metadata
license: mit
base_model: facebook/w2v-bert-2.0
tags:
- generated_from_trainer
datasets:
- common_voice_16_0
metrics:
- wer
model-index:
- name: w2v-bert-2.0-mongolian-colab-CV16.0-test
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: common_voice_16_0
type: common_voice_16_0
config: mn
split: test
args: mn
metrics:
- name: Wer
type: wer
value: 0.872688853671421
w2v-bert-2.0-mongolian-colab-CV16.0-test
This model is a fine-tuned version of facebook/w2v-bert-2.0 on the common_voice_16_0 dataset. It achieves the following results on the evaluation set:
- Loss: 0.5486
- Wer: 0.8727
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: 4
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.7431 | 0.79 | 200 | 0.7963 | 0.9926 |
| 0.4379 | 1.58 | 400 | 0.6480 | 0.9805 |
| 0.3109 | 2.37 | 600 | 0.5584 | 0.9546 |
| 0.2444 | 3.17 | 800 | 0.5261 | 0.9429 |
| 0.2048 | 3.96 | 1000 | 0.5208 | 0.9329 |
| 0.1512 | 4.75 | 1200 | 0.5084 | 0.9229 |
| 0.1161 | 5.54 | 1400 | 0.5248 | 0.9197 |
| 0.0882 | 6.33 | 1600 | 0.5248 | 0.9017 |
| 0.0728 | 7.12 | 1800 | 0.5295 | 0.8885 |
| 0.0608 | 7.91 | 2000 | 0.5178 | 0.8833 |
| 0.0386 | 8.7 | 2200 | 0.5317 | 0.8732 |
| 0.0234 | 9.5 | 2400 | 0.5486 | 0.8727 |
Framework versions
- Transformers 4.38.1
- Pytorch 2.2.1+cu118
- Datasets 2.17.1
- Tokenizers 0.15.2