ft_0202_korean
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.5641
- Cer: 0.0886
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0001
- train_batch_size: 4
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 20
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Cer |
---|---|---|---|---|
24.8667 | 0.1 | 500 | 5.1486 | 1.0 |
4.7502 | 0.2 | 1000 | 4.8664 | 1.0 |
4.6253 | 0.31 | 1500 | 4.6545 | 1.0 |
4.4689 | 0.41 | 2000 | 4.2926 | 0.9784 |
3.5162 | 0.51 | 2500 | 2.6778 | 0.5191 |
2.6355 | 0.61 | 3000 | 2.1095 | 0.4424 |
2.2199 | 0.72 | 3500 | 1.7898 | 0.3896 |
2.0362 | 0.82 | 4000 | 1.5788 | 0.3551 |
1.8016 | 0.92 | 4500 | 1.4385 | 0.3407 |
1.6906 | 1.02 | 5000 | 1.3008 | 0.3072 |
1.5136 | 1.12 | 5500 | 1.1820 | 0.2845 |
1.4168 | 1.23 | 6000 | 1.0870 | 0.2673 |
1.3461 | 1.33 | 6500 | 1.0315 | 0.2589 |
1.3039 | 1.43 | 7000 | 0.9758 | 0.2438 |
1.2616 | 1.53 | 7500 | 0.9589 | 0.2415 |
1.2301 | 1.64 | 8000 | 0.9262 | 0.2275 |
1.1979 | 1.74 | 8500 | 0.8997 | 0.2305 |
1.1518 | 1.84 | 9000 | 0.8420 | 0.2183 |
1.0925 | 1.94 | 9500 | 0.8073 | 0.2103 |
1.051 | 2.04 | 10000 | 0.7964 | 0.2050 |
1.0025 | 2.15 | 10500 | 0.7683 | 0.2017 |
0.9649 | 2.25 | 11000 | 0.7426 | 0.1984 |
0.9574 | 2.35 | 11500 | 0.7213 | 0.1919 |
0.9244 | 2.45 | 12000 | 0.7198 | 0.1939 |
0.9334 | 2.56 | 12500 | 0.6958 | 0.1847 |
0.8901 | 2.66 | 13000 | 0.6925 | 0.1829 |
0.9049 | 2.76 | 13500 | 0.6862 | 0.1819 |
0.8834 | 2.86 | 14000 | 0.6751 | 0.1813 |
0.8691 | 2.97 | 14500 | 0.6659 | 0.1744 |
0.8194 | 3.07 | 15000 | 0.6674 | 0.1740 |
0.7972 | 3.17 | 15500 | 0.6612 | 0.1711 |
0.7737 | 3.27 | 16000 | 0.6474 | 0.1719 |
0.76 | 3.37 | 16500 | 0.6760 | 0.1687 |
0.7744 | 3.48 | 17000 | 0.6232 | 0.1645 |
0.7571 | 3.58 | 17500 | 0.6104 | 0.1607 |
0.753 | 3.68 | 18000 | 0.6099 | 0.1609 |
0.7276 | 3.78 | 18500 | 0.5977 | 0.1562 |
0.7388 | 3.89 | 19000 | 0.5798 | 0.1548 |
0.7173 | 3.99 | 19500 | 0.5894 | 0.1547 |
0.6796 | 4.09 | 20000 | 0.5873 | 0.1537 |
0.6549 | 4.19 | 20500 | 0.5852 | 0.1518 |
0.6534 | 4.29 | 21000 | 0.5859 | 0.1502 |
0.6659 | 4.4 | 21500 | 0.5679 | 0.1474 |
0.654 | 4.5 | 22000 | 0.5804 | 0.1490 |
0.6356 | 4.6 | 22500 | 0.5888 | 0.1504 |
0.6612 | 4.7 | 23000 | 0.5544 | 0.1436 |
0.6344 | 4.81 | 23500 | 0.5641 | 0.1428 |
0.6432 | 4.91 | 24000 | 0.5570 | 0.1417 |
0.6297 | 5.01 | 24500 | 0.5638 | 0.1413 |
0.5746 | 5.11 | 25000 | 0.5808 | 0.1435 |
0.5619 | 5.21 | 25500 | 0.5471 | 0.1403 |
0.5835 | 5.32 | 26000 | 0.5505 | 0.1417 |
0.5867 | 5.42 | 26500 | 0.5446 | 0.1361 |
0.5598 | 5.52 | 27000 | 0.5558 | 0.1358 |
0.5465 | 5.62 | 27500 | 0.5434 | 0.1354 |
0.546 | 5.73 | 28000 | 0.5394 | 0.1362 |
0.5527 | 5.83 | 28500 | 0.5289 | 0.1334 |
0.5712 | 5.93 | 29000 | 0.5175 | 0.1321 |
0.5308 | 6.03 | 29500 | 0.5279 | 0.1318 |
0.5093 | 6.13 | 30000 | 0.5301 | 0.1356 |
0.5129 | 6.24 | 30500 | 0.5252 | 0.1302 |
0.51 | 6.34 | 31000 | 0.5279 | 0.1302 |
0.4938 | 6.44 | 31500 | 0.5234 | 0.1281 |
0.5044 | 6.54 | 32000 | 0.5105 | 0.1290 |
0.4977 | 6.65 | 32500 | 0.5299 | 0.1262 |
0.5046 | 6.75 | 33000 | 0.5164 | 0.1268 |
0.4913 | 6.85 | 33500 | 0.5078 | 0.1260 |
0.4926 | 6.95 | 34000 | 0.5052 | 0.1252 |
0.4636 | 7.06 | 34500 | 0.5200 | 0.1236 |
0.4437 | 7.16 | 35000 | 0.5070 | 0.1217 |
0.4361 | 7.26 | 35500 | 0.5056 | 0.1238 |
0.456 | 7.36 | 36000 | 0.5038 | 0.1212 |
0.4458 | 7.46 | 36500 | 0.5101 | 0.1218 |
0.441 | 7.57 | 37000 | 0.5050 | 0.1211 |
0.4477 | 7.67 | 37500 | 0.5054 | 0.1183 |
0.4521 | 7.77 | 38000 | 0.5029 | 0.1179 |
0.4516 | 7.87 | 38500 | 0.5048 | 0.1211 |
0.4466 | 7.98 | 39000 | 0.4962 | 0.1187 |
0.4103 | 8.08 | 39500 | 0.5037 | 0.1181 |
0.4079 | 8.18 | 40000 | 0.5123 | 0.1188 |
0.3871 | 8.28 | 40500 | 0.5027 | 0.1161 |
0.4012 | 8.38 | 41000 | 0.4979 | 0.1165 |
0.4028 | 8.49 | 41500 | 0.5291 | 0.1190 |
0.405 | 8.59 | 42000 | 0.5006 | 0.1164 |
0.3916 | 8.69 | 42500 | 0.4994 | 0.1136 |
0.4055 | 8.79 | 43000 | 0.4961 | 0.1151 |
0.4098 | 8.9 | 43500 | 0.4934 | 0.1127 |
0.3983 | 9.0 | 44000 | 0.4873 | 0.1122 |
0.3618 | 9.1 | 44500 | 0.4981 | 0.1130 |
0.3504 | 9.2 | 45000 | 0.4963 | 0.1116 |
0.3507 | 9.3 | 45500 | 0.4998 | 0.1132 |
0.37 | 9.41 | 46000 | 0.4787 | 0.1093 |
0.3597 | 9.51 | 46500 | 0.4883 | 0.1095 |
0.3676 | 9.61 | 47000 | 0.4838 | 0.1096 |
0.3469 | 9.71 | 47500 | 0.4967 | 0.1103 |
0.3726 | 9.82 | 48000 | 0.4796 | 0.1081 |
0.3518 | 9.92 | 48500 | 0.4849 | 0.1093 |
0.3444 | 10.02 | 49000 | 0.4902 | 0.1092 |
0.3213 | 10.12 | 49500 | 0.5074 | 0.1082 |
0.3325 | 10.22 | 50000 | 0.5097 | 0.1092 |
0.3289 | 10.33 | 50500 | 0.5087 | 0.1090 |
0.3329 | 10.43 | 51000 | 0.5041 | 0.1090 |
0.325 | 10.53 | 51500 | 0.4926 | 0.1096 |
0.3286 | 10.63 | 52000 | 0.5094 | 0.1058 |
0.3339 | 10.74 | 52500 | 0.4896 | 0.1074 |
0.32 | 10.84 | 53000 | 0.4862 | 0.1069 |
0.3151 | 10.94 | 53500 | 0.4916 | 0.1064 |
0.3262 | 11.04 | 54000 | 0.4836 | 0.1049 |
0.2925 | 11.15 | 54500 | 0.4971 | 0.1056 |
0.2795 | 11.25 | 55000 | 0.5115 | 0.1059 |
0.2937 | 11.35 | 55500 | 0.5015 | 0.1058 |
0.3031 | 11.45 | 56000 | 0.4892 | 0.1039 |
0.3083 | 11.55 | 56500 | 0.5128 | 0.1044 |
0.315 | 11.66 | 57000 | 0.4943 | 0.1029 |
0.2962 | 11.76 | 57500 | 0.4978 | 0.1020 |
0.3018 | 11.86 | 58000 | 0.4841 | 0.1026 |
0.2905 | 11.96 | 58500 | 0.5011 | 0.1032 |
0.2848 | 12.07 | 59000 | 0.5009 | 0.1010 |
0.2618 | 12.17 | 59500 | 0.5169 | 0.1028 |
0.2699 | 12.27 | 60000 | 0.4928 | 0.1012 |
0.2825 | 12.37 | 60500 | 0.5010 | 0.1022 |
0.2723 | 12.47 | 61000 | 0.4960 | 0.1014 |
0.2754 | 12.58 | 61500 | 0.5003 | 0.1002 |
0.267 | 12.68 | 62000 | 0.4775 | 0.0995 |
0.2643 | 12.78 | 62500 | 0.4873 | 0.1003 |
0.2704 | 12.88 | 63000 | 0.5097 | 0.1012 |
0.2611 | 12.99 | 63500 | 0.5054 | 0.1009 |
0.2534 | 13.09 | 64000 | 0.5063 | 0.0996 |
0.247 | 13.19 | 64500 | 0.5079 | 0.0989 |
0.2358 | 13.29 | 65000 | 0.5029 | 0.0986 |
0.2418 | 13.39 | 65500 | 0.5108 | 0.0980 |
0.2536 | 13.5 | 66000 | 0.5140 | 0.0980 |
0.25 | 13.6 | 66500 | 0.4974 | 0.0973 |
0.2489 | 13.7 | 67000 | 0.5047 | 0.0969 |
0.2427 | 13.8 | 67500 | 0.5089 | 0.0980 |
0.2496 | 13.91 | 68000 | 0.5126 | 0.0980 |
0.2572 | 14.01 | 68500 | 0.5071 | 0.0981 |
0.2248 | 14.11 | 69000 | 0.5215 | 0.0985 |
0.2344 | 14.21 | 69500 | 0.5116 | 0.0974 |
0.2245 | 14.31 | 70000 | 0.5185 | 0.0966 |
0.2153 | 14.42 | 70500 | 0.5129 | 0.0962 |
0.2245 | 14.52 | 71000 | 0.5225 | 0.0964 |
0.2229 | 14.62 | 71500 | 0.5194 | 0.0963 |
0.2354 | 14.72 | 72000 | 0.5021 | 0.0957 |
0.2246 | 14.83 | 72500 | 0.5163 | 0.0955 |
0.2154 | 14.93 | 73000 | 0.5114 | 0.0945 |
0.2191 | 15.03 | 73500 | 0.5281 | 0.0947 |
0.2118 | 15.13 | 74000 | 0.5295 | 0.0939 |
0.2161 | 15.24 | 74500 | 0.5131 | 0.0945 |
0.2179 | 15.34 | 75000 | 0.5270 | 0.0946 |
0.2144 | 15.44 | 75500 | 0.5243 | 0.0945 |
0.1999 | 15.54 | 76000 | 0.5212 | 0.0930 |
0.213 | 15.64 | 76500 | 0.5365 | 0.0935 |
0.215 | 15.75 | 77000 | 0.5188 | 0.0937 |
0.2037 | 15.85 | 77500 | 0.5315 | 0.0936 |
0.2122 | 15.95 | 78000 | 0.5267 | 0.0939 |
0.2084 | 16.05 | 78500 | 0.5306 | 0.0919 |
0.1971 | 16.16 | 79000 | 0.5354 | 0.0917 |
0.197 | 16.26 | 79500 | 0.5426 | 0.0920 |
0.1988 | 16.36 | 80000 | 0.5382 | 0.0920 |
0.1904 | 16.46 | 80500 | 0.5396 | 0.0918 |
0.1987 | 16.56 | 81000 | 0.5523 | 0.0918 |
0.1941 | 16.67 | 81500 | 0.5375 | 0.0919 |
0.1879 | 16.77 | 82000 | 0.5460 | 0.0911 |
0.202 | 16.87 | 82500 | 0.5436 | 0.0913 |
0.1932 | 16.97 | 83000 | 0.5389 | 0.0910 |
0.1809 | 17.08 | 83500 | 0.5496 | 0.0912 |
0.1933 | 17.18 | 84000 | 0.5587 | 0.0916 |
0.1802 | 17.28 | 84500 | 0.5466 | 0.0908 |
0.1956 | 17.38 | 85000 | 0.5563 | 0.0907 |
0.1838 | 17.48 | 85500 | 0.5579 | 0.0904 |
0.1787 | 17.59 | 86000 | 0.5520 | 0.0905 |
0.1859 | 17.69 | 86500 | 0.5575 | 0.0901 |
0.1864 | 17.79 | 87000 | 0.5640 | 0.0909 |
0.1887 | 17.89 | 87500 | 0.5594 | 0.0908 |
0.1802 | 18.0 | 88000 | 0.5575 | 0.0905 |
0.1769 | 18.1 | 88500 | 0.5528 | 0.0899 |
0.1772 | 18.2 | 89000 | 0.5672 | 0.0894 |
0.1722 | 18.3 | 89500 | 0.5588 | 0.0897 |
0.1714 | 18.4 | 90000 | 0.5604 | 0.0898 |
0.1632 | 18.51 | 90500 | 0.5576 | 0.0900 |
0.1726 | 18.61 | 91000 | 0.5526 | 0.0897 |
0.167 | 18.71 | 91500 | 0.5652 | 0.0897 |
0.1851 | 18.81 | 92000 | 0.5603 | 0.0899 |
0.1766 | 18.92 | 92500 | 0.5605 | 0.0894 |
0.1685 | 19.02 | 93000 | 0.5620 | 0.0892 |
0.1732 | 19.12 | 93500 | 0.5679 | 0.0889 |
0.166 | 19.22 | 94000 | 0.5655 | 0.0890 |
0.1663 | 19.33 | 94500 | 0.5638 | 0.0892 |
0.1688 | 19.43 | 95000 | 0.5647 | 0.0892 |
0.1613 | 19.53 | 95500 | 0.5661 | 0.0890 |
0.1749 | 19.63 | 96000 | 0.5673 | 0.0889 |
0.1635 | 19.73 | 96500 | 0.5649 | 0.0887 |
0.1647 | 19.84 | 97000 | 0.5641 | 0.0886 |
0.1633 | 19.94 | 97500 | 0.5641 | 0.0886 |
Framework versions
- Transformers 4.36.2
- Pytorch 2.1.2+cu118
- Datasets 2.16.1
- Tokenizers 0.15.0
- Downloads last month
- 4
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.
Model tree for yoon1000/ft_0202_korean
Base model
facebook/wav2vec2-xls-r-300m