metadata
base_model: ylacombe/w2v-bert-2.0
tags:
- generated_from_trainer
metrics:
- wer
model-index:
- name: w2v-bert-2.0-ukrainian-colab-CV16.0
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: mozilla-foundation/common_voice_16_1
type: mozilla-foundation/common_voice_16_1
config: uk
split: test
args: uk
metrics:
- name: Wer
type: wer
value: 0.0987
license: mit
datasets:
- mozilla-foundation/common_voice_16_1
language:
- uk
pipeline_tag: automatic-speech-recognition
library_name: transformers
w2v-bert-2.0-ukrainian-colab-CV16.0
This model is a fine-tuned version of ylacombe/w2v-bert-2.0 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.1438
- Wer: 0.0987
Note: the model was finetuned on Ukrainian alphabet in lowercase plus "'" sign. Therefore this model can't add punctuation or capitalization.
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: 5e-05
- train_batch_size: 64
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 128
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 1.0371 | 1.98 | 525 | 0.1509 | 0.1498 |
| 0.0728 | 3.96 | 1050 | 0.1256 | 0.1279 |
| 0.0382 | 5.94 | 1575 | 0.1260 | 0.1041 |
| 0.0213 | 7.92 | 2100 | 0.1333 | 0.0997 |
| 0.0118 | 9.91 | 2625 | 0.1438 | 0.0987 |
Framework versions
- Transformers 4.37.0.dev0
- Pytorch 1.12.1+cu116
- Datasets 2.4.0
- Tokenizers 0.15.1