{ "best_metric": 0.3600367021867026, "best_model_checkpoint": "t5-small-keyword-generation-inspec\\checkpoint-25000", "epoch": 50.0, "global_step": 50000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.5, "learning_rate": 2.465e-05, "loss": 2.5363, "step": 500 }, { "epoch": 1.0, "learning_rate": 4.965e-05, "loss": 1.6228, "step": 1000 }, { "epoch": 1.5, "learning_rate": 4.9496938775510204e-05, "loss": 1.4329, "step": 1500 }, { "epoch": 2.0, "learning_rate": 4.898673469387755e-05, "loss": 1.3213, "step": 2000 }, { "epoch": 2.5, "learning_rate": 4.8476530612244906e-05, "loss": 1.2251, "step": 2500 }, { "epoch": 3.0, "learning_rate": 4.7967346938775516e-05, "loss": 1.1703, "step": 3000 }, { "epoch": 3.5, "learning_rate": 4.745714285714286e-05, "loss": 1.0748, "step": 3500 }, { "epoch": 4.0, "learning_rate": 4.694693877551021e-05, "loss": 1.1015, "step": 4000 }, { "epoch": 4.5, "learning_rate": 4.6436734693877554e-05, "loss": 0.9512, "step": 4500 }, { "epoch": 5.0, "learning_rate": 4.5927551020408164e-05, "loss": 1.0313, "step": 5000 }, { "epoch": 5.5, "learning_rate": 4.5417346938775515e-05, "loss": 0.9283, "step": 5500 }, { "epoch": 6.0, "learning_rate": 4.490714285714286e-05, "loss": 0.9154, "step": 6000 }, { "epoch": 6.5, "learning_rate": 4.43969387755102e-05, "loss": 0.8623, "step": 6500 }, { "epoch": 7.0, "learning_rate": 4.388673469387755e-05, "loss": 0.8258, "step": 7000 }, { "epoch": 7.5, "learning_rate": 4.337755102040816e-05, "loss": 0.7804, "step": 7500 }, { "epoch": 8.0, "learning_rate": 4.286734693877551e-05, "loss": 0.7949, "step": 8000 }, { "epoch": 8.5, "learning_rate": 4.2357142857142864e-05, "loss": 0.7321, "step": 8500 }, { "epoch": 9.0, "learning_rate": 4.184693877551021e-05, "loss": 0.7393, "step": 9000 }, { "epoch": 9.5, "learning_rate": 4.133775510204082e-05, "loss": 0.6964, "step": 9500 }, { "epoch": 10.0, "learning_rate": 4.082755102040817e-05, "loss": 0.6759, "step": 10000 }, { "epoch": 10.5, "learning_rate": 4.031734693877551e-05, "loss": 0.6642, "step": 10500 }, { "epoch": 11.0, "learning_rate": 3.980714285714286e-05, "loss": 0.6444, "step": 11000 }, { "epoch": 11.5, "learning_rate": 3.929693877551021e-05, "loss": 0.6389, "step": 11500 }, { "epoch": 12.0, "learning_rate": 3.878775510204082e-05, "loss": 0.5822, "step": 12000 }, { "epoch": 12.5, "learning_rate": 3.827755102040816e-05, "loss": 0.5617, "step": 12500 }, { "epoch": 13.0, "learning_rate": 3.776734693877551e-05, "loss": 0.5885, "step": 13000 }, { "epoch": 13.5, "learning_rate": 3.7257142857142856e-05, "loss": 0.5307, "step": 13500 }, { "epoch": 14.0, "learning_rate": 3.6746938775510206e-05, "loss": 0.5597, "step": 14000 }, { "epoch": 14.5, "learning_rate": 3.623673469387756e-05, "loss": 0.4839, "step": 14500 }, { "epoch": 15.0, "learning_rate": 3.57265306122449e-05, "loss": 0.5524, "step": 15000 }, { "epoch": 15.5, "learning_rate": 3.521632653061225e-05, "loss": 0.5058, "step": 15500 }, { "epoch": 16.0, "learning_rate": 3.470714285714286e-05, "loss": 0.4807, "step": 16000 }, { "epoch": 16.5, "learning_rate": 3.4196938775510205e-05, "loss": 0.4749, "step": 16500 }, { "epoch": 17.0, "learning_rate": 3.3686734693877556e-05, "loss": 0.4668, "step": 17000 }, { "epoch": 17.5, "learning_rate": 3.31765306122449e-05, "loss": 0.4512, "step": 17500 }, { "epoch": 18.0, "learning_rate": 3.266734693877551e-05, "loss": 0.4506, "step": 18000 }, { "epoch": 18.5, "learning_rate": 3.215714285714286e-05, "loss": 0.4317, "step": 18500 }, { "epoch": 19.0, "learning_rate": 3.1646938775510204e-05, "loss": 0.4262, "step": 19000 }, { "epoch": 19.5, "learning_rate": 3.113673469387755e-05, "loss": 0.3938, "step": 19500 }, { "epoch": 20.0, "learning_rate": 3.06265306122449e-05, "loss": 0.4165, "step": 20000 }, { "epoch": 20.5, "learning_rate": 3.011734693877551e-05, "loss": 0.4182, "step": 20500 }, { "epoch": 21.0, "learning_rate": 2.960714285714286e-05, "loss": 0.3608, "step": 21000 }, { "epoch": 21.5, "learning_rate": 2.9096938775510207e-05, "loss": 0.377, "step": 21500 }, { "epoch": 22.0, "learning_rate": 2.8586734693877554e-05, "loss": 0.3845, "step": 22000 }, { "epoch": 22.5, "learning_rate": 2.80765306122449e-05, "loss": 0.376, "step": 22500 }, { "epoch": 23.0, "learning_rate": 2.7566326530612245e-05, "loss": 0.3522, "step": 23000 }, { "epoch": 23.5, "learning_rate": 2.7056122448979592e-05, "loss": 0.3495, "step": 23500 }, { "epoch": 24.0, "learning_rate": 2.654591836734694e-05, "loss": 0.3539, "step": 24000 }, { "epoch": 24.5, "learning_rate": 2.6036734693877553e-05, "loss": 0.3357, "step": 24500 }, { "epoch": 25.0, "learning_rate": 2.5526530612244897e-05, "loss": 0.3341, "step": 25000 }, { "epoch": 25.0, "eval_F1@10": 0.22188352801309513, "eval_F1@5": 0.32039403374403397, "eval_F1@M": 0.3600367021867026, "eval_P@10": 0.18379999999999982, "eval_P@5": 0.36759999999999965, "eval_P@M": 0.48766666666666625, "eval_R@10": 0.298240836940837, "eval_R@5": 0.298240836940837, "eval_R@M": 0.298240836940837, "eval_loss": 1.5177357196807861, "eval_model_name": "t5-small-keyword-generation-inspec", "eval_runtime": 282.7, "eval_samples_per_second": 1.769, "eval_steps_per_second": 1.769, "step": 25000 }, { "epoch": 25.5, "learning_rate": 2.5016326530612244e-05, "loss": 0.3203, "step": 25500 }, { "epoch": 26.0, "learning_rate": 2.4506122448979594e-05, "loss": 0.3257, "step": 26000 }, { "epoch": 26.5, "learning_rate": 2.3995918367346938e-05, "loss": 0.3046, "step": 26500 }, { "epoch": 27.0, "learning_rate": 2.3485714285714285e-05, "loss": 0.3236, "step": 27000 }, { "epoch": 27.5, "learning_rate": 2.2975510204081636e-05, "loss": 0.2983, "step": 27500 }, { "epoch": 28.0, "learning_rate": 2.246530612244898e-05, "loss": 0.3063, "step": 28000 }, { "epoch": 28.5, "learning_rate": 2.1956122448979593e-05, "loss": 0.2928, "step": 28500 }, { "epoch": 29.0, "learning_rate": 2.144591836734694e-05, "loss": 0.289, "step": 29000 }, { "epoch": 29.5, "learning_rate": 2.0935714285714288e-05, "loss": 0.2772, "step": 29500 }, { "epoch": 30.0, "learning_rate": 2.042551020408163e-05, "loss": 0.2835, "step": 30000 }, { "epoch": 30.5, "learning_rate": 1.9915306122448982e-05, "loss": 0.2804, "step": 30500 }, { "epoch": 31.0, "learning_rate": 1.940510204081633e-05, "loss": 0.2728, "step": 31000 }, { "epoch": 31.5, "learning_rate": 1.8894897959183673e-05, "loss": 0.2481, "step": 31500 }, { "epoch": 32.0, "learning_rate": 1.838469387755102e-05, "loss": 0.2734, "step": 32000 }, { "epoch": 32.5, "learning_rate": 1.7874489795918368e-05, "loss": 0.2629, "step": 32500 }, { "epoch": 33.0, "learning_rate": 1.7365306122448978e-05, "loss": 0.2546, "step": 33000 }, { "epoch": 33.5, "learning_rate": 1.685510204081633e-05, "loss": 0.2454, "step": 33500 }, { "epoch": 34.0, "learning_rate": 1.6344897959183676e-05, "loss": 0.2456, "step": 34000 }, { "epoch": 34.5, "learning_rate": 1.5835714285714286e-05, "loss": 0.2618, "step": 34500 }, { "epoch": 35.0, "learning_rate": 1.5325510204081633e-05, "loss": 0.2362, "step": 35000 }, { "epoch": 35.5, "learning_rate": 1.481530612244898e-05, "loss": 0.2249, "step": 35500 }, { "epoch": 36.0, "learning_rate": 1.4305102040816326e-05, "loss": 0.2566, "step": 36000 }, { "epoch": 36.5, "learning_rate": 1.3794897959183675e-05, "loss": 0.2247, "step": 36500 }, { "epoch": 37.0, "learning_rate": 1.3285714285714288e-05, "loss": 0.2434, "step": 37000 }, { "epoch": 37.5, "learning_rate": 1.2775510204081634e-05, "loss": 0.2291, "step": 37500 }, { "epoch": 38.0, "learning_rate": 1.2265306122448981e-05, "loss": 0.2301, "step": 38000 }, { "epoch": 38.5, "learning_rate": 1.1755102040816326e-05, "loss": 0.2281, "step": 38500 }, { "epoch": 39.0, "learning_rate": 1.1244897959183674e-05, "loss": 0.2203, "step": 39000 }, { "epoch": 39.5, "learning_rate": 1.073469387755102e-05, "loss": 0.2129, "step": 39500 }, { "epoch": 40.0, "learning_rate": 1.0225510204081633e-05, "loss": 0.2258, "step": 40000 }, { "epoch": 40.5, "learning_rate": 9.71530612244898e-06, "loss": 0.2254, "step": 40500 }, { "epoch": 41.0, "learning_rate": 9.205102040816327e-06, "loss": 0.208, "step": 41000 }, { "epoch": 41.5, "learning_rate": 8.694897959183673e-06, "loss": 0.2295, "step": 41500 }, { "epoch": 42.0, "learning_rate": 8.184693877551021e-06, "loss": 0.1969, "step": 42000 }, { "epoch": 42.5, "learning_rate": 7.674489795918367e-06, "loss": 0.2219, "step": 42500 }, { "epoch": 43.0, "learning_rate": 7.164285714285715e-06, "loss": 0.2032, "step": 43000 }, { "epoch": 43.5, "learning_rate": 6.654081632653061e-06, "loss": 0.2145, "step": 43500 }, { "epoch": 44.0, "learning_rate": 6.144897959183673e-06, "loss": 0.1968, "step": 44000 }, { "epoch": 44.5, "learning_rate": 5.6346938775510205e-06, "loss": 0.1956, "step": 44500 }, { "epoch": 45.0, "learning_rate": 5.124489795918368e-06, "loss": 0.212, "step": 45000 }, { "epoch": 45.5, "learning_rate": 4.614285714285715e-06, "loss": 0.2036, "step": 45500 }, { "epoch": 46.0, "learning_rate": 4.104081632653061e-06, "loss": 0.2101, "step": 46000 }, { "epoch": 46.5, "learning_rate": 3.594897959183673e-06, "loss": 0.1981, "step": 46500 }, { "epoch": 47.0, "learning_rate": 3.0846938775510208e-06, "loss": 0.2008, "step": 47000 }, { "epoch": 47.5, "learning_rate": 2.575510204081633e-06, "loss": 0.209, "step": 47500 }, { "epoch": 48.0, "learning_rate": 2.0653061224489795e-06, "loss": 0.1872, "step": 48000 }, { "epoch": 48.5, "learning_rate": 1.5551020408163267e-06, "loss": 0.1778, "step": 48500 }, { "epoch": 49.0, "learning_rate": 1.0448979591836734e-06, "loss": 0.2174, "step": 49000 }, { "epoch": 49.5, "learning_rate": 5.346938775510204e-07, "loss": 0.1982, "step": 49500 }, { "epoch": 50.0, "learning_rate": 2.4489795918367346e-08, "loss": 0.195, "step": 50000 }, { "epoch": 50.0, "eval_F1@10": 0.21505513330838452, "eval_F1@5": 0.31023088578088626, "eval_F1@M": 0.3480389832389839, "eval_P@10": 0.17839999999999995, "eval_P@5": 0.3567999999999999, "eval_P@M": 0.4749999999999997, "eval_R@10": 0.28769942279942284, "eval_R@5": 0.28769942279942284, "eval_R@M": 0.28769942279942284, "eval_loss": 1.8302260637283325, "eval_model_name": "t5-small-keyword-generation-inspec", "eval_runtime": 260.956, "eval_samples_per_second": 1.916, "eval_steps_per_second": 1.916, "step": 50000 } ], "max_steps": 50000, "num_train_epochs": 50, "total_flos": 6767090073600000.0, "trial_name": null, "trial_params": null }