| { | |
| "best_metric": 0.762901246547699, | |
| "best_model_checkpoint": "/content/drive/MyDrive/Colab Notebooks/models/t5-small-codesearchnet-multilang-python/checkpoint-3750", | |
| "epoch": 13.0, | |
| "global_step": 4875, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "eval_avg_length": 13.785, | |
| "eval_bleu": 0.0003, | |
| "eval_loss": 0.9942636489868164, | |
| "eval_rouge1": 0.1637, | |
| "eval_rouge2": 0.0365, | |
| "eval_runtime": 204.2034, | |
| "eval_samples_per_second": 24.485, | |
| "eval_steps_per_second": 3.061, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 0.011613128706812859, | |
| "loss": 2.445, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_avg_length": 13.0266, | |
| "eval_bleu": 0.0002, | |
| "eval_loss": 0.8991339802742004, | |
| "eval_rouge1": 0.171, | |
| "eval_rouge2": 0.041, | |
| "eval_runtime": 201.2399, | |
| "eval_samples_per_second": 24.846, | |
| "eval_steps_per_second": 3.106, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 0.02327057346701622, | |
| "loss": 0.8324, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_avg_length": 14.9474, | |
| "eval_bleu": 0.001, | |
| "eval_loss": 0.8509177565574646, | |
| "eval_rouge1": 0.1931, | |
| "eval_rouge2": 0.0499, | |
| "eval_runtime": 202.2102, | |
| "eval_samples_per_second": 24.727, | |
| "eval_steps_per_second": 3.091, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 0.03500078245997429, | |
| "loss": 0.7567, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_avg_length": 14.9598, | |
| "eval_bleu": 0.0015, | |
| "eval_loss": 0.8183696866035461, | |
| "eval_rouge1": 0.2019, | |
| "eval_rouge2": 0.0561, | |
| "eval_runtime": 202.119, | |
| "eval_samples_per_second": 24.738, | |
| "eval_steps_per_second": 3.092, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_avg_length": 14.496, | |
| "eval_bleu": 0.0016, | |
| "eval_loss": 0.8002403378486633, | |
| "eval_rouge1": 0.2097, | |
| "eval_rouge2": 0.0608, | |
| "eval_runtime": 192.5036, | |
| "eval_samples_per_second": 25.974, | |
| "eval_steps_per_second": 3.247, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 5.33, | |
| "learning_rate": 0.04683626815676689, | |
| "loss": 0.6947, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_avg_length": 14.6502, | |
| "eval_bleu": 0.0016, | |
| "eval_loss": 0.7792889475822449, | |
| "eval_rouge1": 0.2138, | |
| "eval_rouge2": 0.0631, | |
| "eval_runtime": 203.4751, | |
| "eval_samples_per_second": 24.573, | |
| "eval_steps_per_second": 3.072, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 6.67, | |
| "learning_rate": 0.05882224813103676, | |
| "loss": 0.658, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_avg_length": 15.2, | |
| "eval_bleu": 0.0018, | |
| "eval_loss": 0.7721025347709656, | |
| "eval_rouge1": 0.2104, | |
| "eval_rouge2": 0.0617, | |
| "eval_runtime": 185.635, | |
| "eval_samples_per_second": 26.935, | |
| "eval_steps_per_second": 3.367, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 0.07100344449281693, | |
| "loss": 0.6186, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_avg_length": 15.7472, | |
| "eval_bleu": 0.0023, | |
| "eval_loss": 0.7669309377670288, | |
| "eval_rouge1": 0.2175, | |
| "eval_rouge2": 0.0642, | |
| "eval_runtime": 203.9621, | |
| "eval_samples_per_second": 24.514, | |
| "eval_steps_per_second": 3.064, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_avg_length": 15.862, | |
| "eval_bleu": 0.0027, | |
| "eval_loss": 0.7791835069656372, | |
| "eval_rouge1": 0.2218, | |
| "eval_rouge2": 0.0664, | |
| "eval_runtime": 203.2586, | |
| "eval_samples_per_second": 24.599, | |
| "eval_steps_per_second": 3.075, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 9.33, | |
| "learning_rate": 0.08339222520589828, | |
| "loss": 0.58, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_avg_length": 12.0968, | |
| "eval_bleu": 0.0005, | |
| "eval_loss": 0.762901246547699, | |
| "eval_rouge1": 0.1985, | |
| "eval_rouge2": 0.0591, | |
| "eval_runtime": 199.4481, | |
| "eval_samples_per_second": 25.069, | |
| "eval_steps_per_second": 3.134, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 10.67, | |
| "learning_rate": 0.0960661843419075, | |
| "loss": 0.5533, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_avg_length": 16.9146, | |
| "eval_bleu": 0.0027, | |
| "eval_loss": 0.7826463580131531, | |
| "eval_rouge1": 0.2126, | |
| "eval_rouge2": 0.0631, | |
| "eval_runtime": 199.9109, | |
| "eval_samples_per_second": 25.011, | |
| "eval_steps_per_second": 3.126, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "learning_rate": 0.10908319801092148, | |
| "loss": 0.5279, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_avg_length": 16.656, | |
| "eval_bleu": 0.0025, | |
| "eval_loss": 0.7907007932662964, | |
| "eval_rouge1": 0.2144, | |
| "eval_rouge2": 0.0626, | |
| "eval_runtime": 193.36, | |
| "eval_samples_per_second": 25.859, | |
| "eval_steps_per_second": 3.232, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "eval_avg_length": 12.4734, | |
| "eval_bleu": 0.0007, | |
| "eval_loss": 0.7827274203300476, | |
| "eval_rouge1": 0.2019, | |
| "eval_rouge2": 0.0606, | |
| "eval_runtime": 198.3606, | |
| "eval_samples_per_second": 25.207, | |
| "eval_steps_per_second": 3.151, | |
| "step": 4875 | |
| } | |
| ], | |
| "max_steps": 5625, | |
| "num_train_epochs": 15, | |
| "total_flos": 2.639165128704e+16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |