wav2vec2-large-xls-r-300m-assamese-cv8
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the MOZILLA-FOUNDATION/COMMON_VOICE_8_0 - AS dataset. It achieves the following results on the evaluation set:
- Loss: 0.9814
- Wer: 0.7402
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: 32
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 400
- num_epochs: 100.0
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| No log | 20.0 | 400 | 3.1447 | 1.0 |
| No log | 40.0 | 800 | 1.0074 | 0.8556 |
| 3.1278 | 60.0 | 1200 | 0.9507 | 0.7711 |
| 3.1278 | 80.0 | 1600 | 0.9730 | 0.7630 |
| 0.8247 | 100.0 | 2000 | 0.9814 | 0.7402 |
Framework versions
- Transformers 4.16.0.dev0
- Pytorch 1.10.1+cu102
- Datasets 1.18.3
- Tokenizers 0.11.0