[hops] 2024-09-24 16:01:20.681 | INFO | Initializing a parser from /workspace/configs/exp_camembertv2/camembertav2_base_p2_17k_last_layer.yaml [hops] 2024-09-24 16:01:20.730 | INFO | Generating a FastText model from the treebank [hops] 2024-09-24 16:01:20.745 | INFO | Training fasttext model [hops] 2024-09-24 16:01:28.399 | INFO | Start training on cuda:0 [hops] 2024-09-24 16:01:28.403 | 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 16:01:47.740 | INFO | Epoch 0: train loss 2.7794 dev loss 1.9428 dev tag acc 49.62% dev head acc 27.95% dev deprel acc 53.97% [hops] 2024-09-24 16:01:47.741 | INFO | New best model: head accuracy 27.95% > 0.00% [hops] 2024-09-24 16:02:09.374 | INFO | Epoch 1: train loss 1.5546 dev loss 1.0005 dev tag acc 72.25% dev head acc 61.63% dev deprel acc 80.28% [hops] 2024-09-24 16:02:09.375 | INFO | New best model: head accuracy 61.63% > 27.95% [hops] 2024-09-24 16:02:31.031 | INFO | Epoch 2: train loss 0.9383 dev loss 0.6218 dev tag acc 80.12% dev head acc 79.32% dev deprel acc 87.02% [hops] 2024-09-24 16:02:31.032 | INFO | New best model: head accuracy 79.32% > 61.63% [hops] 2024-09-24 16:02:51.968 | INFO | Epoch 3: train loss 0.6428 dev loss 0.4519 dev tag acc 87.60% dev head acc 84.25% dev deprel acc 90.09% [hops] 2024-09-24 16:02:51.969 | INFO | New best model: head accuracy 84.25% > 79.32% [hops] 2024-09-24 16:03:13.587 | INFO | Epoch 4: train loss 0.4912 dev loss 0.3727 dev tag acc 91.58% dev head acc 85.25% dev deprel acc 92.18% [hops] 2024-09-24 16:03:13.588 | INFO | New best model: head accuracy 85.25% > 84.25% [hops] 2024-09-24 16:03:34.880 | INFO | Epoch 5: train loss 0.3884 dev loss 0.3005 dev tag acc 95.05% dev head acc 88.23% dev deprel acc 93.64% [hops] 2024-09-24 16:03:34.881 | INFO | New best model: head accuracy 88.23% > 85.25% [hops] 2024-09-24 16:03:56.210 | INFO | Epoch 6: train loss 0.3135 dev loss 0.2582 dev tag acc 96.33% dev head acc 90.19% dev deprel acc 94.54% [hops] 2024-09-24 16:03:56.211 | INFO | New best model: head accuracy 90.19% > 88.23% [hops] 2024-09-24 16:04:18.250 | INFO | Epoch 7: train loss 0.2602 dev loss 0.2364 dev tag acc 96.99% dev head acc 90.92% dev deprel acc 95.35% [hops] 2024-09-24 16:04:18.251 | INFO | New best model: head accuracy 90.92% > 90.19% [hops] 2024-09-24 16:04:39.889 | INFO | Epoch 8: train loss 0.2206 dev loss 0.2207 dev tag acc 97.62% dev head acc 91.55% dev deprel acc 95.82% [hops] 2024-09-24 16:04:39.890 | INFO | New best model: head accuracy 91.55% > 90.92% [hops] 2024-09-24 16:05:00.941 | INFO | Epoch 9: train loss 0.1885 dev loss 0.2092 dev tag acc 97.86% dev head acc 92.34% dev deprel acc 96.26% [hops] 2024-09-24 16:05:00.942 | INFO | New best model: head accuracy 92.34% > 91.55% [hops] 2024-09-24 16:05:22.018 | INFO | Epoch 10: train loss 0.1633 dev loss 0.1818 dev tag acc 98.23% dev head acc 92.87% dev deprel acc 96.84% [hops] 2024-09-24 16:05:22.019 | INFO | New best model: head accuracy 92.87% > 92.34% [hops] 2024-09-24 16:05:42.839 | INFO | Epoch 11: train loss 0.1444 dev loss 0.1800 dev tag acc 98.40% dev head acc 93.47% dev deprel acc 96.77% [hops] 2024-09-24 16:05:42.840 | INFO | New best model: head accuracy 93.47% > 92.87% [hops] 2024-09-24 16:06:04.127 | INFO | Epoch 12: train loss 0.1289 dev loss 0.1718 dev tag acc 98.58% dev head acc 93.67% dev deprel acc 97.08% [hops] 2024-09-24 16:06:04.128 | INFO | New best model: head accuracy 93.67% > 93.47% [hops] 2024-09-24 16:06:26.120 | INFO | Epoch 13: train loss 0.1136 dev loss 0.1875 dev tag acc 98.59% dev head acc 93.56% dev deprel acc 97.00% [hops] 2024-09-24 16:06:45.617 | INFO | Epoch 14: train loss 0.1033 dev loss 0.1923 dev tag acc 98.85% dev head acc 93.73% dev deprel acc 97.03% [hops] 2024-09-24 16:06:45.618 | INFO | New best model: head accuracy 93.73% > 93.67% [hops] 2024-09-24 16:07:07.243 | INFO | Epoch 15: train loss 0.0938 dev loss 0.1859 dev tag acc 98.89% dev head acc 94.23% dev deprel acc 97.14% [hops] 2024-09-24 16:07:07.244 | INFO | New best model: head accuracy 94.23% > 93.73% [hops] 2024-09-24 16:07:29.285 | INFO | Epoch 16: train loss 0.0855 dev loss 0.1829 dev tag acc 98.89% dev head acc 94.30% dev deprel acc 97.31% [hops] 2024-09-24 16:07:29.287 | INFO | New best model: head accuracy 94.30% > 94.23% [hops] 2024-09-24 16:07:50.544 | INFO | Epoch 17: train loss 0.0789 dev loss 0.1872 dev tag acc 98.91% dev head acc 94.57% dev deprel acc 97.40% [hops] 2024-09-24 16:07:50.545 | INFO | New best model: head accuracy 94.57% > 94.30% [hops] 2024-09-24 16:08:12.269 | INFO | Epoch 18: train loss 0.0730 dev loss 0.1901 dev tag acc 98.95% dev head acc 94.59% dev deprel acc 97.29% [hops] 2024-09-24 16:08:12.270 | INFO | New best model: head accuracy 94.59% > 94.57% [hops] 2024-09-24 16:08:33.803 | INFO | Epoch 19: train loss 0.0665 dev loss 0.1852 dev tag acc 98.91% dev head acc 94.72% dev deprel acc 97.43% [hops] 2024-09-24 16:08:33.804 | INFO | New best model: head accuracy 94.72% > 94.59% [hops] 2024-09-24 16:08:55.220 | INFO | Epoch 20: train loss 0.0620 dev loss 0.2002 dev tag acc 98.96% dev head acc 95.00% dev deprel acc 97.45% [hops] 2024-09-24 16:08:55.221 | INFO | New best model: head accuracy 95.00% > 94.72% [hops] 2024-09-24 16:09:16.939 | INFO | Epoch 21: train loss 0.0574 dev loss 0.2063 dev tag acc 99.04% dev head acc 94.79% dev deprel acc 97.56% [hops] 2024-09-24 16:09:36.037 | INFO | Epoch 22: train loss 0.0534 dev loss 0.2002 dev tag acc 99.06% dev head acc 95.04% dev deprel acc 97.55% [hops] 2024-09-24 16:09:36.038 | INFO | New best model: head accuracy 95.04% > 95.00% [hops] 2024-09-24 16:09:57.103 | INFO | Epoch 23: train loss 0.0503 dev loss 0.2077 dev tag acc 99.06% dev head acc 95.07% dev deprel acc 97.55% [hops] 2024-09-24 16:09:57.103 | INFO | New best model: head accuracy 95.07% > 95.04% [hops] 2024-09-24 16:10:18.368 | INFO | Epoch 24: train loss 0.0469 dev loss 0.2004 dev tag acc 99.06% dev head acc 95.43% dev deprel acc 97.68% [hops] 2024-09-24 16:10:18.369 | INFO | New best model: head accuracy 95.43% > 95.07% [hops] 2024-09-24 16:10:40.222 | INFO | Epoch 25: train loss 0.0432 dev loss 0.2043 dev tag acc 99.02% dev head acc 95.30% dev deprel acc 97.67% [hops] 2024-09-24 16:10:59.739 | INFO | Epoch 26: train loss 0.0416 dev loss 0.2225 dev tag acc 99.06% dev head acc 95.08% dev deprel acc 97.48% [hops] 2024-09-24 16:11:18.934 | INFO | Epoch 27: train loss 0.0375 dev loss 0.2118 dev tag acc 99.08% dev head acc 95.39% dev deprel acc 97.62% [hops] 2024-09-24 16:11:37.455 | INFO | Epoch 28: train loss 0.0369 dev loss 0.2139 dev tag acc 99.08% dev head acc 95.28% dev deprel acc 97.76% [hops] 2024-09-24 16:11:57.066 | INFO | Epoch 29: train loss 0.0351 dev loss 0.2086 dev tag acc 99.12% dev head acc 95.46% dev deprel acc 97.72% [hops] 2024-09-24 16:11:57.067 | INFO | New best model: head accuracy 95.46% > 95.43% [hops] 2024-09-24 16:12:18.271 | INFO | Epoch 30: train loss 0.0329 dev loss 0.2321 dev tag acc 99.06% dev head acc 95.49% dev deprel acc 97.58% [hops] 2024-09-24 16:12:18.272 | INFO | New best model: head accuracy 95.49% > 95.46% [hops] 2024-09-24 16:12:40.183 | INFO | Epoch 31: train loss 0.0320 dev loss 0.2237 dev tag acc 99.14% dev head acc 95.79% dev deprel acc 97.80% [hops] 2024-09-24 16:12:40.184 | INFO | New best model: head accuracy 95.79% > 95.49% [hops] 2024-09-24 16:13:01.544 | INFO | Epoch 32: train loss 0.0291 dev loss 0.2373 dev tag acc 99.10% dev head acc 95.38% dev deprel acc 97.79% [hops] 2024-09-24 16:13:20.203 | INFO | Epoch 33: train loss 0.0288 dev loss 0.2393 dev tag acc 99.13% dev head acc 95.61% dev deprel acc 97.77% [hops] 2024-09-24 16:13:38.371 | INFO | Epoch 34: train loss 0.0261 dev loss 0.2499 dev tag acc 99.08% dev head acc 95.62% dev deprel acc 97.76% [hops] 2024-09-24 16:13:57.799 | INFO | Epoch 35: train loss 0.0254 dev loss 0.2435 dev tag acc 99.11% dev head acc 95.74% dev deprel acc 97.81% [hops] 2024-09-24 16:14:15.970 | INFO | Epoch 36: train loss 0.0235 dev loss 0.2542 dev tag acc 99.15% dev head acc 95.53% dev deprel acc 97.85% [hops] 2024-09-24 16:14:34.739 | INFO | Epoch 37: train loss 0.0226 dev loss 0.2540 dev tag acc 99.07% dev head acc 95.39% dev deprel acc 97.71% [hops] 2024-09-24 16:14:53.771 | INFO | Epoch 38: train loss 0.0218 dev loss 0.2529 dev tag acc 99.09% dev head acc 95.24% dev deprel acc 97.74% [hops] 2024-09-24 16:15:13.379 | INFO | Epoch 39: train loss 0.0206 dev loss 0.2571 dev tag acc 99.09% dev head acc 95.57% dev deprel acc 97.83% [hops] 2024-09-24 16:15:32.562 | INFO | Epoch 40: train loss 0.0195 dev loss 0.2649 dev tag acc 99.19% dev head acc 95.64% dev deprel acc 97.80% [hops] 2024-09-24 16:15:50.944 | INFO | Epoch 41: train loss 0.0194 dev loss 0.2632 dev tag acc 99.14% dev head acc 95.73% dev deprel acc 97.77% [hops] 2024-09-24 16:16:09.974 | INFO | Epoch 42: train loss 0.0178 dev loss 0.2683 dev tag acc 99.21% dev head acc 95.81% dev deprel acc 97.85% [hops] 2024-09-24 16:16:09.975 | INFO | New best model: head accuracy 95.81% > 95.79% [hops] 2024-09-24 16:16:31.272 | INFO | Epoch 43: train loss 0.0162 dev loss 0.2753 dev tag acc 99.20% dev head acc 95.74% dev deprel acc 97.80% [hops] 2024-09-24 16:16:49.960 | INFO | Epoch 44: train loss 0.0233 dev loss 0.2764 dev tag acc 99.21% dev head acc 95.75% dev deprel acc 97.84% [hops] 2024-09-24 16:17:09.618 | INFO | Epoch 45: train loss 0.0157 dev loss 0.2860 dev tag acc 99.23% dev head acc 95.84% dev deprel acc 97.92% [hops] 2024-09-24 16:17:09.619 | INFO | New best model: head accuracy 95.84% > 95.81% [hops] 2024-09-24 16:17:31.311 | INFO | Epoch 46: train loss 0.0141 dev loss 0.2782 dev tag acc 99.24% dev head acc 95.67% dev deprel acc 97.92% [hops] 2024-09-24 16:17:50.429 | INFO | Epoch 47: train loss 0.0135 dev loss 0.2823 dev tag acc 99.20% dev head acc 95.93% dev deprel acc 97.84% [hops] 2024-09-24 16:17:50.430 | INFO | New best model: head accuracy 95.93% > 95.84% [hops] 2024-09-24 16:18:12.154 | INFO | Epoch 48: train loss 0.0139 dev loss 0.2830 dev tag acc 99.17% dev head acc 95.77% dev deprel acc 97.81% [hops] 2024-09-24 16:18:30.927 | INFO | Epoch 49: train loss 0.0129 dev loss 0.2882 dev tag acc 99.17% dev head acc 95.87% dev deprel acc 97.78% [hops] 2024-09-24 16:18:50.677 | INFO | Epoch 50: train loss 0.0120 dev loss 0.2876 dev tag acc 99.18% dev head acc 95.82% dev deprel acc 97.80% [hops] 2024-09-24 16:19:09.749 | INFO | Epoch 51: train loss 0.0115 dev loss 0.2998 dev tag acc 99.19% dev head acc 95.69% dev deprel acc 97.85% [hops] 2024-09-24 16:19:28.709 | INFO | Epoch 52: train loss 0.0116 dev loss 0.2948 dev tag acc 99.22% dev head acc 95.76% dev deprel acc 97.95% [hops] 2024-09-24 16:19:47.697 | INFO | Epoch 53: train loss 0.0102 dev loss 0.3000 dev tag acc 99.22% dev head acc 95.95% dev deprel acc 97.90% [hops] 2024-09-24 16:19:47.698 | INFO | New best model: head accuracy 95.95% > 95.93% [hops] 2024-09-24 16:20:09.017 | INFO | Epoch 54: train loss 0.0104 dev loss 0.3013 dev tag acc 99.23% dev head acc 96.02% dev deprel acc 97.89% [hops] 2024-09-24 16:20:09.018 | INFO | New best model: head accuracy 96.02% > 95.95% [hops] 2024-09-24 16:20:30.175 | INFO | Epoch 55: train loss 0.0105 dev loss 0.2964 dev tag acc 99.26% dev head acc 96.01% dev deprel acc 97.95% [hops] 2024-09-24 16:20:50.301 | INFO | Epoch 56: train loss 0.0098 dev loss 0.2959 dev tag acc 99.23% dev head acc 95.97% dev deprel acc 97.89% [hops] 2024-09-24 16:21:09.359 | INFO | Epoch 57: train loss 0.0091 dev loss 0.3062 dev tag acc 99.25% dev head acc 95.88% dev deprel acc 97.87% [hops] 2024-09-24 16:21:28.289 | INFO | Epoch 58: train loss 0.0089 dev loss 0.3102 dev tag acc 99.25% dev head acc 95.87% dev deprel acc 97.85% [hops] 2024-09-24 16:21:48.019 | INFO | Epoch 59: train loss 0.0089 dev loss 0.3086 dev tag acc 99.24% dev head acc 95.97% dev deprel acc 97.91% [hops] 2024-09-24 16:22:06.944 | INFO | Epoch 60: train loss 0.0086 dev loss 0.3091 dev tag acc 99.25% dev head acc 95.98% dev deprel acc 97.91% [hops] 2024-09-24 16:22:25.872 | INFO | Epoch 61: train loss 0.0080 dev loss 0.3121 dev tag acc 99.27% dev head acc 95.96% dev deprel acc 97.91% [hops] 2024-09-24 16:22:44.686 | INFO | Epoch 62: train loss 0.0087 dev loss 0.3127 dev tag acc 99.25% dev head acc 96.02% dev deprel acc 97.91% [hops] 2024-09-24 16:23:03.731 | INFO | Epoch 63: train loss 0.0083 dev loss 0.3123 dev tag acc 99.25% dev head acc 96.03% dev deprel acc 97.92% [hops] 2024-09-24 16:23:03.732 | INFO | New best model: head accuracy 96.03% > 96.02% [hops] 2024-09-24 16:23:11.038 | 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 16:23:16.913 | 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 16:23:18.926 | INFO | Metrics for Sequoia-camembertav2_base_p2_17k_last_layer+rand_seed=42 ─────────────────────────────── Split UPOS UAS LAS ─────────────────────────────── Dev 99.25 96.08 94.89 Test 99.42 95.98 94.88 ───────────────────────────────