| { | |
| "best_metric": 0.14330759230951728, | |
| "best_model_checkpoint": "intermediate/deberta-v3-base-aee-classifier-checkpoints/checkpoint-40000", | |
| "epoch": 2.849002849002849, | |
| "eval_steps": 5000, | |
| "global_step": 100000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 4.996058879392213e-05, | |
| "loss": 0.6492, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.991310541310542e-05, | |
| "loss": 0.5813, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.98656220322887e-05, | |
| "loss": 0.5754, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.981813865147199e-05, | |
| "loss": 0.5585, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.9770655270655275e-05, | |
| "loss": 0.5565, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 4.972364672364673e-05, | |
| "loss": 0.545, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 4.9676163342830015e-05, | |
| "loss": 0.5403, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 4.9628679962013296e-05, | |
| "loss": 0.5386, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 4.9581196581196584e-05, | |
| "loss": 0.5482, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 4.953371320037987e-05, | |
| "loss": 0.5305, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 4.9486229819563154e-05, | |
| "loss": 0.5393, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 4.943874643874644e-05, | |
| "loss": 0.5375, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 4.939126305792972e-05, | |
| "loss": 0.5304, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 4.934377967711301e-05, | |
| "loss": 0.5367, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 4.92962962962963e-05, | |
| "loss": 0.5235, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 4.924881291547959e-05, | |
| "loss": 0.5277, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 4.920132953466287e-05, | |
| "loss": 0.5282, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 4.9153846153846157e-05, | |
| "loss": 0.528, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 4.910636277302944e-05, | |
| "loss": 0.5246, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 4.9058879392212726e-05, | |
| "loss": 0.5197, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 4.9011396011396014e-05, | |
| "loss": 0.5207, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 4.8964387464387466e-05, | |
| "loss": 0.5294, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.8916904083570754e-05, | |
| "loss": 0.5155, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.8869420702754035e-05, | |
| "loss": 0.5106, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.8821937321937324e-05, | |
| "loss": 0.5109, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.877445394112061e-05, | |
| "loss": 0.5219, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 4.87269705603039e-05, | |
| "loss": 0.5136, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 4.8680436847103516e-05, | |
| "loss": 0.5234, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 4.8632953466286804e-05, | |
| "loss": 0.5153, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 4.8585470085470085e-05, | |
| "loss": 0.5097, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 4.8537986704653373e-05, | |
| "loss": 0.5153, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 4.849050332383666e-05, | |
| "loss": 0.5053, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 4.844301994301994e-05, | |
| "loss": 0.5136, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 4.839553656220323e-05, | |
| "loss": 0.5131, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 4.834805318138652e-05, | |
| "loss": 0.5137, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 4.83005698005698e-05, | |
| "loss": 0.515, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 4.825308641975309e-05, | |
| "loss": 0.516, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 4.8205603038936376e-05, | |
| "loss": 0.5048, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 4.815811965811966e-05, | |
| "loss": 0.5073, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 4.8110636277302946e-05, | |
| "loss": 0.5088, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 4.806315289648623e-05, | |
| "loss": 0.5109, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 4.801566951566952e-05, | |
| "loss": 0.5053, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 4.79681861348528e-05, | |
| "loss": 0.5001, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 4.792070275403609e-05, | |
| "loss": 0.5121, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 4.787321937321937e-05, | |
| "loss": 0.4984, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 4.782573599240266e-05, | |
| "loss": 0.5027, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 4.777825261158595e-05, | |
| "loss": 0.5043, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 4.7730769230769236e-05, | |
| "loss": 0.498, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 4.768328584995252e-05, | |
| "loss": 0.5147, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 4.7635802469135806e-05, | |
| "loss": 0.502, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "eval_binary_F1_negative": 0.8826562666740798, | |
| "eval_binary_F1_postive": 0.11849080025168, | |
| "eval_loss": 0.43656376004219055, | |
| "eval_macro_F1": 0.5005735334628799, | |
| "eval_micro_F1": 0.7928832623212131, | |
| "eval_runtime": 6165.047, | |
| "eval_samples_per_second": 1537.18, | |
| "eval_steps_per_second": 48.037, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 4.758831908831909e-05, | |
| "loss": 0.5073, | |
| "step": 5100 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 4.754131054131054e-05, | |
| "loss": 0.5098, | |
| "step": 5200 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 4.7493827160493834e-05, | |
| "loss": 0.4923, | |
| "step": 5300 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 4.7446343779677115e-05, | |
| "loss": 0.499, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 4.73988603988604e-05, | |
| "loss": 0.5148, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 4.7351377018043685e-05, | |
| "loss": 0.503, | |
| "step": 5600 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 4.730389363722697e-05, | |
| "loss": 0.5035, | |
| "step": 5700 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 4.725641025641026e-05, | |
| "loss": 0.5, | |
| "step": 5800 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 4.720940170940171e-05, | |
| "loss": 0.5095, | |
| "step": 5900 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 4.7161918328584994e-05, | |
| "loss": 0.5031, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 4.711443494776828e-05, | |
| "loss": 0.4936, | |
| "step": 6100 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 4.706695156695157e-05, | |
| "loss": 0.5046, | |
| "step": 6200 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 4.701946818613486e-05, | |
| "loss": 0.5032, | |
| "step": 6300 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 4.697198480531814e-05, | |
| "loss": 0.4982, | |
| "step": 6400 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 4.692450142450143e-05, | |
| "loss": 0.5027, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 4.687701804368471e-05, | |
| "loss": 0.5035, | |
| "step": 6600 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 4.6829534662868e-05, | |
| "loss": 0.5106, | |
| "step": 6700 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 4.6782051282051285e-05, | |
| "loss": 0.5064, | |
| "step": 6800 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 4.673456790123457e-05, | |
| "loss": 0.5061, | |
| "step": 6900 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 4.6687084520417854e-05, | |
| "loss": 0.4993, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 4.663960113960114e-05, | |
| "loss": 0.4898, | |
| "step": 7100 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 4.6592117758784424e-05, | |
| "loss": 0.5052, | |
| "step": 7200 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 4.654463437796771e-05, | |
| "loss": 0.5087, | |
| "step": 7300 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 4.6497150997151e-05, | |
| "loss": 0.4936, | |
| "step": 7400 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 4.644966761633429e-05, | |
| "loss": 0.4964, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 4.640218423551757e-05, | |
| "loss": 0.4914, | |
| "step": 7600 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 4.635470085470086e-05, | |
| "loss": 0.5074, | |
| "step": 7700 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 4.630721747388414e-05, | |
| "loss": 0.4896, | |
| "step": 7800 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 4.625973409306743e-05, | |
| "loss": 0.4988, | |
| "step": 7900 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 4.6212250712250715e-05, | |
| "loss": 0.4913, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 4.616524216524217e-05, | |
| "loss": 0.4975, | |
| "step": 8100 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 4.6117758784425455e-05, | |
| "loss": 0.4893, | |
| "step": 8200 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 4.6070275403608736e-05, | |
| "loss": 0.502, | |
| "step": 8300 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 4.6022792022792024e-05, | |
| "loss": 0.4926, | |
| "step": 8400 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 4.597530864197531e-05, | |
| "loss": 0.484, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 4.5927825261158594e-05, | |
| "loss": 0.4994, | |
| "step": 8600 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 4.588034188034188e-05, | |
| "loss": 0.4902, | |
| "step": 8700 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 4.583285849952517e-05, | |
| "loss": 0.5019, | |
| "step": 8800 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 4.578537511870846e-05, | |
| "loss": 0.491, | |
| "step": 8900 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 4.573789173789174e-05, | |
| "loss": 0.4921, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 4.569040835707503e-05, | |
| "loss": 0.4983, | |
| "step": 9100 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 4.564292497625831e-05, | |
| "loss": 0.486, | |
| "step": 9200 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 4.5595441595441596e-05, | |
| "loss": 0.494, | |
| "step": 9300 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 4.5547958214624884e-05, | |
| "loss": 0.4975, | |
| "step": 9400 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 4.550047483380817e-05, | |
| "loss": 0.4938, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 4.5452991452991454e-05, | |
| "loss": 0.4918, | |
| "step": 9600 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 4.540550807217474e-05, | |
| "loss": 0.4947, | |
| "step": 9700 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 4.535802469135802e-05, | |
| "loss": 0.4841, | |
| "step": 9800 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 4.5311016144349475e-05, | |
| "loss": 0.4893, | |
| "step": 9900 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 4.526353276353277e-05, | |
| "loss": 0.496, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "eval_binary_F1_negative": 0.8471897003606845, | |
| "eval_binary_F1_postive": 0.10560489532107646, | |
| "eval_loss": 0.4903038442134857, | |
| "eval_macro_F1": 0.4763972978408805, | |
| "eval_micro_F1": 0.738976167658529, | |
| "eval_runtime": 6174.1421, | |
| "eval_samples_per_second": 1534.915, | |
| "eval_steps_per_second": 47.966, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 4.521604938271605e-05, | |
| "loss": 0.4901, | |
| "step": 10100 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 4.516856600189934e-05, | |
| "loss": 0.4952, | |
| "step": 10200 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 4.512108262108262e-05, | |
| "loss": 0.4956, | |
| "step": 10300 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 4.507359924026591e-05, | |
| "loss": 0.4928, | |
| "step": 10400 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 4.50261158594492e-05, | |
| "loss": 0.4859, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 4.497910731244065e-05, | |
| "loss": 0.4953, | |
| "step": 10600 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 4.493162393162394e-05, | |
| "loss": 0.4932, | |
| "step": 10700 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 4.488414055080722e-05, | |
| "loss": 0.504, | |
| "step": 10800 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 4.4836657169990506e-05, | |
| "loss": 0.4975, | |
| "step": 10900 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 4.4789173789173794e-05, | |
| "loss": 0.4982, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 4.4741690408357076e-05, | |
| "loss": 0.4811, | |
| "step": 11100 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 4.4694207027540364e-05, | |
| "loss": 0.4895, | |
| "step": 11200 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 4.4646723646723645e-05, | |
| "loss": 0.4913, | |
| "step": 11300 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 4.459924026590693e-05, | |
| "loss": 0.4818, | |
| "step": 11400 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 4.455175688509022e-05, | |
| "loss": 0.4894, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 4.450427350427351e-05, | |
| "loss": 0.4953, | |
| "step": 11600 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 4.445679012345679e-05, | |
| "loss": 0.4912, | |
| "step": 11700 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 4.440930674264008e-05, | |
| "loss": 0.4949, | |
| "step": 11800 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 4.436182336182336e-05, | |
| "loss": 0.4929, | |
| "step": 11900 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 4.431433998100665e-05, | |
| "loss": 0.4866, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 4.4266856600189936e-05, | |
| "loss": 0.4987, | |
| "step": 12100 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 4.421984805318139e-05, | |
| "loss": 0.4907, | |
| "step": 12200 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 4.4172364672364676e-05, | |
| "loss": 0.4919, | |
| "step": 12300 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 4.412488129154796e-05, | |
| "loss": 0.4843, | |
| "step": 12400 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 4.4077397910731246e-05, | |
| "loss": 0.4963, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 4.4029914529914534e-05, | |
| "loss": 0.4952, | |
| "step": 12600 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 4.398243114909782e-05, | |
| "loss": 0.4874, | |
| "step": 12700 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 4.39349477682811e-05, | |
| "loss": 0.4961, | |
| "step": 12800 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 4.388746438746439e-05, | |
| "loss": 0.4965, | |
| "step": 12900 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 4.383998100664767e-05, | |
| "loss": 0.4972, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 4.379249762583097e-05, | |
| "loss": 0.4913, | |
| "step": 13100 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 4.374501424501425e-05, | |
| "loss": 0.4833, | |
| "step": 13200 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 4.3697530864197536e-05, | |
| "loss": 0.5008, | |
| "step": 13300 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 4.365004748338082e-05, | |
| "loss": 0.4963, | |
| "step": 13400 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 4.3602564102564106e-05, | |
| "loss": 0.489, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 4.355508072174739e-05, | |
| "loss": 0.4997, | |
| "step": 13600 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 4.3507597340930675e-05, | |
| "loss": 0.4894, | |
| "step": 13700 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 4.346011396011396e-05, | |
| "loss": 0.4912, | |
| "step": 13800 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 4.3413105413105415e-05, | |
| "loss": 0.4806, | |
| "step": 13900 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 4.3365622032288697e-05, | |
| "loss": 0.4794, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 4.3318138651471985e-05, | |
| "loss": 0.4955, | |
| "step": 14100 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 4.327065527065527e-05, | |
| "loss": 0.4824, | |
| "step": 14200 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 4.322317188983856e-05, | |
| "loss": 0.4857, | |
| "step": 14300 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 4.317568850902184e-05, | |
| "loss": 0.4962, | |
| "step": 14400 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 4.312820512820513e-05, | |
| "loss": 0.4885, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 4.308072174738841e-05, | |
| "loss": 0.4894, | |
| "step": 14600 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 4.3033238366571706e-05, | |
| "loss": 0.4807, | |
| "step": 14700 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 4.298575498575499e-05, | |
| "loss": 0.4867, | |
| "step": 14800 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 4.293874643874644e-05, | |
| "loss": 0.5073, | |
| "step": 14900 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 4.289126305792973e-05, | |
| "loss": 0.4933, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "eval_binary_F1_negative": 0.9019491373602011, | |
| "eval_binary_F1_postive": 0.13169467244548555, | |
| "eval_loss": 0.4128543734550476, | |
| "eval_macro_F1": 0.5168219049028433, | |
| "eval_micro_F1": 0.823795641264876, | |
| "eval_runtime": 6172.6706, | |
| "eval_samples_per_second": 1535.281, | |
| "eval_steps_per_second": 47.978, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 4.284377967711301e-05, | |
| "loss": 0.4792, | |
| "step": 15100 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 4.2796296296296304e-05, | |
| "loss": 0.4922, | |
| "step": 15200 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 4.2748812915479585e-05, | |
| "loss": 0.4855, | |
| "step": 15300 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 4.270132953466287e-05, | |
| "loss": 0.4915, | |
| "step": 15400 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 4.2653846153846154e-05, | |
| "loss": 0.488, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 4.260636277302944e-05, | |
| "loss": 0.4847, | |
| "step": 15600 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 4.2558879392212724e-05, | |
| "loss": 0.4893, | |
| "step": 15700 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 4.251139601139602e-05, | |
| "loss": 0.4909, | |
| "step": 15800 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 4.24639126305793e-05, | |
| "loss": 0.4852, | |
| "step": 15900 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 4.241642924976259e-05, | |
| "loss": 0.4793, | |
| "step": 16000 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 4.236942070275404e-05, | |
| "loss": 0.4901, | |
| "step": 16100 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 4.232193732193732e-05, | |
| "loss": 0.4894, | |
| "step": 16200 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 4.227445394112061e-05, | |
| "loss": 0.4845, | |
| "step": 16300 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 4.22269705603039e-05, | |
| "loss": 0.4989, | |
| "step": 16400 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 4.217948717948718e-05, | |
| "loss": 0.4913, | |
| "step": 16500 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 4.213200379867047e-05, | |
| "loss": 0.4849, | |
| "step": 16600 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 4.208452041785375e-05, | |
| "loss": 0.4837, | |
| "step": 16700 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 4.203703703703704e-05, | |
| "loss": 0.4971, | |
| "step": 16800 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 4.1989553656220324e-05, | |
| "loss": 0.4951, | |
| "step": 16900 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 4.194207027540361e-05, | |
| "loss": 0.4886, | |
| "step": 17000 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 4.1894586894586893e-05, | |
| "loss": 0.4946, | |
| "step": 17100 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 4.184710351377018e-05, | |
| "loss": 0.4839, | |
| "step": 17200 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 4.179962013295347e-05, | |
| "loss": 0.4913, | |
| "step": 17300 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 4.175213675213676e-05, | |
| "loss": 0.4822, | |
| "step": 17400 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 4.170465337132004e-05, | |
| "loss": 0.4851, | |
| "step": 17500 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 4.165764482431149e-05, | |
| "loss": 0.4873, | |
| "step": 17600 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 4.161016144349478e-05, | |
| "loss": 0.4837, | |
| "step": 17700 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.156267806267806e-05, | |
| "loss": 0.4818, | |
| "step": 17800 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.1515194681861355e-05, | |
| "loss": 0.4767, | |
| "step": 17900 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.1467711301044637e-05, | |
| "loss": 0.4836, | |
| "step": 18000 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.1420227920227925e-05, | |
| "loss": 0.4873, | |
| "step": 18100 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.1372744539411206e-05, | |
| "loss": 0.4777, | |
| "step": 18200 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.1325261158594494e-05, | |
| "loss": 0.4824, | |
| "step": 18300 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.127777777777778e-05, | |
| "loss": 0.4914, | |
| "step": 18400 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.123029439696107e-05, | |
| "loss": 0.4867, | |
| "step": 18500 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.118281101614435e-05, | |
| "loss": 0.4795, | |
| "step": 18600 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.113532763532764e-05, | |
| "loss": 0.4809, | |
| "step": 18700 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.108784425451092e-05, | |
| "loss": 0.4852, | |
| "step": 18800 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.104036087369421e-05, | |
| "loss": 0.4816, | |
| "step": 18900 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.09928774928775e-05, | |
| "loss": 0.4883, | |
| "step": 19000 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.094539411206078e-05, | |
| "loss": 0.4868, | |
| "step": 19100 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.0897910731244066e-05, | |
| "loss": 0.4854, | |
| "step": 19200 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.085042735042735e-05, | |
| "loss": 0.482, | |
| "step": 19300 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.0803418803418806e-05, | |
| "loss": 0.4901, | |
| "step": 19400 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.0755935422602094e-05, | |
| "loss": 0.4861, | |
| "step": 19500 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.0708452041785376e-05, | |
| "loss": 0.4976, | |
| "step": 19600 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.0660968660968664e-05, | |
| "loss": 0.4873, | |
| "step": 19700 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.0613485280151945e-05, | |
| "loss": 0.4787, | |
| "step": 19800 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 4.056600189933523e-05, | |
| "loss": 0.4869, | |
| "step": 19900 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 4.051851851851852e-05, | |
| "loss": 0.4835, | |
| "step": 20000 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "eval_binary_F1_negative": 0.8637453257797093, | |
| "eval_binary_F1_postive": 0.11365188753919629, | |
| "eval_loss": 0.4626865088939667, | |
| "eval_macro_F1": 0.4886986066594528, | |
| "eval_micro_F1": 0.7638006176355572, | |
| "eval_runtime": 6113.144, | |
| "eval_samples_per_second": 1550.231, | |
| "eval_steps_per_second": 48.445, | |
| "step": 20000 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 4.047103513770181e-05, | |
| "loss": 0.4843, | |
| "step": 20100 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 4.042355175688509e-05, | |
| "loss": 0.4828, | |
| "step": 20200 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 4.037606837606838e-05, | |
| "loss": 0.4832, | |
| "step": 20300 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 4.032858499525166e-05, | |
| "loss": 0.4787, | |
| "step": 20400 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 4.0281101614434955e-05, | |
| "loss": 0.4847, | |
| "step": 20500 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 4.0233618233618236e-05, | |
| "loss": 0.4889, | |
| "step": 20600 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 4.0186134852801524e-05, | |
| "loss": 0.4827, | |
| "step": 20700 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 4.0138651471984805e-05, | |
| "loss": 0.4804, | |
| "step": 20800 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 4.009116809116809e-05, | |
| "loss": 0.4862, | |
| "step": 20900 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 4.004368471035138e-05, | |
| "loss": 0.4839, | |
| "step": 21000 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.999620132953467e-05, | |
| "loss": 0.4841, | |
| "step": 21100 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.994871794871795e-05, | |
| "loss": 0.4816, | |
| "step": 21200 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.990123456790124e-05, | |
| "loss": 0.4817, | |
| "step": 21300 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.985375118708452e-05, | |
| "loss": 0.474, | |
| "step": 21400 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.980626780626781e-05, | |
| "loss": 0.4743, | |
| "step": 21500 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.9758784425451096e-05, | |
| "loss": 0.4801, | |
| "step": 21600 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.971130104463438e-05, | |
| "loss": 0.4865, | |
| "step": 21700 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.9663817663817665e-05, | |
| "loss": 0.4763, | |
| "step": 21800 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.9616334283000947e-05, | |
| "loss": 0.4872, | |
| "step": 21900 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 3.956885090218424e-05, | |
| "loss": 0.4776, | |
| "step": 22000 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 3.952136752136752e-05, | |
| "loss": 0.4906, | |
| "step": 22100 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 3.947388414055081e-05, | |
| "loss": 0.4845, | |
| "step": 22200 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 3.942640075973409e-05, | |
| "loss": 0.4797, | |
| "step": 22300 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 3.937891737891738e-05, | |
| "loss": 0.4839, | |
| "step": 22400 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 3.933190883190883e-05, | |
| "loss": 0.4814, | |
| "step": 22500 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 3.928490028490029e-05, | |
| "loss": 0.4836, | |
| "step": 22600 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 3.923741690408357e-05, | |
| "loss": 0.4833, | |
| "step": 22700 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 3.918993352326686e-05, | |
| "loss": 0.4826, | |
| "step": 22800 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 3.914245014245014e-05, | |
| "loss": 0.485, | |
| "step": 22900 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 3.909496676163343e-05, | |
| "loss": 0.4801, | |
| "step": 23000 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 3.904748338081672e-05, | |
| "loss": 0.4924, | |
| "step": 23100 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 3.9000000000000006e-05, | |
| "loss": 0.4845, | |
| "step": 23200 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 3.895251661918329e-05, | |
| "loss": 0.4742, | |
| "step": 23300 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 3.8905033238366575e-05, | |
| "loss": 0.4875, | |
| "step": 23400 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 3.885802469135803e-05, | |
| "loss": 0.4846, | |
| "step": 23500 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 3.8810541310541316e-05, | |
| "loss": 0.4708, | |
| "step": 23600 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 3.876353276353277e-05, | |
| "loss": 0.4889, | |
| "step": 23700 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 3.871604938271605e-05, | |
| "loss": 0.4851, | |
| "step": 23800 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 3.866856600189934e-05, | |
| "loss": 0.4888, | |
| "step": 23900 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 3.862108262108262e-05, | |
| "loss": 0.4761, | |
| "step": 24000 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 3.8573599240265907e-05, | |
| "loss": 0.4884, | |
| "step": 24100 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 3.8526115859449195e-05, | |
| "loss": 0.4722, | |
| "step": 24200 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 3.847910731244065e-05, | |
| "loss": 0.4984, | |
| "step": 24300 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 3.8431623931623935e-05, | |
| "loss": 0.4823, | |
| "step": 24400 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 3.8384140550807216e-05, | |
| "loss": 0.4873, | |
| "step": 24500 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 3.8336657169990504e-05, | |
| "loss": 0.486, | |
| "step": 24600 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 3.828917378917379e-05, | |
| "loss": 0.4843, | |
| "step": 24700 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 3.824169040835708e-05, | |
| "loss": 0.4916, | |
| "step": 24800 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 3.819420702754036e-05, | |
| "loss": 0.478, | |
| "step": 24900 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 3.814672364672365e-05, | |
| "loss": 0.4809, | |
| "step": 25000 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "eval_binary_F1_negative": 0.8819988992692533, | |
| "eval_binary_F1_postive": 0.12081565683024537, | |
| "eval_loss": 0.428783655166626, | |
| "eval_macro_F1": 0.5014072780497494, | |
| "eval_micro_F1": 0.7919249205374059, | |
| "eval_runtime": 6096.9217, | |
| "eval_samples_per_second": 1554.356, | |
| "eval_steps_per_second": 48.574, | |
| "step": 25000 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.809924026590693e-05, | |
| "loss": 0.4845, | |
| "step": 25100 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.8051756885090226e-05, | |
| "loss": 0.4877, | |
| "step": 25200 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.800427350427351e-05, | |
| "loss": 0.4787, | |
| "step": 25300 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.7956790123456795e-05, | |
| "loss": 0.4822, | |
| "step": 25400 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.7909306742640076e-05, | |
| "loss": 0.4899, | |
| "step": 25500 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.7861823361823364e-05, | |
| "loss": 0.4773, | |
| "step": 25600 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.7814339981006646e-05, | |
| "loss": 0.485, | |
| "step": 25700 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.7766856600189934e-05, | |
| "loss": 0.4833, | |
| "step": 25800 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.771937321937322e-05, | |
| "loss": 0.4848, | |
| "step": 25900 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.767188983855651e-05, | |
| "loss": 0.4845, | |
| "step": 26000 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.762440645773979e-05, | |
| "loss": 0.4894, | |
| "step": 26100 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.757692307692308e-05, | |
| "loss": 0.4806, | |
| "step": 26200 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.752943969610637e-05, | |
| "loss": 0.4815, | |
| "step": 26300 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.748195631528965e-05, | |
| "loss": 0.4797, | |
| "step": 26400 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.7434472934472937e-05, | |
| "loss": 0.4782, | |
| "step": 26500 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.738698955365622e-05, | |
| "loss": 0.4766, | |
| "step": 26600 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.7339506172839506e-05, | |
| "loss": 0.4712, | |
| "step": 26700 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.7292022792022794e-05, | |
| "loss": 0.487, | |
| "step": 26800 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 3.724453941120608e-05, | |
| "loss": 0.4697, | |
| "step": 26900 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 3.719705603038936e-05, | |
| "loss": 0.4808, | |
| "step": 27000 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 3.714957264957265e-05, | |
| "loss": 0.4872, | |
| "step": 27100 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 3.710208926875593e-05, | |
| "loss": 0.4828, | |
| "step": 27200 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 3.705460588793923e-05, | |
| "loss": 0.4717, | |
| "step": 27300 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 3.700712250712251e-05, | |
| "loss": 0.4769, | |
| "step": 27400 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 3.69596391263058e-05, | |
| "loss": 0.4776, | |
| "step": 27500 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 3.691215574548908e-05, | |
| "loss": 0.4774, | |
| "step": 27600 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 3.6864672364672366e-05, | |
| "loss": 0.4837, | |
| "step": 27700 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 3.6817188983855654e-05, | |
| "loss": 0.4661, | |
| "step": 27800 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 3.676970560303894e-05, | |
| "loss": 0.4705, | |
| "step": 27900 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 3.6722697056030394e-05, | |
| "loss": 0.486, | |
| "step": 28000 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 3.6675213675213676e-05, | |
| "loss": 0.4775, | |
| "step": 28100 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 3.6627730294396964e-05, | |
| "loss": 0.4763, | |
| "step": 28200 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.6580246913580245e-05, | |
| "loss": 0.4753, | |
| "step": 28300 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.653276353276354e-05, | |
| "loss": 0.4813, | |
| "step": 28400 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.648528015194682e-05, | |
| "loss": 0.4744, | |
| "step": 28500 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.643779677113011e-05, | |
| "loss": 0.4831, | |
| "step": 28600 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.639031339031339e-05, | |
| "loss": 0.4721, | |
| "step": 28700 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.634283000949668e-05, | |
| "loss": 0.4927, | |
| "step": 28800 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.6295346628679966e-05, | |
| "loss": 0.4843, | |
| "step": 28900 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.624786324786325e-05, | |
| "loss": 0.4821, | |
| "step": 29000 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.6200379867046536e-05, | |
| "loss": 0.4766, | |
| "step": 29100 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.615289648622982e-05, | |
| "loss": 0.4697, | |
| "step": 29200 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.6105413105413105e-05, | |
| "loss": 0.4748, | |
| "step": 29300 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 3.605792972459639e-05, | |
| "loss": 0.4701, | |
| "step": 29400 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 3.601044634377968e-05, | |
| "loss": 0.4787, | |
| "step": 29500 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 3.596296296296296e-05, | |
| "loss": 0.4788, | |
| "step": 29600 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 3.591547958214625e-05, | |
| "loss": 0.4732, | |
| "step": 29700 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 3.586799620132953e-05, | |
| "loss": 0.4737, | |
| "step": 29800 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 3.582051282051283e-05, | |
| "loss": 0.4751, | |
| "step": 29900 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 3.577302943969611e-05, | |
| "loss": 0.4746, | |
| "step": 30000 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "eval_binary_F1_negative": 0.8814589763957934, | |
| "eval_binary_F1_postive": 0.1231146706458157, | |
| "eval_loss": 0.44063252210617065, | |
| "eval_macro_F1": 0.5022868235208046, | |
| "eval_micro_F1": 0.7911510294734945, | |
| "eval_runtime": 6110.1654, | |
| "eval_samples_per_second": 1550.987, | |
| "eval_steps_per_second": 48.468, | |
| "step": 30000 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 3.5725546058879396e-05, | |
| "loss": 0.4803, | |
| "step": 30100 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 3.567853751187085e-05, | |
| "loss": 0.4874, | |
| "step": 30200 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 3.563105413105413e-05, | |
| "loss": 0.4888, | |
| "step": 30300 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 3.558404558404558e-05, | |
| "loss": 0.4766, | |
| "step": 30400 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 3.553703703703704e-05, | |
| "loss": 0.4733, | |
| "step": 30500 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 3.548955365622033e-05, | |
| "loss": 0.4759, | |
| "step": 30600 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 3.544207027540361e-05, | |
| "loss": 0.4834, | |
| "step": 30700 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 3.53945868945869e-05, | |
| "loss": 0.4774, | |
| "step": 30800 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 3.534710351377018e-05, | |
| "loss": 0.4777, | |
| "step": 30900 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 3.529962013295347e-05, | |
| "loss": 0.4715, | |
| "step": 31000 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.5252136752136755e-05, | |
| "loss": 0.4784, | |
| "step": 31100 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.520465337132004e-05, | |
| "loss": 0.4706, | |
| "step": 31200 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.5157169990503325e-05, | |
| "loss": 0.476, | |
| "step": 31300 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.510968660968661e-05, | |
| "loss": 0.4732, | |
| "step": 31400 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 3.50622032288699e-05, | |
| "loss": 0.4894, | |
| "step": 31500 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 3.501471984805318e-05, | |
| "loss": 0.4862, | |
| "step": 31600 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 3.496723646723647e-05, | |
| "loss": 0.4799, | |
| "step": 31700 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 3.492022792022792e-05, | |
| "loss": 0.4777, | |
| "step": 31800 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 3.4872744539411204e-05, | |
| "loss": 0.4837, | |
| "step": 31900 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 3.482526115859449e-05, | |
| "loss": 0.4796, | |
| "step": 32000 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 3.477777777777778e-05, | |
| "loss": 0.4756, | |
| "step": 32100 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.473029439696107e-05, | |
| "loss": 0.4794, | |
| "step": 32200 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.468281101614435e-05, | |
| "loss": 0.475, | |
| "step": 32300 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.463532763532764e-05, | |
| "loss": 0.4813, | |
| "step": 32400 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.458784425451092e-05, | |
| "loss": 0.4802, | |
| "step": 32500 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.454036087369421e-05, | |
| "loss": 0.4769, | |
| "step": 32600 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.4492877492877495e-05, | |
| "loss": 0.4793, | |
| "step": 32700 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.444539411206078e-05, | |
| "loss": 0.4838, | |
| "step": 32800 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.4397910731244064e-05, | |
| "loss": 0.48, | |
| "step": 32900 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.435042735042735e-05, | |
| "loss": 0.4758, | |
| "step": 33000 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.430294396961064e-05, | |
| "loss": 0.4825, | |
| "step": 33100 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 3.425546058879393e-05, | |
| "loss": 0.4748, | |
| "step": 33200 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 3.420797720797721e-05, | |
| "loss": 0.4692, | |
| "step": 33300 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 3.41604938271605e-05, | |
| "loss": 0.4714, | |
| "step": 33400 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 3.411301044634378e-05, | |
| "loss": 0.4733, | |
| "step": 33500 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 3.406552706552707e-05, | |
| "loss": 0.4707, | |
| "step": 33600 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 3.4018043684710355e-05, | |
| "loss": 0.4807, | |
| "step": 33700 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 3.3970560303893636e-05, | |
| "loss": 0.4844, | |
| "step": 33800 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.3923076923076924e-05, | |
| "loss": 0.4756, | |
| "step": 33900 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.387559354226021e-05, | |
| "loss": 0.4777, | |
| "step": 34000 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.382811016144349e-05, | |
| "loss": 0.4699, | |
| "step": 34100 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.378062678062678e-05, | |
| "loss": 0.4824, | |
| "step": 34200 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.373314339981007e-05, | |
| "loss": 0.4788, | |
| "step": 34300 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.368566001899335e-05, | |
| "loss": 0.4786, | |
| "step": 34400 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.363817663817664e-05, | |
| "loss": 0.4784, | |
| "step": 34500 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 3.359069325735992e-05, | |
| "loss": 0.4784, | |
| "step": 34600 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 3.3543209876543215e-05, | |
| "loss": 0.4796, | |
| "step": 34700 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 3.3495726495726496e-05, | |
| "loss": 0.4789, | |
| "step": 34800 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 3.3448243114909784e-05, | |
| "loss": 0.4703, | |
| "step": 34900 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 3.3400759734093065e-05, | |
| "loss": 0.4721, | |
| "step": 35000 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_binary_F1_negative": 0.878870945062303, | |
| "eval_binary_F1_postive": 0.12208019620420067, | |
| "eval_loss": 0.41444191336631775, | |
| "eval_macro_F1": 0.5004755706332519, | |
| "eval_micro_F1": 0.7871143233581511, | |
| "eval_runtime": 6169.1187, | |
| "eval_samples_per_second": 1536.165, | |
| "eval_steps_per_second": 48.005, | |
| "step": 35000 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 3.3353276353276354e-05, | |
| "loss": 0.4773, | |
| "step": 35100 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 3.330579297245964e-05, | |
| "loss": 0.4738, | |
| "step": 35200 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 3.325830959164293e-05, | |
| "loss": 0.4691, | |
| "step": 35300 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 3.321082621082621e-05, | |
| "loss": 0.4695, | |
| "step": 35400 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 3.316381766381766e-05, | |
| "loss": 0.4628, | |
| "step": 35500 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 3.311633428300095e-05, | |
| "loss": 0.4748, | |
| "step": 35600 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 3.306885090218424e-05, | |
| "loss": 0.4624, | |
| "step": 35700 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 3.302184235517569e-05, | |
| "loss": 0.4598, | |
| "step": 35800 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 3.297435897435898e-05, | |
| "loss": 0.4714, | |
| "step": 35900 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 3.292687559354226e-05, | |
| "loss": 0.4776, | |
| "step": 36000 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 3.287939221272555e-05, | |
| "loss": 0.472, | |
| "step": 36100 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 3.283190883190883e-05, | |
| "loss": 0.4669, | |
| "step": 36200 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 3.278490028490029e-05, | |
| "loss": 0.466, | |
| "step": 36300 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 3.273741690408357e-05, | |
| "loss": 0.4624, | |
| "step": 36400 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 3.268993352326686e-05, | |
| "loss": 0.4704, | |
| "step": 36500 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 3.264245014245014e-05, | |
| "loss": 0.4612, | |
| "step": 36600 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 3.259496676163343e-05, | |
| "loss": 0.4636, | |
| "step": 36700 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 3.2547483380816716e-05, | |
| "loss": 0.4678, | |
| "step": 36800 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 3.2500000000000004e-05, | |
| "loss": 0.4613, | |
| "step": 36900 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 3.2452516619183285e-05, | |
| "loss": 0.4637, | |
| "step": 37000 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 3.240503323836657e-05, | |
| "loss": 0.4599, | |
| "step": 37100 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 3.2357549857549855e-05, | |
| "loss": 0.4676, | |
| "step": 37200 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 3.2310541310541314e-05, | |
| "loss": 0.4823, | |
| "step": 37300 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 3.22630579297246e-05, | |
| "loss": 0.4643, | |
| "step": 37400 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 3.221557454890788e-05, | |
| "loss": 0.455, | |
| "step": 37500 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 3.216809116809117e-05, | |
| "loss": 0.4588, | |
| "step": 37600 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 3.212060778727445e-05, | |
| "loss": 0.466, | |
| "step": 37700 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 3.207312440645775e-05, | |
| "loss": 0.4606, | |
| "step": 37800 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 3.202564102564103e-05, | |
| "loss": 0.4663, | |
| "step": 37900 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 3.1978157644824316e-05, | |
| "loss": 0.4631, | |
| "step": 38000 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 3.19306742640076e-05, | |
| "loss": 0.4588, | |
| "step": 38100 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 3.1883190883190886e-05, | |
| "loss": 0.465, | |
| "step": 38200 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 3.183570750237417e-05, | |
| "loss": 0.4595, | |
| "step": 38300 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 3.1788224121557455e-05, | |
| "loss": 0.4648, | |
| "step": 38400 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 3.174074074074074e-05, | |
| "loss": 0.4607, | |
| "step": 38500 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 3.169325735992403e-05, | |
| "loss": 0.4589, | |
| "step": 38600 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 3.164577397910731e-05, | |
| "loss": 0.4634, | |
| "step": 38700 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 3.15982905982906e-05, | |
| "loss": 0.4671, | |
| "step": 38800 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 3.155080721747389e-05, | |
| "loss": 0.4595, | |
| "step": 38900 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 3.150332383665717e-05, | |
| "loss": 0.4578, | |
| "step": 39000 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 3.145584045584046e-05, | |
| "loss": 0.459, | |
| "step": 39100 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 3.140835707502374e-05, | |
| "loss": 0.4602, | |
| "step": 39200 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 3.136087369420703e-05, | |
| "loss": 0.4546, | |
| "step": 39300 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 3.1313390313390315e-05, | |
| "loss": 0.4531, | |
| "step": 39400 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 3.126638176638177e-05, | |
| "loss": 0.4612, | |
| "step": 39500 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 3.1218898385565055e-05, | |
| "loss": 0.4519, | |
| "step": 39600 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 3.117141500474834e-05, | |
| "loss": 0.4563, | |
| "step": 39700 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 3.1123931623931625e-05, | |
| "loss": 0.457, | |
| "step": 39800 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 3.107644824311491e-05, | |
| "loss": 0.4514, | |
| "step": 39900 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 3.1029439696106365e-05, | |
| "loss": 0.4661, | |
| "step": 40000 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "eval_binary_F1_negative": 0.9133590402149661, | |
| "eval_binary_F1_postive": 0.14330759230951728, | |
| "eval_loss": 0.3721241354942322, | |
| "eval_macro_F1": 0.5283333162622417, | |
| "eval_micro_F1": 0.842633251399789, | |
| "eval_runtime": 6347.9789, | |
| "eval_samples_per_second": 1492.882, | |
| "eval_steps_per_second": 46.653, | |
| "step": 40000 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 3.098195631528965e-05, | |
| "loss": 0.4573, | |
| "step": 40100 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 3.0934472934472934e-05, | |
| "loss": 0.456, | |
| "step": 40200 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 3.088698955365622e-05, | |
| "loss": 0.4631, | |
| "step": 40300 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 3.0839506172839504e-05, | |
| "loss": 0.4463, | |
| "step": 40400 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 3.07920227920228e-05, | |
| "loss": 0.457, | |
| "step": 40500 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 3.074453941120608e-05, | |
| "loss": 0.465, | |
| "step": 40600 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 3.069705603038937e-05, | |
| "loss": 0.457, | |
| "step": 40700 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 3.064957264957265e-05, | |
| "loss": 0.4625, | |
| "step": 40800 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 3.060208926875594e-05, | |
| "loss": 0.4528, | |
| "step": 40900 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 3.0554605887939225e-05, | |
| "loss": 0.4646, | |
| "step": 41000 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 3.050712250712251e-05, | |
| "loss": 0.4601, | |
| "step": 41100 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 3.0459639126305795e-05, | |
| "loss": 0.4541, | |
| "step": 41200 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 3.041215574548908e-05, | |
| "loss": 0.4572, | |
| "step": 41300 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 3.0364672364672364e-05, | |
| "loss": 0.4579, | |
| "step": 41400 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 3.0317188983855655e-05, | |
| "loss": 0.4566, | |
| "step": 41500 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 3.026970560303894e-05, | |
| "loss": 0.4622, | |
| "step": 41600 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 3.0222222222222225e-05, | |
| "loss": 0.4588, | |
| "step": 41700 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 3.017473884140551e-05, | |
| "loss": 0.4631, | |
| "step": 41800 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 3.0127255460588794e-05, | |
| "loss": 0.4651, | |
| "step": 41900 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 3.0079772079772085e-05, | |
| "loss": 0.4594, | |
| "step": 42000 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 3.003228869895537e-05, | |
| "loss": 0.4587, | |
| "step": 42100 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 2.9984805318138655e-05, | |
| "loss": 0.4486, | |
| "step": 42200 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 2.993732193732194e-05, | |
| "loss": 0.4592, | |
| "step": 42300 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 2.9889838556505224e-05, | |
| "loss": 0.4656, | |
| "step": 42400 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 2.984235517568851e-05, | |
| "loss": 0.4544, | |
| "step": 42500 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 2.9794871794871797e-05, | |
| "loss": 0.4525, | |
| "step": 42600 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 2.974738841405508e-05, | |
| "loss": 0.4411, | |
| "step": 42700 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 2.9700379867046534e-05, | |
| "loss": 0.4679, | |
| "step": 42800 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 2.9652896486229818e-05, | |
| "loss": 0.4485, | |
| "step": 42900 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 2.9605413105413103e-05, | |
| "loss": 0.455, | |
| "step": 43000 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 2.9557929724596394e-05, | |
| "loss": 0.4481, | |
| "step": 43100 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 2.951044634377968e-05, | |
| "loss": 0.4556, | |
| "step": 43200 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 2.9462962962962964e-05, | |
| "loss": 0.4459, | |
| "step": 43300 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 2.941547958214625e-05, | |
| "loss": 0.4632, | |
| "step": 43400 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 2.9367996201329533e-05, | |
| "loss": 0.4524, | |
| "step": 43500 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 2.9320512820512824e-05, | |
| "loss": 0.4394, | |
| "step": 43600 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 2.927302943969611e-05, | |
| "loss": 0.4606, | |
| "step": 43700 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 2.9225546058879394e-05, | |
| "loss": 0.4457, | |
| "step": 43800 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 2.917806267806268e-05, | |
| "loss": 0.4616, | |
| "step": 43900 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 2.9130579297245963e-05, | |
| "loss": 0.4559, | |
| "step": 44000 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 2.9083095916429255e-05, | |
| "loss": 0.4539, | |
| "step": 44100 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 2.903561253561254e-05, | |
| "loss": 0.4553, | |
| "step": 44200 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 2.8988129154795824e-05, | |
| "loss": 0.4452, | |
| "step": 44300 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 2.894064577397911e-05, | |
| "loss": 0.4512, | |
| "step": 44400 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 2.8893162393162393e-05, | |
| "loss": 0.4651, | |
| "step": 44500 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 2.8845679012345678e-05, | |
| "loss": 0.4481, | |
| "step": 44600 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 2.879819563152897e-05, | |
| "loss": 0.4515, | |
| "step": 44700 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 2.8750712250712254e-05, | |
| "loss": 0.4492, | |
| "step": 44800 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 2.870322886989554e-05, | |
| "loss": 0.4439, | |
| "step": 44900 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 2.8655745489078823e-05, | |
| "loss": 0.4471, | |
| "step": 45000 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "eval_binary_F1_negative": 0.8891971728549981, | |
| "eval_binary_F1_postive": 0.12781955239145038, | |
| "eval_loss": 0.4176163971424103, | |
| "eval_macro_F1": 0.5085083626232243, | |
| "eval_micro_F1": 0.8033739497757995, | |
| "eval_runtime": 6153.163, | |
| "eval_samples_per_second": 1540.149, | |
| "eval_steps_per_second": 48.13, | |
| "step": 45000 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 2.8608262108262108e-05, | |
| "loss": 0.4577, | |
| "step": 45100 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 2.8561253561253564e-05, | |
| "loss": 0.4507, | |
| "step": 45200 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 2.8513770180436848e-05, | |
| "loss": 0.452, | |
| "step": 45300 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 2.8466286799620133e-05, | |
| "loss": 0.4532, | |
| "step": 45400 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 2.8418803418803418e-05, | |
| "loss": 0.4559, | |
| "step": 45500 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 2.8371320037986702e-05, | |
| "loss": 0.4411, | |
| "step": 45600 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 2.8323836657169994e-05, | |
| "loss": 0.4516, | |
| "step": 45700 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 2.827635327635328e-05, | |
| "loss": 0.4486, | |
| "step": 45800 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 2.822934472934473e-05, | |
| "loss": 0.4618, | |
| "step": 45900 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 2.8181861348528015e-05, | |
| "loss": 0.4604, | |
| "step": 46000 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 2.81343779677113e-05, | |
| "loss": 0.4586, | |
| "step": 46100 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 2.808689458689459e-05, | |
| "loss": 0.4429, | |
| "step": 46200 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 2.8039411206077876e-05, | |
| "loss": 0.4513, | |
| "step": 46300 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 2.799192782526116e-05, | |
| "loss": 0.4526, | |
| "step": 46400 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 2.7944444444444445e-05, | |
| "loss": 0.4423, | |
| "step": 46500 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 2.789696106362773e-05, | |
| "loss": 0.454, | |
| "step": 46600 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 2.7849477682811015e-05, | |
| "loss": 0.4471, | |
| "step": 46700 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 2.7801994301994306e-05, | |
| "loss": 0.4477, | |
| "step": 46800 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 2.775451092117759e-05, | |
| "loss": 0.4615, | |
| "step": 46900 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 2.7707027540360875e-05, | |
| "loss": 0.4612, | |
| "step": 47000 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 2.765954415954416e-05, | |
| "loss": 0.4461, | |
| "step": 47100 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 2.7612060778727445e-05, | |
| "loss": 0.4588, | |
| "step": 47200 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 2.7564577397910736e-05, | |
| "loss": 0.4518, | |
| "step": 47300 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 2.751709401709402e-05, | |
| "loss": 0.4511, | |
| "step": 47400 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 2.7469610636277305e-05, | |
| "loss": 0.4452, | |
| "step": 47500 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 2.742212725546059e-05, | |
| "loss": 0.4621, | |
| "step": 47600 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 2.7374643874643875e-05, | |
| "loss": 0.4574, | |
| "step": 47700 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 2.7327160493827163e-05, | |
| "loss": 0.4472, | |
| "step": 47800 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 2.7279677113010447e-05, | |
| "loss": 0.4573, | |
| "step": 47900 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 2.7232193732193732e-05, | |
| "loss": 0.4561, | |
| "step": 48000 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 2.7184710351377017e-05, | |
| "loss": 0.4571, | |
| "step": 48100 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 2.7137226970560305e-05, | |
| "loss": 0.4553, | |
| "step": 48200 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 2.7089743589743593e-05, | |
| "loss": 0.4414, | |
| "step": 48300 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 2.7042260208926878e-05, | |
| "loss": 0.4598, | |
| "step": 48400 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 2.6994776828110162e-05, | |
| "loss": 0.4544, | |
| "step": 48500 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 2.6947293447293447e-05, | |
| "loss": 0.4509, | |
| "step": 48600 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 2.689981006647673e-05, | |
| "loss": 0.4583, | |
| "step": 48700 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 2.6852326685660016e-05, | |
| "loss": 0.455, | |
| "step": 48800 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 2.6805318138651475e-05, | |
| "loss": 0.4513, | |
| "step": 48900 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 2.675783475783476e-05, | |
| "loss": 0.4416, | |
| "step": 49000 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 2.6710351377018045e-05, | |
| "loss": 0.4435, | |
| "step": 49100 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 2.666286799620133e-05, | |
| "loss": 0.4602, | |
| "step": 49200 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 2.6615384615384614e-05, | |
| "loss": 0.4406, | |
| "step": 49300 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 2.6567901234567905e-05, | |
| "loss": 0.4437, | |
| "step": 49400 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 2.652041785375119e-05, | |
| "loss": 0.4594, | |
| "step": 49500 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 2.6472934472934475e-05, | |
| "loss": 0.4455, | |
| "step": 49600 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 2.642545109211776e-05, | |
| "loss": 0.4486, | |
| "step": 49700 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 2.6377967711301044e-05, | |
| "loss": 0.4428, | |
| "step": 49800 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 2.6330484330484335e-05, | |
| "loss": 0.4406, | |
| "step": 49900 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 2.628300094966762e-05, | |
| "loss": 0.4584, | |
| "step": 50000 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "eval_binary_F1_negative": 0.8831693513653097, | |
| "eval_binary_F1_postive": 0.12474272160887594, | |
| "eval_loss": 0.4173279404640198, | |
| "eval_macro_F1": 0.5039560364870929, | |
| "eval_micro_F1": 0.7938552163148983, | |
| "eval_runtime": 6080.6805, | |
| "eval_samples_per_second": 1558.507, | |
| "eval_steps_per_second": 48.703, | |
| "step": 50000 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 2.6235517568850905e-05, | |
| "loss": 0.4532, | |
| "step": 50100 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 2.618803418803419e-05, | |
| "loss": 0.4427, | |
| "step": 50200 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 2.6140550807217474e-05, | |
| "loss": 0.4545, | |
| "step": 50300 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 2.6093067426400762e-05, | |
| "loss": 0.4506, | |
| "step": 50400 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 2.6046058879392214e-05, | |
| "loss": 0.4551, | |
| "step": 50500 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 2.59985754985755e-05, | |
| "loss": 0.4551, | |
| "step": 50600 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 2.5951092117758784e-05, | |
| "loss": 0.4546, | |
| "step": 50700 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 2.590360873694207e-05, | |
| "loss": 0.4546, | |
| "step": 50800 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 2.5856125356125356e-05, | |
| "loss": 0.4535, | |
| "step": 50900 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 2.5808641975308644e-05, | |
| "loss": 0.4506, | |
| "step": 51000 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 2.576115859449193e-05, | |
| "loss": 0.4493, | |
| "step": 51100 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 2.571415004748338e-05, | |
| "loss": 0.4525, | |
| "step": 51200 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 2.5666666666666666e-05, | |
| "loss": 0.4493, | |
| "step": 51300 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 2.561918328584995e-05, | |
| "loss": 0.4549, | |
| "step": 51400 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 2.5571699905033242e-05, | |
| "loss": 0.4661, | |
| "step": 51500 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 2.5524216524216527e-05, | |
| "loss": 0.4558, | |
| "step": 51600 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 2.547673314339981e-05, | |
| "loss": 0.4498, | |
| "step": 51700 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 2.5429724596391264e-05, | |
| "loss": 0.4462, | |
| "step": 51800 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 2.538224121557455e-05, | |
| "loss": 0.461, | |
| "step": 51900 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 2.533475783475784e-05, | |
| "loss": 0.4597, | |
| "step": 52000 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 2.5287274453941124e-05, | |
| "loss": 0.4527, | |
| "step": 52100 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 2.523979107312441e-05, | |
| "loss": 0.4574, | |
| "step": 52200 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 2.5192307692307694e-05, | |
| "loss": 0.4466, | |
| "step": 52300 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 2.514482431149098e-05, | |
| "loss": 0.4524, | |
| "step": 52400 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 2.5097340930674266e-05, | |
| "loss": 0.4449, | |
| "step": 52500 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 2.504985754985755e-05, | |
| "loss": 0.4492, | |
| "step": 52600 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 2.500237416904084e-05, | |
| "loss": 0.4515, | |
| "step": 52700 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 2.4954890788224124e-05, | |
| "loss": 0.4459, | |
| "step": 52800 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 2.490740740740741e-05, | |
| "loss": 0.4449, | |
| "step": 52900 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 2.4859924026590693e-05, | |
| "loss": 0.4385, | |
| "step": 53000 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 2.4812440645773978e-05, | |
| "loss": 0.4509, | |
| "step": 53100 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 2.4764957264957266e-05, | |
| "loss": 0.4543, | |
| "step": 53200 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 2.471747388414055e-05, | |
| "loss": 0.4381, | |
| "step": 53300 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 2.4669990503323835e-05, | |
| "loss": 0.4484, | |
| "step": 53400 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 2.4622507122507123e-05, | |
| "loss": 0.4529, | |
| "step": 53500 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 2.4575023741690408e-05, | |
| "loss": 0.4497, | |
| "step": 53600 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 2.4527540360873696e-05, | |
| "loss": 0.4384, | |
| "step": 53700 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 2.448005698005698e-05, | |
| "loss": 0.4417, | |
| "step": 53800 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 2.4432573599240265e-05, | |
| "loss": 0.4481, | |
| "step": 53900 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 2.4385090218423553e-05, | |
| "loss": 0.4485, | |
| "step": 54000 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 2.4337606837606838e-05, | |
| "loss": 0.4532, | |
| "step": 54100 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 2.4290123456790126e-05, | |
| "loss": 0.4446, | |
| "step": 54200 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 2.424264007597341e-05, | |
| "loss": 0.4462, | |
| "step": 54300 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 2.4195156695156695e-05, | |
| "loss": 0.4463, | |
| "step": 54400 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 2.4147673314339983e-05, | |
| "loss": 0.4546, | |
| "step": 54500 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 2.4100189933523268e-05, | |
| "loss": 0.455, | |
| "step": 54600 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 2.4052706552706553e-05, | |
| "loss": 0.466, | |
| "step": 54700 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 2.400522317188984e-05, | |
| "loss": 0.4458, | |
| "step": 54800 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 2.3957739791073125e-05, | |
| "loss": 0.4397, | |
| "step": 54900 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 2.3910256410256413e-05, | |
| "loss": 0.4461, | |
| "step": 55000 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "eval_binary_F1_negative": 0.903273104173285, | |
| "eval_binary_F1_postive": 0.13748510112399526, | |
| "eval_loss": 0.3925754725933075, | |
| "eval_macro_F1": 0.5203791026486402, | |
| "eval_micro_F1": 0.8260534742474929, | |
| "eval_runtime": 6018.9013, | |
| "eval_samples_per_second": 1574.504, | |
| "eval_steps_per_second": 49.203, | |
| "step": 55000 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 2.3862773029439698e-05, | |
| "loss": 0.4471, | |
| "step": 55100 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 2.3815289648622983e-05, | |
| "loss": 0.4476, | |
| "step": 55200 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 2.376828110161444e-05, | |
| "loss": 0.4426, | |
| "step": 55300 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 2.3720797720797723e-05, | |
| "loss": 0.4433, | |
| "step": 55400 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 2.3673314339981008e-05, | |
| "loss": 0.4433, | |
| "step": 55500 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 2.3625830959164292e-05, | |
| "loss": 0.4514, | |
| "step": 55600 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 2.357834757834758e-05, | |
| "loss": 0.4489, | |
| "step": 55700 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 2.3530864197530865e-05, | |
| "loss": 0.4438, | |
| "step": 55800 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 2.348338081671415e-05, | |
| "loss": 0.4422, | |
| "step": 55900 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 2.3436372269705602e-05, | |
| "loss": 0.449, | |
| "step": 56000 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 2.338888888888889e-05, | |
| "loss": 0.4517, | |
| "step": 56100 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 2.3341405508072175e-05, | |
| "loss": 0.4482, | |
| "step": 56200 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 2.3294396961063627e-05, | |
| "loss": 0.4483, | |
| "step": 56300 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 2.3246913580246915e-05, | |
| "loss": 0.4441, | |
| "step": 56400 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 2.31994301994302e-05, | |
| "loss": 0.4383, | |
| "step": 56500 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 2.3151946818613488e-05, | |
| "loss": 0.4412, | |
| "step": 56600 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 2.3104463437796772e-05, | |
| "loss": 0.4454, | |
| "step": 56700 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 2.3056980056980057e-05, | |
| "loss": 0.4504, | |
| "step": 56800 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 2.3009496676163345e-05, | |
| "loss": 0.4367, | |
| "step": 56900 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 2.296201329534663e-05, | |
| "loss": 0.4478, | |
| "step": 57000 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 2.2914529914529918e-05, | |
| "loss": 0.4382, | |
| "step": 57100 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 2.2867046533713202e-05, | |
| "loss": 0.45, | |
| "step": 57200 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 2.2819563152896487e-05, | |
| "loss": 0.4465, | |
| "step": 57300 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 2.2772079772079775e-05, | |
| "loss": 0.4453, | |
| "step": 57400 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 2.272459639126306e-05, | |
| "loss": 0.4501, | |
| "step": 57500 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 2.2677113010446345e-05, | |
| "loss": 0.4412, | |
| "step": 57600 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 2.2629629629629633e-05, | |
| "loss": 0.4433, | |
| "step": 57700 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 2.2582146248812917e-05, | |
| "loss": 0.4464, | |
| "step": 57800 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 2.2534662867996202e-05, | |
| "loss": 0.4455, | |
| "step": 57900 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 2.248717948717949e-05, | |
| "loss": 0.4472, | |
| "step": 58000 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 2.2439696106362775e-05, | |
| "loss": 0.4446, | |
| "step": 58100 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 2.239221272554606e-05, | |
| "loss": 0.4561, | |
| "step": 58200 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 2.2344729344729344e-05, | |
| "loss": 0.4442, | |
| "step": 58300 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 2.2297245963912632e-05, | |
| "loss": 0.4352, | |
| "step": 58400 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 2.2249762583095917e-05, | |
| "loss": 0.4517, | |
| "step": 58500 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 2.220275403608737e-05, | |
| "loss": 0.4532, | |
| "step": 58600 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 2.2155270655270657e-05, | |
| "loss": 0.4364, | |
| "step": 58700 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 2.210826210826211e-05, | |
| "loss": 0.4576, | |
| "step": 58800 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 2.2060778727445394e-05, | |
| "loss": 0.4488, | |
| "step": 58900 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 2.2013295346628682e-05, | |
| "loss": 0.4516, | |
| "step": 59000 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 2.1965811965811967e-05, | |
| "loss": 0.4396, | |
| "step": 59100 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 2.191832858499525e-05, | |
| "loss": 0.453, | |
| "step": 59200 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 2.187084520417854e-05, | |
| "loss": 0.4387, | |
| "step": 59300 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 2.1823361823361824e-05, | |
| "loss": 0.459, | |
| "step": 59400 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 2.1775878442545112e-05, | |
| "loss": 0.4446, | |
| "step": 59500 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 2.1728395061728397e-05, | |
| "loss": 0.4515, | |
| "step": 59600 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 2.168091168091168e-05, | |
| "loss": 0.4532, | |
| "step": 59700 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 2.163342830009497e-05, | |
| "loss": 0.4519, | |
| "step": 59800 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 2.1585944919278254e-05, | |
| "loss": 0.4587, | |
| "step": 59900 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 2.1538461538461542e-05, | |
| "loss": 0.4459, | |
| "step": 60000 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "eval_binary_F1_negative": 0.8868438433738486, | |
| "eval_binary_F1_postive": 0.1270438701597026, | |
| "eval_loss": 0.41710054874420166, | |
| "eval_macro_F1": 0.5069438567667756, | |
| "eval_micro_F1": 0.7996569723110768, | |
| "eval_runtime": 6097.117, | |
| "eval_samples_per_second": 1554.306, | |
| "eval_steps_per_second": 48.572, | |
| "step": 60000 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 2.1490978157644827e-05, | |
| "loss": 0.444, | |
| "step": 60100 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 2.144349477682811e-05, | |
| "loss": 0.452, | |
| "step": 60200 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 2.13960113960114e-05, | |
| "loss": 0.46, | |
| "step": 60300 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 2.1348528015194684e-05, | |
| "loss": 0.4513, | |
| "step": 60400 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 2.130104463437797e-05, | |
| "loss": 0.4543, | |
| "step": 60500 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 2.1253561253561253e-05, | |
| "loss": 0.4554, | |
| "step": 60600 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 2.1206077872744538e-05, | |
| "loss": 0.4479, | |
| "step": 60700 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 2.1158594491927826e-05, | |
| "loss": 0.4563, | |
| "step": 60800 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 2.111111111111111e-05, | |
| "loss": 0.4525, | |
| "step": 60900 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 2.1063627730294395e-05, | |
| "loss": 0.4515, | |
| "step": 61000 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 2.1016144349477683e-05, | |
| "loss": 0.4542, | |
| "step": 61100 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 2.0969135802469136e-05, | |
| "loss": 0.4634, | |
| "step": 61200 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 2.0921652421652424e-05, | |
| "loss": 0.4491, | |
| "step": 61300 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 2.087416904083571e-05, | |
| "loss": 0.4521, | |
| "step": 61400 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 2.0826685660018993e-05, | |
| "loss": 0.4495, | |
| "step": 61500 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 2.077920227920228e-05, | |
| "loss": 0.4469, | |
| "step": 61600 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 2.0732193732193733e-05, | |
| "loss": 0.446, | |
| "step": 61700 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 2.0684710351377018e-05, | |
| "loss": 0.4416, | |
| "step": 61800 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 2.0637226970560306e-05, | |
| "loss": 0.4544, | |
| "step": 61900 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 2.058974358974359e-05, | |
| "loss": 0.4422, | |
| "step": 62000 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 2.054226020892688e-05, | |
| "loss": 0.4489, | |
| "step": 62100 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 2.0494776828110163e-05, | |
| "loss": 0.4528, | |
| "step": 62200 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 2.0447293447293448e-05, | |
| "loss": 0.4574, | |
| "step": 62300 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 2.0399810066476736e-05, | |
| "loss": 0.447, | |
| "step": 62400 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 2.035232668566002e-05, | |
| "loss": 0.4434, | |
| "step": 62500 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 2.0304843304843306e-05, | |
| "loss": 0.449, | |
| "step": 62600 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 2.0257359924026594e-05, | |
| "loss": 0.4447, | |
| "step": 62700 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 2.0209876543209878e-05, | |
| "loss": 0.4474, | |
| "step": 62800 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 2.0162393162393163e-05, | |
| "loss": 0.4354, | |
| "step": 62900 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 2.0114909781576448e-05, | |
| "loss": 0.4361, | |
| "step": 63000 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 2.0067426400759736e-05, | |
| "loss": 0.4521, | |
| "step": 63100 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 2.001994301994302e-05, | |
| "loss": 0.4446, | |
| "step": 63200 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 1.9972459639126305e-05, | |
| "loss": 0.4442, | |
| "step": 63300 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 1.9924976258309593e-05, | |
| "loss": 0.4444, | |
| "step": 63400 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 1.9877492877492878e-05, | |
| "loss": 0.4477, | |
| "step": 63500 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 1.9830009496676162e-05, | |
| "loss": 0.4449, | |
| "step": 63600 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 1.978252611585945e-05, | |
| "loss": 0.4495, | |
| "step": 63700 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 1.9735042735042735e-05, | |
| "loss": 0.4433, | |
| "step": 63800 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 1.9688034188034187e-05, | |
| "loss": 0.4592, | |
| "step": 63900 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 1.9640550807217475e-05, | |
| "loss": 0.4553, | |
| "step": 64000 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 1.959306742640076e-05, | |
| "loss": 0.45, | |
| "step": 64100 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 1.9545584045584048e-05, | |
| "loss": 0.4397, | |
| "step": 64200 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 1.9498100664767333e-05, | |
| "loss": 0.441, | |
| "step": 64300 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 1.9450617283950617e-05, | |
| "loss": 0.4417, | |
| "step": 64400 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 1.9403133903133905e-05, | |
| "loss": 0.4389, | |
| "step": 64500 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 1.935565052231719e-05, | |
| "loss": 0.4433, | |
| "step": 64600 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 1.9308167141500475e-05, | |
| "loss": 0.4426, | |
| "step": 64700 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 1.9260683760683763e-05, | |
| "loss": 0.4397, | |
| "step": 64800 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 1.9213200379867047e-05, | |
| "loss": 0.4381, | |
| "step": 64900 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 1.9165716999050335e-05, | |
| "loss": 0.4448, | |
| "step": 65000 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "eval_binary_F1_negative": 0.8923812371624389, | |
| "eval_binary_F1_postive": 0.13096124463792627, | |
| "eval_loss": 0.4017094373703003, | |
| "eval_macro_F1": 0.5116712409001826, | |
| "eval_micro_F1": 0.808479689211089, | |
| "eval_runtime": 6068.1256, | |
| "eval_samples_per_second": 1561.732, | |
| "eval_steps_per_second": 48.804, | |
| "step": 65000 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 1.911823361823362e-05, | |
| "loss": 0.4439, | |
| "step": 65100 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 1.9070750237416905e-05, | |
| "loss": 0.453, | |
| "step": 65200 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 1.9023266856600193e-05, | |
| "loss": 0.4544, | |
| "step": 65300 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 1.8976258309591645e-05, | |
| "loss": 0.4609, | |
| "step": 65400 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 1.892877492877493e-05, | |
| "loss": 0.4479, | |
| "step": 65500 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 1.8881291547958214e-05, | |
| "loss": 0.4445, | |
| "step": 65600 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 1.88338081671415e-05, | |
| "loss": 0.4449, | |
| "step": 65700 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 1.8786324786324787e-05, | |
| "loss": 0.4522, | |
| "step": 65800 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 1.8738841405508072e-05, | |
| "loss": 0.4489, | |
| "step": 65900 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 1.8691358024691356e-05, | |
| "loss": 0.4516, | |
| "step": 66000 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 1.8643874643874644e-05, | |
| "loss": 0.44, | |
| "step": 66100 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 1.859639126305793e-05, | |
| "loss": 0.4482, | |
| "step": 66200 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 1.8548907882241217e-05, | |
| "loss": 0.4332, | |
| "step": 66300 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 1.850189933523267e-05, | |
| "loss": 0.4459, | |
| "step": 66400 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 1.8454415954415954e-05, | |
| "loss": 0.4442, | |
| "step": 66500 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 1.8406932573599242e-05, | |
| "loss": 0.457, | |
| "step": 66600 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 1.8359449192782527e-05, | |
| "loss": 0.4556, | |
| "step": 66700 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 1.831196581196581e-05, | |
| "loss": 0.45, | |
| "step": 66800 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 1.82644824311491e-05, | |
| "loss": 0.4496, | |
| "step": 66900 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 1.8217473884140552e-05, | |
| "loss": 0.4568, | |
| "step": 67000 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 1.816999050332384e-05, | |
| "loss": 0.4474, | |
| "step": 67100 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 1.8122507122507124e-05, | |
| "loss": 0.4452, | |
| "step": 67200 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 1.807502374169041e-05, | |
| "loss": 0.4524, | |
| "step": 67300 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 1.8027540360873697e-05, | |
| "loss": 0.4444, | |
| "step": 67400 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 1.7980056980056982e-05, | |
| "loss": 0.4551, | |
| "step": 67500 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 1.7932573599240267e-05, | |
| "loss": 0.451, | |
| "step": 67600 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 1.7885090218423555e-05, | |
| "loss": 0.4405, | |
| "step": 67700 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 1.783760683760684e-05, | |
| "loss": 0.4516, | |
| "step": 67800 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 1.7790123456790124e-05, | |
| "loss": 0.453, | |
| "step": 67900 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 1.774264007597341e-05, | |
| "loss": 0.4472, | |
| "step": 68000 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 1.7695156695156697e-05, | |
| "loss": 0.4419, | |
| "step": 68100 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 1.764767331433998e-05, | |
| "loss": 0.4488, | |
| "step": 68200 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 1.7600189933523266e-05, | |
| "loss": 0.4447, | |
| "step": 68300 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 1.7552706552706554e-05, | |
| "loss": 0.4404, | |
| "step": 68400 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 1.750522317188984e-05, | |
| "loss": 0.4364, | |
| "step": 68500 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 1.745821462488129e-05, | |
| "loss": 0.4407, | |
| "step": 68600 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 1.741073124406458e-05, | |
| "loss": 0.4412, | |
| "step": 68700 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 1.7363247863247864e-05, | |
| "loss": 0.45, | |
| "step": 68800 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 1.7315764482431148e-05, | |
| "loss": 0.455, | |
| "step": 68900 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 1.7268281101614436e-05, | |
| "loss": 0.4452, | |
| "step": 69000 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 1.722079772079772e-05, | |
| "loss": 0.441, | |
| "step": 69100 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 1.717331433998101e-05, | |
| "loss": 0.4394, | |
| "step": 69200 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 1.7125830959164294e-05, | |
| "loss": 0.4526, | |
| "step": 69300 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 1.707834757834758e-05, | |
| "loss": 0.4489, | |
| "step": 69400 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 1.7030864197530866e-05, | |
| "loss": 0.4473, | |
| "step": 69500 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 1.698338081671415e-05, | |
| "loss": 0.4472, | |
| "step": 69600 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 1.6935897435897436e-05, | |
| "loss": 0.4499, | |
| "step": 69700 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 1.6888414055080724e-05, | |
| "loss": 0.4471, | |
| "step": 69800 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 1.684093067426401e-05, | |
| "loss": 0.4486, | |
| "step": 69900 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 1.6793447293447296e-05, | |
| "loss": 0.4408, | |
| "step": 70000 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "eval_binary_F1_negative": 0.8824636576561534, | |
| "eval_binary_F1_postive": 0.12512000185320726, | |
| "eval_loss": 0.42592427134513855, | |
| "eval_macro_F1": 0.5037918297546804, | |
| "eval_micro_F1": 0.7927680333817816, | |
| "eval_runtime": 6059.415, | |
| "eval_samples_per_second": 1563.977, | |
| "eval_steps_per_second": 48.874, | |
| "step": 70000 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 1.674596391263058e-05, | |
| "loss": 0.4405, | |
| "step": 70100 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 1.6698480531813866e-05, | |
| "loss": 0.4459, | |
| "step": 70200 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 1.6650997150997154e-05, | |
| "loss": 0.4429, | |
| "step": 70300 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 1.660351377018044e-05, | |
| "loss": 0.4395, | |
| "step": 70400 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 1.6556030389363723e-05, | |
| "loss": 0.4405, | |
| "step": 70500 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 1.650854700854701e-05, | |
| "loss": 0.4326, | |
| "step": 70600 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 1.6461063627730296e-05, | |
| "loss": 0.4455, | |
| "step": 70700 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 1.641358024691358e-05, | |
| "loss": 0.4324, | |
| "step": 70800 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 1.6366096866096865e-05, | |
| "loss": 0.4234, | |
| "step": 70900 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 1.6319088319088317e-05, | |
| "loss": 0.4387, | |
| "step": 71000 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 1.6272079772079773e-05, | |
| "loss": 0.448, | |
| "step": 71100 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 1.6224596391263058e-05, | |
| "loss": 0.4339, | |
| "step": 71200 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 1.6177113010446346e-05, | |
| "loss": 0.4398, | |
| "step": 71300 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 1.612962962962963e-05, | |
| "loss": 0.4359, | |
| "step": 71400 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 1.6082146248812915e-05, | |
| "loss": 0.4351, | |
| "step": 71500 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 1.6034662867996203e-05, | |
| "loss": 0.448, | |
| "step": 71600 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 1.5987654320987655e-05, | |
| "loss": 0.438, | |
| "step": 71700 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 1.594017094017094e-05, | |
| "loss": 0.4377, | |
| "step": 71800 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 1.5892687559354228e-05, | |
| "loss": 0.4424, | |
| "step": 71900 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 1.5845204178537513e-05, | |
| "loss": 0.4379, | |
| "step": 72000 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 1.57977207977208e-05, | |
| "loss": 0.4339, | |
| "step": 72100 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 1.5750237416904085e-05, | |
| "loss": 0.4291, | |
| "step": 72200 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 1.570275403608737e-05, | |
| "loss": 0.4372, | |
| "step": 72300 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 1.5655270655270658e-05, | |
| "loss": 0.4475, | |
| "step": 72400 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 1.5607787274453943e-05, | |
| "loss": 0.4366, | |
| "step": 72500 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 1.5560303893637227e-05, | |
| "loss": 0.4315, | |
| "step": 72600 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 1.5512820512820516e-05, | |
| "loss": 0.4326, | |
| "step": 72700 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 1.54653371320038e-05, | |
| "loss": 0.4463, | |
| "step": 72800 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 1.5417853751187085e-05, | |
| "loss": 0.4379, | |
| "step": 72900 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 1.537037037037037e-05, | |
| "loss": 0.436, | |
| "step": 73000 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 1.5322886989553654e-05, | |
| "loss": 0.438, | |
| "step": 73100 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 1.5275403608736942e-05, | |
| "loss": 0.4343, | |
| "step": 73200 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 1.5227920227920229e-05, | |
| "loss": 0.4447, | |
| "step": 73300 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 1.5180436847103515e-05, | |
| "loss": 0.4361, | |
| "step": 73400 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 1.5132953466286801e-05, | |
| "loss": 0.4378, | |
| "step": 73500 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 1.5085470085470086e-05, | |
| "loss": 0.4378, | |
| "step": 73600 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 1.5037986704653372e-05, | |
| "loss": 0.4315, | |
| "step": 73700 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 1.4990503323836657e-05, | |
| "loss": 0.438, | |
| "step": 73800 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 1.4943019943019942e-05, | |
| "loss": 0.4371, | |
| "step": 73900 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 1.489553656220323e-05, | |
| "loss": 0.43, | |
| "step": 74000 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 1.4848053181386514e-05, | |
| "loss": 0.4254, | |
| "step": 74100 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 1.4800569800569802e-05, | |
| "loss": 0.4311, | |
| "step": 74200 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 1.4753086419753087e-05, | |
| "loss": 0.4243, | |
| "step": 74300 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 1.4705603038936372e-05, | |
| "loss": 0.4274, | |
| "step": 74400 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 1.465811965811966e-05, | |
| "loss": 0.4223, | |
| "step": 74500 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 1.4610636277302944e-05, | |
| "loss": 0.4318, | |
| "step": 74600 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 1.4563152896486232e-05, | |
| "loss": 0.4183, | |
| "step": 74700 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 1.4515669515669517e-05, | |
| "loss": 0.4271, | |
| "step": 74800 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 1.446866096866097e-05, | |
| "loss": 0.4278, | |
| "step": 74900 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 1.4421177587844256e-05, | |
| "loss": 0.4213, | |
| "step": 75000 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "eval_binary_F1_negative": 0.8895014744009567, | |
| "eval_binary_F1_postive": 0.1294856799712583, | |
| "eval_loss": 0.4064924120903015, | |
| "eval_macro_F1": 0.5094935771861075, | |
| "eval_micro_F1": 0.8038954345914323, | |
| "eval_runtime": 6121.7432, | |
| "eval_samples_per_second": 1548.053, | |
| "eval_steps_per_second": 48.377, | |
| "step": 75000 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 1.437369420702754e-05, | |
| "loss": 0.4396, | |
| "step": 75100 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 1.4326685660018993e-05, | |
| "loss": 0.4253, | |
| "step": 75200 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 1.427920227920228e-05, | |
| "loss": 0.4231, | |
| "step": 75300 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 1.4231718898385565e-05, | |
| "loss": 0.4348, | |
| "step": 75400 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 1.4184235517568853e-05, | |
| "loss": 0.4167, | |
| "step": 75500 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 1.4136752136752138e-05, | |
| "loss": 0.4221, | |
| "step": 75600 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 1.4089268755935423e-05, | |
| "loss": 0.4321, | |
| "step": 75700 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 1.404178537511871e-05, | |
| "loss": 0.4282, | |
| "step": 75800 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 1.3994301994301995e-05, | |
| "loss": 0.4317, | |
| "step": 75900 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 1.394681861348528e-05, | |
| "loss": 0.4209, | |
| "step": 76000 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 1.3899335232668566e-05, | |
| "loss": 0.4364, | |
| "step": 76100 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 1.3851851851851853e-05, | |
| "loss": 0.4315, | |
| "step": 76200 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 1.380436847103514e-05, | |
| "loss": 0.4261, | |
| "step": 76300 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 1.3756885090218424e-05, | |
| "loss": 0.4251, | |
| "step": 76400 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 1.3709401709401708e-05, | |
| "loss": 0.4287, | |
| "step": 76500 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 1.3661918328584997e-05, | |
| "loss": 0.4275, | |
| "step": 76600 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 1.3614434947768281e-05, | |
| "loss": 0.4326, | |
| "step": 76700 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 1.3566951566951566e-05, | |
| "loss": 0.425, | |
| "step": 76800 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 1.3519468186134854e-05, | |
| "loss": 0.4261, | |
| "step": 76900 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 1.3471984805318139e-05, | |
| "loss": 0.4371, | |
| "step": 77000 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 1.3424501424501427e-05, | |
| "loss": 0.4285, | |
| "step": 77100 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 1.3377018043684711e-05, | |
| "loss": 0.4308, | |
| "step": 77200 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 1.3329534662867996e-05, | |
| "loss": 0.4214, | |
| "step": 77300 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 1.3282051282051284e-05, | |
| "loss": 0.4276, | |
| "step": 77400 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 1.3234567901234569e-05, | |
| "loss": 0.4306, | |
| "step": 77500 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 1.3187084520417855e-05, | |
| "loss": 0.4211, | |
| "step": 77600 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 1.3140075973409307e-05, | |
| "loss": 0.4259, | |
| "step": 77700 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 1.3092592592592592e-05, | |
| "loss": 0.41, | |
| "step": 77800 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 1.304510921177588e-05, | |
| "loss": 0.4367, | |
| "step": 77900 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 1.2997625830959165e-05, | |
| "loss": 0.4161, | |
| "step": 78000 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 1.295014245014245e-05, | |
| "loss": 0.4236, | |
| "step": 78100 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 1.2902659069325737e-05, | |
| "loss": 0.418, | |
| "step": 78200 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 1.2855175688509022e-05, | |
| "loss": 0.4205, | |
| "step": 78300 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 1.280769230769231e-05, | |
| "loss": 0.4155, | |
| "step": 78400 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 1.2760208926875595e-05, | |
| "loss": 0.4344, | |
| "step": 78500 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 1.271272554605888e-05, | |
| "loss": 0.4164, | |
| "step": 78600 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 1.2665716999050333e-05, | |
| "loss": 0.4077, | |
| "step": 78700 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 1.2618708452041786e-05, | |
| "loss": 0.4327, | |
| "step": 78800 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 1.257122507122507e-05, | |
| "loss": 0.4159, | |
| "step": 78900 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 1.2523741690408358e-05, | |
| "loss": 0.4347, | |
| "step": 79000 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 1.2476258309591643e-05, | |
| "loss": 0.4229, | |
| "step": 79100 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 1.242877492877493e-05, | |
| "loss": 0.4228, | |
| "step": 79200 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 1.2381291547958216e-05, | |
| "loss": 0.4244, | |
| "step": 79300 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 1.2333808167141502e-05, | |
| "loss": 0.4178, | |
| "step": 79400 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 1.2286324786324787e-05, | |
| "loss": 0.4178, | |
| "step": 79500 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 1.2238841405508073e-05, | |
| "loss": 0.4348, | |
| "step": 79600 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 1.219135802469136e-05, | |
| "loss": 0.4149, | |
| "step": 79700 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 1.2143874643874644e-05, | |
| "loss": 0.4212, | |
| "step": 79800 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 1.209639126305793e-05, | |
| "loss": 0.4162, | |
| "step": 79900 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 1.2048907882241215e-05, | |
| "loss": 0.4153, | |
| "step": 80000 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "eval_binary_F1_negative": 0.8925881895571441, | |
| "eval_binary_F1_postive": 0.130868243395573, | |
| "eval_loss": 0.4078446924686432, | |
| "eval_macro_F1": 0.5117282164763586, | |
| "eval_micro_F1": 0.808805221517084, | |
| "eval_runtime": 5859.7848, | |
| "eval_samples_per_second": 1617.258, | |
| "eval_steps_per_second": 50.539, | |
| "step": 80000 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 1.2001424501424501e-05, | |
| "loss": 0.4134, | |
| "step": 80100 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 1.1954415954415955e-05, | |
| "loss": 0.4295, | |
| "step": 80200 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 1.1906932573599242e-05, | |
| "loss": 0.4195, | |
| "step": 80300 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 1.1859449192782526e-05, | |
| "loss": 0.4215, | |
| "step": 80400 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 1.1811965811965813e-05, | |
| "loss": 0.4188, | |
| "step": 80500 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 1.1764482431149099e-05, | |
| "loss": 0.4287, | |
| "step": 80600 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 1.1716999050332385e-05, | |
| "loss": 0.4088, | |
| "step": 80700 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 1.166951566951567e-05, | |
| "loss": 0.4208, | |
| "step": 80800 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 1.1622032288698956e-05, | |
| "loss": 0.4119, | |
| "step": 80900 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 1.1574548907882243e-05, | |
| "loss": 0.4337, | |
| "step": 81000 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 1.1527065527065527e-05, | |
| "loss": 0.4261, | |
| "step": 81100 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 1.1479582146248812e-05, | |
| "loss": 0.4244, | |
| "step": 81200 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 1.1432098765432098e-05, | |
| "loss": 0.4059, | |
| "step": 81300 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 1.1384615384615385e-05, | |
| "loss": 0.4165, | |
| "step": 81400 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 1.1337132003798671e-05, | |
| "loss": 0.4186, | |
| "step": 81500 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 1.1289648622981956e-05, | |
| "loss": 0.4068, | |
| "step": 81600 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 1.1242165242165242e-05, | |
| "loss": 0.4214, | |
| "step": 81700 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 1.1194681861348529e-05, | |
| "loss": 0.4126, | |
| "step": 81800 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 1.1147198480531815e-05, | |
| "loss": 0.4129, | |
| "step": 81900 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 1.1099715099715101e-05, | |
| "loss": 0.4294, | |
| "step": 82000 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 1.1052231718898386e-05, | |
| "loss": 0.432, | |
| "step": 82100 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 1.1004748338081672e-05, | |
| "loss": 0.4131, | |
| "step": 82200 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 1.0957739791073125e-05, | |
| "loss": 0.4313, | |
| "step": 82300 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 1.0910256410256411e-05, | |
| "loss": 0.4144, | |
| "step": 82400 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 1.0862773029439696e-05, | |
| "loss": 0.4142, | |
| "step": 82500 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 1.0815289648622982e-05, | |
| "loss": 0.4122, | |
| "step": 82600 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 1.0767806267806268e-05, | |
| "loss": 0.4345, | |
| "step": 82700 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 1.0720322886989555e-05, | |
| "loss": 0.4269, | |
| "step": 82800 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 1.067283950617284e-05, | |
| "loss": 0.4135, | |
| "step": 82900 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 1.0625356125356126e-05, | |
| "loss": 0.4234, | |
| "step": 83000 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 1.0577872744539412e-05, | |
| "loss": 0.423, | |
| "step": 83100 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 1.0530389363722698e-05, | |
| "loss": 0.4201, | |
| "step": 83200 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 1.0482905982905983e-05, | |
| "loss": 0.4302, | |
| "step": 83300 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 1.043542260208927e-05, | |
| "loss": 0.402, | |
| "step": 83400 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 1.0387939221272556e-05, | |
| "loss": 0.4262, | |
| "step": 83500 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 1.0340455840455842e-05, | |
| "loss": 0.4249, | |
| "step": 83600 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 1.0292972459639127e-05, | |
| "loss": 0.4204, | |
| "step": 83700 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 1.0245489078822413e-05, | |
| "loss": 0.4268, | |
| "step": 83800 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 1.0198005698005698e-05, | |
| "loss": 0.4194, | |
| "step": 83900 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 1.0150522317188984e-05, | |
| "loss": 0.4186, | |
| "step": 84000 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 1.010303893637227e-05, | |
| "loss": 0.4044, | |
| "step": 84100 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 1.0056030389363723e-05, | |
| "loss": 0.4113, | |
| "step": 84200 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 1.0008547008547009e-05, | |
| "loss": 0.4323, | |
| "step": 84300 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 9.961063627730295e-06, | |
| "loss": 0.4072, | |
| "step": 84400 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 9.913580246913582e-06, | |
| "loss": 0.4075, | |
| "step": 84500 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 9.866096866096866e-06, | |
| "loss": 0.4222, | |
| "step": 84600 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 9.818613485280153e-06, | |
| "loss": 0.4133, | |
| "step": 84700 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 9.771130104463439e-06, | |
| "loss": 0.4124, | |
| "step": 84800 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 9.723646723646725e-06, | |
| "loss": 0.4083, | |
| "step": 84900 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 9.67616334283001e-06, | |
| "loss": 0.4054, | |
| "step": 85000 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "eval_binary_F1_negative": 0.9003478851520047, | |
| "eval_binary_F1_postive": 0.1351646421298386, | |
| "eval_loss": 0.3821626603603363, | |
| "eval_macro_F1": 0.5177562636409216, | |
| "eval_micro_F1": 0.8212881455801576, | |
| "eval_runtime": 5802.2246, | |
| "eval_samples_per_second": 1633.302, | |
| "eval_steps_per_second": 51.041, | |
| "step": 85000 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 9.628679962013295e-06, | |
| "loss": 0.4289, | |
| "step": 85100 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 9.581196581196581e-06, | |
| "loss": 0.4184, | |
| "step": 85200 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 9.534188034188035e-06, | |
| "loss": 0.4084, | |
| "step": 85300 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 9.48670465337132e-06, | |
| "loss": 0.4204, | |
| "step": 85400 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 9.439221272554606e-06, | |
| "loss": 0.4249, | |
| "step": 85500 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 9.391737891737892e-06, | |
| "loss": 0.4246, | |
| "step": 85600 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 9.344254510921179e-06, | |
| "loss": 0.4209, | |
| "step": 85700 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 9.296771130104463e-06, | |
| "loss": 0.4246, | |
| "step": 85800 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 9.24928774928775e-06, | |
| "loss": 0.4244, | |
| "step": 85900 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 9.201804368471036e-06, | |
| "loss": 0.4206, | |
| "step": 86000 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 9.154320987654323e-06, | |
| "loss": 0.4166, | |
| "step": 86100 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 9.106837606837607e-06, | |
| "loss": 0.4187, | |
| "step": 86200 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 9.05982905982906e-06, | |
| "loss": 0.4176, | |
| "step": 86300 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 9.012345679012346e-06, | |
| "loss": 0.4159, | |
| "step": 86400 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 8.9653371320038e-06, | |
| "loss": 0.418, | |
| "step": 86500 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 8.917853751187084e-06, | |
| "loss": 0.4373, | |
| "step": 86600 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 8.87037037037037e-06, | |
| "loss": 0.4244, | |
| "step": 86700 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 8.822886989553657e-06, | |
| "loss": 0.4146, | |
| "step": 86800 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 8.775403608736943e-06, | |
| "loss": 0.4156, | |
| "step": 86900 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 8.727920227920228e-06, | |
| "loss": 0.4252, | |
| "step": 87000 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 8.680436847103514e-06, | |
| "loss": 0.4254, | |
| "step": 87100 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 8.632953466286801e-06, | |
| "loss": 0.4196, | |
| "step": 87200 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 8.585470085470086e-06, | |
| "loss": 0.423, | |
| "step": 87300 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 8.537986704653372e-06, | |
| "loss": 0.4097, | |
| "step": 87400 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 8.490503323836657e-06, | |
| "loss": 0.4229, | |
| "step": 87500 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 8.44349477682811e-06, | |
| "loss": 0.4127, | |
| "step": 87600 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 8.396011396011397e-06, | |
| "loss": 0.4163, | |
| "step": 87700 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 8.348528015194683e-06, | |
| "loss": 0.4222, | |
| "step": 87800 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 8.301044634377968e-06, | |
| "loss": 0.4099, | |
| "step": 87900 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 8.253561253561254e-06, | |
| "loss": 0.4159, | |
| "step": 88000 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 8.20607787274454e-06, | |
| "loss": 0.4086, | |
| "step": 88100 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 8.158594491927827e-06, | |
| "loss": 0.4167, | |
| "step": 88200 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 8.111111111111112e-06, | |
| "loss": 0.4225, | |
| "step": 88300 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 8.063627730294398e-06, | |
| "loss": 0.4057, | |
| "step": 88400 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 8.016144349477683e-06, | |
| "loss": 0.419, | |
| "step": 88500 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 7.968660968660969e-06, | |
| "loss": 0.4164, | |
| "step": 88600 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 7.921177587844254e-06, | |
| "loss": 0.4185, | |
| "step": 88700 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 7.87369420702754e-06, | |
| "loss": 0.4082, | |
| "step": 88800 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 7.826210826210826e-06, | |
| "loss": 0.4059, | |
| "step": 88900 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 7.778727445394113e-06, | |
| "loss": 0.4189, | |
| "step": 89000 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 7.731244064577399e-06, | |
| "loss": 0.4194, | |
| "step": 89100 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 7.683760683760684e-06, | |
| "loss": 0.4207, | |
| "step": 89200 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 7.63627730294397e-06, | |
| "loss": 0.4158, | |
| "step": 89300 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 7.589268755935423e-06, | |
| "loss": 0.4108, | |
| "step": 89400 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 7.541785375118708e-06, | |
| "loss": 0.417, | |
| "step": 89500 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 7.494301994301994e-06, | |
| "loss": 0.4213, | |
| "step": 89600 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 7.4468186134852805e-06, | |
| "loss": 0.4297, | |
| "step": 89700 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 7.399335232668567e-06, | |
| "loss": 0.4309, | |
| "step": 89800 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 7.351851851851853e-06, | |
| "loss": 0.4122, | |
| "step": 89900 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 7.304368471035138e-06, | |
| "loss": 0.4036, | |
| "step": 90000 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "eval_binary_F1_negative": 0.8948790984171406, | |
| "eval_binary_F1_postive": 0.1321842841084109, | |
| "eval_loss": 0.39195898175239563, | |
| "eval_macro_F1": 0.5135316912627758, | |
| "eval_micro_F1": 0.8124737648396829, | |
| "eval_runtime": 6154.1657, | |
| "eval_samples_per_second": 1539.898, | |
| "eval_steps_per_second": 48.122, | |
| "step": 90000 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 7.256885090218423e-06, | |
| "loss": 0.4112, | |
| "step": 90100 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 7.20940170940171e-06, | |
| "loss": 0.4153, | |
| "step": 90200 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 7.161918328584996e-06, | |
| "loss": 0.4187, | |
| "step": 90300 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 7.114434947768281e-06, | |
| "loss": 0.4102, | |
| "step": 90400 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 7.066951566951567e-06, | |
| "loss": 0.4078, | |
| "step": 90500 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 7.0194681861348535e-06, | |
| "loss": 0.4098, | |
| "step": 90600 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 6.97198480531814e-06, | |
| "loss": 0.4168, | |
| "step": 90700 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 6.9245014245014245e-06, | |
| "loss": 0.4196, | |
| "step": 90800 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 6.87701804368471e-06, | |
| "loss": 0.4103, | |
| "step": 90900 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 6.829534662867996e-06, | |
| "loss": 0.4144, | |
| "step": 91000 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 6.782051282051283e-06, | |
| "loss": 0.4175, | |
| "step": 91100 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 6.734567901234569e-06, | |
| "loss": 0.4237, | |
| "step": 91200 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 6.687084520417854e-06, | |
| "loss": 0.4203, | |
| "step": 91300 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 6.63960113960114e-06, | |
| "loss": 0.4215, | |
| "step": 91400 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 6.592117758784426e-06, | |
| "loss": 0.4129, | |
| "step": 91500 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 6.544634377967712e-06, | |
| "loss": 0.407, | |
| "step": 91600 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 6.497150997150997e-06, | |
| "loss": 0.4108, | |
| "step": 91700 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 6.449667616334283e-06, | |
| "loss": 0.4147, | |
| "step": 91800 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 6.402184235517569e-06, | |
| "loss": 0.4177, | |
| "step": 91900 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 6.354700854700856e-06, | |
| "loss": 0.4011, | |
| "step": 92000 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 6.30721747388414e-06, | |
| "loss": 0.4131, | |
| "step": 92100 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 6.260208926875593e-06, | |
| "loss": 0.4043, | |
| "step": 92200 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 6.21272554605888e-06, | |
| "loss": 0.4122, | |
| "step": 92300 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 6.165242165242165e-06, | |
| "loss": 0.4152, | |
| "step": 92400 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 6.117758784425452e-06, | |
| "loss": 0.4156, | |
| "step": 92500 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 6.070275403608737e-06, | |
| "loss": 0.4181, | |
| "step": 92600 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 6.0227920227920235e-06, | |
| "loss": 0.4099, | |
| "step": 92700 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 5.975783475783476e-06, | |
| "loss": 0.4121, | |
| "step": 92800 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 5.928300094966762e-06, | |
| "loss": 0.4169, | |
| "step": 92900 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 5.880816714150048e-06, | |
| "loss": 0.4084, | |
| "step": 93000 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 5.833333333333334e-06, | |
| "loss": 0.4169, | |
| "step": 93100 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 5.7858499525166195e-06, | |
| "loss": 0.4154, | |
| "step": 93200 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 5.738366571699905e-06, | |
| "loss": 0.422, | |
| "step": 93300 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 5.6908831908831905e-06, | |
| "loss": 0.416, | |
| "step": 93400 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 5.643399810066477e-06, | |
| "loss": 0.4032, | |
| "step": 93500 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 5.595916429249763e-06, | |
| "loss": 0.4195, | |
| "step": 93600 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 5.548433048433049e-06, | |
| "loss": 0.4212, | |
| "step": 93700 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 5.501424501424502e-06, | |
| "loss": 0.405, | |
| "step": 93800 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 5.454415954415955e-06, | |
| "loss": 0.4209, | |
| "step": 93900 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 5.40693257359924e-06, | |
| "loss": 0.4215, | |
| "step": 94000 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 5.359449192782527e-06, | |
| "loss": 0.4202, | |
| "step": 94100 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 5.311965811965812e-06, | |
| "loss": 0.4085, | |
| "step": 94200 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 5.264482431149098e-06, | |
| "loss": 0.4274, | |
| "step": 94300 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 5.216999050332383e-06, | |
| "loss": 0.4071, | |
| "step": 94400 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 5.16951566951567e-06, | |
| "loss": 0.4336, | |
| "step": 94500 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 5.122032288698956e-06, | |
| "loss": 0.4166, | |
| "step": 94600 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 5.0745489078822415e-06, | |
| "loss": 0.4193, | |
| "step": 94700 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 5.027065527065528e-06, | |
| "loss": 0.4233, | |
| "step": 94800 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 4.979582146248813e-06, | |
| "loss": 0.4219, | |
| "step": 94900 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 4.932098765432099e-06, | |
| "loss": 0.4339, | |
| "step": 95000 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "eval_binary_F1_negative": 0.8967303701812424, | |
| "eval_binary_F1_postive": 0.13340118465635153, | |
| "eval_loss": 0.39675113558769226, | |
| "eval_macro_F1": 0.515065777418797, | |
| "eval_micro_F1": 0.8154526228618014, | |
| "eval_runtime": 5940.4524, | |
| "eval_samples_per_second": 1595.297, | |
| "eval_steps_per_second": 49.853, | |
| "step": 95000 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 4.884615384615384e-06, | |
| "loss": 0.4174, | |
| "step": 95100 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 4.837132003798671e-06, | |
| "loss": 0.4187, | |
| "step": 95200 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 4.789648622981956e-06, | |
| "loss": 0.4258, | |
| "step": 95300 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 4.742165242165243e-06, | |
| "loss": 0.4266, | |
| "step": 95400 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 4.694681861348528e-06, | |
| "loss": 0.4193, | |
| "step": 95500 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 4.6471984805318144e-06, | |
| "loss": 0.4311, | |
| "step": 95600 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 4.5997150997151e-06, | |
| "loss": 0.425, | |
| "step": 95700 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 4.5522317188983855e-06, | |
| "loss": 0.4163, | |
| "step": 95800 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 4.504748338081671e-06, | |
| "loss": 0.4283, | |
| "step": 95900 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 4.457264957264957e-06, | |
| "loss": 0.4251, | |
| "step": 96000 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 4.409781576448244e-06, | |
| "loss": 0.4199, | |
| "step": 96100 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 4.362298195631529e-06, | |
| "loss": 0.4236, | |
| "step": 96200 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 4.3148148148148155e-06, | |
| "loss": 0.4366, | |
| "step": 96300 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 4.267331433998101e-06, | |
| "loss": 0.4157, | |
| "step": 96400 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 4.2198480531813866e-06, | |
| "loss": 0.4204, | |
| "step": 96500 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 4.172364672364672e-06, | |
| "loss": 0.4174, | |
| "step": 96600 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 4.124881291547958e-06, | |
| "loss": 0.4164, | |
| "step": 96700 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 4.077397910731244e-06, | |
| "loss": 0.4189, | |
| "step": 96800 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 4.02991452991453e-06, | |
| "loss": 0.4125, | |
| "step": 96900 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 3.982431149097816e-06, | |
| "loss": 0.4219, | |
| "step": 97000 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 3.934947768281102e-06, | |
| "loss": 0.4136, | |
| "step": 97100 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 3.887464387464388e-06, | |
| "loss": 0.4212, | |
| "step": 97200 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 3.839981006647673e-06, | |
| "loss": 0.4262, | |
| "step": 97300 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 3.7924976258309595e-06, | |
| "loss": 0.4292, | |
| "step": 97400 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 3.745014245014245e-06, | |
| "loss": 0.4249, | |
| "step": 97500 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 3.6975308641975314e-06, | |
| "loss": 0.4154, | |
| "step": 97600 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 3.650047483380817e-06, | |
| "loss": 0.4187, | |
| "step": 97700 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 3.602564102564103e-06, | |
| "loss": 0.4166, | |
| "step": 97800 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 3.5550807217473883e-06, | |
| "loss": 0.4155, | |
| "step": 97900 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 3.5075973409306747e-06, | |
| "loss": 0.4123, | |
| "step": 98000 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 3.46011396011396e-06, | |
| "loss": 0.4051, | |
| "step": 98100 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 3.4131054131054137e-06, | |
| "loss": 0.4209, | |
| "step": 98200 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 3.3656220322886992e-06, | |
| "loss": 0.4149, | |
| "step": 98300 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 3.318138651471985e-06, | |
| "loss": 0.414, | |
| "step": 98400 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 3.2706552706552707e-06, | |
| "loss": 0.4165, | |
| "step": 98500 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 3.223171889838557e-06, | |
| "loss": 0.4196, | |
| "step": 98600 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 3.1756885090218425e-06, | |
| "loss": 0.4152, | |
| "step": 98700 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 3.1282051282051284e-06, | |
| "loss": 0.4246, | |
| "step": 98800 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 3.0807217473884144e-06, | |
| "loss": 0.413, | |
| "step": 98900 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 3.0332383665717003e-06, | |
| "loss": 0.4314, | |
| "step": 99000 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 2.985754985754986e-06, | |
| "loss": 0.4263, | |
| "step": 99100 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 2.938746438746439e-06, | |
| "loss": 0.4219, | |
| "step": 99200 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 2.891263057929725e-06, | |
| "loss": 0.4135, | |
| "step": 99300 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 2.8437796771130108e-06, | |
| "loss": 0.4145, | |
| "step": 99400 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 2.7967711301044635e-06, | |
| "loss": 0.4165, | |
| "step": 99500 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 2.7492877492877494e-06, | |
| "loss": 0.4118, | |
| "step": 99600 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 2.7018043684710353e-06, | |
| "loss": 0.42, | |
| "step": 99700 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 2.6543209876543212e-06, | |
| "loss": 0.4135, | |
| "step": 99800 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 2.606837606837607e-06, | |
| "loss": 0.4166, | |
| "step": 99900 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 2.5593542260208927e-06, | |
| "loss": 0.4114, | |
| "step": 100000 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "eval_binary_F1_negative": 0.893749650986467, | |
| "eval_binary_F1_postive": 0.13183102844111674, | |
| "eval_loss": 0.3945052921772003, | |
| "eval_macro_F1": 0.5127903397137918, | |
| "eval_micro_F1": 0.8106703053123707, | |
| "eval_runtime": 5994.5555, | |
| "eval_samples_per_second": 1580.899, | |
| "eval_steps_per_second": 49.403, | |
| "step": 100000 | |
| } | |
| ], | |
| "logging_steps": 100, | |
| "max_steps": 105300, | |
| "num_train_epochs": 3, | |
| "save_steps": 10000, | |
| "total_flos": 1.4208252178857984e+19, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |