755432, + "memory/device_reserved (GiB)": 76.47, + "memory/max_active (GiB)": 73.68, + "memory/max_allocated (GiB)": 73.68, + "ppl": 1.34467, + "step": 552, + "tokens/total": 136275776, + "tokens/train_per_sec_per_gpu": 12.53, + "tokens/trainable": 1093406 + }, + { + "epoch": 1.1285714285714286, + "grad_norm": 0.2742115557193756, + "learning_rate": 4.218643383265771e-05, + "loss": 0.28198349475860596, + "memory/device_reserved (GiB)": 76.47, + "memory/max_active (GiB)": 74.28, + "memory/max_allocated (GiB)": 74.28, + "ppl": 1.32576, + "step": 553, + "tokens/total": 136531904, + "tokens/train_per_sec_per_gpu": 14.66, + "tokens/trainable": 1095695 + }, + { + "epoch": 1.1306122448979592, + "grad_norm": 0.2290259450674057, + "learning_rate": 4.202333293769959e-05, + "loss": 0.2121150642633438, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 76.65, + "memory/max_allocated (GiB)": 76.65, + "ppl": 1.23629, + "step": 554, + "tokens/total": 136812672, + "tokens/train_per_sec_per_gpu": 10.85, + "tokens/trainable": 1097777 + }, + { + "epoch": 1.1326530612244898, + "grad_norm": 0.3228800892829895, + "learning_rate": 4.186031909093884e-05, + "loss": 0.2622196078300476, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 74.49, + "memory/max_allocated (GiB)": 74.49, + "ppl": 1.29981, + "step": 555, + "tokens/total": 137069184, + "tokens/train_per_sec_per_gpu": 12.82, + "tokens/trainable": 1099760 + }, + { + "epoch": 1.1346938775510205, + "grad_norm": 0.26782989501953125, + "learning_rate": 4.1697394071321674e-05, + "loss": 0.2277393639087677, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 75.01, + "memory/max_allocated (GiB)": 75.01, + "ppl": 1.25576, + "step": 556, + "tokens/total": 137327488, + "tokens/train_per_sec_per_gpu": 11.36, + "tokens/trainable": 1101664 + }, + { + "epoch": 1.136734693877551, + "grad_norm": 0.32837456464767456, + "learning_rate": 4.1534559656824884e-05, + "loss": 0.2760929465293884, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 75.07, + "memory/max_allocated (GiB)": 75.07, + "ppl": 1.31797, + "step": 557, + "tokens/total": 137583488, + "tokens/train_per_sec_per_gpu": 11.85, + "tokens/trainable": 1103549 + }, + { + "epoch": 1.1387755102040815, + "grad_norm": 0.24754345417022705, + "learning_rate": 4.1371817624436574e-05, + "loss": 0.23856747150421143, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 75.28, + "memory/max_allocated (GiB)": 75.28, + "ppl": 1.26943, + "step": 558, + "tokens/total": 137861696, + "tokens/train_per_sec_per_gpu": 11.08, + "tokens/trainable": 1105655 + }, + { + "epoch": 1.1408163265306122, + "grad_norm": 0.23945318162441254, + "learning_rate": 4.1209169750136675e-05, + "loss": 0.2171681970357895, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 75.77, + "memory/max_allocated (GiB)": 75.77, + "ppl": 1.24255, + "step": 559, + "tokens/total": 138095552, + "tokens/train_per_sec_per_gpu": 12.87, + "tokens/trainable": 1107799 + }, + { + "epoch": 1.1428571428571428, + "grad_norm": 0.27181726694107056, + "learning_rate": 4.104661780887755e-05, + "loss": 0.2639198303222656, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 72.81, + "memory/max_allocated (GiB)": 72.81, + "ppl": 1.30202, + "step": 560, + "tokens/total": 138342272, + "tokens/train_per_sec_per_gpu": 13.07, + "tokens/trainable": 1109843 + }, + { + "epoch": 1.1448979591836734, + "grad_norm": 0.31380489468574524, + "learning_rate": 4.088416357456471e-05, + "loss": 0.2358066290616989, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 74.66, + "memory/max_allocated (GiB)": 74.66, + "ppl": 1.26593, + "step": 561, + "tokens/total": 138597248, + "tokens/train_per_sec_per_gpu": 14.1, + "tokens/trainable": 1112009 + }, + { + "epoch": 1.146938775510204, + "grad_norm": 0.2777253985404968, + "learning_rate": 4.072180882003735e-05, + "loss": 0.22062432765960693, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 75.53, + "memory/max_allocated (GiB)": 75.53, + "ppl": 1.24685, + "step": 562, + "tokens/total": 138868224, + "tokens/train_per_sec_per_gpu": 11.77, + "tokens/trainable": 1114095 + }, + { + "epoch": 1.1489795918367347, + "grad_norm": 0.26274609565734863, + "learning_rate": 4.055955531704912e-05, + "loss": 0.18324688076972961, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 73.46, + "memory/max_allocated (GiB)": 73.46, + "ppl": 1.20111, + "step": 563, + "tokens/total": 139090368, + "tokens/train_per_sec_per_gpu": 10.65, + "tokens/trainable": 1115690 + }, + { + "epoch": 1.1510204081632653, + "grad_norm": 0.21677400171756744, + "learning_rate": 4.0397404836248684e-05, + "loss": 0.17965775728225708, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 73.76, + "memory/max_allocated (GiB)": 73.76, + "ppl": 1.19681, + "step": 564, + "tokens/total": 139344000, + "tokens/train_per_sec_per_gpu": 10.36, + "tokens/trainable": 1117422 + }, + { + "epoch": 1.153061224489796, + "grad_norm": 0.28004777431488037, + "learning_rate": 4.023535914716042e-05, + "loss": 0.27366122603416443, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 72.75, + "memory/max_allocated (GiB)": 72.75, + "ppl": 1.31477, + "step": 565, + "tokens/total": 139563008, + "tokens/train_per_sec_per_gpu": 19.1, + "tokens/trainable": 1119744 + }, + { + "epoch": 1.1551020408163266, + "grad_norm": 0.32491323351860046, + "learning_rate": 4.00734200181652e-05, + "loss": 0.25971972942352295, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 75.99, + "memory/max_allocated (GiB)": 75.99, + "ppl": 1.29657, + "step": 566, + "tokens/total": 139785344, + "tokens/train_per_sec_per_gpu": 13.7, + "tokens/trainable": 1121809 + }, + { + "epoch": 1.157142857142857, + "grad_norm": 0.2867746651172638, + "learning_rate": 3.991158921648096e-05, + "loss": 0.2492719143629074, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 76.65, + "memory/max_allocated (GiB)": 76.65, + "ppl": 1.28309, + "step": 567, + "tokens/total": 140050880, + "tokens/train_per_sec_per_gpu": 10.76, + "tokens/trainable": 1123838 + }, + { + "epoch": 1.1591836734693877, + "grad_norm": 0.3497733473777771, + "learning_rate": 3.974986850814352e-05, + "loss": 0.28724974393844604, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 74.93, + "memory/max_allocated (GiB)": 74.93, + "ppl": 1.33276, + "step": 568, + "tokens/total": 140296192, + "tokens/train_per_sec_per_gpu": 12.77, + "tokens/trainable": 1125948 + }, + { + "epoch": 1.1612244897959183, + "grad_norm": 0.41115349531173706, + "learning_rate": 3.95882596579872e-05, + "loss": 0.2272345870733261, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 73.54, + "memory/max_allocated (GiB)": 73.54, + "ppl": 1.25512, + "step": 569, + "tokens/total": 140532864, + "tokens/train_per_sec_per_gpu": 13.27, + "tokens/trainable": 1128041 + }, + { + "epoch": 1.163265306122449, + "grad_norm": 0.3145340085029602, + "learning_rate": 3.94267644296257e-05, + "loss": 0.22401602566242218, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 75.01, + "memory/max_allocated (GiB)": 75.01, + "ppl": 1.25109, + "step": 570, + "tokens/total": 140783744, + "tokens/train_per_sec_per_gpu": 11.88, + "tokens/trainable": 1129943 + }, + { + "epoch": 1.1653061224489796, + "grad_norm": 0.2436593770980835, + "learning_rate": 3.926538458543275e-05, + "loss": 0.2128709852695465, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 72.53, + "memory/max_allocated (GiB)": 72.53, + "ppl": 1.23723, + "step": 571, + "tokens/total": 141002688, + "tokens/train_per_sec_per_gpu": 14.24, + "tokens/trainable": 1131939 + }, + { + "epoch": 1.1673469387755102, + "grad_norm": 0.26945507526397705, + "learning_rate": 3.910412188652291e-05, + "loss": 0.2309986799955368, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 75.12, + "memory/max_allocated (GiB)": 75.12, + "ppl": 1.25986, + "step": 572, + "tokens/total": 141257728, + "tokens/train_per_sec_per_gpu": 12.08, + "tokens/trainable": 1133781 + }, + { + "epoch": 1.1693877551020408, + "grad_norm": 0.304496705532074, + "learning_rate": 3.894297809273237e-05, + "loss": 0.22200755774974823, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 74.22, + "memory/max_allocated (GiB)": 74.22, + "ppl": 1.24858, + "step": 573, + "tokens/total": 141491072, + "tokens/train_per_sec_per_gpu": 13.24, + "tokens/trainable": 1135751 + }, + { + "epoch": 1.1714285714285715, + "grad_norm": 0.2798304557800293, + "learning_rate": 3.878195496259967e-05, + "loss": 0.26520785689353943, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 75.12, + "memory/max_allocated (GiB)": 75.12, + "ppl": 1.3037, + "step": 574, + "tokens/total": 141758720, + "tokens/train_per_sec_per_gpu": 12.59, + "tokens/trainable": 1137846 + }, + { + "epoch": 1.1734693877551021, + "grad_norm": 0.2676425278186798, + "learning_rate": 3.862105425334664e-05, + "loss": 0.2218177169561386, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 76.54, + "memory/max_allocated (GiB)": 76.54, + "ppl": 1.24834, + "step": 575, + "tokens/total": 141998464, + "tokens/train_per_sec_per_gpu": 11.51, + "tokens/trainable": 1139692 + }, + { + "epoch": 1.1755102040816325, + "grad_norm": 0.217097669839859, + "learning_rate": 3.8460277720859115e-05, + "loss": 0.16239887475967407, + "memory/device_reserved (GiB)": 77.55, + "memory/max_active (GiB)": 75.83, + "memory/max_allocated (GiB)": 75.83, + "ppl": 1.17633, + "step": 576, + "tokens/total": 142244352, + "tokens/train_per_sec_per_gpu": 8.79, + "tokens/trainable": 1141577 + }, + { + "epoch": 1.1775510204081632, + "grad_norm": 0.21695883572101593, + "learning_rate": 3.8299627119667794e-05, + "loss": 0.16566182672977448, + "memory/device_reserved (GiB)": 76.8, + "memory/max_active (GiB)": 75.72, + "memory/max_allocated (GiB)": 75.72, + "ppl": 1.18017, + "step": 577, + "tokens/total": 142494144, + "tokens/train_per_sec_per_gpu": 10.3, + "tokens/trainable": 1143365 + }, + { + "epoch": 1.1795918367346938, + "grad_norm": 0.2727644741535187, + "learning_rate": 3.8139104202929104e-05, + "loss": 0.21703991293907166, + "memory/device_reserved (GiB)": 76.8, + "memory/max_active (GiB)": 75.28, + "memory/max_allocated (GiB)": 75.28, + "ppl": 1.24239, + "step": 578, + "tokens/total": 142735232, + "tokens/train_per_sec_per_gpu": 13.21, + "tokens/trainable": 1145370 + }, + { + "epoch": 1.1816326530612244, + "grad_norm": 0.2928217053413391, + "learning_rate": 3.797871072240611e-05, + "loss": 0.25232720375061035, + "memory/device_reserved (GiB)": 77.33, + "memory/max_active (GiB)": 76.18, + "memory/max_allocated (GiB)": 76.18, + "ppl": 1.28702, + "step": 579, + "tokens/total": 142994112, + "tokens/train_per_sec_per_gpu": 12.21, + "tokens/trainable": 1147339 + }, + { + "epoch": 1.183673469387755, + "grad_norm": 0.2661975920200348, + "learning_rate": 3.7818448428449324e-05, + "loss": 0.24778707325458527, + "memory/device_reserved (GiB)": 77.33, + "memory/max_active (GiB)": 72.45, + "memory/max_allocated (GiB)": 72.45, + "ppl": 1.28119, + "step": 580, + "tokens/total": 143236992, + "tokens/train_per_sec_per_gpu": 13.06, + "tokens/trainable": 1149347 + }, + { + "epoch": 1.1857142857142857, + "grad_norm": 0.31016600131988525, + "learning_rate": 3.765831906997765e-05, + "loss": 0.2796646058559418, + "memory/device_reserved (GiB)": 77.33, + "memory/max_active (GiB)": 75.56, + "memory/max_allocated (GiB)": 75.56, + "ppl": 1.32269, + "step": 581, + "tokens/total": 143458816, + "tokens/train_per_sec_per_gpu": 16.4, + "tokens/trainable": 1151496 + }, + { + "epoch": 1.1877551020408164, + "grad_norm": 0.4069809317588806, + "learning_rate": 3.749832439445925e-05, + "loss": 0.26152700185775757, + "memory/device_reserved (GiB)": 77.68, + "memory/max_active (GiB)": 76.78, + "memory/max_allocated (GiB)": 76.78, + "ppl": 1.29891, + "step": 582, + "tokens/total": 143711040, + "tokens/train_per_sec_per_gpu": 10.22, + "tokens/trainable": 1153402 + }, + { + "epoch": 1.189795918367347, + "grad_norm": 0.2766212522983551, + "learning_rate": 3.7338466147892605e-05, + "loss": 0.1950775682926178, + "memory/device_reserved (GiB)": 77.68, + "memory/max_active (GiB)": 73.62, + "memory/max_allocated (GiB)": 73.62, + "ppl": 1.21541, + "step": 583, + "tokens/total": 143945088, + "tokens/train_per_sec_per_gpu": 14.36, + "tokens/trainable": 1155443 + }, + { + "epoch": 1.1918367346938776, + "grad_norm": 0.20926155149936676, + "learning_rate": 3.7178746074787266e-05, + "loss": 0.16814126074314117, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 74.49, + "memory/max_allocated (GiB)": 74.49, + "ppl": 1.1831, + "step": 584, + "tokens/total": 144180864, + "tokens/train_per_sec_per_gpu": 12.26, + "tokens/trainable": 1157320 + }, + { + "epoch": 1.193877551020408, + "grad_norm": 0.27808478474617004, + "learning_rate": 3.701916591814497e-05, + "loss": 0.2127000242471695, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 74.68, + "memory/max_allocated (GiB)": 74.68, + "ppl": 1.23701, + "step": 585, + "tokens/total": 144407616, + "tokens/train_per_sec_per_gpu": 13.83, + "tokens/trainable": 1159287 + }, + { + "epoch": 1.1959183673469387, + "grad_norm": 0.27595117688179016, + "learning_rate": 3.685972741944057e-05, + "loss": 0.2882786691188812, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 75.91, + "memory/max_allocated (GiB)": 75.91, + "ppl": 1.33413, + "step": 586, + "tokens/total": 144669824, + "tokens/train_per_sec_per_gpu": 12.35, + "tokens/trainable": 1161554 + }, + { + "epoch": 1.1979591836734693, + "grad_norm": 0.2914755046367645, + "learning_rate": 3.6700432318602965e-05, + "loss": 0.23543119430541992, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 73.08, + "memory/max_allocated (GiB)": 73.08, + "ppl": 1.26545, + "step": 587, + "tokens/total": 144902208, + "tokens/train_per_sec_per_gpu": 12.73, + "tokens/trainable": 1163368 + }, + { + "epoch": 1.2, + "grad_norm": 0.2846899926662445, + "learning_rate": 3.654128235399628e-05, + "loss": 0.2750670611858368, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 73.54, + "memory/max_allocated (GiB)": 73.54, + "ppl": 1.31662, + "step": 588, + "tokens/total": 145133760, + "tokens/train_per_sec_per_gpu": 12.99, + "tokens/trainable": 1165385 + }, + { + "epoch": 1.2020408163265306, + "grad_norm": 0.25691884756088257, + "learning_rate": 3.6382279262400665e-05, + "loss": 0.22056685388088226, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 72.92, + "memory/max_allocated (GiB)": 72.92, + "ppl": 1.24678, + "step": 589, + "tokens/total": 145372736, + "tokens/train_per_sec_per_gpu": 11.53, + "tokens/trainable": 1167408 + }, + { + "epoch": 1.2040816326530612, + "grad_norm": 0.2667784094810486, + "learning_rate": 3.622342477899358e-05, + "loss": 0.23244816064834595, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 74.11, + "memory/max_allocated (GiB)": 74.11, + "ppl": 1.26169, + "step": 590, + "tokens/total": 145618176, + "tokens/train_per_sec_per_gpu": 13.62, + "tokens/trainable": 1169425 + }, + { + "epoch": 1.2061224489795919, + "grad_norm": 0.20627418160438538, + "learning_rate": 3.606472063733067e-05, + "loss": 0.14769814908504486, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 76.62, + "memory/max_allocated (GiB)": 76.62, + "ppl": 1.15916, + "step": 591, + "tokens/total": 145860672, + "tokens/train_per_sec_per_gpu": 10.59, + "tokens/trainable": 1171231 + }, + { + "epoch": 1.2081632653061225, + "grad_norm": 0.23447421193122864, + "learning_rate": 3.590616856932697e-05, + "loss": 0.19406414031982422, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 76.51, + "memory/max_allocated (GiB)": 76.51, + "ppl": 1.21417, + "step": 592, + "tokens/total": 146099968, + "tokens/train_per_sec_per_gpu": 12.48, + "tokens/trainable": 1173366 + }, + { + "epoch": 1.2102040816326531, + "grad_norm": 0.2896268367767334, + "learning_rate": 3.574777030523791e-05, + "loss": 0.26103439927101135, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 76.37, + "memory/max_allocated (GiB)": 76.37, + "ppl": 1.29827, + "step": 593, + "tokens/total": 146342848, + "tokens/train_per_sec_per_gpu": 13.25, + "tokens/trainable": 1175393 + }, + { + "epoch": 1.2122448979591836, + "grad_norm": 0.2959030866622925, + "learning_rate": 3.558952757364054e-05, + "loss": 0.2680799663066864, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 75.04, + "memory/max_allocated (GiB)": 75.04, + "ppl": 1.30745, + "step": 594, + "tokens/total": 146590400, + "tokens/train_per_sec_per_gpu": 12.98, + "tokens/trainable": 1177580 + }, + { + "epoch": 1.2142857142857142, + "grad_norm": 0.2562417685985565, + "learning_rate": 3.543144210141454e-05, + "loss": 0.22213822603225708, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 71.85, + "memory/max_allocated (GiB)": 71.85, + "ppl": 1.24874, + "step": 595, + "tokens/total": 146817216, + "tokens/train_per_sec_per_gpu": 13.01, + "tokens/trainable": 1179498 + }, + { + "epoch": 1.2163265306122448, + "grad_norm": 0.2862449884414673, + "learning_rate": 3.5273515613723503e-05, + "loss": 0.22280345857143402, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 76.43, + "memory/max_allocated (GiB)": 76.43, + "ppl": 1.24957, + "step": 596, + "tokens/total": 147047296, + "tokens/train_per_sec_per_gpu": 12.99, + "tokens/trainable": 1181390 + }, + { + "epoch": 1.2183673469387755, + "grad_norm": 0.224739670753479, + "learning_rate": 3.511574983399599e-05, + "loss": 0.17955514788627625, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 76.05, + "memory/max_allocated (GiB)": 76.05, + "ppl": 1.19668, + "step": 597, + "tokens/total": 147315840, + "tokens/train_per_sec_per_gpu": 11.56, + "tokens/trainable": 1183391 + }, + { + "epoch": 1.220408163265306, + "grad_norm": 0.4748108983039856, + "learning_rate": 3.495814648390684e-05, + "loss": 0.20630517601966858, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 75.56, + "memory/max_allocated (GiB)": 75.56, + "ppl": 1.22913, + "step": 598, + "tokens/total": 147590208, + "tokens/train_per_sec_per_gpu": 12.83, + "tokens/trainable": 1185669 + }, + { + "epoch": 1.2224489795918367, + "grad_norm": 0.28958484530448914, + "learning_rate": 3.480070728335825e-05, + "loss": 0.25522685050964355, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 75.2, + "memory/max_allocated (GiB)": 75.2, + "ppl": 1.29075, + "step": 599, + "tokens/total": 147862528, + "tokens/train_per_sec_per_gpu": 12.66, + "tokens/trainable": 1187855 + }, + { + "epoch": 1.2244897959183674, + "grad_norm": 0.4274929463863373, + "learning_rate": 3.464343395046118e-05, + "loss": 0.2889910936355591, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 74.41, + "memory/max_allocated (GiB)": 74.41, + "ppl": 1.33508, + "step": 600, + "tokens/total": 148102016, + "tokens/train_per_sec_per_gpu": 11.94, + "tokens/trainable": 1189789 + }, + { + "epoch": 1.226530612244898, + "grad_norm": 0.2663540840148926, + "learning_rate": 3.4486328201516374e-05, + "loss": 0.20353618264198303, + "memory/device_reserved (GiB)": 77.69, + "memory/max_active (GiB)": 72.29, + "memory/max_allocated (GiB)": 72.29, + "ppl": 1.22573, + "step": 601, + "tokens/total": 148338560, + "tokens/train_per_sec_per_gpu": 9.19, + "tokens/trainable": 1191771 + }, + { + "epoch": 1.2285714285714286, + "grad_norm": 0.26682329177856445, + "learning_rate": 3.4329391750995856e-05, + "loss": 0.2231263816356659, + "memory/device_reserved (GiB)": 77.29, + "memory/max_active (GiB)": 76.45, + "memory/max_allocated (GiB)": 76.45, + "ppl": 1.24998, + "step": 602, + "tokens/total": 148592640, + "tokens/train_per_sec_per_gpu": 16.22, + "tokens/trainable": 1194404 + }, + { + "epoch": 1.230612244897959, + "grad_norm": 0.24795755743980408, + "learning_rate": 3.417262631152409e-05, + "loss": 0.2027260661125183, + "memory/device_reserved (GiB)": 77.29, + "memory/max_active (GiB)": 76.07, + "memory/max_allocated (GiB)": 76.07, + "ppl": 1.22474, + "step": 603, + "tokens/total": 148835648, + "tokens/train_per_sec_per_gpu": 11.52, + "tokens/trainable": 1196291 + }, + { + "epoch": 1.2326530612244897, + "grad_norm": 0.2479274868965149, + "learning_rate": 3.401603359385929e-05, + "loss": 0.25008320808410645, + "memory/device_reserved (GiB)": 77.29, + "memory/max_active (GiB)": 75.96, + "memory/max_allocated (GiB)": 75.96, + "ppl": 1.28413, + "step": 604, + "tokens/total": 149085120, + "tokens/train_per_sec_per_gpu": 12.65, + "tokens/trainable": 1198617 + }, + { + "epoch": 1.2346938775510203, + "grad_norm": 0.27939435839653015, + "learning_rate": 3.3859615306874824e-05, + "loss": 0.26484474539756775, + "memory/device_reserved (GiB)": 77.29, + "memory/max_active (GiB)": 74.58, + "memory/max_allocated (GiB)": 74.58, + "ppl": 1.30323, + "step": 605, + "tokens/total": 149340736, + "tokens/train_per_sec_per_gpu": 12.93, + "tokens/trainable": 1200602 + }, + { + "epoch": 1.236734693877551, + "grad_norm": 0.2564728856086731, + "learning_rate": 3.3703373157540525e-05, + "loss": 0.21744893491268158, + "memory/device_reserved (GiB)": 77.29, + "memory/max_active (GiB)": 73.81, + "memory/max_allocated (GiB)": 73.81, + "ppl": 1.2429, + "step": 606, + "tokens/total": 149593152, + "tokens/train_per_sec_per_gpu": 12.69, + "tokens/trainable": 1202522 + }, + { + "epoch": 1.2387755102040816, + "grad_norm": 0.2831055819988251, + "learning_rate": 3.3547308850904015e-05, + "loss": 0.20425094664096832, + "memory/device_reserved (GiB)": 77.29, + "memory/max_active (GiB)": 74.85, + "memory/max_allocated (GiB)": 74.85, + "ppl": 1.22661, + "step": 607, + "tokens/total": 149825536, + "tokens/train_per_sec_per_gpu": 10.48, + "tokens/trainable": 1204200 + }, + { + "epoch": 1.2408163265306122, + "grad_norm": 0.25006458163261414, + "learning_rate": 3.339142409007221e-05, + "loss": 0.18129464983940125, + "memory/device_reserved (GiB)": 77.29, + "memory/max_active (GiB)": 76.16, + "memory/max_allocated (GiB)": 76.16, + "ppl": 1.19877, + "step": 608, + "tokens/total": 150078400, + "tokens/train_per_sec_per_gpu": 11.6, + "tokens/trainable": 1206089 + }, + { + "epoch": 1.2428571428571429, + "grad_norm": 0.27177634835243225, + "learning_rate": 3.3235720576192596e-05, + "loss": 0.21773484349250793, + "memory/device_reserved (GiB)": 77.29, + "memory/max_active (GiB)": 72.64, + "memory/max_allocated (GiB)": 72.64, + "ppl": 1.24326, + "step": 609, + "tokens/total": 150285056, + "tokens/train_per_sec_per_gpu": 16.5, + "tokens/trainable": 1207977 + }, + { + "epoch": 1.2448979591836735, + "grad_norm": 0.2669946849346161, + "learning_rate": 3.3080200008434804e-05, + "loss": 0.23606613278388977, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 76.43, + "memory/max_allocated (GiB)": 76.43, + "ppl": 1.26626, + "step": 610, + "tokens/total": 150553216, + "tokens/train_per_sec_per_gpu": 11.75, + "tokens/trainable": 1210024 + }, + { + "epoch": 1.2469387755102042, + "grad_norm": 0.2924709916114807, + "learning_rate": 3.2924864083971975e-05, + "loss": 0.24505622684955597, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 75.04, + "memory/max_allocated (GiB)": 75.04, + "ppl": 1.27769, + "step": 611, + "tokens/total": 150809600, + "tokens/train_per_sec_per_gpu": 11.65, + "tokens/trainable": 1211818 + }, + { + "epoch": 1.2489795918367346, + "grad_norm": 0.3352468013763428, + "learning_rate": 3.276971449796224e-05, + "loss": 0.291441947221756, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 74.36, + "memory/max_allocated (GiB)": 74.36, + "ppl": 1.33836, + "step": 612, + "tokens/total": 151047424, + "tokens/train_per_sec_per_gpu": 11.4, + "tokens/trainable": 1213661 + }, + { + "epoch": 1.2510204081632654, + "grad_norm": 0.2645813822746277, + "learning_rate": 3.261475294353028e-05, + "loss": 0.18487977981567383, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 74.11, + "memory/max_allocated (GiB)": 74.11, + "ppl": 1.20307, + "step": 613, + "tokens/total": 151264512, + "tokens/train_per_sec_per_gpu": 12.27, + "tokens/trainable": 1215443 + }, + { + "epoch": 1.2530612244897958, + "grad_norm": 0.21285974979400635, + "learning_rate": 3.245998111174882e-05, + "loss": 0.17316727340221405, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 75.31, + "memory/max_allocated (GiB)": 75.31, + "ppl": 1.18906, + "step": 614, + "tokens/total": 151545344, + "tokens/train_per_sec_per_gpu": 11.54, + "tokens/trainable": 1217513 + }, + { + "epoch": 1.2551020408163265, + "grad_norm": 0.2863808870315552, + "learning_rate": 3.2305400691620125e-05, + "loss": 0.2364487648010254, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 74.49, + "memory/max_allocated (GiB)": 74.49, + "ppl": 1.26674, + "step": 615, + "tokens/total": 151808640, + "tokens/train_per_sec_per_gpu": 12.56, + "tokens/trainable": 1219818 + }, + { + "epoch": 1.2571428571428571, + "grad_norm": 0.28447479009628296, + "learning_rate": 3.215101337005765e-05, + "loss": 0.2561761736869812, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 74.28, + "memory/max_allocated (GiB)": 74.28, + "ppl": 1.29198, + "step": 616, + "tokens/total": 152055296, + "tokens/train_per_sec_per_gpu": 15.03, + "tokens/trainable": 1222014 + }, + { + "epoch": 1.2591836734693878, + "grad_norm": 0.2917903661727905, + "learning_rate": 3.199682083186759e-05, + "loss": 0.3035793900489807, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 73.0, + "memory/max_allocated (GiB)": 73.0, + "ppl": 1.3547, + "step": 617, + "tokens/total": 152284672, + "tokens/train_per_sec_per_gpu": 14.91, + "tokens/trainable": 1224226 + }, + { + "epoch": 1.2612244897959184, + "grad_norm": 0.23619985580444336, + "learning_rate": 3.1842824759730513e-05, + "loss": 0.18813863396644592, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 75.91, + "memory/max_allocated (GiB)": 75.91, + "ppl": 1.207, + "step": 618, + "tokens/total": 152545856, + "tokens/train_per_sec_per_gpu": 12.64, + "tokens/trainable": 1226341 + }, + { + "epoch": 1.263265306122449, + "grad_norm": 0.35663172602653503, + "learning_rate": 3.168902683418292e-05, + "loss": 0.261131227016449, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 71.99, + "memory/max_allocated (GiB)": 71.99, + "ppl": 1.2984, + "step": 619, + "tokens/total": 152784896, + "tokens/train_per_sec_per_gpu": 12.57, + "tokens/trainable": 1228267 + }, + { + "epoch": 1.2653061224489797, + "grad_norm": 0.34750106930732727, + "learning_rate": 3.153542873359904e-05, + "loss": 0.28193962574005127, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 73.76, + "memory/max_allocated (GiB)": 73.76, + "ppl": 1.3257, + "step": 620, + "tokens/total": 153017984, + "tokens/train_per_sec_per_gpu": 13.63, + "tokens/trainable": 1230100 + }, + { + "epoch": 1.26734693877551, + "grad_norm": 0.296609491109848, + "learning_rate": 3.13820321341724e-05, + "loss": 0.28097981214523315, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 71.96, + "memory/max_allocated (GiB)": 71.96, + "ppl": 1.32443, + "step": 621, + "tokens/total": 153261888, + "tokens/train_per_sec_per_gpu": 15.19, + "tokens/trainable": 1232372 + }, + { + "epoch": 1.269387755102041, + "grad_norm": 0.2779256999492645, + "learning_rate": 3.1228838709897567e-05, + "loss": 0.21719186007976532, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 71.96, + "memory/max_allocated (GiB)": 71.96, + "ppl": 1.24258, + "step": 622, + "tokens/total": 153507264, + "tokens/train_per_sec_per_gpu": 12.66, + "tokens/trainable": 1234368 + }, + { + "epoch": 1.2714285714285714, + "grad_norm": 0.26468080282211304, + "learning_rate": 3.107585013255194e-05, + "loss": 0.19213882088661194, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 73.49, + "memory/max_allocated (GiB)": 73.49, + "ppl": 1.21184, + "step": 623, + "tokens/total": 153749312, + "tokens/train_per_sec_per_gpu": 13.02, + "tokens/trainable": 1236487 + }, + { + "epoch": 1.273469387755102, + "grad_norm": 0.33944669365882874, + "learning_rate": 3.092306807167738e-05, + "loss": 0.20130395889282227, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 76.43, + "memory/max_allocated (GiB)": 76.43, + "ppl": 1.223, + "step": 624, + "tokens/total": 154038080, + "tokens/train_per_sec_per_gpu": 10.86, + "tokens/trainable": 1238427 + }, + { + "epoch": 1.2755102040816326, + "grad_norm": 0.2875191867351532, + "learning_rate": 3.077049419456212e-05, + "loss": 0.21526801586151123, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 75.64, + "memory/max_allocated (GiB)": 75.64, + "ppl": 1.24019, + "step": 625, + "tokens/total": 154299136, + "tokens/train_per_sec_per_gpu": 13.62, + "tokens/trainable": 1240636 + }, + { + "epoch": 1.2775510204081633, + "grad_norm": 0.3573671579360962, + "learning_rate": 3.061813016622251e-05, + "loss": 0.20149338245391846, + "memory/device_reserved (GiB)": 77.65, + "memory/max_active (GiB)": 75.04, + "memory/max_allocated (GiB)": 75.04, + "ppl": 1.22323, + "step": 626, + "tokens/total": 154545408, + "tokens/train_per_sec_per_gpu": 11.57, + "tokens/trainable": 1242889 + }, + { + "epoch": 1.279591836734694, + "grad_norm": 0.26556330919265747, + "learning_rate": 3.0465977649384812e-05, + "loss": 0.2197554111480713, + "memory/device_reserved (GiB)": 75.93, + "memory/max_active (GiB)": 74.41, + "memory/max_allocated (GiB)": 74.41, + "ppl": 1.24577, + "step": 627, + "tokens/total": 154810240, + "tokens/train_per_sec_per_gpu": 12.22, + "tokens/trainable": 1245046 + }, + { + "epoch": 1.2816326530612245, + "grad_norm": 0.34495222568511963, + "learning_rate": 3.031403830446713e-05, + "loss": 0.2639291286468506, + "memory/device_reserved (GiB)": 75.93, + "memory/max_active (GiB)": 72.21, + "memory/max_allocated (GiB)": 72.21, + "ppl": 1.30204, + "step": 628, + "tokens/total": 155036544, + "tokens/train_per_sec_per_gpu": 13.03, + "tokens/trainable": 1247065 + }, + { + "epoch": 1.2836734693877552, + "grad_norm": 0.24303406476974487, + "learning_rate": 3.016231378956126e-05, + "loss": 0.18169401586055756, + "memory/device_reserved (GiB)": 77.35, + "memory/max_active (GiB)": 76.51, + "memory/max_allocated (GiB)": 76.51, + "ppl": 1.19925, + "step": 629, + "tokens/total": 155315328, + "tokens/train_per_sec_per_gpu": 9.16, + "tokens/trainable": 1248833 + }, + { + "epoch": 1.2857142857142856, + "grad_norm": 0.26993343234062195, + "learning_rate": 3.0010805760414546e-05, + "loss": 0.2202262282371521, + "memory/device_reserved (GiB)": 77.86, + "memory/max_active (GiB)": 76.73, + "memory/max_allocated (GiB)": 76.73, + "ppl": 1.24636, + "step": 630, + "tokens/total": 155576768, + "tokens/train_per_sec_per_gpu": 11.43, + "tokens/trainable": 1250891 + }, + { + "epoch": 1.2877551020408164, + "grad_norm": 0.26272836327552795, + "learning_rate": 2.9859515870411902e-05, + "loss": 0.1943410485982895, + "memory/device_reserved (GiB)": 77.86, + "memory/max_active (GiB)": 75.58, + "memory/max_allocated (GiB)": 75.58, + "ppl": 1.21451, + "step": 631, + "tokens/total": 155834240, + "tokens/train_per_sec_per_gpu": 12.76, + "tokens/trainable": 1252910 + }, + { + "epoch": 1.2897959183673469, + "grad_norm": 0.2472894787788391, + "learning_rate": 2.9708445770557685e-05, + "loss": 0.2008545845746994, + "memory/device_reserved (GiB)": 77.86, + "memory/max_active (GiB)": 74.88, + "memory/max_allocated (GiB)": 74.88, + "ppl": 1.22245, + "step": 632, + "tokens/total": 156075392, + "tokens/train_per_sec_per_gpu": 9.47, + "tokens/trainable": 1254505 + }, + { + "epoch": 1.2918367346938775, + "grad_norm": 0.36615756154060364, + "learning_rate": 2.955759710945773e-05, + "loss": 0.300974041223526, + "memory/device_reserved (GiB)": 77.86, + "memory/max_active (GiB)": 76.81, + "memory/max_allocated (GiB)": 76.81, + "ppl": 1.35117, + "step": 633, + "tokens/total": 156328832, + "tokens/train_per_sec_per_gpu": 12.96, + "tokens/trainable": 1256633 + }, + { + "epoch": 1.2938775510204081, + "grad_norm": 0.2761453092098236, + "learning_rate": 2.9406971533301365e-05, + "loss": 0.23298825323581696, + "memory/device_reserved (GiB)": 77.86, + "memory/max_active (GiB)": 75.07, + "memory/max_allocated (GiB)": 75.07, + "ppl": 1.26237, + "step": 634, + "tokens/total": 156574848, + "tokens/train_per_sec_per_gpu": 11.78, + "tokens/trainable": 1258644 + }, + { + "epoch": 1.2959183673469388, + "grad_norm": 0.37995588779449463, + "learning_rate": 2.92565706858434e-05, + "loss": 0.2130972146987915, + "memory/device_reserved (GiB)": 77.86, + "memory/max_active (GiB)": 73.22, + "memory/max_allocated (GiB)": 73.22, + "ppl": 1.2375, + "step": 635, + "tokens/total": 156804416, + "tokens/train_per_sec_per_gpu": 10.94, + "tokens/trainable": 1260469 + }, + { + "epoch": 1.2979591836734694, + "grad_norm": 0.28451839089393616, + "learning_rate": 2.9106396208386187e-05, + "loss": 0.21629148721694946, + "memory/device_reserved (GiB)": 77.86, + "memory/max_active (GiB)": 73.54, + "memory/max_allocated (GiB)": 73.54, + "ppl": 1.24146, + "step": 636, + "tokens/total": 157044224, + "tokens/train_per_sec_per_gpu": 12.55, + "tokens/trainable": 1262307 + }, + { + "epoch": 1.3, + "grad_norm": 0.3277578055858612, + "learning_rate": 2.8956449739761815e-05, + "loss": 0.2166963368654251, + "memory/device_reserved (GiB)": 77.86, + "memory/max_active (GiB)": 75.18, + "memory/max_allocated (GiB)": 75.18, + "ppl": 1.24197, + "step": 637, + "tokens/total": 157291776, + "tokens/train_per_sec_per_gpu": 10.51, + "tokens/trainable": 1264036 + }, + { + "epoch": 1.3020408163265307, + "grad_norm": 0.22155879437923431, + "learning_rate": 2.880673291631406e-05, + "loss": 0.16716590523719788, + "memory/device_reserved (GiB)": 77.86, + "memory/max_active (GiB)": 75.2, + "memory/max_allocated (GiB)": 75.2, + "ppl": 1.18195, + "step": 638, + "tokens/total": 157554304, + "tokens/train_per_sec_per_gpu": 11.64, + "tokens/trainable": 1266040 + }, + { + "epoch": 1.304081632653061, + "grad_norm": 0.3041877746582031, + "learning_rate": 2.865724737188067e-05, + "loss": 0.18923434615135193, + "memory/device_reserved (GiB)": 77.86, + "memory/max_active (GiB)": 71.77, + "memory/max_allocated (GiB)": 71.77, + "ppl": 1.20832, + "step": 639, + "tokens/total": 157765120, + "tokens/train_per_sec_per_gpu": 12.45, + "tokens/trainable": 1267803 + }, + { + "epoch": 1.306122448979592, + "grad_norm": 0.2440873682498932, + "learning_rate": 2.8507994737775434e-05, + "loss": 0.20218439400196075, + "memory/device_reserved (GiB)": 77.86, + "memory/max_active (GiB)": 73.49, + "memory/max_allocated (GiB)": 73.49, + "ppl": 1.22407, + "step": 640, + "tokens/total": 158030592, + "tokens/train_per_sec_per_gpu": 10.64, + "tokens/trainable": 1269729 + }, + { + "epoch": 1.3081632653061224, + "grad_norm": 0.2696002423763275, + "learning_rate": 2.8358976642770486e-05, + "loss": 0.21480688452720642, + "memory/device_reserved (GiB)": 77.86, + "memory/max_active (GiB)": 73.16, + "memory/max_allocated (GiB)": 73.16, + "ppl": 1.23962, + "step": 641, + "tokens/total": 158293504, + "tokens/train_per_sec_per_gpu": 11.69, + "tokens/trainable": 1271734 + }, + { + "epoch": 1.310204081632653, + "grad_norm": 0.28075316548347473, + "learning_rate": 2.8210194713078407e-05, + "loss": 0.2665201723575592, + "memory/device_reserved (GiB)": 77.86, + "memory/max_active (GiB)": 72.89, + "memory/max_allocated (GiB)": 72.89, + "ppl": 1.30541, + "step": 642, + "tokens/total": 158555392, + "tokens/train_per_sec_per_gpu": 11.69, + "tokens/trainable": 1273820 + }, + { + "epoch": 1.3122448979591836, + "grad_norm": 0.3203342854976654, + "learning_rate": 2.8061650572334568e-05, + "loss": 0.2859092950820923, + "memory/device_reserved (GiB)": 77.86, + "memory/max_active (GiB)": 74.82, + "memory/max_allocated (GiB)": 74.82, + "ppl": 1.33097, + "step": 643, + "tokens/total": 158787328, + "tokens/train_per_sec_per_gpu": 12.96, + "tokens/trainable": 1275866 + }, + { + "epoch": 1.3142857142857143, + "grad_norm": 0.3039790689945221, + "learning_rate": 2.791334584157941e-05, + "loss": 0.289364218711853, + "memory/device_reserved (GiB)": 77.86, + "memory/max_active (GiB)": 76.59, + "memory/max_allocated (GiB)": 76.59, + "ppl": 1.33558, + "step": 644, + "tokens/total": 159051712, + "tokens/train_per_sec_per_gpu": 13.26, + "tokens/trainable": 1277917 + }, + { + "epoch": 1.316326530612245, + "grad_norm": 0.2710573673248291, + "learning_rate": 2.7765282139240677e-05, + "loss": 0.23384779691696167, + "memory/device_reserved (GiB)": 77.86, + "memory/max_active (GiB)": 75.28, + "memory/max_allocated (GiB)": 75.28, + "ppl": 1.26345, + "step": 645, + "tokens/total": 159294208, + "tokens/train_per_sec_per_gpu": 11.84, + "tokens/trainable": 1279852 + }, + { + "epoch": 1.3183673469387756, + "grad_norm": 0.2765846848487854, + "learning_rate": 2.7617461081115827e-05, + "loss": 0.20561407506465912, + "memory/device_reserved (GiB)": 77.86, + "memory/max_active (GiB)": 72.32, + "memory/max_allocated (GiB)": 72.32, + "ppl": 1.22828, + "step": 646, + "tokens/total": 159547328, + "tokens/train_per_sec_per_gpu": 10.48, + "tokens/trainable": 1281744 + }, + { + "epoch": 1.3204081632653062, + "grad_norm": 0.29625484347343445, + "learning_rate": 2.7469884280354403e-05, + "loss": 0.22659578919410706, + "memory/device_reserved (GiB)": 77.86, + "memory/max_active (GiB)": 74.49, + "memory/max_allocated (GiB)": 74.49, + "ppl": 1.25432, + "step": 647, + "tokens/total": 159774784, + "tokens/train_per_sec_per_gpu": 12.04, + "tokens/trainable": 1283532 + }, + { + "epoch": 1.3224489795918366, + "grad_norm": 0.29532772302627563, + "learning_rate": 2.7322553347440365e-05, + "loss": 0.2580396831035614, + "memory/device_reserved (GiB)": 77.86, + "memory/max_active (GiB)": 74.58, + "memory/max_allocated (GiB)": 74.58, + "ppl": 1.29439, + "step": 648, + "tokens/total": 159993728, + "tokens/train_per_sec_per_gpu": 15.97, + "tokens/trainable": 1285501 + }, + { + "epoch": 1.3244897959183675, + "grad_norm": 0.2813834249973297, + "learning_rate": 2.7175469890174554e-05, + "loss": 0.20868343114852905, + "memory/device_reserved (GiB)": 77.86, + "memory/max_active (GiB)": 73.98, + "memory/max_allocated (GiB)": 73.98, + "ppl": 1.23205, + "step": 649, + "tokens/total": 160241024, + "tokens/train_per_sec_per_gpu": 9.88, + "tokens/trainable": 1287090 + }, + { + "epoch": 1.3265306122448979, + "grad_norm": 0.2811291217803955, + "learning_rate": 2.7028635513657174e-05, + "loss": 0.21979570388793945, + "memory/device_reserved (GiB)": 77.86, + "memory/max_active (GiB)": 75.72, + "memory/max_allocated (GiB)": 75.72, + "ppl": 1.24582, + "step": 650, + "tokens/total": 160476416, + "tokens/train_per_sec_per_gpu": 12.44, + "tokens/trainable": 1289083 + }, + { + "epoch": 1.3285714285714285, + "grad_norm": 0.3985142707824707, + "learning_rate": 2.688205182027026e-05, + "loss": 0.25182634592056274, + "memory/device_reserved (GiB)": 77.86, + "memory/max_active (GiB)": 75.77, + "memory/max_allocated (GiB)": 75.77, + "ppl": 1.28637, + "step": 651, + "tokens/total": 160738944, + "tokens/train_per_sec_per_gpu": 9.82, + "tokens/trainable": 1291235 + }, + { + "epoch": 1.3306122448979592, + "grad_norm": 0.2896854281425476, + "learning_rate": 2.6735720409660126e-05, + "loss": 0.25489988923072815, + "memory/device_reserved (GiB)": 76.67, + "memory/max_active (GiB)": 74.17, + "memory/max_allocated (GiB)": 74.17, + "ppl": 1.29033, + "step": 652, + "tokens/total": 160988096, + "tokens/train_per_sec_per_gpu": 15.33, + "tokens/trainable": 1293598 + }, + { + "epoch": 1.3326530612244898, + "grad_norm": 0.24566292762756348, + "learning_rate": 2.6589642878719967e-05, + "loss": 0.18562713265419006, + "memory/device_reserved (GiB)": 76.67, + "memory/max_active (GiB)": 75.72, + "memory/max_allocated (GiB)": 75.72, + "ppl": 1.20397, + "step": 653, + "tokens/total": 161212928, + "tokens/train_per_sec_per_gpu": 12.97, + "tokens/trainable": 1295558 + }, + { + "epoch": 1.3346938775510204, + "grad_norm": 0.20682737231254578, + "learning_rate": 2.6443820821572496e-05, + "loss": 0.16127759218215942, + "memory/device_reserved (GiB)": 76.67, + "memory/max_active (GiB)": 75.07, + "memory/max_allocated (GiB)": 75.07, + "ppl": 1.17501, + "step": 654, + "tokens/total": 161465792, + "tokens/train_per_sec_per_gpu": 12.05, + "tokens/trainable": 1297561 + }, + { + "epoch": 1.336734693877551, + "grad_norm": 0.2621935307979584, + "learning_rate": 2.6298255829552392e-05, + "loss": 0.18108519911766052, + "memory/device_reserved (GiB)": 77.29, + "memory/max_active (GiB)": 76.26, + "memory/max_allocated (GiB)": 76.26, + "ppl": 1.19852, + "step": 655, + "tokens/total": 161685632, + "tokens/train_per_sec_per_gpu": 12.76, + "tokens/trainable": 1299570 + }, + { + "epoch": 1.3387755102040817, + "grad_norm": 0.29202499985694885, + "learning_rate": 2.615294949118905e-05, + "loss": 0.23991014063358307, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 76.7, + "memory/max_allocated (GiB)": 76.7, + "ppl": 1.27113, + "step": 656, + "tokens/total": 161938304, + "tokens/train_per_sec_per_gpu": 10.47, + "tokens/trainable": 1301396 + }, + { + "epoch": 1.3408163265306121, + "grad_norm": 0.3090575635433197, + "learning_rate": 2.600790339218926e-05, + "loss": 0.24309760332107544, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 73.62, + "memory/max_allocated (GiB)": 73.62, + "ppl": 1.27519, + "step": 657, + "tokens/total": 162193408, + "tokens/train_per_sec_per_gpu": 13.02, + "tokens/trainable": 1303459 + }, + { + "epoch": 1.342857142857143, + "grad_norm": 0.2368037849664688, + "learning_rate": 2.5863119115419798e-05, + "loss": 0.1915193498134613, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 75.96, + "memory/max_allocated (GiB)": 75.96, + "ppl": 1.21109, + "step": 658, + "tokens/total": 162436800, + "tokens/train_per_sec_per_gpu": 15.0, + "tokens/trainable": 1305821 + }, + { + "epoch": 1.3448979591836734, + "grad_norm": 0.22970855236053467, + "learning_rate": 2.5718598240890213e-05, + "loss": 0.15314342081546783, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 76.1, + "memory/max_allocated (GiB)": 76.1, + "ppl": 1.16549, + "step": 659, + "tokens/total": 162673472, + "tokens/train_per_sec_per_gpu": 10.67, + "tokens/trainable": 1307582 + }, + { + "epoch": 1.346938775510204, + "grad_norm": 0.3147336542606354, + "learning_rate": 2.5574342345735652e-05, + "loss": 0.25489136576652527, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 75.66, + "memory/max_allocated (GiB)": 75.66, + "ppl": 1.29032, + "step": 660, + "tokens/total": 162920512, + "tokens/train_per_sec_per_gpu": 11.79, + "tokens/trainable": 1309512 + }, + { + "epoch": 1.3489795918367347, + "grad_norm": 0.2854621410369873, + "learning_rate": 2.543035300419951e-05, + "loss": 0.20861515402793884, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 74.68, + "memory/max_allocated (GiB)": 74.68, + "ppl": 1.23197, + "step": 661, + "tokens/total": 163142336, + "tokens/train_per_sec_per_gpu": 15.28, + "tokens/trainable": 1311474 + }, + { + "epoch": 1.3510204081632653, + "grad_norm": 0.2864938974380493, + "learning_rate": 2.5286631787616345e-05, + "loss": 0.2276490479707718, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 73.02, + "memory/max_allocated (GiB)": 73.02, + "ppl": 1.25564, + "step": 662, + "tokens/total": 163376768, + "tokens/train_per_sec_per_gpu": 14.01, + "tokens/trainable": 1313613 + }, + { + "epoch": 1.353061224489796, + "grad_norm": 0.2498278170824051, + "learning_rate": 2.514318026439469e-05, + "loss": 0.19553984701633453, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 74.63, + "memory/max_allocated (GiB)": 74.63, + "ppl": 1.21597, + "step": 663, + "tokens/total": 163615744, + "tokens/train_per_sec_per_gpu": 10.1, + "tokens/trainable": 1315250 + }, + { + "epoch": 1.3551020408163266, + "grad_norm": 0.2796618938446045, + "learning_rate": 2.500000000000001e-05, + "loss": 0.1958717703819275, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 75.5, + "memory/max_allocated (GiB)": 75.5, + "ppl": 1.21637, + "step": 664, + "tokens/total": 163870592, + "tokens/train_per_sec_per_gpu": 10.4, + "tokens/trainable": 1316976 + }, + { + "epoch": 1.3571428571428572, + "grad_norm": 0.3069683909416199, + "learning_rate": 2.4857092556937506e-05, + "loss": 0.24876132607460022, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 75.5, + "memory/max_allocated (GiB)": 75.5, + "ppl": 1.28244, + "step": 665, + "tokens/total": 164108480, + "tokens/train_per_sec_per_gpu": 12.29, + "tokens/trainable": 1318933 + }, + { + "epoch": 1.3591836734693876, + "grad_norm": 0.2928718328475952, + "learning_rate": 2.471445949473512e-05, + "loss": 0.21668942272663116, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 75.56, + "memory/max_allocated (GiB)": 75.56, + "ppl": 1.24196, + "step": 666, + "tokens/total": 164325696, + "tokens/train_per_sec_per_gpu": 12.94, + "tokens/trainable": 1320755 + }, + { + "epoch": 1.3612244897959185, + "grad_norm": 0.2910647988319397, + "learning_rate": 2.4572102369926546e-05, + "loss": 0.2385576367378235, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 72.07, + "memory/max_allocated (GiB)": 72.07, + "ppl": 1.26942, + "step": 667, + "tokens/total": 164552064, + "tokens/train_per_sec_per_gpu": 15.33, + "tokens/trainable": 1322827 + }, + { + "epoch": 1.363265306122449, + "grad_norm": 0.31465256214141846, + "learning_rate": 2.443002273603423e-05, + "loss": 0.2538492977619171, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 76.37, + "memory/max_allocated (GiB)": 76.37, + "ppl": 1.28898, + "step": 668, + "tokens/total": 164815680, + "tokens/train_per_sec_per_gpu": 12.84, + "tokens/trainable": 1325014 + }, + { + "epoch": 1.3653061224489795, + "grad_norm": 0.36235368251800537, + "learning_rate": 2.4288222143552348e-05, + "loss": 0.27942681312561035, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 75.28, + "memory/max_allocated (GiB)": 75.28, + "ppl": 1.32237, + "step": 669, + "tokens/total": 165064192, + "tokens/train_per_sec_per_gpu": 11.53, + "tokens/trainable": 1326988 + }, + { + "epoch": 1.3673469387755102, + "grad_norm": 0.2722671627998352, + "learning_rate": 2.414670213992994e-05, + "loss": 0.18563473224639893, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 74.66, + "memory/max_allocated (GiB)": 74.66, + "ppl": 1.20398, + "step": 670, + "tokens/total": 165317056, + "tokens/train_per_sec_per_gpu": 9.62, + "tokens/trainable": 1328637 + }, + { + "epoch": 1.3693877551020408, + "grad_norm": 0.3143163025379181, + "learning_rate": 2.4005464269554077e-05, + "loss": 0.2682737410068512, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 73.84, + "memory/max_allocated (GiB)": 73.84, + "ppl": 1.30771, + "step": 671, + "tokens/total": 165535552, + "tokens/train_per_sec_per_gpu": 15.59, + "tokens/trainable": 1330709 + }, + { + "epoch": 1.3714285714285714, + "grad_norm": 0.22773534059524536, + "learning_rate": 2.3864510073732916e-05, + "loss": 0.16950935125350952, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 74.14, + "memory/max_allocated (GiB)": 74.14, + "ppl": 1.18472, + "step": 672, + "tokens/total": 165769856, + "tokens/train_per_sec_per_gpu": 9.85, + "tokens/trainable": 1332400 + }, + { + "epoch": 1.373469387755102, + "grad_norm": 0.2484290599822998, + "learning_rate": 2.372384109067888e-05, + "loss": 0.2010881006717682, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 76.26, + "memory/max_allocated (GiB)": 76.26, + "ppl": 1.22273, + "step": 673, + "tokens/total": 166044928, + "tokens/train_per_sec_per_gpu": 9.53, + "tokens/trainable": 1334176 + }, + { + "epoch": 1.3755102040816327, + "grad_norm": 0.3130978047847748, + "learning_rate": 2.3583458855491985e-05, + "loss": 0.24413397908210754, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 71.55, + "memory/max_allocated (GiB)": 71.55, + "ppl": 1.27652, + "step": 674, + "tokens/total": 166266048, + "tokens/train_per_sec_per_gpu": 10.93, + "tokens/trainable": 1335964 + }, + { + "epoch": 1.3775510204081631, + "grad_norm": 0.27505946159362793, + "learning_rate": 2.344336490014295e-05, + "loss": 0.18199363350868225, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 75.28, + "memory/max_allocated (GiB)": 75.28, + "ppl": 1.19961, + "step": 675, + "tokens/total": 166551296, + "tokens/train_per_sec_per_gpu": 11.2, + "tokens/trainable": 1337919 + }, + { + "epoch": 1.379591836734694, + "grad_norm": 0.2739170789718628, + "learning_rate": 2.3303560753456537e-05, + "loss": 0.1982268989086151, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 76.4, + "memory/max_allocated (GiB)": 76.4, + "ppl": 1.21924, + "step": 676, + "tokens/total": 166793088, + "tokens/train_per_sec_per_gpu": 9.73, + "tokens/trainable": 1340058 + }, + { + "epoch": 1.3816326530612244, + "grad_norm": 0.2594189941883087, + "learning_rate": 2.316404794109493e-05, + "loss": 0.22886978089809418, + "memory/device_reserved (GiB)": 77.35, + "memory/max_active (GiB)": 74.0, + "memory/max_allocated (GiB)": 74.0, + "ppl": 1.25718, + "step": 677, + "tokens/total": 167026816, + "tokens/train_per_sec_per_gpu": 16.13, + "tokens/trainable": 1342590 + }, + { + "epoch": 1.383673469387755, + "grad_norm": 0.29924502968788147, + "learning_rate": 2.3024827985540958e-05, + "loss": 0.1979307383298874, + "memory/device_reserved (GiB)": 77.35, + "memory/max_active (GiB)": 75.26, + "memory/max_allocated (GiB)": 75.26, + "ppl": 1.21888, + "step": 678, + "tokens/total": 167277952, + "tokens/train_per_sec_per_gpu": 10.27, + "tokens/trainable": 1344339 + }, + { + "epoch": 1.3857142857142857, + "grad_norm": 0.3069194257259369, + "learning_rate": 2.288590240608155e-05, + "loss": 0.2513231039047241, + "memory/device_reserved (GiB)": 77.35, + "memory/max_active (GiB)": 73.24, + "memory/max_allocated (GiB)": 73.24, + "ppl": 1.28573, + "step": 679, + "tokens/total": 167528448, + "tokens/train_per_sec_per_gpu": 12.6, + "tokens/trainable": 1346292 + }, + { + "epoch": 1.3877551020408163, + "grad_norm": 0.25783997774124146, + "learning_rate": 2.274727271879122e-05, + "loss": 0.18607079982757568, + "memory/device_reserved (GiB)": 77.35, + "memory/max_active (GiB)": 74.66, + "memory/max_allocated (GiB)": 74.66, + "ppl": 1.20451, + "step": 680, + "tokens/total": 167771584, + "tokens/train_per_sec_per_gpu": 11.84, + "tokens/trainable": 1348124 + }, + { + "epoch": 1.389795918367347, + "grad_norm": 0.262175977230072, + "learning_rate": 2.260894043651537e-05, + "loss": 0.16938456892967224, + "memory/device_reserved (GiB)": 77.35, + "memory/max_active (GiB)": 76.21, + "memory/max_allocated (GiB)": 76.21, + "ppl": 1.18458, + "step": 681, + "tokens/total": 167996800, + "tokens/train_per_sec_per_gpu": 9.26, + "tokens/trainable": 1349876 + }, + { + "epoch": 1.3918367346938776, + "grad_norm": 0.24788179993629456, + "learning_rate": 2.247090706885392e-05, + "loss": 0.17150065302848816, + "memory/device_reserved (GiB)": 77.35, + "memory/max_active (GiB)": 74.9, + "memory/max_allocated (GiB)": 74.9, + "ppl": 1.18708, + "step": 682, + "tokens/total": 168252608, + "tokens/train_per_sec_per_gpu": 11.26, + "tokens/trainable": 1351839 + }, + { + "epoch": 1.3938775510204082, + "grad_norm": 0.3262105882167816, + "learning_rate": 2.233317412214473e-05, + "loss": 0.2576940059661865, + "memory/device_reserved (GiB)": 77.35, + "memory/max_active (GiB)": 73.43, + "memory/max_allocated (GiB)": 73.43, + "ppl": 1.29394, + "step": 683, + "tokens/total": 168460928, + "tokens/train_per_sec_per_gpu": 15.16, + "tokens/trainable": 1353674 + }, + { + "epoch": 1.3959183673469386, + "grad_norm": 0.30838677287101746, + "learning_rate": 2.2195743099447258e-05, + "loss": 0.2625044584274292, + "memory/device_reserved (GiB)": 77.35, + "memory/max_active (GiB)": 75.88, + "memory/max_allocated (GiB)": 75.88, + "ppl": 1.30018, + "step": 684, + "tokens/total": 168714688, + "tokens/train_per_sec_per_gpu": 12.79, + "tokens/trainable": 1355649 + }, + { + "epoch": 1.3979591836734695, + "grad_norm": 0.25927993655204773, + "learning_rate": 2.205861550052612e-05, + "loss": 0.17148219048976898, + "memory/device_reserved (GiB)": 77.35, + "memory/max_active (GiB)": 73.68, + "memory/max_allocated (GiB)": 73.68, + "ppl": 1.18706, + "step": 685, + "tokens/total": 168948672, + "tokens/train_per_sec_per_gpu": 12.34, + "tokens/trainable": 1357500 + }, + { + "epoch": 1.4, + "grad_norm": 0.28541332483291626, + "learning_rate": 2.192179282183464e-05, + "loss": 0.24442759156227112, + "memory/device_reserved (GiB)": 77.37, + "memory/max_active (GiB)": 76.45, + "memory/max_allocated (GiB)": 76.45, + "ppl": 1.27689, + "step": 686, + "tokens/total": 169205312, + "tokens/train_per_sec_per_gpu": 14.34, + "tokens/trainable": 1359802 + }, + { + "epoch": 1.4020408163265305, + "grad_norm": 0.31135135889053345, + "learning_rate": 2.178527655649868e-05, + "loss": 0.24949350953102112, + "memory/device_reserved (GiB)": 77.37, + "memory/max_active (GiB)": 74.03, + "memory/max_allocated (GiB)": 74.03, + "ppl": 1.28338, + "step": 687, + "tokens/total": 169457472, + "tokens/train_per_sec_per_gpu": 14.09, + "tokens/trainable": 1362100 + }, + { + "epoch": 1.4040816326530612, + "grad_norm": 0.28527817130088806, + "learning_rate": 2.164906819430019e-05, + "loss": 0.2309095859527588, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 76.73, + "memory/max_allocated (GiB)": 76.73, + "ppl": 1.25975, + "step": 688, + "tokens/total": 169712000, + "tokens/train_per_sec_per_gpu": 10.59, + "tokens/trainable": 1363921 + }, + { + "epoch": 1.4061224489795918, + "grad_norm": 0.24128185212612152, + "learning_rate": 2.1513169221661046e-05, + "loss": 0.16157332062721252, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 71.99, + "memory/max_allocated (GiB)": 71.99, + "ppl": 1.17536, + "step": 689, + "tokens/total": 169941760, + "tokens/train_per_sec_per_gpu": 13.22, + "tokens/trainable": 1366092 + }, + { + "epoch": 1.4081632653061225, + "grad_norm": 0.2898857593536377, + "learning_rate": 2.1377581121626777e-05, + "loss": 0.19659003615379333, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 74.39, + "memory/max_allocated (GiB)": 74.39, + "ppl": 1.21724, + "step": 690, + "tokens/total": 170169792, + "tokens/train_per_sec_per_gpu": 12.79, + "tokens/trainable": 1367877 + }, + { + "epoch": 1.410204081632653, + "grad_norm": 0.27626797556877136, + "learning_rate": 2.1242305373850468e-05, + "loss": 0.19579342007637024, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 75.69, + "memory/max_allocated (GiB)": 75.69, + "ppl": 1.21628, + "step": 691, + "tokens/total": 170437504, + "tokens/train_per_sec_per_gpu": 10.26, + "tokens/trainable": 1369673 + }, + { + "epoch": 1.4122448979591837, + "grad_norm": 0.2580835223197937, + "learning_rate": 2.1107343454576485e-05, + "loss": 0.1920934021472931, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 76.32, + "memory/max_allocated (GiB)": 76.32, + "ppl": 1.21178, + "step": 692, + "tokens/total": 170700096, + "tokens/train_per_sec_per_gpu": 12.77, + "tokens/trainable": 1371976 + }, + { + "epoch": 1.4142857142857144, + "grad_norm": 0.28825631737709045, + "learning_rate": 2.097269683662444e-05, + "loss": 0.2126019299030304, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 74.52, + "memory/max_allocated (GiB)": 74.52, + "ppl": 1.23689, + "step": 693, + "tokens/total": 170981440, + "tokens/train_per_sec_per_gpu": 10.83, + "tokens/trainable": 1373867 + }, + { + "epoch": 1.416326530612245, + "grad_norm": 0.27239957451820374, + "learning_rate": 2.083836698937315e-05, + "loss": 0.19974026083946228, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 74.2, + "memory/max_allocated (GiB)": 74.2, + "ppl": 1.22109, + "step": 694, + "tokens/total": 171222720, + "tokens/train_per_sec_per_gpu": 11.8, + "tokens/trainable": 1375736 + }, + { + "epoch": 1.4183673469387754, + "grad_norm": 0.32045778632164, + "learning_rate": 2.0704355378744512e-05, + "loss": 0.259243369102478, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 73.54, + "memory/max_allocated (GiB)": 73.54, + "ppl": 1.29595, + "step": 695, + "tokens/total": 171449664, + "tokens/train_per_sec_per_gpu": 15.34, + "tokens/trainable": 1377910 + }, + { + "epoch": 1.420408163265306, + "grad_norm": 0.3079512417316437, + "learning_rate": 2.0570663467187556e-05, + "loss": 0.2490934133529663, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 75.47, + "memory/max_allocated (GiB)": 75.47, + "ppl": 1.28286, + "step": 696, + "tokens/total": 171684288, + "tokens/train_per_sec_per_gpu": 10.9, + "tokens/trainable": 1379708 + }, + { + "epoch": 1.4224489795918367, + "grad_norm": 0.2902098298072815, + "learning_rate": 2.0437292713662537e-05, + "loss": 0.22740019857883453, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 74.11, + "memory/max_allocated (GiB)": 74.11, + "ppl": 1.25533, + "step": 697, + "tokens/total": 171910144, + "tokens/train_per_sec_per_gpu": 11.14, + "tokens/trainable": 1381456 + }, + { + "epoch": 1.4244897959183673, + "grad_norm": 0.32826298475265503, + "learning_rate": 2.0304244573624896e-05, + "loss": 0.2717878222465515, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 73.6, + "memory/max_allocated (GiB)": 73.6, + "ppl": 1.31231, + "step": 698, + "tokens/total": 172170112, + "tokens/train_per_sec_per_gpu": 13.05, + "tokens/trainable": 1383633 + }, + { + "epoch": 1.426530612244898, + "grad_norm": 0.2709813714027405, + "learning_rate": 2.0171520499009457e-05, + "loss": 0.20220345258712769, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 75.04, + "memory/max_allocated (GiB)": 75.04, + "ppl": 1.2241, + "step": 699, + "tokens/total": 172408320, + "tokens/train_per_sec_per_gpu": 13.46, + "tokens/trainable": 1385927 + }, + { + "epoch": 1.4285714285714286, + "grad_norm": 0.3553300201892853, + "learning_rate": 2.0039121938214604e-05, + "loss": 0.2955847680568695, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 75.39, + "memory/max_allocated (GiB)": 75.39, + "ppl": 1.34391, + "step": 700, + "tokens/total": 172664704, + "tokens/train_per_sec_per_gpu": 14.12, + "tokens/trainable": 1388090 + }, + { + "epoch": 1.4306122448979592, + "grad_norm": 0.2816275358200073, + "learning_rate": 1.9907050336086368e-05, + "loss": 0.18583577871322632, + "memory/device_reserved (GiB)": 77.67, + "memory/max_active (GiB)": 74.28, + "memory/max_allocated (GiB)": 74.28, + "ppl": 1.20422, + "step": 701, + "tokens/total": 172917696, + "tokens/train_per_sec_per_gpu": 9.25, + "tokens/trainable": 1389960 + }, + { + "epoch": 1.4326530612244899, + "grad_norm": 0.3052232563495636, + "learning_rate": 1.977530713390281e-05, + "loss": 0.241130992770195, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 76.59, + "memory/max_allocated (GiB)": 76.59, + "ppl": 1.27269, + "step": 702, + "tokens/total": 173162368, + "tokens/train_per_sec_per_gpu": 13.36, + "tokens/trainable": 1392017 + }, + { + "epoch": 1.4346938775510205, + "grad_norm": 0.24402208626270294, + "learning_rate": 1.9643893769358123e-05, + "loss": 0.190439373254776, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 72.64, + "memory/max_allocated (GiB)": 72.64, + "ppl": 1.20978, + "step": 703, + "tokens/total": 173396672, + "tokens/train_per_sec_per_gpu": 12.18, + "tokens/trainable": 1393982 + }, + { + "epoch": 1.436734693877551, + "grad_norm": 0.27753302454948425, + "learning_rate": 1.9512811676547117e-05, + "loss": 0.21275126934051514, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 72.75, + "memory/max_allocated (GiB)": 72.75, + "ppl": 1.23708, + "step": 704, + "tokens/total": 173640896, + "tokens/train_per_sec_per_gpu": 11.89, + "tokens/trainable": 1395924 + }, + { + "epoch": 1.4387755102040816, + "grad_norm": 0.3034632205963135, + "learning_rate": 1.9382062285949414e-05, + "loss": 0.2784818410873413, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 75.94, + "memory/max_allocated (GiB)": 75.94, + "ppl": 1.32112, + "step": 705, + "tokens/total": 173897984, + "tokens/train_per_sec_per_gpu": 11.73, + "tokens/trainable": 1398062 + }, + { + "epoch": 1.4408163265306122, + "grad_norm": 0.2592516243457794, + "learning_rate": 1.925164702441392e-05, + "loss": 0.20499449968338013, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 74.77, + "memory/max_allocated (GiB)": 74.77, + "ppl": 1.22752, + "step": 706, + "tokens/total": 174148288, + "tokens/train_per_sec_per_gpu": 11.4, + "tokens/trainable": 1400059 + }, + { + "epoch": 1.4428571428571428, + "grad_norm": 0.27760761976242065, + "learning_rate": 1.9121567315143283e-05, + "loss": 0.1814465969800949, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 73.46, + "memory/max_allocated (GiB)": 73.46, + "ppl": 1.19895, + "step": 707, + "tokens/total": 174384640, + "tokens/train_per_sec_per_gpu": 14.0, + "tokens/trainable": 1402130 + }, + { + "epoch": 1.4448979591836735, + "grad_norm": 0.3253805637359619, + "learning_rate": 1.8991824577678268e-05, + "loss": 0.2989218831062317, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 76.32, + "memory/max_allocated (GiB)": 76.32, + "ppl": 1.3484, + "step": 708, + "tokens/total": 174627328, + "tokens/train_per_sec_per_gpu": 13.36, + "tokens/trainable": 1404177 + }, + { + "epoch": 1.4469387755102041, + "grad_norm": 0.29065045714378357, + "learning_rate": 1.8862420227882338e-05, + "loss": 0.21384821832180023, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 73.43, + "memory/max_allocated (GiB)": 73.43, + "ppl": 1.23843, + "step": 709, + "tokens/total": 174854272, + "tokens/train_per_sec_per_gpu": 15.0, + "tokens/trainable": 1406360 + }, + { + "epoch": 1.4489795918367347, + "grad_norm": 0.26728355884552, + "learning_rate": 1.8733355677926156e-05, + "loss": 0.23495501279830933, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 74.49, + "memory/max_allocated (GiB)": 74.49, + "ppl": 1.26485, + "step": 710, + "tokens/total": 175097024, + "tokens/train_per_sec_per_gpu": 13.64, + "tokens/trainable": 1408395 + }, + { + "epoch": 1.4510204081632654, + "grad_norm": 0.268636554479599, + "learning_rate": 1.860463233627225e-05, + "loss": 0.22195157408714294, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 74.52, + "memory/max_allocated (GiB)": 74.52, + "ppl": 1.24851, + "step": 711, + "tokens/total": 175331200, + "tokens/train_per_sec_per_gpu": 12.03, + "tokens/trainable": 1410217 + }, + { + "epoch": 1.453061224489796, + "grad_norm": 0.28556984663009644, + "learning_rate": 1.847625160765955e-05, + "loss": 0.20533058047294617, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 75.37, + "memory/max_allocated (GiB)": 75.37, + "ppl": 1.22793, + "step": 712, + "tokens/total": 175566336, + "tokens/train_per_sec_per_gpu": 13.15, + "tokens/trainable": 1412105 + }, + { + "epoch": 1.4551020408163264, + "grad_norm": 0.2985648214817047, + "learning_rate": 1.83482148930881e-05, + "loss": 0.23746563494205475, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 75.83, + "memory/max_allocated (GiB)": 75.83, + "ppl": 1.26803, + "step": 713, + "tokens/total": 175808192, + "tokens/train_per_sec_per_gpu": 13.46, + "tokens/trainable": 1414284 + }, + { + "epoch": 1.457142857142857, + "grad_norm": 0.27680814266204834, + "learning_rate": 1.8220523589803806e-05, + "loss": 0.2486806958913803, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 75.45, + "memory/max_allocated (GiB)": 75.45, + "ppl": 1.28233, + "step": 714, + "tokens/total": 176064320, + "tokens/train_per_sec_per_gpu": 14.22, + "tokens/trainable": 1416434 + }, + { + "epoch": 1.4591836734693877, + "grad_norm": 0.24422398209571838, + "learning_rate": 1.809317909128312e-05, + "loss": 0.18191733956336975, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 75.31, + "memory/max_allocated (GiB)": 75.31, + "ppl": 1.19952, + "step": 715, + "tokens/total": 176313792, + "tokens/train_per_sec_per_gpu": 10.91, + "tokens/trainable": 1418217 + }, + { + "epoch": 1.4612244897959183, + "grad_norm": 0.30382683873176575, + "learning_rate": 1.7966182787217844e-05, + "loss": 0.27327001094818115, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 75.47, + "memory/max_allocated (GiB)": 75.47, + "ppl": 1.31426, + "step": 716, + "tokens/total": 176582784, + "tokens/train_per_sec_per_gpu": 11.94, + "tokens/trainable": 1420273 + }, + { + "epoch": 1.463265306122449, + "grad_norm": 0.24096240103244781, + "learning_rate": 1.783953606350005e-05, + "loss": 0.15613946318626404, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 73.11, + "memory/max_allocated (GiB)": 73.11, + "ppl": 1.16899, + "step": 717, + "tokens/total": 176807296, + "tokens/train_per_sec_per_gpu": 10.97, + "tokens/trainable": 1421948 + }, + { + "epoch": 1.4653061224489796, + "grad_norm": 0.30400145053863525, + "learning_rate": 1.7713240302206825e-05, + "loss": 0.2681472897529602, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 75.88, + "memory/max_allocated (GiB)": 75.88, + "ppl": 1.30754, + "step": 718, + "tokens/total": 177066368, + "tokens/train_per_sec_per_gpu": 13.28, + "tokens/trainable": 1424352 + }, + { + "epoch": 1.4673469387755103, + "grad_norm": 0.28764045238494873, + "learning_rate": 1.7587296881585308e-05, + "loss": 0.218339741230011, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 74.27, + "memory/max_allocated (GiB)": 74.27, + "ppl": 1.24401, + "step": 719, + "tokens/total": 177298304, + "tokens/train_per_sec_per_gpu": 14.73, + "tokens/trainable": 1426362 + }, + { + "epoch": 1.469387755102041, + "grad_norm": 0.288003146648407, + "learning_rate": 1.7461707176037545e-05, + "loss": 0.2564609944820404, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 75.28, + "memory/max_allocated (GiB)": 75.28, + "ppl": 1.29235, + "step": 720, + "tokens/total": 177551872, + "tokens/train_per_sec_per_gpu": 12.2, + "tokens/trainable": 1428307 + }, + { + "epoch": 1.4714285714285715, + "grad_norm": 0.22934523224830627, + "learning_rate": 1.7336472556105593e-05, + "loss": 0.18992646038532257, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 75.45, + "memory/max_allocated (GiB)": 75.45, + "ppl": 1.20916, + "step": 721, + "tokens/total": 177836416, + "tokens/train_per_sec_per_gpu": 11.76, + "tokens/trainable": 1430435 + }, + { + "epoch": 1.473469387755102, + "grad_norm": 0.2770795226097107, + "learning_rate": 1.721159438845647e-05, + "loss": 0.21522647142410278, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 75.18, + "memory/max_allocated (GiB)": 75.18, + "ppl": 1.24014, + "step": 722, + "tokens/total": 178120896, + "tokens/train_per_sec_per_gpu": 9.83, + "tokens/trainable": 1432272 + }, + { + "epoch": 1.4755102040816326, + "grad_norm": 0.28103068470954895, + "learning_rate": 1.7087074035867285e-05, + "loss": 0.2046436369419098, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 72.64, + "memory/max_allocated (GiB)": 72.64, + "ppl": 1.22709, + "step": 723, + "tokens/total": 178354112, + "tokens/train_per_sec_per_gpu": 12.44, + "tokens/trainable": 1434276 + }, + { + "epoch": 1.4775510204081632, + "grad_norm": 0.2643071711063385, + "learning_rate": 1.696291285721041e-05, + "loss": 0.1900281012058258, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 76.24, + "memory/max_allocated (GiB)": 76.24, + "ppl": 1.20928, + "step": 724, + "tokens/total": 178627392, + "tokens/train_per_sec_per_gpu": 11.05, + "tokens/trainable": 1436325 + }, + { + "epoch": 1.4795918367346939, + "grad_norm": 0.28737112879753113, + "learning_rate": 1.6839112207438557e-05, + "loss": 0.20797918736934662, + "memory/device_reserved (GiB)": 77.47, + "memory/max_active (GiB)": 73.95, + "memory/max_allocated (GiB)": 73.95, + "ppl": 1.23119, + "step": 725, + "tokens/total": 178873280, + "tokens/train_per_sec_per_gpu": 13.52, + "tokens/trainable": 1438389 + } + ], + "logging_steps": 1, + "max_steps": 980, + "num_input_tokens_seen": 0, + "num_train_epochs": 2, + "save_steps": 25, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": false + }, + "attributes": {} + } + }, + "total_flos": 3.3123903498996417e+19, + "train_batch_size": 1, + "trial_name": null, + "trial_params": null +}