xls-r-asr_af-run3
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the asr_af dataset. It achieves the following results:
- Wer (Validation): 43.06%
- Wer (Test): 43.66%
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0003
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- gradient_accumulation_steps: 3
- total_train_batch_size: 12
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 30
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer (Train) |
|---|---|---|---|---|
| No log | 1.1 | 250 | 2.9664 | 1.0 |
| 3.6929 | 2.2 | 500 | 1.1311 | 0.8448 |
| 3.6929 | 3.3 | 750 | 0.6159 | 0.5333 |
| 0.5976 | 4.41 | 1000 | 0.5717 | 0.4752 |
| 0.5976 | 5.51 | 1250 | 0.4962 | 0.4306 |
| 0.2559 | 6.61 | 1500 | 0.5146 | 0.4137 |
| 0.2559 | 7.71 | 1750 | 0.5407 | 0.4123 |
| 0.1545 | 8.81 | 2000 | 0.5217 | 0.4069 |
Framework versions
- Transformers 4.28.0
- Pytorch 2.0.1+cu117
- Datasets 2.14.4
- Tokenizers 0.13.3
- Downloads last month
- 6
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.