torgo_xlsr_finetune_M03
This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 1.1905
- Wer: 0.2097
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: 0.0001
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 8
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 1000
- num_epochs: 20
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
3.5007 | 0.85 | 1000 | 3.2973 | 1.0 |
2.2459 | 1.71 | 2000 | 1.9925 | 0.8829 |
0.9013 | 2.56 | 3000 | 1.1537 | 0.6138 |
0.6388 | 3.41 | 4000 | 1.2210 | 0.5017 |
0.5391 | 4.27 | 5000 | 1.2570 | 0.4032 |
0.4528 | 5.12 | 6000 | 1.1298 | 0.3718 |
0.3892 | 5.97 | 7000 | 1.1642 | 0.3090 |
0.3382 | 6.83 | 8000 | 1.0970 | 0.3149 |
0.3279 | 7.68 | 9000 | 1.1686 | 0.3107 |
0.2816 | 8.53 | 10000 | 1.3912 | 0.3107 |
0.2667 | 9.39 | 11000 | 1.2643 | 0.2776 |
0.2517 | 10.24 | 12000 | 1.2157 | 0.2504 |
0.2312 | 11.09 | 13000 | 1.2624 | 0.2640 |
0.2239 | 11.95 | 14000 | 1.2676 | 0.2640 |
0.1849 | 12.8 | 15000 | 1.1427 | 0.2623 |
0.1841 | 13.65 | 16000 | 1.2277 | 0.2547 |
0.1793 | 14.51 | 17000 | 1.3833 | 0.2572 |
0.1704 | 15.36 | 18000 | 1.3813 | 0.2691 |
0.1688 | 16.21 | 19000 | 1.3418 | 0.2589 |
0.1527 | 17.06 | 20000 | 1.2787 | 0.2343 |
0.1304 | 17.92 | 21000 | 1.2078 | 0.2190 |
0.1332 | 18.77 | 22000 | 1.2041 | 0.2105 |
0.1253 | 19.62 | 23000 | 1.1905 | 0.2097 |
Framework versions
- Transformers 4.26.1
- Pytorch 2.2.1
- Datasets 2.18.0
- Tokenizers 0.13.3
- Downloads last month
- 4
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.