metadata
license: mit
tags:
- generated_from_trainer
metrics:
- wer
base_model: facebook/w2v-bert-2.0
model-index:
- name: w2v-bert-2.0-nonstudio_and_studioRecords
results: []
w2v-bert-2.0-nonstudio_and_studioRecords
This model is a fine-tuned version of facebook/w2v-bert-2.0 on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.1624
- Wer: 0.1242
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: 5e-05
- train_batch_size: 16
- eval_batch_size: 8
- 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: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 1.1627 | 0.46 | 600 | 0.3628 | 0.4690 |
| 0.1739 | 0.92 | 1200 | 0.2809 | 0.3429 |
| 0.1254 | 1.38 | 1800 | 0.1997 | 0.2876 |
| 0.1083 | 1.84 | 2400 | 0.2066 | 0.2735 |
| 0.0917 | 2.3 | 3000 | 0.1918 | 0.2600 |
| 0.0778 | 2.76 | 3600 | 0.1739 | 0.2346 |
| 0.0685 | 3.22 | 4200 | 0.1629 | 0.2210 |
| 0.058 | 3.68 | 4800 | 0.1641 | 0.1983 |
| 0.053 | 4.14 | 5400 | 0.1557 | 0.1841 |
| 0.0417 | 4.6 | 6000 | 0.1464 | 0.1928 |
| 0.0385 | 5.06 | 6600 | 0.1448 | 0.1610 |
| 0.0303 | 5.52 | 7200 | 0.1591 | 0.1510 |
| 0.0303 | 5.98 | 7800 | 0.1352 | 0.1510 |
| 0.0212 | 6.44 | 8400 | 0.1376 | 0.1473 |
| 0.0208 | 6.9 | 9000 | 0.1335 | 0.1314 |
| 0.0153 | 7.36 | 9600 | 0.1455 | 0.1411 |
| 0.0134 | 7.82 | 10200 | 0.1414 | 0.1403 |
| 0.0109 | 8.28 | 10800 | 0.1440 | 0.1361 |
| 0.008 | 8.74 | 11400 | 0.1499 | 0.1286 |
| 0.006 | 9.2 | 12000 | 0.1597 | 0.1212 |
| 0.0039 | 9.66 | 12600 | 0.1624 | 0.1242 |
Framework versions
- Transformers 4.39.3
- Pytorch 2.1.1+cu121
- Datasets 2.16.1
- Tokenizers 0.15.1