wav2vec2-large-mms-1b-korean-colab_v2
This model is a fine-tuned version of facebook/mms-1b-l1107 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.1650
- Wer: 0.3776
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: 0.005
- train_batch_size: 8
- 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: 100
- num_epochs: 3
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 4.6667 | 0.18 | 100 | 0.8024 | 0.8379 |
| 0.5754 | 0.36 | 200 | 0.3907 | 0.6495 |
| 0.4658 | 0.53 | 300 | 0.3620 | 0.6224 |
| 0.4321 | 0.71 | 400 | 0.3184 | 0.5842 |
| 0.399 | 0.89 | 500 | 0.2930 | 0.5120 |
| 0.3538 | 1.07 | 600 | 0.2446 | 0.4698 |
| 0.3379 | 1.24 | 700 | 0.2341 | 0.4692 |
| 0.3333 | 1.42 | 800 | 0.2121 | 0.4488 |
| 0.31 | 1.6 | 900 | 0.2054 | 0.4297 |
| 0.3049 | 1.78 | 1000 | 0.1958 | 0.4180 |
| 0.2885 | 1.95 | 1100 | 0.1885 | 0.4143 |
| 0.2632 | 2.13 | 1200 | 0.1865 | 0.4094 |
| 0.2592 | 2.31 | 1300 | 0.1774 | 0.3853 |
| 0.2591 | 2.49 | 1400 | 0.1700 | 0.3924 |
| 0.2605 | 2.66 | 1500 | 0.1701 | 0.3789 |
| 0.2361 | 2.84 | 1600 | 0.1650 | 0.3776 |
Framework versions
- Transformers 4.31.0.dev0
- Pytorch 2.0.1+cu118
- Datasets 2.13.1
- Tokenizers 0.13.3