| { | |
| "best_metric": 0.759707510471344, | |
| "best_model_checkpoint": "/home2/s5431786/jb-internship/results/Salesforce-codet5-small-CodeXGLUE-CONCODE-selected/checkpoint-12000", | |
| "epoch": 9.6, | |
| "global_step": 30000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.3808e-05, | |
| "loss": 1.45, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "eval_bleu": 0.24715996940066084, | |
| "eval_exact_match": 0.1615, | |
| "eval_loss": 0.8543930649757385, | |
| "eval_rouge1": 0.5607398648042461, | |
| "eval_rouge2": 0.376714337460786, | |
| "eval_rougeL": 0.5368883844476564, | |
| "eval_runtime": 233.0119, | |
| "eval_samples_per_second": 8.583, | |
| "eval_steps_per_second": 0.27, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 9.853642105263158e-05, | |
| "loss": 0.7186, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "eval_bleu": 0.2686100325072973, | |
| "eval_exact_match": 0.1755, | |
| "eval_loss": 0.8113107681274414, | |
| "eval_rouge1": 0.5759137545878418, | |
| "eval_rouge2": 0.3934978008929957, | |
| "eval_rougeL": 0.5514257399531202, | |
| "eval_runtime": 222.3222, | |
| "eval_samples_per_second": 8.996, | |
| "eval_steps_per_second": 0.283, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 9.516800000000001e-05, | |
| "loss": 0.6385, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "eval_bleu": 0.2859251068334253, | |
| "eval_exact_match": 0.1795, | |
| "eval_loss": 0.7783799767494202, | |
| "eval_rouge1": 0.5758903457879601, | |
| "eval_rouge2": 0.4004382979016302, | |
| "eval_rougeL": 0.5524759373364955, | |
| "eval_runtime": 259.7789, | |
| "eval_samples_per_second": 7.699, | |
| "eval_steps_per_second": 0.243, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 9.179957894736842e-05, | |
| "loss": 0.562, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "eval_bleu": 0.2967822969771582, | |
| "eval_exact_match": 0.179, | |
| "eval_loss": 0.7699160575866699, | |
| "eval_rouge1": 0.5769004117490721, | |
| "eval_rouge2": 0.400017695314681, | |
| "eval_rougeL": 0.5526377939756941, | |
| "eval_runtime": 274.6962, | |
| "eval_samples_per_second": 7.281, | |
| "eval_steps_per_second": 0.229, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "learning_rate": 8.843115789473684e-05, | |
| "loss": 0.5314, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "eval_bleu": 0.29975143100679225, | |
| "eval_exact_match": 0.1815, | |
| "eval_loss": 0.7741894125938416, | |
| "eval_rouge1": 0.595198816050763, | |
| "eval_rouge2": 0.4127741738801316, | |
| "eval_rougeL": 0.5696024067817068, | |
| "eval_runtime": 220.4946, | |
| "eval_samples_per_second": 9.071, | |
| "eval_steps_per_second": 0.286, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 3.84, | |
| "learning_rate": 8.506273684210527e-05, | |
| "loss": 0.4921, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 3.84, | |
| "eval_bleu": 0.29459469300176405, | |
| "eval_exact_match": 0.1835, | |
| "eval_loss": 0.759707510471344, | |
| "eval_rouge1": 0.5964669156896572, | |
| "eval_rouge2": 0.41676711356974905, | |
| "eval_rougeL": 0.5727323739644565, | |
| "eval_runtime": 247.9977, | |
| "eval_samples_per_second": 8.065, | |
| "eval_steps_per_second": 0.254, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 4.48, | |
| "learning_rate": 8.169431578947368e-05, | |
| "loss": 0.4575, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 4.48, | |
| "eval_bleu": 0.3082674516281405, | |
| "eval_exact_match": 0.181, | |
| "eval_loss": 0.7730007171630859, | |
| "eval_rouge1": 0.59579739088828, | |
| "eval_rouge2": 0.4140226910138735, | |
| "eval_rougeL": 0.5699065207669394, | |
| "eval_runtime": 246.1487, | |
| "eval_samples_per_second": 8.125, | |
| "eval_steps_per_second": 0.256, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 5.12, | |
| "learning_rate": 7.832757894736842e-05, | |
| "loss": 0.4376, | |
| "step": 16000 | |
| }, | |
| { | |
| "epoch": 5.12, | |
| "eval_bleu": 0.3122947893870329, | |
| "eval_exact_match": 0.1765, | |
| "eval_loss": 0.7814369797706604, | |
| "eval_rouge1": 0.6033397512051191, | |
| "eval_rouge2": 0.41418290168088967, | |
| "eval_rougeL": 0.5773324606448702, | |
| "eval_runtime": 206.0798, | |
| "eval_samples_per_second": 9.705, | |
| "eval_steps_per_second": 0.306, | |
| "step": 16000 | |
| }, | |
| { | |
| "epoch": 5.76, | |
| "learning_rate": 7.496084210526317e-05, | |
| "loss": 0.4123, | |
| "step": 18000 | |
| }, | |
| { | |
| "epoch": 5.76, | |
| "eval_bleu": 0.3161064543245236, | |
| "eval_exact_match": 0.185, | |
| "eval_loss": 0.7642457485198975, | |
| "eval_rouge1": 0.6004680977665069, | |
| "eval_rouge2": 0.4177038650507878, | |
| "eval_rougeL": 0.5740117301579993, | |
| "eval_runtime": 216.7206, | |
| "eval_samples_per_second": 9.228, | |
| "eval_steps_per_second": 0.291, | |
| "step": 18000 | |
| }, | |
| { | |
| "epoch": 6.4, | |
| "learning_rate": 7.159242105263159e-05, | |
| "loss": 0.3899, | |
| "step": 20000 | |
| }, | |
| { | |
| "epoch": 6.4, | |
| "eval_bleu": 0.3059724707400374, | |
| "eval_exact_match": 0.1805, | |
| "eval_loss": 0.7763121128082275, | |
| "eval_rouge1": 0.6026149641849021, | |
| "eval_rouge2": 0.41838415223593517, | |
| "eval_rougeL": 0.5767998545492816, | |
| "eval_runtime": 194.5368, | |
| "eval_samples_per_second": 10.281, | |
| "eval_steps_per_second": 0.324, | |
| "step": 20000 | |
| }, | |
| { | |
| "epoch": 7.04, | |
| "learning_rate": 6.8224e-05, | |
| "loss": 0.3803, | |
| "step": 22000 | |
| }, | |
| { | |
| "epoch": 7.04, | |
| "eval_bleu": 0.3162298613031694, | |
| "eval_exact_match": 0.1805, | |
| "eval_loss": 0.7929842472076416, | |
| "eval_rouge1": 0.6026097008816469, | |
| "eval_rouge2": 0.416157662903899, | |
| "eval_rougeL": 0.5752187985145643, | |
| "eval_runtime": 178.6941, | |
| "eval_samples_per_second": 11.192, | |
| "eval_steps_per_second": 0.353, | |
| "step": 22000 | |
| }, | |
| { | |
| "epoch": 7.68, | |
| "learning_rate": 6.485726315789475e-05, | |
| "loss": 0.3533, | |
| "step": 24000 | |
| }, | |
| { | |
| "epoch": 7.68, | |
| "eval_bleu": 0.3155753382171417, | |
| "eval_exact_match": 0.1795, | |
| "eval_loss": 0.7889729738235474, | |
| "eval_rouge1": 0.6082820845852519, | |
| "eval_rouge2": 0.41955238841842035, | |
| "eval_rougeL": 0.5807883221370937, | |
| "eval_runtime": 167.7479, | |
| "eval_samples_per_second": 11.923, | |
| "eval_steps_per_second": 0.376, | |
| "step": 24000 | |
| }, | |
| { | |
| "epoch": 8.32, | |
| "learning_rate": 6.148884210526316e-05, | |
| "loss": 0.3424, | |
| "step": 26000 | |
| }, | |
| { | |
| "epoch": 8.32, | |
| "eval_bleu": 0.3195565466494748, | |
| "eval_exact_match": 0.1785, | |
| "eval_loss": 0.811191737651825, | |
| "eval_rouge1": 0.6034126690137341, | |
| "eval_rouge2": 0.4117122080617083, | |
| "eval_rougeL": 0.5747807479180939, | |
| "eval_runtime": 177.8803, | |
| "eval_samples_per_second": 11.244, | |
| "eval_steps_per_second": 0.354, | |
| "step": 26000 | |
| }, | |
| { | |
| "epoch": 8.96, | |
| "learning_rate": 5.81221052631579e-05, | |
| "loss": 0.3313, | |
| "step": 28000 | |
| }, | |
| { | |
| "epoch": 8.96, | |
| "eval_bleu": 0.31867689638786295, | |
| "eval_exact_match": 0.182, | |
| "eval_loss": 0.8014870285987854, | |
| "eval_rouge1": 0.6103585439964291, | |
| "eval_rouge2": 0.4227833245285987, | |
| "eval_rougeL": 0.5822662636066496, | |
| "eval_runtime": 174.5922, | |
| "eval_samples_per_second": 11.455, | |
| "eval_steps_per_second": 0.361, | |
| "step": 28000 | |
| }, | |
| { | |
| "epoch": 9.6, | |
| "learning_rate": 5.475536842105263e-05, | |
| "loss": 0.3119, | |
| "step": 30000 | |
| }, | |
| { | |
| "epoch": 9.6, | |
| "eval_bleu": 0.32475336720503345, | |
| "eval_exact_match": 0.18, | |
| "eval_loss": 0.8151144981384277, | |
| "eval_rouge1": 0.610219724333067, | |
| "eval_rouge2": 0.41998657811516843, | |
| "eval_rougeL": 0.5813340970149274, | |
| "eval_runtime": 153.9026, | |
| "eval_samples_per_second": 12.995, | |
| "eval_steps_per_second": 0.409, | |
| "step": 30000 | |
| } | |
| ], | |
| "max_steps": 62500, | |
| "num_train_epochs": 20, | |
| "total_flos": 1.0815604062368563e+17, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |