File size: 2,218 Bytes
e93b51d 7b4c9f5 e93b51d 7b4c9f5 e93b51d 7b4c9f5 e93b51d 7b4c9f5 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 |
---
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 |