metadata
language:
- sv-SE
license: apache-2.0
tags:
- automatic-speech-recognition
- mozilla-foundation/common_voice_7_0
- generated_from_trainer
datasets:
- common_voice
model-index:
- name: ''
results: []
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the MOZILLA-FOUNDATION/COMMON_VOICE_7_0 - SV-SE dataset. It achieves the following results on the evaluation set:
- Loss: 0.8004
- Wer: 0.7139
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: 7.5e-05
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- gradient_accumulation_steps: 8
- 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: 2000
- num_epochs: 10.0
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 2.6683 | 1.45 | 500 | 1.7698 | 1.0041 |
| 1.9548 | 2.91 | 1000 | 1.0890 | 0.8602 |
| 1.9568 | 4.36 | 1500 | 1.0878 | 0.8680 |
| 1.9497 | 5.81 | 2000 | 1.1501 | 0.8838 |
| 1.8453 | 7.27 | 2500 | 1.0452 | 0.8418 |
| 1.6952 | 8.72 | 3000 | 0.9153 | 0.7823 |
Framework versions
- Transformers 4.16.0.dev0
- Pytorch 1.10.0+cu113
- Datasets 1.18.1.dev0
- Tokenizers 0.10.3