[hops] 2024-09-24 14:34:15.377 | INFO | Initializing a parser from /workspace/configs/exp_camembertv2/camembertav2_base_p2_17k_last_layer.yaml [hops] 2024-09-24 14:34:15.407 | INFO | Generating a FastText model from the treebank [hops] 2024-09-24 14:34:15.414 | INFO | Training fasttext model [hops] 2024-09-24 14:34:22.479 | INFO | Start training on cuda:1 [hops] 2024-09-24 14:34:22.483 | WARNING | You're using a RobertaTokenizerFast tokenizer. Please note that with a fast tokenizer, using the `__call__` method is faster than using a method to encode the text followed by a call to the `pad` method to get a padded encoding. [hops] 2024-09-24 14:34:38.000 | INFO | Epoch 0: train loss 3.1853 dev loss 2.6151 dev tag acc 15.14% dev head acc 29.90% dev deprel acc 43.21% [hops] 2024-09-24 14:34:38.002 | INFO | New best model: head accuracy 29.90% > 0.00% [hops] 2024-09-24 14:34:55.908 | INFO | Epoch 1: train loss 2.2397 dev loss 1.8289 dev tag acc 29.62% dev head acc 50.56% dev deprel acc 69.99% [hops] 2024-09-24 14:34:55.909 | INFO | New best model: head accuracy 50.56% > 29.90% [hops] 2024-09-24 14:35:13.698 | INFO | Epoch 2: train loss 1.7309 dev loss 1.4351 dev tag acc 50.91% dev head acc 60.57% dev deprel acc 77.66% [hops] 2024-09-24 14:35:13.699 | INFO | New best model: head accuracy 60.57% > 50.56% [hops] 2024-09-24 14:35:31.267 | INFO | Epoch 3: train loss 1.3681 dev loss 1.1524 dev tag acc 69.08% dev head acc 66.75% dev deprel acc 82.88% [hops] 2024-09-24 14:35:31.268 | INFO | New best model: head accuracy 66.75% > 60.57% [hops] 2024-09-24 14:35:47.681 | INFO | Epoch 4: train loss 1.1018 dev loss 0.9730 dev tag acc 74.09% dev head acc 72.10% dev deprel acc 83.64% [hops] 2024-09-24 14:35:47.682 | INFO | New best model: head accuracy 72.10% > 66.75% [hops] 2024-09-24 14:36:05.527 | INFO | Epoch 5: train loss 0.9098 dev loss 0.8205 dev tag acc 78.00% dev head acc 75.25% dev deprel acc 85.24% [hops] 2024-09-24 14:36:05.528 | INFO | New best model: head accuracy 75.25% > 72.10% [hops] 2024-09-24 14:36:22.354 | INFO | Epoch 6: train loss 0.7515 dev loss 0.7259 dev tag acc 83.95% dev head acc 77.12% dev deprel acc 86.37% [hops] 2024-09-24 14:36:22.355 | INFO | New best model: head accuracy 77.12% > 75.25% [hops] 2024-09-24 14:36:40.825 | INFO | Epoch 7: train loss 0.6351 dev loss 0.6541 dev tag acc 87.20% dev head acc 79.35% dev deprel acc 86.70% [hops] 2024-09-24 14:36:40.825 | INFO | New best model: head accuracy 79.35% > 77.12% [hops] 2024-09-24 14:36:58.663 | INFO | Epoch 8: train loss 0.5378 dev loss 0.6082 dev tag acc 89.21% dev head acc 81.08% dev deprel acc 87.17% [hops] 2024-09-24 14:36:58.664 | INFO | New best model: head accuracy 81.08% > 79.35% [hops] 2024-09-24 14:37:16.472 | INFO | Epoch 9: train loss 0.4691 dev loss 0.5612 dev tag acc 91.07% dev head acc 80.99% dev deprel acc 88.50% [hops] 2024-09-24 14:37:31.564 | INFO | Epoch 10: train loss 0.4153 dev loss 0.5578 dev tag acc 92.18% dev head acc 82.23% dev deprel acc 88.90% [hops] 2024-09-24 14:37:31.565 | INFO | New best model: head accuracy 82.23% > 81.08% [hops] 2024-09-24 14:37:48.571 | INFO | Epoch 11: train loss 0.3739 dev loss 0.5293 dev tag acc 92.55% dev head acc 83.24% dev deprel acc 89.12% [hops] 2024-09-24 14:37:48.572 | INFO | New best model: head accuracy 83.24% > 82.23% [hops] 2024-09-24 14:38:06.670 | INFO | Epoch 12: train loss 0.3366 dev loss 0.5339 dev tag acc 93.24% dev head acc 83.53% dev deprel acc 89.48% [hops] 2024-09-24 14:38:06.671 | INFO | New best model: head accuracy 83.53% > 83.24% [hops] 2024-09-24 14:38:24.365 | INFO | Epoch 13: train loss 0.3117 dev loss 0.5259 dev tag acc 93.54% dev head acc 83.93% dev deprel acc 89.86% [hops] 2024-09-24 14:38:24.366 | INFO | New best model: head accuracy 83.93% > 83.53% [hops] 2024-09-24 14:38:41.914 | INFO | Epoch 14: train loss 0.2863 dev loss 0.5402 dev tag acc 93.51% dev head acc 83.92% dev deprel acc 89.87% [hops] 2024-09-24 14:38:57.081 | INFO | Epoch 15: train loss 0.2669 dev loss 0.5419 dev tag acc 93.89% dev head acc 84.29% dev deprel acc 89.89% [hops] 2024-09-24 14:38:57.082 | INFO | New best model: head accuracy 84.29% > 83.93% [hops] 2024-09-24 14:39:14.548 | INFO | Epoch 16: train loss 0.2440 dev loss 0.5344 dev tag acc 94.02% dev head acc 84.68% dev deprel acc 90.56% [hops] 2024-09-24 14:39:14.549 | INFO | New best model: head accuracy 84.68% > 84.29% [hops] 2024-09-24 14:39:32.313 | INFO | Epoch 17: train loss 0.2342 dev loss 0.5615 dev tag acc 93.97% dev head acc 84.92% dev deprel acc 90.37% [hops] 2024-09-24 14:39:32.314 | INFO | New best model: head accuracy 84.92% > 84.68% [hops] 2024-09-24 14:39:49.332 | INFO | Epoch 18: train loss 0.2163 dev loss 0.5847 dev tag acc 94.21% dev head acc 84.47% dev deprel acc 90.54% [hops] 2024-09-24 14:40:04.450 | INFO | Epoch 19: train loss 0.2017 dev loss 0.5894 dev tag acc 94.22% dev head acc 85.00% dev deprel acc 90.76% [hops] 2024-09-24 14:40:04.451 | INFO | New best model: head accuracy 85.00% > 84.92% [hops] 2024-09-24 14:40:22.784 | INFO | Epoch 20: train loss 0.1901 dev loss 0.5888 dev tag acc 94.34% dev head acc 85.31% dev deprel acc 90.99% [hops] 2024-09-24 14:40:22.784 | INFO | New best model: head accuracy 85.31% > 85.00% [hops] 2024-09-24 14:40:40.285 | INFO | Epoch 21: train loss 0.1786 dev loss 0.6046 dev tag acc 94.35% dev head acc 85.54% dev deprel acc 90.91% [hops] 2024-09-24 14:40:40.286 | INFO | New best model: head accuracy 85.54% > 85.31% [hops] 2024-09-24 14:40:58.101 | INFO | Epoch 22: train loss 0.1680 dev loss 0.6209 dev tag acc 94.51% dev head acc 85.56% dev deprel acc 90.92% [hops] 2024-09-24 14:40:58.102 | INFO | New best model: head accuracy 85.56% > 85.54% [hops] 2024-09-24 14:41:16.021 | INFO | Epoch 23: train loss 0.1592 dev loss 0.6549 dev tag acc 94.57% dev head acc 85.32% dev deprel acc 91.15% [hops] 2024-09-24 14:41:31.264 | INFO | Epoch 24: train loss 0.1496 dev loss 0.6540 dev tag acc 94.63% dev head acc 85.78% dev deprel acc 91.18% [hops] 2024-09-24 14:41:31.265 | INFO | New best model: head accuracy 85.78% > 85.56% [hops] 2024-09-24 14:41:48.616 | INFO | Epoch 25: train loss 0.1423 dev loss 0.6558 dev tag acc 94.63% dev head acc 85.49% dev deprel acc 91.17% [hops] 2024-09-24 14:42:04.233 | INFO | Epoch 26: train loss 0.1339 dev loss 0.6730 dev tag acc 94.75% dev head acc 86.01% dev deprel acc 91.23% [hops] 2024-09-24 14:42:04.234 | INFO | New best model: head accuracy 86.01% > 85.78% [hops] 2024-09-24 14:42:22.418 | INFO | Epoch 27: train loss 0.1301 dev loss 0.7089 dev tag acc 94.78% dev head acc 85.80% dev deprel acc 91.28% [hops] 2024-09-24 14:42:37.688 | INFO | Epoch 28: train loss 0.1210 dev loss 0.7102 dev tag acc 94.85% dev head acc 86.12% dev deprel acc 91.34% [hops] 2024-09-24 14:42:37.689 | INFO | New best model: head accuracy 86.12% > 86.01% [hops] 2024-09-24 14:42:54.569 | INFO | Epoch 29: train loss 0.1140 dev loss 0.7184 dev tag acc 94.85% dev head acc 85.87% dev deprel acc 91.29% [hops] 2024-09-24 14:43:10.080 | INFO | Epoch 30: train loss 0.1121 dev loss 0.7474 dev tag acc 94.74% dev head acc 85.97% dev deprel acc 91.67% [hops] 2024-09-24 14:43:25.793 | INFO | Epoch 31: train loss 0.1082 dev loss 0.7457 dev tag acc 94.88% dev head acc 85.99% dev deprel acc 91.58% [hops] 2024-09-24 14:43:40.730 | INFO | Epoch 32: train loss 0.1038 dev loss 0.7704 dev tag acc 94.81% dev head acc 86.14% dev deprel acc 91.37% [hops] 2024-09-24 14:43:40.731 | INFO | New best model: head accuracy 86.14% > 86.12% [hops] 2024-09-24 14:43:58.212 | INFO | Epoch 33: train loss 0.0975 dev loss 0.7758 dev tag acc 94.87% dev head acc 86.52% dev deprel acc 91.39% [hops] 2024-09-24 14:43:58.213 | INFO | New best model: head accuracy 86.52% > 86.14% [hops] 2024-09-24 14:44:16.413 | INFO | Epoch 34: train loss 0.0951 dev loss 0.7914 dev tag acc 94.84% dev head acc 86.33% dev deprel acc 91.41% [hops] 2024-09-24 14:44:31.475 | INFO | Epoch 35: train loss 0.0904 dev loss 0.8350 dev tag acc 94.91% dev head acc 86.53% dev deprel acc 91.53% [hops] 2024-09-24 14:44:31.476 | INFO | New best model: head accuracy 86.53% > 86.52% [hops] 2024-09-24 14:44:48.600 | INFO | Epoch 36: train loss 0.0855 dev loss 0.8383 dev tag acc 94.94% dev head acc 86.34% dev deprel acc 91.45% [hops] 2024-09-24 14:45:04.629 | INFO | Epoch 37: train loss 0.0852 dev loss 0.8443 dev tag acc 94.91% dev head acc 86.42% dev deprel acc 91.62% [hops] 2024-09-24 14:45:19.577 | INFO | Epoch 38: train loss 0.0821 dev loss 0.8490 dev tag acc 94.98% dev head acc 86.45% dev deprel acc 91.47% [hops] 2024-09-24 14:45:34.440 | INFO | Epoch 39: train loss 0.0788 dev loss 0.8424 dev tag acc 94.92% dev head acc 86.29% dev deprel acc 91.43% [hops] 2024-09-24 14:45:48.748 | INFO | Epoch 40: train loss 0.0768 dev loss 0.8570 dev tag acc 95.08% dev head acc 86.55% dev deprel acc 91.52% [hops] 2024-09-24 14:45:48.750 | INFO | New best model: head accuracy 86.55% > 86.53% [hops] 2024-09-24 14:46:06.299 | INFO | Epoch 41: train loss 0.0740 dev loss 0.8655 dev tag acc 95.11% dev head acc 86.34% dev deprel acc 91.52% [hops] 2024-09-24 14:46:22.078 | INFO | Epoch 42: train loss 0.0709 dev loss 0.8882 dev tag acc 95.04% dev head acc 86.49% dev deprel acc 91.47% [hops] 2024-09-24 14:46:37.401 | INFO | Epoch 43: train loss 0.0685 dev loss 0.8956 dev tag acc 95.09% dev head acc 86.32% dev deprel acc 91.63% [hops] 2024-09-24 14:46:52.792 | INFO | Epoch 44: train loss 0.0673 dev loss 0.9303 dev tag acc 95.06% dev head acc 86.42% dev deprel acc 91.53% [hops] 2024-09-24 14:47:08.151 | INFO | Epoch 45: train loss 0.0652 dev loss 0.9314 dev tag acc 95.10% dev head acc 86.33% dev deprel acc 91.47% [hops] 2024-09-24 14:47:23.787 | INFO | Epoch 46: train loss 0.0623 dev loss 0.9163 dev tag acc 95.10% dev head acc 86.43% dev deprel acc 91.48% [hops] 2024-09-24 14:47:39.059 | INFO | Epoch 47: train loss 0.0599 dev loss 0.9738 dev tag acc 95.04% dev head acc 86.47% dev deprel acc 91.62% [hops] 2024-09-24 14:47:54.002 | INFO | Epoch 48: train loss 0.0587 dev loss 0.9811 dev tag acc 95.12% dev head acc 86.31% dev deprel acc 91.50% [hops] 2024-09-24 14:48:08.900 | INFO | Epoch 49: train loss 0.0564 dev loss 0.9841 dev tag acc 95.02% dev head acc 86.55% dev deprel acc 91.57% [hops] 2024-09-24 14:48:08.902 | INFO | New best model: head accuracy 86.55% > 86.55% [hops] 2024-09-24 14:48:26.237 | INFO | Epoch 50: train loss 0.0543 dev loss 0.9959 dev tag acc 95.05% dev head acc 86.51% dev deprel acc 91.54% [hops] 2024-09-24 14:48:41.164 | INFO | Epoch 51: train loss 0.0539 dev loss 0.9901 dev tag acc 95.10% dev head acc 86.47% dev deprel acc 91.64% [hops] 2024-09-24 14:48:57.119 | INFO | Epoch 52: train loss 0.0535 dev loss 1.0037 dev tag acc 95.09% dev head acc 86.47% dev deprel acc 91.67% [hops] 2024-09-24 14:49:11.881 | INFO | Epoch 53: train loss 0.0515 dev loss 1.0083 dev tag acc 95.17% dev head acc 86.53% dev deprel acc 91.63% [hops] 2024-09-24 14:49:27.377 | INFO | Epoch 54: train loss 0.0495 dev loss 1.0323 dev tag acc 95.18% dev head acc 86.59% dev deprel acc 91.55% [hops] 2024-09-24 14:49:27.378 | INFO | New best model: head accuracy 86.59% > 86.55% [hops] 2024-09-24 14:49:44.663 | INFO | Epoch 55: train loss 0.0494 dev loss 1.0093 dev tag acc 95.16% dev head acc 86.47% dev deprel acc 91.53% [hops] 2024-09-24 14:49:59.239 | INFO | Epoch 56: train loss 0.0489 dev loss 1.0157 dev tag acc 95.13% dev head acc 86.55% dev deprel acc 91.57% [hops] 2024-09-24 14:50:13.539 | INFO | Epoch 57: train loss 0.0475 dev loss 1.0208 dev tag acc 95.20% dev head acc 86.52% dev deprel acc 91.57% [hops] 2024-09-24 14:50:28.650 | INFO | Epoch 58: train loss 0.0465 dev loss 1.0348 dev tag acc 95.16% dev head acc 86.50% dev deprel acc 91.65% [hops] 2024-09-24 14:50:44.170 | INFO | Epoch 59: train loss 0.0459 dev loss 1.0435 dev tag acc 95.20% dev head acc 86.56% dev deprel acc 91.66% [hops] 2024-09-24 14:50:59.437 | INFO | Epoch 60: train loss 0.0445 dev loss 1.0493 dev tag acc 95.21% dev head acc 86.59% dev deprel acc 91.74% [hops] 2024-09-24 14:51:13.819 | INFO | Epoch 61: train loss 0.0447 dev loss 1.0545 dev tag acc 95.21% dev head acc 86.69% dev deprel acc 91.68% [hops] 2024-09-24 14:51:13.820 | INFO | New best model: head accuracy 86.69% > 86.59% [hops] 2024-09-24 14:51:31.044 | INFO | Epoch 62: train loss 0.0437 dev loss 1.0525 dev tag acc 95.21% dev head acc 86.64% dev deprel acc 91.65% [hops] 2024-09-24 14:51:46.773 | INFO | Epoch 63: train loss 0.0427 dev loss 1.0558 dev tag acc 95.21% dev head acc 86.60% dev deprel acc 91.66% [hops] 2024-09-24 14:51:52.241 | WARNING | You're using a RobertaTokenizerFast tokenizer. Please note that with a fast tokenizer, using the `__call__` method is faster than using a method to encode the text followed by a call to the `pad` method to get a padded encoding. [hops] 2024-09-24 14:51:59.722 | WARNING | You're using a RobertaTokenizerFast tokenizer. Please note that with a fast tokenizer, using the `__call__` method is faster than using a method to encode the text followed by a call to the `pad` method to get a padded encoding. [hops] 2024-09-24 14:52:02.962 | INFO | Metrics for FSMB-camembertav2_base_p2_17k_last_layer+rand_seed=123 ─────────────────────────────── Split UPOS UAS LAS ─────────────────────────────── Dev 95.12 86.90 81.38 Test 95.09 86.76 81.67 ───────────────────────────────