metadata
library_name: transformers
license: mit
base_model: facebook/w2v-bert-2.0
tags:
- automatic-speech-recognition
- CLEAR-Global/chichewa_34h
- generated_from_trainer
metrics:
- wer
model-index:
- name: w2v-bert-2.0-chichewa_34h
results: []
w2v-bert-2.0-chichewa_34h
This model is a fine-tuned version of facebook/w2v-bert-2.0 on the CLEAR-GLOBAL/CHICHEWA_34H - NA dataset. It achieves the following results on the evaluation set:
- Loss: 0.3389
- Wer: 0.4045
- Cer: 0.1148
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: 3e-05
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 64
- optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- training_steps: 100000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 0.4609 | 5.6197 | 1000 | 0.7327 | 0.6746 | 0.1953 |
| 0.1207 | 11.2366 | 2000 | 0.4130 | 0.4797 | 0.1341 |
| 0.1104 | 16.8563 | 3000 | 0.3404 | 0.4165 | 0.1182 |
| 0.0417 | 22.4732 | 4000 | 0.3389 | 0.4046 | 0.1149 |
| 0.0849 | 28.0901 | 5000 | 0.3593 | 0.3860 | 0.1110 |
| 0.0169 | 33.7099 | 6000 | 0.4053 | 0.3799 | 0.1086 |
| 0.0625 | 39.3268 | 7000 | 0.4394 | 0.3820 | 0.1103 |
| 0.0226 | 44.9465 | 8000 | 0.4477 | 0.3922 | 0.1099 |
| 0.0236 | 50.5634 | 9000 | 0.4660 | 0.3855 | 0.1101 |
Framework versions
- Transformers 4.48.1
- Pytorch 2.6.0+cu124
- Datasets 3.5.0
- Tokenizers 0.21.1