2025-04-08 12:39:00 INFO: Running parser in train mode 2025-04-08 12:39:00 INFO: Directory saved_models/depparse does not exist; creating... 2025-04-08 12:39:00 INFO: Loading data with batch size 5000... 2025-04-08 12:39:01 INFO: Original data size: 5089 2025-04-08 12:39:01 WARNING: Found no sentences which matched can_augment_predicate 2025-04-08 12:39:01 INFO: Augmented data size: 5089 2025-04-08 12:39:02 WARNING: la_giuseppe is not a known dataset. Examining the data to choose which xpos vocab to use 2025-04-08 12:39:02 INFO: Original length = 5089 2025-04-08 12:39:02 INFO: Filtered length = 5089 2025-04-08 12:39:03 WARNING: Chose XPOSDescription(xpos_type=, sep='') for the xpos factory for la_giuseppe 2025-04-08 12:39:05 DEBUG: 19 batches created. 2025-04-08 12:39:05 DEBUG: 3 batches created. 2025-04-08 12:39:05 INFO: Training parser... Some weights of RobertaModel were not initialized from the model checkpoint at bowphs/LaBerta and are newly initialized: ['roberta.pooler.dense.bias', 'roberta.pooler.dense.weight'] You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. 2025-04-08 12:39:07 DEBUG: Building Adam with lr=0.003000, betas=(0.9, 0.95), eps=0.000001 2025-04-08 12:39:08 DEBUG: Building Adam with lr=0.003000, betas=(0.9, 0.95), eps=0.000001, weight_decay=0.0 2025-04-08 12:39:22 INFO: Finished STEP 20/50000, loss = 5.539350 (0.812 sec/batch), lr: 0.003000 2025-04-08 12:39:35 INFO: Finished STEP 40/50000, loss = 5.667335 (0.964 sec/batch), lr: 0.003000 2025-04-08 12:39:47 INFO: Finished STEP 60/50000, loss = 7.796917 (0.509 sec/batch), lr: 0.003000 2025-04-08 12:40:00 INFO: Finished STEP 80/50000, loss = 8.955844 (0.499 sec/batch), lr: 0.003000 2025-04-08 12:40:13 INFO: Finished STEP 100/50000, loss = 6.841305 (0.480 sec/batch), lr: 0.003000 2025-04-08 12:40:13 INFO: Evaluating on dev set... 2025-04-08 12:40:19 INFO: LAS MLAS BLEX 2025-04-08 12:40:19 INFO: 5.66 0.00 0.00 2025-04-08 12:40:19 INFO: step 100: train_loss = 12.155895, dev_score = 0.0566 2025-04-08 12:40:20 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 12:40:20 INFO: new best model saved. 2025-04-08 12:40:20 INFO: scheduler bert_scheduler learning rate: [0.0] 2025-04-08 12:40:20 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 12:40:20 INFO: new model checkpoint saved. 2025-04-08 12:40:34 INFO: Finished STEP 120/50000, loss = 6.639511 (0.490 sec/batch), lr: 0.003000 2025-04-08 12:40:47 INFO: Finished STEP 140/50000, loss = 6.906423 (0.602 sec/batch), lr: 0.003000 2025-04-08 12:41:00 INFO: Finished STEP 160/50000, loss = 4.782283 (0.574 sec/batch), lr: 0.003000 2025-04-08 12:41:13 INFO: Finished STEP 180/50000, loss = 4.495800 (0.574 sec/batch), lr: 0.003000 /gpfs/bwfor/home/hd/hd_hd/hd_ec199/reproduce/evalatin2024-latinpipe/.venv/lib/python3.12/site-packages/torch/optim/lr_scheduler.py:240: UserWarning: The epoch parameter in `scheduler.step()` was not necessary and is being deprecated where possible. Please use `scheduler.step()` to step the scheduler. During the deprecation, if epoch is different from None, the closed form is used instead of the new chainable form, where available. Please open an issue if you are unable to replicate your use case: https://github.com/pytorch/pytorch/issues/new/choose. warnings.warn(EPOCH_DEPRECATION_WARNING, UserWarning) 2025-04-08 12:41:26 INFO: Finished STEP 200/50000, loss = 3.751415 (0.638 sec/batch), lr: 0.003000 2025-04-08 12:41:26 INFO: Evaluating on dev set... 2025-04-08 12:41:30 INFO: LAS MLAS BLEX 2025-04-08 12:41:30 INFO: 32.20 0.00 0.00 2025-04-08 12:41:30 INFO: step 200: train_loss = 5.704251, dev_score = 0.3220 2025-04-08 12:41:30 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 12:41:30 INFO: new best model saved. 2025-04-08 12:41:30 INFO: scheduler bert_scheduler learning rate: [0.0] 2025-04-08 12:41:31 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 12:41:31 INFO: new model checkpoint saved. 2025-04-08 12:41:44 INFO: Finished STEP 220/50000, loss = 5.551241 (0.487 sec/batch), lr: 0.003000 2025-04-08 12:41:57 INFO: Finished STEP 240/50000, loss = 3.565381 (0.572 sec/batch), lr: 0.003000 2025-04-08 12:42:10 INFO: Finished STEP 260/50000, loss = 4.956666 (0.489 sec/batch), lr: 0.003000 2025-04-08 12:42:23 INFO: Finished STEP 280/50000, loss = 2.984083 (0.613 sec/batch), lr: 0.003000 2025-04-08 12:42:37 INFO: Finished STEP 300/50000, loss = 6.103778 (0.500 sec/batch), lr: 0.003000 2025-04-08 12:42:37 INFO: Evaluating on dev set... 2025-04-08 12:42:39 INFO: LAS MLAS BLEX 2025-04-08 12:42:39 INFO: 45.78 0.00 0.00 2025-04-08 12:42:39 INFO: step 300: train_loss = 3.965885, dev_score = 0.4578 2025-04-08 12:42:39 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 12:42:39 INFO: new best model saved. 2025-04-08 12:42:39 INFO: scheduler bert_scheduler learning rate: [0.0015] 2025-04-08 12:42:40 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 12:42:40 INFO: new model checkpoint saved. 2025-04-08 12:42:53 INFO: Finished STEP 320/50000, loss = 2.712356 (0.611 sec/batch), lr: 0.003000 2025-04-08 12:43:06 INFO: Finished STEP 340/50000, loss = 3.037747 (0.566 sec/batch), lr: 0.003000 2025-04-08 12:43:18 INFO: Finished STEP 360/50000, loss = 3.229224 (0.585 sec/batch), lr: 0.003000 2025-04-08 12:43:32 INFO: Finished STEP 380/50000, loss = 5.710863 (0.527 sec/batch), lr: 0.003000 2025-04-08 12:43:45 INFO: Finished STEP 400/50000, loss = 3.193926 (0.579 sec/batch), lr: 0.003000 2025-04-08 12:43:45 INFO: Evaluating on dev set... 2025-04-08 12:43:48 INFO: LAS MLAS BLEX 2025-04-08 12:43:48 INFO: 41.43 0.00 0.00 2025-04-08 12:43:48 INFO: step 400: train_loss = 3.817218, dev_score = 0.4143 2025-04-08 12:43:48 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 12:43:49 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 12:43:49 INFO: new model checkpoint saved. 2025-04-08 12:44:01 INFO: Finished STEP 420/50000, loss = 3.967075 (0.520 sec/batch), lr: 0.003000 2025-04-08 12:44:15 INFO: Finished STEP 440/50000, loss = 4.157637 (0.501 sec/batch), lr: 0.003000 2025-04-08 12:44:27 INFO: Finished STEP 460/50000, loss = 3.213076 (0.593 sec/batch), lr: 0.003000 2025-04-08 12:44:41 INFO: Finished STEP 480/50000, loss = 2.960643 (0.579 sec/batch), lr: 0.003000 2025-04-08 12:44:54 INFO: Finished STEP 500/50000, loss = 2.068380 (0.804 sec/batch), lr: 0.003000 2025-04-08 12:44:54 INFO: Evaluating on dev set... 2025-04-08 12:44:57 INFO: LAS MLAS BLEX 2025-04-08 12:44:57 INFO: 48.52 0.00 0.00 2025-04-08 12:44:57 INFO: step 500: train_loss = 3.894343, dev_score = 0.4852 2025-04-08 12:44:57 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 12:44:57 INFO: new best model saved. 2025-04-08 12:44:57 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 12:44:58 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 12:44:58 INFO: new model checkpoint saved. 2025-04-08 12:45:10 INFO: Finished STEP 520/50000, loss = 3.490720 (0.527 sec/batch), lr: 0.003000 2025-04-08 12:45:23 INFO: Finished STEP 540/50000, loss = 3.225158 (0.543 sec/batch), lr: 0.003000 2025-04-08 12:45:37 INFO: Finished STEP 560/50000, loss = 3.686380 (0.518 sec/batch), lr: 0.003000 2025-04-08 12:45:50 INFO: Finished STEP 580/50000, loss = 3.711233 (0.500 sec/batch), lr: 0.003000 2025-04-08 12:46:02 INFO: Finished STEP 600/50000, loss = 1.564660 (0.974 sec/batch), lr: 0.003000 2025-04-08 12:46:02 INFO: Evaluating on dev set... 2025-04-08 12:46:05 INFO: LAS MLAS BLEX 2025-04-08 12:46:05 INFO: 51.38 0.00 0.00 2025-04-08 12:46:05 INFO: step 600: train_loss = 3.569887, dev_score = 0.5138 2025-04-08 12:46:05 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 12:46:05 INFO: new best model saved. 2025-04-08 12:46:05 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 12:46:06 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 12:46:06 INFO: new model checkpoint saved. 2025-04-08 12:46:19 INFO: Finished STEP 620/50000, loss = 7.548009 (0.589 sec/batch), lr: 0.003000 2025-04-08 12:46:31 INFO: Finished STEP 640/50000, loss = 4.128086 (0.486 sec/batch), lr: 0.003000 2025-04-08 12:46:45 INFO: Finished STEP 660/50000, loss = 3.399152 (0.490 sec/batch), lr: 0.003000 2025-04-08 12:46:58 INFO: Finished STEP 680/50000, loss = 3.657129 (0.484 sec/batch), lr: 0.003000 2025-04-08 12:47:11 INFO: Finished STEP 700/50000, loss = 3.181570 (0.508 sec/batch), lr: 0.003000 2025-04-08 12:47:11 INFO: Evaluating on dev set... 2025-04-08 12:47:13 INFO: LAS MLAS BLEX 2025-04-08 12:47:13 INFO: 53.25 0.00 0.00 2025-04-08 12:47:13 INFO: step 700: train_loss = 3.260934, dev_score = 0.5325 2025-04-08 12:47:14 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 12:47:14 INFO: new best model saved. 2025-04-08 12:47:14 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 12:47:14 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 12:47:14 INFO: new model checkpoint saved. 2025-04-08 12:47:27 INFO: Finished STEP 720/50000, loss = 2.654041 (0.564 sec/batch), lr: 0.003000 2025-04-08 12:47:40 INFO: Finished STEP 740/50000, loss = 1.538671 (0.963 sec/batch), lr: 0.003000 2025-04-08 12:47:53 INFO: Finished STEP 760/50000, loss = 2.957897 (0.555 sec/batch), lr: 0.003000 2025-04-08 12:48:06 INFO: Finished STEP 780/50000, loss = 2.886534 (0.549 sec/batch), lr: 0.003000 2025-04-08 12:48:19 INFO: Finished STEP 800/50000, loss = 2.972443 (0.545 sec/batch), lr: 0.003000 2025-04-08 12:48:19 INFO: Evaluating on dev set... 2025-04-08 12:48:21 INFO: LAS MLAS BLEX 2025-04-08 12:48:21 INFO: 53.55 0.00 0.00 2025-04-08 12:48:21 INFO: step 800: train_loss = 3.219666, dev_score = 0.5355 2025-04-08 12:48:21 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 12:48:21 INFO: new best model saved. 2025-04-08 12:48:21 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 12:48:22 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 12:48:22 INFO: new model checkpoint saved. 2025-04-08 12:48:35 INFO: Finished STEP 820/50000, loss = 3.050849 (0.548 sec/batch), lr: 0.003000 2025-04-08 12:48:48 INFO: Finished STEP 840/50000, loss = 6.544877 (0.366 sec/batch), lr: 0.003000 2025-04-08 12:49:01 INFO: Finished STEP 860/50000, loss = 1.277311 (0.964 sec/batch), lr: 0.003000 2025-04-08 12:49:14 INFO: Finished STEP 880/50000, loss = 4.767632 (0.500 sec/batch), lr: 0.003000 2025-04-08 12:49:27 INFO: Finished STEP 900/50000, loss = 4.074880 (0.488 sec/batch), lr: 0.003000 2025-04-08 12:49:27 INFO: Evaluating on dev set... 2025-04-08 12:49:29 INFO: LAS MLAS BLEX 2025-04-08 12:49:29 INFO: 54.61 0.00 0.00 2025-04-08 12:49:29 INFO: step 900: train_loss = 3.083053, dev_score = 0.5461 2025-04-08 12:49:30 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 12:49:30 INFO: new best model saved. 2025-04-08 12:49:30 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 12:49:30 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 12:49:30 INFO: new model checkpoint saved. 2025-04-08 12:49:44 INFO: Finished STEP 920/50000, loss = 2.859793 (0.570 sec/batch), lr: 0.003000 2025-04-08 12:49:56 INFO: Finished STEP 940/50000, loss = 2.462156 (0.576 sec/batch), lr: 0.003000 2025-04-08 12:50:09 INFO: Finished STEP 960/50000, loss = 1.037470 (0.633 sec/batch), lr: 0.003000 2025-04-08 12:50:22 INFO: Finished STEP 980/50000, loss = 2.250489 (0.608 sec/batch), lr: 0.003000 2025-04-08 12:50:35 INFO: Finished STEP 1000/50000, loss = 3.635698 (0.482 sec/batch), lr: 0.003000 2025-04-08 12:50:35 INFO: Evaluating on dev set... 2025-04-08 12:50:37 INFO: LAS MLAS BLEX 2025-04-08 12:50:37 INFO: 55.65 0.00 0.00 2025-04-08 12:50:37 INFO: step 1000: train_loss = 3.043254, dev_score = 0.5565 2025-04-08 12:50:37 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 12:50:37 INFO: new best model saved. 2025-04-08 12:50:37 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 12:50:38 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 12:50:38 INFO: new model checkpoint saved. 2025-04-08 12:50:51 INFO: Finished STEP 1020/50000, loss = 2.816247 (0.493 sec/batch), lr: 0.003000 2025-04-08 12:51:04 INFO: Finished STEP 1040/50000, loss = 2.291480 (0.568 sec/batch), lr: 0.003000 2025-04-08 12:51:17 INFO: Finished STEP 1060/50000, loss = 2.112345 (0.578 sec/batch), lr: 0.003000 2025-04-08 12:51:30 INFO: Finished STEP 1080/50000, loss = 1.477807 (0.798 sec/batch), lr: 0.003000 2025-04-08 12:51:43 INFO: Finished STEP 1100/50000, loss = 2.729558 (0.525 sec/batch), lr: 0.003000 2025-04-08 12:51:43 INFO: Evaluating on dev set... 2025-04-08 12:51:45 INFO: LAS MLAS BLEX 2025-04-08 12:51:45 INFO: 56.91 0.00 0.00 2025-04-08 12:51:45 INFO: step 1100: train_loss = 2.878591, dev_score = 0.5691 2025-04-08 12:51:45 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 12:51:45 INFO: new best model saved. 2025-04-08 12:51:45 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 12:51:46 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 12:51:46 INFO: new model checkpoint saved. 2025-04-08 12:51:59 INFO: Finished STEP 1120/50000, loss = 2.260900 (0.575 sec/batch), lr: 0.003000 2025-04-08 12:52:12 INFO: Finished STEP 1140/50000, loss = 2.533880 (0.569 sec/batch), lr: 0.003000 2025-04-08 12:52:25 INFO: Finished STEP 1160/50000, loss = 1.987008 (0.608 sec/batch), lr: 0.003000 2025-04-08 12:52:38 INFO: Finished STEP 1180/50000, loss = 2.895975 (0.501 sec/batch), lr: 0.003000 2025-04-08 12:52:51 INFO: Finished STEP 1200/50000, loss = 3.675969 (0.484 sec/batch), lr: 0.003000 2025-04-08 12:52:51 INFO: Evaluating on dev set... 2025-04-08 12:52:53 INFO: LAS MLAS BLEX 2025-04-08 12:52:53 INFO: 57.55 0.00 0.00 2025-04-08 12:52:53 INFO: step 1200: train_loss = 2.809461, dev_score = 0.5755 2025-04-08 12:52:53 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 12:52:53 INFO: new best model saved. 2025-04-08 12:52:53 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 12:52:54 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 12:52:54 INFO: new model checkpoint saved. 2025-04-08 12:53:06 INFO: Finished STEP 1220/50000, loss = 2.549118 (0.547 sec/batch), lr: 0.003000 2025-04-08 12:53:19 INFO: Finished STEP 1240/50000, loss = 1.974401 (0.577 sec/batch), lr: 0.003000 2025-04-08 12:53:32 INFO: Finished STEP 1260/50000, loss = 3.038520 (0.492 sec/batch), lr: 0.003000 2025-04-08 12:53:45 INFO: Finished STEP 1280/50000, loss = 1.310572 (0.811 sec/batch), lr: 0.003000 2025-04-08 12:53:58 INFO: Finished STEP 1300/50000, loss = 4.426167 (0.498 sec/batch), lr: 0.003000 2025-04-08 12:53:58 INFO: Evaluating on dev set... 2025-04-08 12:54:00 INFO: LAS MLAS BLEX 2025-04-08 12:54:00 INFO: 58.68 0.00 0.00 2025-04-08 12:54:00 INFO: step 1300: train_loss = 2.730646, dev_score = 0.5868 2025-04-08 12:54:01 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 12:54:01 INFO: new best model saved. 2025-04-08 12:54:01 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 12:54:01 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 12:54:01 INFO: new model checkpoint saved. 2025-04-08 12:54:14 INFO: Finished STEP 1320/50000, loss = 2.054265 (0.578 sec/batch), lr: 0.003000 2025-04-08 12:54:27 INFO: Finished STEP 1340/50000, loss = 2.539223 (0.506 sec/batch), lr: 0.003000 2025-04-08 12:54:40 INFO: Finished STEP 1360/50000, loss = 2.670833 (0.512 sec/batch), lr: 0.003000 2025-04-08 12:54:53 INFO: Finished STEP 1380/50000, loss = 2.105981 (0.568 sec/batch), lr: 0.003000 2025-04-08 12:55:06 INFO: Finished STEP 1400/50000, loss = 1.710022 (0.641 sec/batch), lr: 0.003000 2025-04-08 12:55:06 INFO: Evaluating on dev set... 2025-04-08 12:55:08 INFO: LAS MLAS BLEX 2025-04-08 12:55:08 INFO: 57.54 0.00 0.00 2025-04-08 12:55:08 INFO: step 1400: train_loss = 2.666066, dev_score = 0.5754 2025-04-08 12:55:08 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 12:55:09 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 12:55:09 INFO: new model checkpoint saved. 2025-04-08 12:55:22 INFO: Finished STEP 1420/50000, loss = 2.763271 (0.519 sec/batch), lr: 0.003000 2025-04-08 12:55:34 INFO: Finished STEP 1440/50000, loss = 3.339348 (0.485 sec/batch), lr: 0.003000 2025-04-08 12:55:47 INFO: Finished STEP 1460/50000, loss = 1.654047 (0.678 sec/batch), lr: 0.003000 2025-04-08 12:56:00 INFO: Finished STEP 1480/50000, loss = 2.732059 (0.519 sec/batch), lr: 0.003000 2025-04-08 12:56:13 INFO: Finished STEP 1500/50000, loss = 3.039102 (0.511 sec/batch), lr: 0.003000 2025-04-08 12:56:13 INFO: Evaluating on dev set... 2025-04-08 12:56:15 INFO: LAS MLAS BLEX 2025-04-08 12:56:15 INFO: 58.79 0.00 0.00 2025-04-08 12:56:15 INFO: step 1500: train_loss = 2.703031, dev_score = 0.5879 2025-04-08 12:56:15 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 12:56:15 INFO: new best model saved. 2025-04-08 12:56:15 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 12:56:16 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 12:56:16 INFO: new model checkpoint saved. 2025-04-08 12:56:29 INFO: Finished STEP 1520/50000, loss = 1.318498 (0.816 sec/batch), lr: 0.003000 2025-04-08 12:56:42 INFO: Finished STEP 1540/50000, loss = 1.286851 (0.814 sec/batch), lr: 0.003000 2025-04-08 12:56:55 INFO: Finished STEP 1560/50000, loss = 1.444117 (0.800 sec/batch), lr: 0.003000 2025-04-08 12:57:08 INFO: Finished STEP 1580/50000, loss = 2.130930 (0.577 sec/batch), lr: 0.003000 2025-04-08 12:57:21 INFO: Finished STEP 1600/50000, loss = 2.605561 (0.536 sec/batch), lr: 0.003000 2025-04-08 12:57:21 INFO: Evaluating on dev set... 2025-04-08 12:57:23 INFO: LAS MLAS BLEX 2025-04-08 12:57:23 INFO: 58.78 0.00 0.00 2025-04-08 12:57:23 INFO: step 1600: train_loss = 2.673654, dev_score = 0.5878 2025-04-08 12:57:23 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 12:57:24 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 12:57:24 INFO: new model checkpoint saved. 2025-04-08 12:57:37 INFO: Finished STEP 1620/50000, loss = 3.006736 (0.483 sec/batch), lr: 0.003000 2025-04-08 12:57:49 INFO: Finished STEP 1640/50000, loss = 2.466913 (0.539 sec/batch), lr: 0.003000 2025-04-08 12:58:02 INFO: Finished STEP 1660/50000, loss = 3.121765 (0.499 sec/batch), lr: 0.003000 2025-04-08 12:58:15 INFO: Finished STEP 1680/50000, loss = 2.677148 (0.519 sec/batch), lr: 0.003000 2025-04-08 12:58:28 INFO: Finished STEP 1700/50000, loss = 2.791246 (0.492 sec/batch), lr: 0.003000 2025-04-08 12:58:28 INFO: Evaluating on dev set... 2025-04-08 12:58:30 INFO: LAS MLAS BLEX 2025-04-08 12:58:30 INFO: 59.24 0.00 0.00 2025-04-08 12:58:30 INFO: step 1700: train_loss = 2.587400, dev_score = 0.5924 2025-04-08 12:58:30 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 12:58:30 INFO: new best model saved. 2025-04-08 12:58:30 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 12:58:31 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 12:58:31 INFO: new model checkpoint saved. 2025-04-08 12:58:44 INFO: Finished STEP 1720/50000, loss = 3.274817 (0.501 sec/batch), lr: 0.003000 2025-04-08 12:58:56 INFO: Finished STEP 1740/50000, loss = 2.317755 (0.543 sec/batch), lr: 0.003000 2025-04-08 12:59:10 INFO: Finished STEP 1760/50000, loss = 2.167442 (0.577 sec/batch), lr: 0.003000 2025-04-08 12:59:22 INFO: Finished STEP 1780/50000, loss = 2.197192 (0.554 sec/batch), lr: 0.003000 2025-04-08 12:59:36 INFO: Finished STEP 1800/50000, loss = 1.999892 (0.568 sec/batch), lr: 0.003000 2025-04-08 12:59:36 INFO: Evaluating on dev set... 2025-04-08 12:59:38 INFO: LAS MLAS BLEX 2025-04-08 12:59:38 INFO: 59.53 0.00 0.00 2025-04-08 12:59:38 INFO: step 1800: train_loss = 2.634204, dev_score = 0.5953 2025-04-08 12:59:38 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 12:59:38 INFO: new best model saved. 2025-04-08 12:59:38 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 12:59:39 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 12:59:39 INFO: new model checkpoint saved. 2025-04-08 12:59:52 INFO: Finished STEP 1820/50000, loss = 2.443938 (0.507 sec/batch), lr: 0.003000 2025-04-08 13:00:05 INFO: Finished STEP 1840/50000, loss = 1.621668 (0.644 sec/batch), lr: 0.003000 2025-04-08 13:00:18 INFO: Finished STEP 1860/50000, loss = 5.905592 (0.367 sec/batch), lr: 0.003000 2025-04-08 13:00:30 INFO: Finished STEP 1880/50000, loss = 0.991068 (0.636 sec/batch), lr: 0.003000 2025-04-08 13:00:44 INFO: Finished STEP 1900/50000, loss = 6.288648 (0.590 sec/batch), lr: 0.003000 2025-04-08 13:00:44 INFO: Evaluating on dev set... 2025-04-08 13:00:45 INFO: LAS MLAS BLEX 2025-04-08 13:00:45 INFO: 59.80 0.00 0.00 2025-04-08 13:00:45 INFO: step 1900: train_loss = 2.470928, dev_score = 0.5980 2025-04-08 13:00:46 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 13:00:46 INFO: new best model saved. 2025-04-08 13:00:46 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:00:46 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:00:46 INFO: new model checkpoint saved. 2025-04-08 13:00:59 INFO: Finished STEP 1920/50000, loss = 2.760041 (0.485 sec/batch), lr: 0.003000 2025-04-08 13:01:12 INFO: Finished STEP 1940/50000, loss = 3.123771 (0.485 sec/batch), lr: 0.003000 2025-04-08 13:01:25 INFO: Finished STEP 1960/50000, loss = 2.070040 (0.555 sec/batch), lr: 0.003000 2025-04-08 13:01:38 INFO: Finished STEP 1980/50000, loss = 3.560690 (0.605 sec/batch), lr: 0.003000 2025-04-08 13:01:51 INFO: Finished STEP 2000/50000, loss = 2.555387 (0.518 sec/batch), lr: 0.003000 2025-04-08 13:01:51 INFO: Evaluating on dev set... 2025-04-08 13:01:53 INFO: LAS MLAS BLEX 2025-04-08 13:01:53 INFO: 59.14 0.00 0.00 2025-04-08 13:01:53 INFO: step 2000: train_loss = 2.457844, dev_score = 0.5914 2025-04-08 13:01:53 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:01:53 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:01:53 INFO: new model checkpoint saved. 2025-04-08 13:02:06 INFO: Finished STEP 2020/50000, loss = 1.170316 (0.812 sec/batch), lr: 0.003000 2025-04-08 13:02:19 INFO: Finished STEP 2040/50000, loss = 1.228714 (0.800 sec/batch), lr: 0.003000 2025-04-08 13:02:32 INFO: Finished STEP 2060/50000, loss = 1.513821 (0.678 sec/batch), lr: 0.003000 2025-04-08 13:02:45 INFO: Finished STEP 2080/50000, loss = 2.254135 (0.527 sec/batch), lr: 0.003000 2025-04-08 13:02:58 INFO: Finished STEP 2100/50000, loss = 2.065131 (0.555 sec/batch), lr: 0.003000 2025-04-08 13:02:58 INFO: Evaluating on dev set... 2025-04-08 13:03:00 INFO: LAS MLAS BLEX 2025-04-08 13:03:00 INFO: 60.29 0.00 0.00 2025-04-08 13:03:00 INFO: step 2100: train_loss = 2.437357, dev_score = 0.6029 2025-04-08 13:03:00 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 13:03:00 INFO: new best model saved. 2025-04-08 13:03:00 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:03:01 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:03:01 INFO: new model checkpoint saved. 2025-04-08 13:03:14 INFO: Finished STEP 2120/50000, loss = 3.186567 (0.484 sec/batch), lr: 0.003000 2025-04-08 13:03:26 INFO: Finished STEP 2140/50000, loss = 2.698305 (0.491 sec/batch), lr: 0.003000 2025-04-08 13:03:39 INFO: Finished STEP 2160/50000, loss = 2.479464 (0.496 sec/batch), lr: 0.003000 2025-04-08 13:03:52 INFO: Finished STEP 2180/50000, loss = 2.747345 (0.485 sec/batch), lr: 0.003000 2025-04-08 13:04:06 INFO: Finished STEP 2200/50000, loss = 2.107197 (0.538 sec/batch), lr: 0.003000 2025-04-08 13:04:06 INFO: Evaluating on dev set... 2025-04-08 13:04:07 INFO: LAS MLAS BLEX 2025-04-08 13:04:07 INFO: 59.84 0.00 0.00 2025-04-08 13:04:07 INFO: step 2200: train_loss = 2.478039, dev_score = 0.5984 2025-04-08 13:04:07 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:04:08 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:04:08 INFO: new model checkpoint saved. 2025-04-08 13:04:20 INFO: Finished STEP 2220/50000, loss = 2.436221 (0.500 sec/batch), lr: 0.003000 2025-04-08 13:04:33 INFO: Finished STEP 2240/50000, loss = 2.060035 (0.555 sec/batch), lr: 0.003000 2025-04-08 13:04:46 INFO: Finished STEP 2260/50000, loss = 1.641157 (0.617 sec/batch), lr: 0.003000 2025-04-08 13:05:00 INFO: Finished STEP 2280/50000, loss = 1.223841 (1.153 sec/batch), lr: 0.003000 2025-04-08 13:05:12 INFO: Finished STEP 2300/50000, loss = 2.068623 (0.538 sec/batch), lr: 0.003000 2025-04-08 13:05:12 INFO: Evaluating on dev set... 2025-04-08 13:05:14 INFO: LAS MLAS BLEX 2025-04-08 13:05:14 INFO: 59.66 0.00 0.00 2025-04-08 13:05:14 INFO: step 2300: train_loss = 2.374079, dev_score = 0.5966 2025-04-08 13:05:14 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:05:15 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:05:15 INFO: new model checkpoint saved. 2025-04-08 13:05:28 INFO: Finished STEP 2320/50000, loss = 6.100393 (0.589 sec/batch), lr: 0.003000 2025-04-08 13:05:41 INFO: Finished STEP 2340/50000, loss = 1.715900 (0.582 sec/batch), lr: 0.003000 2025-04-08 13:05:53 INFO: Finished STEP 2360/50000, loss = 2.767299 (0.510 sec/batch), lr: 0.003000 2025-04-08 13:06:07 INFO: Finished STEP 2380/50000, loss = 3.209242 (0.485 sec/batch), lr: 0.003000 2025-04-08 13:06:20 INFO: Finished STEP 2400/50000, loss = 5.401347 (0.367 sec/batch), lr: 0.003000 2025-04-08 13:06:20 INFO: Evaluating on dev set... 2025-04-08 13:06:22 INFO: LAS MLAS BLEX 2025-04-08 13:06:22 INFO: 59.59 0.00 0.00 2025-04-08 13:06:22 INFO: step 2400: train_loss = 2.386782, dev_score = 0.5959 2025-04-08 13:06:22 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:06:22 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:06:22 INFO: new model checkpoint saved. 2025-04-08 13:06:35 INFO: Finished STEP 2420/50000, loss = 1.718473 (0.591 sec/batch), lr: 0.003000 2025-04-08 13:06:48 INFO: Finished STEP 2440/50000, loss = 2.033537 (0.569 sec/batch), lr: 0.003000 2025-04-08 13:07:01 INFO: Finished STEP 2460/50000, loss = 2.534352 (0.501 sec/batch), lr: 0.003000 2025-04-08 13:07:15 INFO: Finished STEP 2480/50000, loss = 1.899252 (0.574 sec/batch), lr: 0.003000 2025-04-08 13:07:27 INFO: Finished STEP 2500/50000, loss = 3.087384 (0.607 sec/batch), lr: 0.003000 2025-04-08 13:07:27 INFO: Evaluating on dev set... 2025-04-08 13:07:29 INFO: LAS MLAS BLEX 2025-04-08 13:07:29 INFO: 60.51 0.00 0.00 2025-04-08 13:07:29 INFO: step 2500: train_loss = 2.430344, dev_score = 0.6051 2025-04-08 13:07:29 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 13:07:29 INFO: new best model saved. 2025-04-08 13:07:29 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:07:29 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:07:29 INFO: new model checkpoint saved. 2025-04-08 13:07:43 INFO: Finished STEP 2520/50000, loss = 3.308458 (0.562 sec/batch), lr: 0.003000 2025-04-08 13:07:55 INFO: Finished STEP 2540/50000, loss = 5.937185 (0.591 sec/batch), lr: 0.003000 2025-04-08 13:08:09 INFO: Finished STEP 2560/50000, loss = 1.817277 (0.575 sec/batch), lr: 0.003000 2025-04-08 13:08:21 INFO: Finished STEP 2580/50000, loss = 3.219316 (0.604 sec/batch), lr: 0.003000 2025-04-08 13:08:34 INFO: Finished STEP 2600/50000, loss = 1.785019 (0.571 sec/batch), lr: 0.003000 2025-04-08 13:08:34 INFO: Evaluating on dev set... 2025-04-08 13:08:36 INFO: LAS MLAS BLEX 2025-04-08 13:08:36 INFO: 60.61 0.00 0.00 2025-04-08 13:08:36 INFO: step 2600: train_loss = 2.295300, dev_score = 0.6061 2025-04-08 13:08:36 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 13:08:36 INFO: new best model saved. 2025-04-08 13:08:36 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:08:37 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:08:37 INFO: new model checkpoint saved. 2025-04-08 13:08:50 INFO: Finished STEP 2620/50000, loss = 1.825521 (0.569 sec/batch), lr: 0.003000 2025-04-08 13:09:03 INFO: Finished STEP 2640/50000, loss = 1.665753 (0.580 sec/batch), lr: 0.003000 2025-04-08 13:09:16 INFO: Finished STEP 2660/50000, loss = 2.888885 (0.508 sec/batch), lr: 0.003000 2025-04-08 13:09:29 INFO: Finished STEP 2680/50000, loss = 1.443666 (0.676 sec/batch), lr: 0.003000 2025-04-08 13:09:42 INFO: Finished STEP 2700/50000, loss = 2.018549 (0.544 sec/batch), lr: 0.003000 2025-04-08 13:09:42 INFO: Evaluating on dev set... 2025-04-08 13:09:43 INFO: LAS MLAS BLEX 2025-04-08 13:09:43 INFO: 60.85 0.00 0.00 2025-04-08 13:09:43 INFO: step 2700: train_loss = 2.271099, dev_score = 0.6085 2025-04-08 13:09:44 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 13:09:44 INFO: new best model saved. 2025-04-08 13:09:44 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:09:44 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:09:44 INFO: new model checkpoint saved. 2025-04-08 13:09:57 INFO: Finished STEP 2720/50000, loss = 3.243386 (0.564 sec/batch), lr: 0.003000 2025-04-08 13:10:10 INFO: Finished STEP 2740/50000, loss = 0.909023 (0.964 sec/batch), lr: 0.003000 2025-04-08 13:10:24 INFO: Finished STEP 2760/50000, loss = 0.843779 (0.961 sec/batch), lr: 0.003000 2025-04-08 13:10:36 INFO: Finished STEP 2780/50000, loss = 1.016577 (0.815 sec/batch), lr: 0.003000 2025-04-08 13:10:49 INFO: Finished STEP 2800/50000, loss = 1.194479 (0.799 sec/batch), lr: 0.003000 2025-04-08 13:10:49 INFO: Evaluating on dev set... 2025-04-08 13:10:51 INFO: LAS MLAS BLEX 2025-04-08 13:10:51 INFO: 60.02 0.00 0.00 2025-04-08 13:10:51 INFO: step 2800: train_loss = 2.273083, dev_score = 0.6002 2025-04-08 13:10:51 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:10:52 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:10:52 INFO: new model checkpoint saved. 2025-04-08 13:11:05 INFO: Finished STEP 2820/50000, loss = 1.778186 (0.574 sec/batch), lr: 0.003000 2025-04-08 13:11:18 INFO: Finished STEP 2840/50000, loss = 0.842333 (0.964 sec/batch), lr: 0.003000 2025-04-08 13:11:31 INFO: Finished STEP 2860/50000, loss = 0.689533 (0.619 sec/batch), lr: 0.003000 2025-04-08 13:11:43 INFO: Finished STEP 2880/50000, loss = 2.071353 (0.548 sec/batch), lr: 0.003000 2025-04-08 13:11:56 INFO: Finished STEP 2900/50000, loss = 1.157779 (0.718 sec/batch), lr: 0.003000 2025-04-08 13:11:56 INFO: Evaluating on dev set... 2025-04-08 13:11:58 INFO: LAS MLAS BLEX 2025-04-08 13:11:58 INFO: 61.00 0.00 0.00 2025-04-08 13:11:58 INFO: step 2900: train_loss = 2.271078, dev_score = 0.6100 2025-04-08 13:11:58 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 13:11:58 INFO: new best model saved. 2025-04-08 13:11:58 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:11:59 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:11:59 INFO: new model checkpoint saved. 2025-04-08 13:12:11 INFO: Finished STEP 2920/50000, loss = 1.744012 (0.567 sec/batch), lr: 0.003000 2025-04-08 13:12:25 INFO: Finished STEP 2940/50000, loss = 2.185216 (0.509 sec/batch), lr: 0.003000 2025-04-08 13:12:38 INFO: Finished STEP 2960/50000, loss = 1.065106 (0.816 sec/batch), lr: 0.003000 2025-04-08 13:12:51 INFO: Finished STEP 2980/50000, loss = 3.869857 (0.501 sec/batch), lr: 0.003000 2025-04-08 13:13:03 INFO: Finished STEP 3000/50000, loss = 1.721972 (0.586 sec/batch), lr: 0.003000 2025-04-08 13:13:03 INFO: Evaluating on dev set... 2025-04-08 13:13:05 INFO: LAS MLAS BLEX 2025-04-08 13:13:05 INFO: 60.57 0.00 0.00 2025-04-08 13:13:05 INFO: step 3000: train_loss = 2.231438, dev_score = 0.6057 2025-04-08 13:13:05 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:13:06 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:13:06 INFO: new model checkpoint saved. 2025-04-08 13:13:19 INFO: Finished STEP 3020/50000, loss = 2.263343 (0.509 sec/batch), lr: 0.003000 2025-04-08 13:13:32 INFO: Finished STEP 3040/50000, loss = 1.841634 (0.538 sec/batch), lr: 0.003000 2025-04-08 13:13:45 INFO: Finished STEP 3060/50000, loss = 1.553981 (0.577 sec/batch), lr: 0.003000 2025-04-08 13:13:58 INFO: Finished STEP 3080/50000, loss = 2.995255 (0.604 sec/batch), lr: 0.003000 2025-04-08 13:14:11 INFO: Finished STEP 3100/50000, loss = 2.299805 (0.479 sec/batch), lr: 0.003000 2025-04-08 13:14:11 INFO: Evaluating on dev set... 2025-04-08 13:14:13 INFO: LAS MLAS BLEX 2025-04-08 13:14:13 INFO: 60.87 0.00 0.00 2025-04-08 13:14:13 INFO: step 3100: train_loss = 2.168434, dev_score = 0.6087 2025-04-08 13:14:13 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:14:13 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:14:13 INFO: new model checkpoint saved. 2025-04-08 13:14:26 INFO: Finished STEP 3120/50000, loss = 2.187892 (0.496 sec/batch), lr: 0.003000 2025-04-08 13:14:38 INFO: Finished STEP 3140/50000, loss = 1.582949 (0.574 sec/batch), lr: 0.003000 2025-04-08 13:14:52 INFO: Finished STEP 3160/50000, loss = 5.442513 (0.365 sec/batch), lr: 0.003000 2025-04-08 13:15:05 INFO: Finished STEP 3180/50000, loss = 2.115753 (0.509 sec/batch), lr: 0.003000 2025-04-08 13:15:17 INFO: Finished STEP 3200/50000, loss = 5.479081 (0.365 sec/batch), lr: 0.003000 2025-04-08 13:15:17 INFO: Evaluating on dev set... 2025-04-08 13:15:19 INFO: LAS MLAS BLEX 2025-04-08 13:15:19 INFO: 60.37 0.00 0.00 2025-04-08 13:15:19 INFO: step 3200: train_loss = 2.243722, dev_score = 0.6037 2025-04-08 13:15:19 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:15:20 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:15:20 INFO: new model checkpoint saved. 2025-04-08 13:15:33 INFO: Finished STEP 3220/50000, loss = 0.965773 (0.961 sec/batch), lr: 0.003000 2025-04-08 13:15:46 INFO: Finished STEP 3240/50000, loss = 2.728976 (0.501 sec/batch), lr: 0.003000 2025-04-08 13:15:59 INFO: Finished STEP 3260/50000, loss = 2.640983 (0.493 sec/batch), lr: 0.003000 2025-04-08 13:16:12 INFO: Finished STEP 3280/50000, loss = 2.001231 (0.542 sec/batch), lr: 0.003000 2025-04-08 13:16:25 INFO: Finished STEP 3300/50000, loss = 2.039145 (0.491 sec/batch), lr: 0.003000 2025-04-08 13:16:25 INFO: Evaluating on dev set... 2025-04-08 13:16:26 INFO: LAS MLAS BLEX 2025-04-08 13:16:26 INFO: 60.91 0.00 0.00 2025-04-08 13:16:26 INFO: step 3300: train_loss = 2.207767, dev_score = 0.6091 2025-04-08 13:16:26 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:16:27 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:16:27 INFO: new model checkpoint saved. 2025-04-08 13:16:40 INFO: Finished STEP 3320/50000, loss = 0.940702 (0.814 sec/batch), lr: 0.003000 2025-04-08 13:16:52 INFO: Finished STEP 3340/50000, loss = 2.510640 (0.485 sec/batch), lr: 0.003000 2025-04-08 13:17:06 INFO: Finished STEP 3360/50000, loss = 1.854818 (0.536 sec/batch), lr: 0.003000 2025-04-08 13:17:18 INFO: Finished STEP 3380/50000, loss = 1.540765 (0.580 sec/batch), lr: 0.003000 2025-04-08 13:17:31 INFO: Finished STEP 3400/50000, loss = 1.423092 (0.602 sec/batch), lr: 0.003000 2025-04-08 13:17:31 INFO: Evaluating on dev set... 2025-04-08 13:17:33 INFO: LAS MLAS BLEX 2025-04-08 13:17:33 INFO: 60.42 0.00 0.00 2025-04-08 13:17:33 INFO: step 3400: train_loss = 2.187651, dev_score = 0.6042 2025-04-08 13:17:33 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:17:34 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:17:34 INFO: new model checkpoint saved. 2025-04-08 13:17:47 INFO: Finished STEP 3420/50000, loss = 3.392330 (0.525 sec/batch), lr: 0.003000 2025-04-08 13:18:00 INFO: Finished STEP 3440/50000, loss = 5.625112 (0.589 sec/batch), lr: 0.003000 2025-04-08 13:18:13 INFO: Finished STEP 3460/50000, loss = 1.151558 (0.811 sec/batch), lr: 0.003000 2025-04-08 13:18:26 INFO: Finished STEP 3480/50000, loss = 1.158133 (0.714 sec/batch), lr: 0.003000 2025-04-08 13:18:38 INFO: Finished STEP 3500/50000, loss = 1.242396 (0.677 sec/batch), lr: 0.003000 2025-04-08 13:18:38 INFO: Evaluating on dev set... 2025-04-08 13:18:40 INFO: LAS MLAS BLEX 2025-04-08 13:18:40 INFO: 61.07 0.00 0.00 2025-04-08 13:18:40 INFO: step 3500: train_loss = 2.134372, dev_score = 0.6107 2025-04-08 13:18:40 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 13:18:40 INFO: new best model saved. 2025-04-08 13:18:40 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:18:41 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:18:41 INFO: new model checkpoint saved. 2025-04-08 13:18:54 INFO: Finished STEP 3520/50000, loss = 0.634599 (0.633 sec/batch), lr: 0.003000 2025-04-08 13:19:07 INFO: Finished STEP 3540/50000, loss = 1.534561 (0.583 sec/batch), lr: 0.003000 2025-04-08 13:19:20 INFO: Finished STEP 3560/50000, loss = 2.300629 (0.492 sec/batch), lr: 0.003000 2025-04-08 13:19:33 INFO: Finished STEP 3580/50000, loss = 1.500450 (0.578 sec/batch), lr: 0.003000 2025-04-08 13:19:46 INFO: Finished STEP 3600/50000, loss = 1.586440 (0.583 sec/batch), lr: 0.003000 2025-04-08 13:19:46 INFO: Evaluating on dev set... 2025-04-08 13:19:48 INFO: LAS MLAS BLEX 2025-04-08 13:19:48 INFO: 61.18 0.00 0.00 2025-04-08 13:19:48 INFO: step 3600: train_loss = 2.094013, dev_score = 0.6118 2025-04-08 13:19:48 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 13:19:48 INFO: new best model saved. 2025-04-08 13:19:48 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:19:49 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:19:49 INFO: new model checkpoint saved. 2025-04-08 13:20:01 INFO: Finished STEP 3620/50000, loss = 1.883033 (0.508 sec/batch), lr: 0.003000 2025-04-08 13:20:15 INFO: Finished STEP 3640/50000, loss = 1.676175 (0.565 sec/batch), lr: 0.003000 2025-04-08 13:20:27 INFO: Finished STEP 3660/50000, loss = 1.904340 (0.505 sec/batch), lr: 0.003000 2025-04-08 13:20:40 INFO: Finished STEP 3680/50000, loss = 5.045929 (0.366 sec/batch), lr: 0.003000 2025-04-08 13:20:53 INFO: Finished STEP 3700/50000, loss = 1.096014 (0.703 sec/batch), lr: 0.003000 2025-04-08 13:20:53 INFO: Evaluating on dev set... 2025-04-08 13:20:54 INFO: LAS MLAS BLEX 2025-04-08 13:20:54 INFO: 61.09 0.00 0.00 2025-04-08 13:20:54 INFO: step 3700: train_loss = 2.108466, dev_score = 0.6109 2025-04-08 13:20:54 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:20:55 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:20:55 INFO: new model checkpoint saved. 2025-04-08 13:21:08 INFO: Finished STEP 3720/50000, loss = 1.981587 (0.546 sec/batch), lr: 0.003000 2025-04-08 13:21:21 INFO: Finished STEP 3740/50000, loss = 1.463764 (0.574 sec/batch), lr: 0.003000 2025-04-08 13:21:34 INFO: Finished STEP 3760/50000, loss = 3.789796 (0.498 sec/batch), lr: 0.003000 2025-04-08 13:21:47 INFO: Finished STEP 3780/50000, loss = 1.662740 (0.547 sec/batch), lr: 0.003000 2025-04-08 13:22:00 INFO: Finished STEP 3800/50000, loss = 1.294607 (0.677 sec/batch), lr: 0.003000 2025-04-08 13:22:00 INFO: Evaluating on dev set... 2025-04-08 13:22:02 INFO: LAS MLAS BLEX 2025-04-08 13:22:02 INFO: 61.05 0.00 0.00 2025-04-08 13:22:02 INFO: step 3800: train_loss = 2.080286, dev_score = 0.6105 2025-04-08 13:22:02 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:22:02 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:22:02 INFO: new model checkpoint saved. 2025-04-08 13:22:15 INFO: Finished STEP 3820/50000, loss = 1.808879 (0.505 sec/batch), lr: 0.003000 2025-04-08 13:22:28 INFO: Finished STEP 3840/50000, loss = 2.020971 (0.509 sec/batch), lr: 0.003000 2025-04-08 13:22:42 INFO: Finished STEP 3860/50000, loss = 1.030083 (0.799 sec/batch), lr: 0.003000 2025-04-08 13:22:54 INFO: Finished STEP 3880/50000, loss = 3.481009 (0.527 sec/batch), lr: 0.003000 2025-04-08 13:23:07 INFO: Finished STEP 3900/50000, loss = 1.689104 (0.572 sec/batch), lr: 0.003000 2025-04-08 13:23:07 INFO: Evaluating on dev set... 2025-04-08 13:23:09 INFO: LAS MLAS BLEX 2025-04-08 13:23:09 INFO: 61.73 0.00 0.00 2025-04-08 13:23:09 INFO: step 3900: train_loss = 2.102922, dev_score = 0.6173 2025-04-08 13:23:09 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 13:23:09 INFO: new best model saved. 2025-04-08 13:23:09 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:23:10 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:23:10 INFO: new model checkpoint saved. 2025-04-08 13:23:23 INFO: Finished STEP 3920/50000, loss = 1.173690 (0.677 sec/batch), lr: 0.003000 2025-04-08 13:23:36 INFO: Finished STEP 3940/50000, loss = 3.014637 (0.488 sec/batch), lr: 0.003000 2025-04-08 13:23:49 INFO: Finished STEP 3960/50000, loss = 2.497069 (0.510 sec/batch), lr: 0.003000 2025-04-08 13:24:01 INFO: Finished STEP 3980/50000, loss = 1.470442 (0.607 sec/batch), lr: 0.003000 2025-04-08 13:24:15 INFO: Finished STEP 4000/50000, loss = 1.697983 (0.573 sec/batch), lr: 0.003000 2025-04-08 13:24:15 INFO: Evaluating on dev set... 2025-04-08 13:24:16 INFO: LAS MLAS BLEX 2025-04-08 13:24:16 INFO: 61.38 0.00 0.00 2025-04-08 13:24:16 INFO: step 4000: train_loss = 2.104181, dev_score = 0.6138 2025-04-08 13:24:16 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:24:17 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:24:17 INFO: new model checkpoint saved. 2025-04-08 13:24:30 INFO: Finished STEP 4020/50000, loss = 1.896750 (0.514 sec/batch), lr: 0.003000 2025-04-08 13:24:43 INFO: Finished STEP 4040/50000, loss = 2.108132 (0.500 sec/batch), lr: 0.003000 2025-04-08 13:24:55 INFO: Finished STEP 4060/50000, loss = 2.048178 (0.484 sec/batch), lr: 0.003000 2025-04-08 13:25:08 INFO: Finished STEP 4080/50000, loss = 2.519763 (0.500 sec/batch), lr: 0.003000 2025-04-08 13:25:21 INFO: Finished STEP 4100/50000, loss = 1.603151 (0.552 sec/batch), lr: 0.003000 2025-04-08 13:25:21 INFO: Evaluating on dev set... 2025-04-08 13:25:23 INFO: LAS MLAS BLEX 2025-04-08 13:25:23 INFO: 61.70 0.00 0.00 2025-04-08 13:25:23 INFO: step 4100: train_loss = 2.098191, dev_score = 0.6170 2025-04-08 13:25:23 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:25:24 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:25:24 INFO: new model checkpoint saved. 2025-04-08 13:25:37 INFO: Finished STEP 4120/50000, loss = 1.165612 (0.680 sec/batch), lr: 0.003000 2025-04-08 13:25:50 INFO: Finished STEP 4140/50000, loss = 2.668312 (0.484 sec/batch), lr: 0.003000 2025-04-08 13:26:03 INFO: Finished STEP 4160/50000, loss = 2.389383 (0.489 sec/batch), lr: 0.003000 2025-04-08 13:26:16 INFO: Finished STEP 4180/50000, loss = 2.645849 (0.487 sec/batch), lr: 0.003000 2025-04-08 13:26:29 INFO: Finished STEP 4200/50000, loss = 1.524088 (0.559 sec/batch), lr: 0.003000 2025-04-08 13:26:29 INFO: Evaluating on dev set... 2025-04-08 13:26:30 INFO: LAS MLAS BLEX 2025-04-08 13:26:30 INFO: 62.04 0.00 0.00 2025-04-08 13:26:30 INFO: step 4200: train_loss = 2.025912, dev_score = 0.6204 2025-04-08 13:26:30 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 13:26:30 INFO: new best model saved. 2025-04-08 13:26:30 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:26:31 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:26:31 INFO: new model checkpoint saved. 2025-04-08 13:26:44 INFO: Finished STEP 4220/50000, loss = 1.443073 (0.575 sec/batch), lr: 0.003000 2025-04-08 13:26:57 INFO: Finished STEP 4240/50000, loss = 2.192273 (0.485 sec/batch), lr: 0.003000 2025-04-08 13:27:10 INFO: Finished STEP 4260/50000, loss = 1.348808 (0.578 sec/batch), lr: 0.003000 2025-04-08 13:27:24 INFO: Finished STEP 4280/50000, loss = 2.059426 (0.517 sec/batch), lr: 0.003000 2025-04-08 13:27:36 INFO: Finished STEP 4300/50000, loss = 1.114165 (0.733 sec/batch), lr: 0.003000 2025-04-08 13:27:36 INFO: Evaluating on dev set... 2025-04-08 13:27:38 INFO: LAS MLAS BLEX 2025-04-08 13:27:38 INFO: 62.06 0.00 0.00 2025-04-08 13:27:38 INFO: step 4300: train_loss = 1.998292, dev_score = 0.6206 2025-04-08 13:27:38 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 13:27:38 INFO: new best model saved. 2025-04-08 13:27:38 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:27:39 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:27:39 INFO: new model checkpoint saved. 2025-04-08 13:27:52 INFO: Finished STEP 4320/50000, loss = 1.588072 (0.563 sec/batch), lr: 0.003000 2025-04-08 13:28:04 INFO: Finished STEP 4340/50000, loss = 1.597782 (0.562 sec/batch), lr: 0.003000 2025-04-08 13:28:17 INFO: Finished STEP 4360/50000, loss = 2.731332 (0.604 sec/batch), lr: 0.003000 2025-04-08 13:28:30 INFO: Finished STEP 4380/50000, loss = 2.199270 (0.480 sec/batch), lr: 0.003000 2025-04-08 13:28:43 INFO: Finished STEP 4400/50000, loss = 2.586877 (0.484 sec/batch), lr: 0.003000 2025-04-08 13:28:43 INFO: Evaluating on dev set... 2025-04-08 13:28:45 INFO: LAS MLAS BLEX 2025-04-08 13:28:45 INFO: 61.76 0.00 0.00 2025-04-08 13:28:45 INFO: step 4400: train_loss = 2.092441, dev_score = 0.6176 2025-04-08 13:28:45 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:28:46 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:28:46 INFO: new model checkpoint saved. 2025-04-08 13:28:59 INFO: Finished STEP 4420/50000, loss = 3.615049 (0.525 sec/batch), lr: 0.003000 2025-04-08 13:29:12 INFO: Finished STEP 4440/50000, loss = 3.296819 (0.523 sec/batch), lr: 0.003000 2025-04-08 13:29:24 INFO: Finished STEP 4460/50000, loss = 1.305707 (0.611 sec/batch), lr: 0.003000 2025-04-08 13:29:37 INFO: Finished STEP 4480/50000, loss = 2.342412 (0.510 sec/batch), lr: 0.003000 2025-04-08 13:29:50 INFO: Finished STEP 4500/50000, loss = 2.343141 (0.509 sec/batch), lr: 0.003000 2025-04-08 13:29:50 INFO: Evaluating on dev set... 2025-04-08 13:29:52 INFO: LAS MLAS BLEX 2025-04-08 13:29:52 INFO: 61.01 0.00 0.00 2025-04-08 13:29:52 INFO: step 4500: train_loss = 1.996532, dev_score = 0.6101 2025-04-08 13:29:52 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:29:52 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:29:52 INFO: new model checkpoint saved. 2025-04-08 13:30:06 INFO: Finished STEP 4520/50000, loss = 2.593366 (0.484 sec/batch), lr: 0.003000 2025-04-08 13:30:19 INFO: Finished STEP 4540/50000, loss = 0.654038 (0.618 sec/batch), lr: 0.003000 2025-04-08 13:30:32 INFO: Finished STEP 4560/50000, loss = 2.453485 (0.503 sec/batch), lr: 0.003000 2025-04-08 13:30:44 INFO: Finished STEP 4580/50000, loss = 1.998242 (0.500 sec/batch), lr: 0.003000 2025-04-08 13:30:57 INFO: Finished STEP 4600/50000, loss = 1.492977 (0.548 sec/batch), lr: 0.003000 2025-04-08 13:30:57 INFO: Evaluating on dev set... 2025-04-08 13:30:59 INFO: LAS MLAS BLEX 2025-04-08 13:30:59 INFO: 61.52 0.00 0.00 2025-04-08 13:30:59 INFO: step 4600: train_loss = 2.024841, dev_score = 0.6152 2025-04-08 13:30:59 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:30:59 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:30:59 INFO: new model checkpoint saved. 2025-04-08 13:31:12 INFO: Finished STEP 4620/50000, loss = 1.766703 (0.545 sec/batch), lr: 0.003000 2025-04-08 13:31:26 INFO: Finished STEP 4640/50000, loss = 1.019174 (0.720 sec/batch), lr: 0.003000 2025-04-08 13:31:38 INFO: Finished STEP 4660/50000, loss = 1.247105 (0.609 sec/batch), lr: 0.003000 2025-04-08 13:31:51 INFO: Finished STEP 4680/50000, loss = 1.598421 (0.569 sec/batch), lr: 0.003000 2025-04-08 13:32:04 INFO: Finished STEP 4700/50000, loss = 1.799268 (0.538 sec/batch), lr: 0.003000 2025-04-08 13:32:04 INFO: Evaluating on dev set... 2025-04-08 13:32:05 INFO: LAS MLAS BLEX 2025-04-08 13:32:05 INFO: 61.49 0.00 0.00 2025-04-08 13:32:05 INFO: step 4700: train_loss = 1.982152, dev_score = 0.6149 2025-04-08 13:32:05 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:32:06 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:32:06 INFO: new model checkpoint saved. 2025-04-08 13:32:20 INFO: Finished STEP 4720/50000, loss = 0.848733 (0.810 sec/batch), lr: 0.003000 2025-04-08 13:32:32 INFO: Finished STEP 4740/50000, loss = 2.304163 (0.486 sec/batch), lr: 0.003000 2025-04-08 13:32:45 INFO: Finished STEP 4760/50000, loss = 2.046792 (0.495 sec/batch), lr: 0.003000 2025-04-08 13:32:58 INFO: Finished STEP 4780/50000, loss = 1.957442 (0.520 sec/batch), lr: 0.003000 2025-04-08 13:33:12 INFO: Finished STEP 4800/50000, loss = 1.195087 (0.640 sec/batch), lr: 0.003000 2025-04-08 13:33:12 INFO: Evaluating on dev set... 2025-04-08 13:33:13 INFO: LAS MLAS BLEX 2025-04-08 13:33:13 INFO: 61.44 0.00 0.00 2025-04-08 13:33:13 INFO: step 4800: train_loss = 1.968694, dev_score = 0.6144 2025-04-08 13:33:13 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:33:14 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:33:14 INFO: new model checkpoint saved. 2025-04-08 13:33:27 INFO: Finished STEP 4820/50000, loss = 2.109784 (0.500 sec/batch), lr: 0.003000 2025-04-08 13:33:40 INFO: Finished STEP 4840/50000, loss = 2.421183 (0.509 sec/batch), lr: 0.003000 2025-04-08 13:33:52 INFO: Finished STEP 4860/50000, loss = 1.517776 (0.568 sec/batch), lr: 0.003000 2025-04-08 13:34:05 INFO: Finished STEP 4880/50000, loss = 1.548072 (0.537 sec/batch), lr: 0.003000 2025-04-08 13:34:18 INFO: Finished STEP 4900/50000, loss = 1.394600 (0.576 sec/batch), lr: 0.003000 2025-04-08 13:34:18 INFO: Evaluating on dev set... 2025-04-08 13:34:20 INFO: LAS MLAS BLEX 2025-04-08 13:34:20 INFO: 61.49 0.00 0.00 2025-04-08 13:34:20 INFO: step 4900: train_loss = 2.021885, dev_score = 0.6149 2025-04-08 13:34:20 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:34:20 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:34:20 INFO: new model checkpoint saved. 2025-04-08 13:34:33 INFO: Finished STEP 4920/50000, loss = 1.252762 (0.609 sec/batch), lr: 0.003000 2025-04-08 13:34:46 INFO: Finished STEP 4940/50000, loss = 1.572381 (0.584 sec/batch), lr: 0.003000 2025-04-08 13:34:59 INFO: Finished STEP 4960/50000, loss = 2.260174 (0.484 sec/batch), lr: 0.003000 2025-04-08 13:35:12 INFO: Finished STEP 4980/50000, loss = 2.992162 (0.566 sec/batch), lr: 0.003000 2025-04-08 13:35:25 INFO: Finished STEP 5000/50000, loss = 1.868307 (0.546 sec/batch), lr: 0.003000 2025-04-08 13:35:25 INFO: Evaluating on dev set... 2025-04-08 13:35:27 INFO: LAS MLAS BLEX 2025-04-08 13:35:27 INFO: 61.67 0.00 0.00 2025-04-08 13:35:27 INFO: step 5000: train_loss = 1.973462, dev_score = 0.6167 2025-04-08 13:35:27 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:35:27 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:35:27 INFO: new model checkpoint saved. 2025-04-08 13:35:40 INFO: Finished STEP 5020/50000, loss = 0.809553 (0.813 sec/batch), lr: 0.003000 2025-04-08 13:35:53 INFO: Finished STEP 5040/50000, loss = 1.082677 (0.679 sec/batch), lr: 0.003000 2025-04-08 13:36:06 INFO: Finished STEP 5060/50000, loss = 0.603558 (0.619 sec/batch), lr: 0.003000 2025-04-08 13:36:19 INFO: Finished STEP 5080/50000, loss = 5.230546 (0.589 sec/batch), lr: 0.003000 2025-04-08 13:36:32 INFO: Finished STEP 5100/50000, loss = 0.946424 (0.813 sec/batch), lr: 0.003000 2025-04-08 13:36:32 INFO: Evaluating on dev set... 2025-04-08 13:36:34 INFO: LAS MLAS BLEX 2025-04-08 13:36:34 INFO: 61.26 0.00 0.00 2025-04-08 13:36:34 INFO: step 5100: train_loss = 1.921387, dev_score = 0.6126 2025-04-08 13:36:34 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:36:35 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:36:35 INFO: new model checkpoint saved. 2025-04-08 13:36:47 INFO: Finished STEP 5120/50000, loss = 0.896935 (0.814 sec/batch), lr: 0.003000 2025-04-08 13:37:00 INFO: Finished STEP 5140/50000, loss = 1.976615 (0.501 sec/batch), lr: 0.003000 2025-04-08 13:37:13 INFO: Finished STEP 5160/50000, loss = 5.266327 (0.588 sec/batch), lr: 0.003000 2025-04-08 13:37:26 INFO: Finished STEP 5180/50000, loss = 1.412925 (0.575 sec/batch), lr: 0.003000 2025-04-08 13:37:39 INFO: Finished STEP 5200/50000, loss = 1.107314 (0.680 sec/batch), lr: 0.003000 2025-04-08 13:37:39 INFO: Evaluating on dev set... 2025-04-08 13:37:41 INFO: LAS MLAS BLEX 2025-04-08 13:37:41 INFO: 61.63 0.00 0.00 2025-04-08 13:37:41 INFO: step 5200: train_loss = 1.938719, dev_score = 0.6163 2025-04-08 13:37:41 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:37:42 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:37:42 INFO: new model checkpoint saved. 2025-04-08 13:37:54 INFO: Finished STEP 5220/50000, loss = 2.433164 (0.511 sec/batch), lr: 0.003000 2025-04-08 13:38:07 INFO: Finished STEP 5240/50000, loss = 1.273141 (0.577 sec/batch), lr: 0.003000 2025-04-08 13:38:20 INFO: Finished STEP 5260/50000, loss = 1.427732 (0.568 sec/batch), lr: 0.003000 2025-04-08 13:38:33 INFO: Finished STEP 5280/50000, loss = 1.685317 (0.506 sec/batch), lr: 0.003000 2025-04-08 13:38:46 INFO: Finished STEP 5300/50000, loss = 1.559889 (0.564 sec/batch), lr: 0.003000 2025-04-08 13:38:46 INFO: Evaluating on dev set... 2025-04-08 13:38:48 INFO: LAS MLAS BLEX 2025-04-08 13:38:48 INFO: 62.17 0.00 0.00 2025-04-08 13:38:48 INFO: step 5300: train_loss = 1.923150, dev_score = 0.6217 2025-04-08 13:38:48 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 13:38:48 INFO: new best model saved. 2025-04-08 13:38:48 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:38:49 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:38:49 INFO: new model checkpoint saved. 2025-04-08 13:39:01 INFO: Finished STEP 5320/50000, loss = 2.189765 (0.490 sec/batch), lr: 0.003000 2025-04-08 13:39:14 INFO: Finished STEP 5340/50000, loss = 1.927550 (0.501 sec/batch), lr: 0.003000 2025-04-08 13:39:27 INFO: Finished STEP 5360/50000, loss = 3.106446 (0.526 sec/batch), lr: 0.003000 2025-04-08 13:39:40 INFO: Finished STEP 5380/50000, loss = 1.995113 (0.479 sec/batch), lr: 0.003000 2025-04-08 13:39:53 INFO: Finished STEP 5400/50000, loss = 1.273235 (0.599 sec/batch), lr: 0.003000 2025-04-08 13:39:53 INFO: Evaluating on dev set... 2025-04-08 13:39:55 INFO: LAS MLAS BLEX 2025-04-08 13:39:55 INFO: 61.85 0.00 0.00 2025-04-08 13:39:55 INFO: step 5400: train_loss = 1.932986, dev_score = 0.6185 2025-04-08 13:39:55 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:39:55 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:39:55 INFO: new model checkpoint saved. 2025-04-08 13:40:08 INFO: Finished STEP 5420/50000, loss = 1.796276 (0.496 sec/batch), lr: 0.003000 2025-04-08 13:40:21 INFO: Finished STEP 5440/50000, loss = 1.973434 (0.497 sec/batch), lr: 0.003000 2025-04-08 13:40:34 INFO: Finished STEP 5460/50000, loss = 4.900478 (0.364 sec/batch), lr: 0.003000 2025-04-08 13:40:47 INFO: Finished STEP 5480/50000, loss = 1.295357 (0.591 sec/batch), lr: 0.003000 2025-04-08 13:41:00 INFO: Finished STEP 5500/50000, loss = 1.169220 (0.642 sec/batch), lr: 0.003000 2025-04-08 13:41:00 INFO: Evaluating on dev set... 2025-04-08 13:41:02 INFO: LAS MLAS BLEX 2025-04-08 13:41:02 INFO: 61.53 0.00 0.00 2025-04-08 13:41:02 INFO: step 5500: train_loss = 1.955204, dev_score = 0.6153 2025-04-08 13:41:02 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:41:03 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:41:03 INFO: new model checkpoint saved. 2025-04-08 13:41:16 INFO: Finished STEP 5520/50000, loss = 1.432669 (0.567 sec/batch), lr: 0.003000 2025-04-08 13:41:28 INFO: Finished STEP 5540/50000, loss = 4.762352 (0.367 sec/batch), lr: 0.003000 2025-04-08 13:41:42 INFO: Finished STEP 5560/50000, loss = 2.128884 (0.483 sec/batch), lr: 0.003000 2025-04-08 13:41:54 INFO: Finished STEP 5580/50000, loss = 1.483930 (0.568 sec/batch), lr: 0.003000 2025-04-08 13:42:07 INFO: Finished STEP 5600/50000, loss = 2.523759 (0.485 sec/batch), lr: 0.003000 2025-04-08 13:42:07 INFO: Evaluating on dev set... 2025-04-08 13:42:09 INFO: LAS MLAS BLEX 2025-04-08 13:42:09 INFO: 61.40 0.00 0.00 2025-04-08 13:42:09 INFO: step 5600: train_loss = 1.925072, dev_score = 0.6140 2025-04-08 13:42:09 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:42:09 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:42:09 INFO: new model checkpoint saved. 2025-04-08 13:42:23 INFO: Finished STEP 5620/50000, loss = 0.715229 (0.961 sec/batch), lr: 0.003000 2025-04-08 13:42:36 INFO: Finished STEP 5640/50000, loss = 1.625232 (0.513 sec/batch), lr: 0.003000 2025-04-08 13:42:48 INFO: Finished STEP 5660/50000, loss = 3.101979 (0.525 sec/batch), lr: 0.003000 2025-04-08 13:43:01 INFO: Finished STEP 5680/50000, loss = 2.241370 (0.485 sec/batch), lr: 0.003000 2025-04-08 13:43:14 INFO: Finished STEP 5700/50000, loss = 0.633499 (0.638 sec/batch), lr: 0.003000 2025-04-08 13:43:14 INFO: Evaluating on dev set... 2025-04-08 13:43:16 INFO: LAS MLAS BLEX 2025-04-08 13:43:16 INFO: 61.70 0.00 0.00 2025-04-08 13:43:16 INFO: step 5700: train_loss = 1.856325, dev_score = 0.6170 2025-04-08 13:43:16 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:43:17 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:43:17 INFO: new model checkpoint saved. 2025-04-08 13:43:30 INFO: Finished STEP 5720/50000, loss = 1.710941 (0.507 sec/batch), lr: 0.003000 2025-04-08 13:43:43 INFO: Finished STEP 5740/50000, loss = 1.406241 (0.584 sec/batch), lr: 0.003000 2025-04-08 13:43:56 INFO: Finished STEP 5760/50000, loss = 4.840482 (0.368 sec/batch), lr: 0.003000 2025-04-08 13:44:09 INFO: Finished STEP 5780/50000, loss = 1.119718 (0.626 sec/batch), lr: 0.003000 2025-04-08 13:44:21 INFO: Finished STEP 5800/50000, loss = 1.237176 (0.607 sec/batch), lr: 0.003000 2025-04-08 13:44:21 INFO: Evaluating on dev set... 2025-04-08 13:44:23 INFO: LAS MLAS BLEX 2025-04-08 13:44:23 INFO: 61.46 0.00 0.00 2025-04-08 13:44:23 INFO: step 5800: train_loss = 1.846373, dev_score = 0.6146 2025-04-08 13:44:23 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:44:24 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:44:24 INFO: new model checkpoint saved. 2025-04-08 13:44:36 INFO: Finished STEP 5820/50000, loss = 2.147384 (0.510 sec/batch), lr: 0.003000 2025-04-08 13:44:49 INFO: Finished STEP 5840/50000, loss = 1.418395 (0.551 sec/batch), lr: 0.003000 2025-04-08 13:45:02 INFO: Finished STEP 5860/50000, loss = 1.687887 (0.514 sec/batch), lr: 0.003000 2025-04-08 13:45:15 INFO: Finished STEP 5880/50000, loss = 1.752961 (0.510 sec/batch), lr: 0.003000 2025-04-08 13:45:28 INFO: Finished STEP 5900/50000, loss = 1.195682 (0.617 sec/batch), lr: 0.003000 2025-04-08 13:45:28 INFO: Evaluating on dev set... 2025-04-08 13:45:30 INFO: LAS MLAS BLEX 2025-04-08 13:45:30 INFO: 62.51 0.00 0.00 2025-04-08 13:45:30 INFO: step 5900: train_loss = 1.861884, dev_score = 0.6251 2025-04-08 13:45:30 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser.pt 2025-04-08 13:45:30 INFO: new best model saved. 2025-04-08 13:45:30 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:45:31 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:45:31 INFO: new model checkpoint saved. 2025-04-08 13:45:44 INFO: Finished STEP 5920/50000, loss = 3.188910 (0.524 sec/batch), lr: 0.003000 2025-04-08 13:45:57 INFO: Finished STEP 5940/50000, loss = 1.461580 (0.546 sec/batch), lr: 0.003000 2025-04-08 13:46:10 INFO: Finished STEP 5960/50000, loss = 1.683458 (0.547 sec/batch), lr: 0.003000 2025-04-08 13:46:23 INFO: Finished STEP 5980/50000, loss = 0.874793 (0.798 sec/batch), lr: 0.003000 2025-04-08 13:46:35 INFO: Finished STEP 6000/50000, loss = 1.792540 (0.504 sec/batch), lr: 0.003000 2025-04-08 13:46:35 INFO: Evaluating on dev set... 2025-04-08 13:46:37 INFO: LAS MLAS BLEX 2025-04-08 13:46:37 INFO: 61.50 0.00 0.00 2025-04-08 13:46:37 INFO: step 6000: train_loss = 1.896360, dev_score = 0.6150 2025-04-08 13:46:37 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:46:38 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:46:38 INFO: new model checkpoint saved. 2025-04-08 13:46:50 INFO: Finished STEP 6020/50000, loss = 2.381431 (0.504 sec/batch), lr: 0.003000 2025-04-08 13:47:04 INFO: Finished STEP 6040/50000, loss = 2.891144 (0.489 sec/batch), lr: 0.003000 2025-04-08 13:47:17 INFO: Finished STEP 6060/50000, loss = 0.700361 (0.624 sec/batch), lr: 0.003000 2025-04-08 13:47:29 INFO: Finished STEP 6080/50000, loss = 4.845391 (0.367 sec/batch), lr: 0.003000 2025-04-08 13:47:42 INFO: Finished STEP 6100/50000, loss = 1.441089 (0.574 sec/batch), lr: 0.003000 2025-04-08 13:47:42 INFO: Evaluating on dev set... 2025-04-08 13:47:44 INFO: LAS MLAS BLEX 2025-04-08 13:47:44 INFO: 62.09 0.00 0.00 2025-04-08 13:47:44 INFO: step 6100: train_loss = 1.881823, dev_score = 0.6209 2025-04-08 13:47:44 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:47:45 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:47:45 INFO: new model checkpoint saved. 2025-04-08 13:47:58 INFO: Finished STEP 6120/50000, loss = 1.571468 (0.542 sec/batch), lr: 0.003000 2025-04-08 13:48:10 INFO: Finished STEP 6140/50000, loss = 1.458384 (0.574 sec/batch), lr: 0.003000 2025-04-08 13:48:23 INFO: Finished STEP 6160/50000, loss = 1.783120 (0.509 sec/batch), lr: 0.003000 2025-04-08 13:48:36 INFO: Finished STEP 6180/50000, loss = 1.135499 (0.606 sec/batch), lr: 0.003000 2025-04-08 13:48:49 INFO: Finished STEP 6200/50000, loss = 1.985002 (0.490 sec/batch), lr: 0.003000 2025-04-08 13:48:49 INFO: Evaluating on dev set... 2025-04-08 13:48:50 INFO: LAS MLAS BLEX 2025-04-08 13:48:50 INFO: 62.04 0.00 0.00 2025-04-08 13:48:50 INFO: step 6200: train_loss = 1.886429, dev_score = 0.6204 2025-04-08 13:48:50 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:48:51 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:48:51 INFO: new model checkpoint saved. 2025-04-08 13:49:04 INFO: Finished STEP 6220/50000, loss = 1.763013 (0.498 sec/batch), lr: 0.003000 2025-04-08 13:49:18 INFO: Finished STEP 6240/50000, loss = 1.970363 (0.488 sec/batch), lr: 0.003000 2025-04-08 13:49:31 INFO: Finished STEP 6260/50000, loss = 1.518250 (0.552 sec/batch), lr: 0.003000 2025-04-08 13:49:43 INFO: Finished STEP 6280/50000, loss = 1.237253 (0.589 sec/batch), lr: 0.003000 2025-04-08 13:49:56 INFO: Finished STEP 6300/50000, loss = 1.651991 (0.491 sec/batch), lr: 0.003000 2025-04-08 13:49:56 INFO: Evaluating on dev set... 2025-04-08 13:49:58 INFO: LAS MLAS BLEX 2025-04-08 13:49:58 INFO: 62.26 0.00 0.00 2025-04-08 13:49:58 INFO: step 6300: train_loss = 1.792855, dev_score = 0.6226 2025-04-08 13:49:58 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:49:59 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:49:59 INFO: new model checkpoint saved. 2025-04-08 13:50:12 INFO: Finished STEP 6320/50000, loss = 1.068430 (0.675 sec/batch), lr: 0.003000 2025-04-08 13:50:24 INFO: Finished STEP 6340/50000, loss = 1.836678 (0.500 sec/batch), lr: 0.003000 2025-04-08 13:50:37 INFO: Finished STEP 6360/50000, loss = 0.986036 (0.719 sec/batch), lr: 0.003000 2025-04-08 13:50:50 INFO: Finished STEP 6380/50000, loss = 1.551772 (0.542 sec/batch), lr: 0.003000 2025-04-08 13:51:03 INFO: Finished STEP 6400/50000, loss = 2.760977 (0.565 sec/batch), lr: 0.003000 2025-04-08 13:51:03 INFO: Evaluating on dev set... 2025-04-08 13:51:05 INFO: LAS MLAS BLEX 2025-04-08 13:51:05 INFO: 61.57 0.00 0.00 2025-04-08 13:51:05 INFO: step 6400: train_loss = 1.800208, dev_score = 0.6157 2025-04-08 13:51:05 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:51:05 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:51:05 INFO: new model checkpoint saved. 2025-04-08 13:51:18 INFO: Finished STEP 6420/50000, loss = 2.623710 (0.483 sec/batch), lr: 0.003000 2025-04-08 13:51:31 INFO: Finished STEP 6440/50000, loss = 1.520910 (0.547 sec/batch), lr: 0.003000 2025-04-08 13:51:44 INFO: Finished STEP 6460/50000, loss = 1.476026 (0.570 sec/batch), lr: 0.003000 2025-04-08 13:51:57 INFO: Finished STEP 6480/50000, loss = 1.787203 (0.511 sec/batch), lr: 0.003000 2025-04-08 13:52:10 INFO: Finished STEP 6500/50000, loss = 1.375769 (0.552 sec/batch), lr: 0.003000 2025-04-08 13:52:10 INFO: Evaluating on dev set... 2025-04-08 13:52:12 INFO: LAS MLAS BLEX 2025-04-08 13:52:12 INFO: 60.96 0.00 0.00 2025-04-08 13:52:12 INFO: step 6500: train_loss = 1.843865, dev_score = 0.6096 2025-04-08 13:52:12 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:52:13 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:52:13 INFO: new model checkpoint saved. 2025-04-08 13:52:25 INFO: Finished STEP 6520/50000, loss = 2.425041 (0.488 sec/batch), lr: 0.003000 2025-04-08 13:52:38 INFO: Finished STEP 6540/50000, loss = 1.927385 (0.494 sec/batch), lr: 0.003000 2025-04-08 13:52:51 INFO: Finished STEP 6560/50000, loss = 3.019066 (0.524 sec/batch), lr: 0.003000 2025-04-08 13:53:04 INFO: Finished STEP 6580/50000, loss = 1.109605 (0.611 sec/batch), lr: 0.003000 2025-04-08 13:53:17 INFO: Finished STEP 6600/50000, loss = 0.553289 (0.621 sec/batch), lr: 0.003000 2025-04-08 13:53:17 INFO: Evaluating on dev set... 2025-04-08 13:53:19 INFO: LAS MLAS BLEX 2025-04-08 13:53:19 INFO: 61.02 0.00 0.00 2025-04-08 13:53:19 INFO: step 6600: train_loss = 1.840983, dev_score = 0.6102 2025-04-08 13:53:19 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:53:19 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:53:19 INFO: new model checkpoint saved. 2025-04-08 13:53:33 INFO: Finished STEP 6620/50000, loss = 1.595263 (0.506 sec/batch), lr: 0.003000 2025-04-08 13:53:45 INFO: Finished STEP 6640/50000, loss = 1.947466 (0.490 sec/batch), lr: 0.003000 2025-04-08 13:53:58 INFO: Finished STEP 6660/50000, loss = 3.379229 (0.524 sec/batch), lr: 0.003000 2025-04-08 13:54:11 INFO: Finished STEP 6680/50000, loss = 2.645754 (0.564 sec/batch), lr: 0.003000 2025-04-08 13:54:24 INFO: Finished STEP 6700/50000, loss = 0.765848 (0.956 sec/batch), lr: 0.003000 2025-04-08 13:54:24 INFO: Evaluating on dev set... 2025-04-08 13:54:26 INFO: LAS MLAS BLEX 2025-04-08 13:54:26 INFO: 61.73 0.00 0.00 2025-04-08 13:54:26 INFO: step 6700: train_loss = 1.805882, dev_score = 0.6173 2025-04-08 13:54:26 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:54:26 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:54:26 INFO: new model checkpoint saved. 2025-04-08 13:54:39 INFO: Finished STEP 6720/50000, loss = 2.045537 (0.490 sec/batch), lr: 0.003000 2025-04-08 13:54:52 INFO: Finished STEP 6740/50000, loss = 1.147556 (0.640 sec/batch), lr: 0.003000 2025-04-08 13:55:05 INFO: Finished STEP 6760/50000, loss = 1.021457 (0.733 sec/batch), lr: 0.003000 2025-04-08 13:55:18 INFO: Finished STEP 6780/50000, loss = 0.585405 (0.619 sec/batch), lr: 0.003000 2025-04-08 13:55:31 INFO: Finished STEP 6800/50000, loss = 0.926011 (0.799 sec/batch), lr: 0.003000 2025-04-08 13:55:31 INFO: Evaluating on dev set... 2025-04-08 13:55:32 INFO: LAS MLAS BLEX 2025-04-08 13:55:32 INFO: 62.47 0.00 0.00 2025-04-08 13:55:32 INFO: step 6800: train_loss = 1.896006, dev_score = 0.6247 2025-04-08 13:55:32 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:55:33 INFO: Model saved to saved_models/depparse/la_giuseppe_transformer_parser_checkpoint.pt 2025-04-08 13:55:33 INFO: new model checkpoint saved. 2025-04-08 13:55:46 INFO: Finished STEP 6820/50000, loss = 1.319737 (0.608 sec/batch), lr: 0.003000 2025-04-08 13:55:59 INFO: Finished STEP 6840/50000, loss = 0.905927 (0.769 sec/batch), lr: 0.003000 2025-04-08 13:56:12 INFO: Finished STEP 6860/50000, loss = 1.316787 (0.584 sec/batch), lr: 0.003000 2025-04-08 13:56:25 INFO: Finished STEP 6880/50000, loss = 1.661569 (0.506 sec/batch), lr: 0.003000 2025-04-08 13:56:38 INFO: Finished STEP 6900/50000, loss = 0.823536 (0.814 sec/batch), lr: 0.003000 2025-04-08 13:56:38 INFO: Evaluating on dev set... 2025-04-08 13:56:40 INFO: LAS MLAS BLEX 2025-04-08 13:56:40 INFO: 61.96 0.00 0.00 2025-04-08 13:56:40 INFO: step 6900: train_loss = 1.866068, dev_score = 0.6196 2025-04-08 13:56:40 INFO: scheduler bert_scheduler learning rate: [0.003] 2025-04-08 13:56:40 INFO: Training ended with 6900 steps. 2025-04-08 13:56:40 INFO: Best dev F1 = 62.51, at iteration = 5900