<!-- 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. -->
med_v1_M05
This model is a fine-tuned version of facebook/wav2vec2-base on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 3.6774
- Wer: 1.0
- Cer: 0.8996
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: 0.0002
- train_batch_size: 10
- 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: 2575
- num_epochs: 100
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
---|---|---|---|---|---|
44.6314 | 5.0 | 2575 | 3.8187 | 1.0 | 1.0 |
3.2116 | 10.0 | 5150 | 3.9616 | 1.0 | 0.9957 |
2.9315 | 15.0 | 7725 | 3.7222 | 1.0 | 0.9844 |
2.6931 | 20.0 | 10300 | 3.6458 | 1.0 | 0.9775 |
2.557 | 25.0 | 12875 | 3.6697 | 1.0 | 0.9576 |
2.4757 | 30.0 | 15450 | 3.6671 | 1.0 | 0.9446 |
2.4201 | 35.0 | 18025 | 3.7050 | 1.0 | 0.9403 |
2.3807 | 40.0 | 20600 | 3.6993 | 1.0 | 0.9359 |
2.3523 | 45.0 | 23175 | 3.6753 | 1.0 | 0.9212 |
2.3284 | 50.0 | 25750 | 3.6720 | 1.0 | 0.9212 |
2.3104 | 55.0 | 28325 | 3.6742 | 1.0 | 0.9152 |
2.2964 | 60.0 | 30900 | 3.6970 | 1.0 | 0.9100 |
2.2833 | 65.0 | 33475 | 3.6954 | 1.0 | 0.9108 |
2.2713 | 70.0 | 36050 | 3.7118 | 1.0 | 0.9065 |
2.2667 | 75.0 | 38625 | 3.7006 | 1.0 | 0.9022 |
2.2606 | 80.0 | 41200 | 3.6913 | 1.0 | 0.9074 |
2.252 | 85.0 | 43775 | 3.6884 | 1.0 | 0.9013 |
2.2513 | 90.0 | 46350 | 3.6858 | 1.0 | 0.9039 |
2.251 | 95.0 | 48925 | 3.6737 | 1.0 | 0.8978 |
2.2455 | 100.0 | 51500 | 3.6774 | 1.0 | 0.8996 |
Framework versions
- Transformers 4.18.0
- Pytorch 1.10.2+cu102
- Datasets 2.3.2
- Tokenizers 0.12.1