|
--- |
|
license: apache-2.0 |
|
tags: |
|
- generated_from_trainer |
|
metrics: |
|
- wer |
|
model-index: |
|
- name: xls-r-asr_af-run2 |
|
results: [] |
|
datasets: |
|
- lucas-meyer/asr_af |
|
--- |
|
|
|
<!-- This model card has been generated automatically according to the information the Trainer had access to. You |
|
should probably proofread and complete it, then remove this comment. --> |
|
|
|
# xls-r-asr_af-run2 |
|
|
|
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the asr_af dataset. |
|
It achieves the following results: |
|
- Wer (Validation): 42.47% |
|
- Wer (Test): 43.42% |
|
|
|
### 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) | |
|
|:-------------:|:-----:|:----:|:---------------:|:------:| |
|
| 8.6534 | 0.44 | 100 | 4.1962 | 1.0 | |
|
| 3.3291 | 0.88 | 200 | 2.9968 | 1.0 | |
|
| 2.9771 | 1.32 | 300 | 2.9616 | 1.0 | |
|
| 2.6613 | 1.76 | 400 | 1.7252 | 0.9456 | |
|
| 1.2954 | 2.2 | 500 | 1.0669 | 0.8175 | |
|
| 0.8676 | 2.64 | 600 | 0.7491 | 0.6426 | |
|
| 0.6715 | 3.08 | 700 | 0.5918 | 0.5330 | |
|
| 0.4916 | 3.52 | 800 | 0.5411 | 0.4762 | |
|
| 0.4443 | 3.96 | 900 | 0.5167 | 0.4767 | |
|
| 0.3304 | 4.41 | 1000 | 0.5264 | 0.4533 | |
|
| 0.3162 | 4.85 | 1100 | 0.5299 | 0.4675 | |
|
| 0.2931 | 5.29 | 1200 | 0.4696 | 0.4192 | |
|
| 0.2472 | 5.73 | 1300 | 0.4630 | 0.4252 | |
|
| 0.2312 | 6.17 | 1400 | 0.4824 | 0.4164 | |
|
| 0.2007 | 6.61 | 1500 | 0.4637 | 0.4035 | |
|
| 0.2036 | 7.05 | 1600 | 0.4802 | 0.3983 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.28.0 |
|
- Pytorch 2.0.1+cu117 |
|
- Datasets 2.14.4 |
|
- Tokenizers 0.13.3 |
