<!-- 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-xlsr-ft-cy-btb
This model is a fine-tuned version of DewiBrynJones/wav2vec2-xlsr-ft-enc-cy on the TECHIAITH/BANC-TRAWSGRIFIADAU-BANGOR - NA dataset. It achieves the following results on the evaluation set:
- Loss: 0.4005
- Wer: 0.2866
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: 16
- eval_batch_size: 128
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 5.0
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
No log | 0.21 | 100 | 3.1359 | 1.0 |
No log | 0.41 | 200 | 1.5312 | 0.8469 |
No log | 0.62 | 300 | 0.7244 | 0.4368 |
No log | 0.83 | 400 | 0.5974 | 0.4147 |
2.4175 | 1.03 | 500 | 0.5503 | 0.3684 |
2.4175 | 1.24 | 600 | 0.5284 | 0.3704 |
2.4175 | 1.44 | 700 | 0.5080 | 0.3570 |
2.4175 | 1.65 | 800 | 0.4968 | 0.3511 |
2.4175 | 1.86 | 900 | 0.4704 | 0.3358 |
0.6101 | 2.06 | 1000 | 0.4415 | 0.3254 |
0.6101 | 2.27 | 1100 | 0.4388 | 0.3219 |
0.6101 | 2.48 | 1200 | 0.4338 | 0.3167 |
0.6101 | 2.68 | 1300 | 0.4196 | 0.3129 |
0.6101 | 2.89 | 1400 | 0.4169 | 0.3074 |
0.4711 | 3.1 | 1500 | 0.4233 | 0.3103 |
0.4711 | 3.3 | 1600 | 0.4156 | 0.3094 |
0.4711 | 3.51 | 1700 | 0.4225 | 0.3061 |
0.4711 | 3.72 | 1800 | 0.4198 | 0.3028 |
0.4711 | 3.92 | 1900 | 0.4020 | 0.2958 |
0.3831 | 4.13 | 2000 | 0.4087 | 0.2961 |
0.3831 | 4.33 | 2100 | 0.4141 | 0.2928 |
0.3831 | 4.54 | 2200 | 0.4077 | 0.2876 |
0.3831 | 4.75 | 2300 | 0.4022 | 0.2871 |
0.3831 | 4.95 | 2400 | 0.4008 | 0.2872 |
Framework versions
- Transformers 4.28.1
- Pytorch 2.0.0+cu117
- Datasets 2.11.0
- Tokenizers 0.13.3