wav2vec2-xls-r-300m-th-v7_0
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: 3.4099
- Wer: 0.9988
- Cer: 0.7861
- Clean Cer: 0.7617
- Learning Rate: 0.0000
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: 5e-05
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 64
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer | Cer | Clean Cer | Rate |
---|---|---|---|---|---|---|---|
8.5484 | 0.4 | 500 | 3.6234 | 1.0 | 1.0 | 1.0 | 0.0000 |
3.2275 | 0.8 | 1000 | 2.2960 | 0.9998 | 0.7081 | 0.6540 | 0.0000 |
0.9955 | 1.2 | 1500 | 1.2224 | 0.9549 | 0.4327 | 0.3756 | 0.0000 |
0.66 | 1.61 | 2000 | 0.9559 | 0.9232 | 0.3651 | 0.3040 | 0.0000 |
0.546 | 2.01 | 2500 | 0.9207 | 0.9481 | 0.3585 | 0.2826 | 0.0000 |
0.4459 | 2.41 | 3000 | 0.7701 | 0.8693 | 0.2940 | 0.2383 | 0.0000 |
0.4041 | 2.81 | 3500 | 0.7756 | 0.8224 | 0.2949 | 0.2634 | 0.0000 |
0.3637 | 3.21 | 4000 | 0.6015 | 0.7015 | 0.2064 | 0.1807 | 0.0000 |
0.334 | 3.61 | 4500 | 0.5615 | 0.6675 | 0.1907 | 0.1638 | 0.0000 |
0.3283 | 4.02 | 5000 | 0.6205 | 0.7073 | 0.2092 | 0.1803 | 0.0000 |
0.3762 | 4.42 | 5500 | 0.7517 | 0.6366 | 0.1778 | 0.1600 | 0.0000 |
0.4954 | 4.82 | 6000 | 0.9374 | 0.7073 | 0.2023 | 0.1735 | 0.0000 |
0.5568 | 5.22 | 6500 | 0.8859 | 0.7027 | 0.1982 | 0.1666 | 0.0000 |
0.6756 | 5.62 | 7000 | 1.0252 | 0.6802 | 0.1920 | 0.1628 | 0.0000 |
0.7752 | 6.02 | 7500 | 1.1259 | 0.7657 | 0.2309 | 0.1908 | 0.0000 |
0.8305 | 6.43 | 8000 | 1.3857 | 0.9029 | 0.3252 | 0.2668 | 0.0000 |
1.7385 | 6.83 | 8500 | 3.2320 | 0.9998 | 0.9234 | 0.9114 | 0.0000 |
2.7839 | 7.23 | 9000 | 3.3238 | 0.9999 | 0.9400 | 0.9306 | 0.0000 |
2.8307 | 7.63 | 9500 | 3.2678 | 0.9998 | 0.9167 | 0.9053 | 0.0000 |
2.7672 | 8.03 | 10000 | 3.2435 | 0.9995 | 0.8992 | 0.8867 | 0.0000 |
2.7426 | 8.43 | 10500 | 3.2396 | 0.9995 | 0.8720 | 0.8561 | 0.0000 |
2.7608 | 8.84 | 11000 | 3.2689 | 0.9993 | 0.8399 | 0.8202 | 0.0000 |
2.8195 | 9.24 | 11500 | 3.3283 | 0.9989 | 0.8084 | 0.7865 | 0.0000 |
2.9044 | 9.64 | 12000 | 3.4099 | 0.9988 | 0.7861 | 0.7617 | 0.0000 |
Framework versions
- Transformers 4.27.0.dev0
- Pytorch 1.13.1+cu116
- Datasets 2.9.0
- Tokenizers 0.13.2
- Downloads last month
- 3
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.