<!-- 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. -->
wav2vec2-common_voice-tr-demo
This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the COMMON_VOICE - TR dataset. It achieves the following results on the evaluation set:
- Loss: 0.3943
- Wer: 0.3340
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: 0.0003
- train_batch_size: 4
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 8
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 15.0
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| No log | 0.23 | 100 | 3.6296 | 1.0 |
| No log | 0.46 | 200 | 3.1588 | 0.9999 |
| No log | 0.69 | 300 | 2.3111 | 1.0083 |
| No log | 0.92 | 400 | 0.9852 | 0.7981 |
| 3.6643 | 1.15 | 500 | 0.7056 | 0.7363 |
| 3.6643 | 1.38 | 600 | 0.6146 | 0.6287 |
| 3.6643 | 1.61 | 700 | 0.5583 | 0.6195 |
| 3.6643 | 1.84 | 800 | 0.5529 | 0.5678 |
| 3.6643 | 2.07 | 900 | 0.5280 | 0.5373 |
| 0.5896 | 2.3 | 1000 | 0.5253 | 0.5349 |
| 0.5896 | 2.53 | 1100 | 0.4803 | 0.5057 |
| 0.5896 | 2.76 | 1200 | 0.4562 | 0.5132 |
| 0.5896 | 2.99 | 1300 | 0.4252 | 0.4873 |
| 0.5896 | 3.22 | 1400 | 0.4428 | 0.4831 |
| 0.368 | 3.45 | 1500 | 0.4510 | 0.4779 |
| 0.368 | 3.68 | 1600 | 0.4404 | 0.4946 |
| 0.368 | 3.91 | 1700 | 0.4330 | 0.4785 |
| 0.368 | 4.14 | 1800 | 0.4358 | 0.4558 |
| 0.368 | 4.37 | 1900 | 0.4126 | 0.4643 |
| 0.2629 | 4.6 | 2000 | 0.4197 | 0.4529 |
| 0.2629 | 4.83 | 2100 | 0.4064 | 0.4409 |
| 0.2629 | 5.06 | 2200 | 0.4285 | 0.4514 |
| 0.2629 | 5.29 | 2300 | 0.4193 | 0.4204 |
| 0.2629 | 5.52 | 2400 | 0.4301 | 0.4219 |
| 0.2072 | 5.75 | 2500 | 0.4222 | 0.4335 |
| 0.2072 | 5.98 | 2600 | 0.4077 | 0.4231 |
| 0.2072 | 6.21 | 2700 | 0.4132 | 0.4121 |
| 0.2072 | 6.44 | 2800 | 0.4113 | 0.4220 |
| 0.2072 | 6.67 | 2900 | 0.4101 | 0.4175 |
| 0.1731 | 6.9 | 3000 | 0.4240 | 0.4122 |
| 0.1731 | 7.13 | 3100 | 0.4309 | 0.4023 |
| 0.1731 | 7.36 | 3200 | 0.4275 | 0.3987 |
| 0.1731 | 7.59 | 3300 | 0.4289 | 0.4063 |
| 0.1731 | 7.82 | 3400 | 0.4181 | 0.4025 |
| 0.1397 | 8.05 | 3500 | 0.4490 | 0.3885 |
| 0.1397 | 8.28 | 3600 | 0.4198 | 0.3872 |
| 0.1397 | 8.51 | 3700 | 0.3980 | 0.3842 |
| 0.1397 | 8.74 | 3800 | 0.4051 | 0.3876 |
| 0.1397 | 8.97 | 3900 | 0.4080 | 0.3912 |
| 0.1224 | 9.2 | 4000 | 0.4180 | 0.3774 |
| 0.1224 | 9.43 | 4100 | 0.4102 | 0.3820 |
| 0.1224 | 9.66 | 4200 | 0.3978 | 0.3880 |
| 0.1224 | 9.89 | 4300 | 0.4157 | 0.3731 |
| 0.1224 | 10.11 | 4400 | 0.4175 | 0.3741 |
| 0.1012 | 10.34 | 4500 | 0.3887 | 0.3705 |
| 0.1012 | 10.57 | 4600 | 0.4064 | 0.3774 |
| 0.1012 | 10.8 | 4700 | 0.3961 | 0.3622 |
| 0.1012 | 11.03 | 4800 | 0.3912 | 0.3574 |
| 0.1012 | 11.26 | 4900 | 0.4020 | 0.3638 |
| 0.088 | 11.49 | 5000 | 0.4117 | 0.3560 |
| 0.088 | 11.72 | 5100 | 0.3916 | 0.3524 |
| 0.088 | 11.95 | 5200 | 0.4012 | 0.3533 |
| 0.088 | 12.18 | 5300 | 0.4085 | 0.3584 |
| 0.088 | 12.41 | 5400 | 0.4000 | 0.3547 |
| 0.0775 | 12.64 | 5500 | 0.4137 | 0.3525 |
| 0.0775 | 12.87 | 5600 | 0.4005 | 0.3466 |
| 0.0775 | 13.1 | 5700 | 0.3986 | 0.3479 |
| 0.0775 | 13.33 | 5800 | 0.3983 | 0.3470 |
| 0.0775 | 13.56 | 5900 | 0.3940 | 0.3429 |
| 0.0716 | 13.79 | 6000 | 0.3872 | 0.3383 |
| 0.0716 | 14.02 | 6100 | 0.4005 | 0.3384 |
| 0.0716 | 14.25 | 6200 | 0.4005 | 0.3363 |
| 0.0716 | 14.48 | 6300 | 0.3973 | 0.3357 |
| 0.0716 | 14.71 | 6400 | 0.3957 | 0.3347 |
| 0.0639 | 14.94 | 6500 | 0.3942 | 0.3340 |
Framework versions
- Transformers 4.29.2
- Pytorch 2.0.1+cu117
- Datasets 2.12.0
- Tokenizers 0.11.0
