<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
exp18-F04-both
This model is a fine-tuned version of yongjian/wav2vec2-large-a on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.4137
- Wer: 0.4647
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: 1000
- num_epochs: 30
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
41.5777 | 0.34 | 500 | 3.0940 | 1.0188 |
3.2064 | 0.68 | 1000 | 2.8577 | 1.0157 |
2.997 | 1.02 | 1500 | 2.7604 | 1.0126 |
2.8537 | 1.36 | 2000 | 2.7305 | 1.0 |
2.6677 | 1.7 | 2500 | 2.3201 | 1.2512 |
2.4414 | 2.04 | 3000 | 2.1550 | 1.2575 |
2.2113 | 2.38 | 3500 | 2.0825 | 1.2433 |
2.0619 | 2.72 | 4000 | 2.0245 | 1.2245 |
1.921 | 3.07 | 4500 | 1.6541 | 1.2057 |
1.8182 | 3.41 | 5000 | 1.3678 | 1.1962 |
1.759 | 3.75 | 5500 | 1.1805 | 1.2214 |
1.6229 | 4.09 | 6000 | 1.0100 | 1.1695 |
1.4557 | 4.43 | 6500 | 0.8956 | 1.1287 |
1.4799 | 4.77 | 7000 | 0.7858 | 1.0801 |
1.3277 | 5.11 | 7500 | 0.7306 | 1.0267 |
1.2419 | 5.45 | 8000 | 0.6326 | 0.9262 |
1.1537 | 5.79 | 8500 | 0.6280 | 0.8901 |
1.0972 | 6.13 | 9000 | 0.5639 | 0.9027 |
1.0375 | 6.47 | 9500 | 0.7219 | 0.8352 |
0.9301 | 6.81 | 10000 | 0.4786 | 0.7881 |
0.9423 | 7.15 | 10500 | 0.4969 | 0.7441 |
0.8276 | 7.49 | 11000 | 0.4640 | 0.7551 |
0.8674 | 7.83 | 11500 | 0.5401 | 0.7582 |
0.7633 | 8.17 | 12000 | 0.4610 | 0.6970 |
0.7314 | 8.51 | 12500 | 0.4026 | 0.6923 |
0.7259 | 8.86 | 13000 | 0.4874 | 0.6970 |
0.6591 | 9.2 | 13500 | 0.4701 | 0.6546 |
0.615 | 9.54 | 14000 | 0.4259 | 0.6421 |
0.6098 | 9.88 | 14500 | 0.4206 | 0.6122 |
0.554 | 10.22 | 15000 | 0.4550 | 0.6201 |
0.5521 | 10.56 | 15500 | 0.4777 | 0.6154 |
0.5726 | 10.9 | 16000 | 0.3307 | 0.5997 |
0.5301 | 11.24 | 16500 | 0.4095 | 0.5777 |
0.5098 | 11.58 | 17000 | 0.4914 | 0.5934 |
0.5174 | 11.92 | 17500 | 0.4223 | 0.5981 |
0.4674 | 12.26 | 18000 | 0.3593 | 0.5651 |
0.4574 | 12.6 | 18500 | 0.3951 | 0.5651 |
0.4182 | 12.94 | 19000 | 0.4727 | 0.5808 |
0.388 | 13.28 | 19500 | 0.4737 | 0.5447 |
0.3924 | 13.62 | 20000 | 0.4047 | 0.5322 |
0.3752 | 13.96 | 20500 | 0.3499 | 0.5306 |
0.3374 | 14.31 | 21000 | 0.2930 | 0.5243 |
0.3239 | 14.65 | 21500 | 0.4708 | 0.5338 |
0.3609 | 14.99 | 22000 | 0.3415 | 0.5118 |
0.309 | 15.33 | 22500 | 0.4738 | 0.5149 |
0.2987 | 15.67 | 23000 | 0.4351 | 0.5275 |
0.3726 | 16.01 | 23500 | 0.4305 | 0.5306 |
0.3075 | 16.35 | 24000 | 0.3290 | 0.5212 |
0.2995 | 16.69 | 24500 | 0.3386 | 0.4976 |
0.3262 | 17.03 | 25000 | 0.5279 | 0.5165 |
0.2607 | 17.37 | 25500 | 0.3836 | 0.5008 |
0.2664 | 17.71 | 26000 | 0.4128 | 0.4961 |
0.2578 | 18.05 | 26500 | 0.3517 | 0.4945 |
0.2443 | 18.39 | 27000 | 0.3126 | 0.4804 |
0.2488 | 18.73 | 27500 | 0.3895 | 0.4976 |
0.2382 | 19.07 | 28000 | 0.5097 | 0.5055 |
0.2684 | 19.41 | 28500 | 0.4171 | 0.5071 |
0.2038 | 19.75 | 29000 | 0.4126 | 0.4851 |
0.2273 | 20.1 | 29500 | 0.4142 | 0.4898 |
0.2144 | 20.44 | 30000 | 0.5022 | 0.4961 |
0.2274 | 20.78 | 30500 | 0.4640 | 0.4819 |
0.2055 | 21.12 | 31000 | 0.5124 | 0.4851 |
0.1814 | 21.46 | 31500 | 0.4745 | 0.4804 |
0.201 | 21.8 | 32000 | 0.4669 | 0.4835 |
0.1788 | 22.14 | 32500 | 0.5168 | 0.4851 |
0.2206 | 22.48 | 33000 | 0.4279 | 0.4772 |
0.1847 | 22.82 | 33500 | 0.3862 | 0.4772 |
0.1875 | 23.16 | 34000 | 0.4506 | 0.4851 |
0.1546 | 23.5 | 34500 | 0.4411 | 0.4867 |
0.1768 | 23.84 | 35000 | 0.3386 | 0.4584 |
0.1601 | 24.18 | 35500 | 0.3914 | 0.4678 |
0.1815 | 24.52 | 36000 | 0.3449 | 0.4600 |
0.1495 | 24.86 | 36500 | 0.4789 | 0.4819 |
0.1347 | 25.2 | 37000 | 0.4584 | 0.4741 |
0.1516 | 25.54 | 37500 | 0.3993 | 0.4678 |
0.1514 | 25.89 | 38000 | 0.3898 | 0.4662 |
0.1288 | 26.23 | 38500 | 0.4486 | 0.4819 |
0.1414 | 26.57 | 39000 | 0.4233 | 0.4835 |
0.1407 | 26.91 | 39500 | 0.4119 | 0.4710 |
0.1383 | 27.25 | 40000 | 0.4084 | 0.4788 |
0.1391 | 27.59 | 40500 | 0.4254 | 0.4757 |
0.1302 | 27.93 | 41000 | 0.4208 | 0.4741 |
0.1335 | 28.27 | 41500 | 0.3952 | 0.4662 |
0.1426 | 28.61 | 42000 | 0.4086 | 0.4647 |
0.1303 | 28.95 | 42500 | 0.4071 | 0.4615 |
0.1148 | 29.29 | 43000 | 0.4220 | 0.4662 |
0.1131 | 29.63 | 43500 | 0.4170 | 0.4662 |
0.0998 | 29.97 | 44000 | 0.4137 | 0.4647 |
Framework versions
- Transformers 4.23.1
- Pytorch 1.12.1+cu113
- Datasets 1.18.3
- Tokenizers 0.13.2