metadata
library_name: transformers
language:
- ne
license: mit
base_model: facebook/w2v-bert-2.0
tags:
- generated_from_trainer
datasets:
- kiranpantha/OpenSLR54-Balanced-Nepali
metrics:
- wer
model-index:
- name: Wave2Vec2-Bert2.0 - Kiran Pantha
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: kiranpantha/OpenSLR54-Balanced-Nepali
type: kiranpantha/OpenSLR54-Balanced-Nepali
args: 'config: ne, split: train,test'
metrics:
- name: Wer
type: wer
value: 0.45406330196749356
Wave2Vec2-Bert2.0 - Kiran Pantha
This model is a fine-tuned version of facebook/w2v-bert-2.0 on the kiranpantha/OpenSLR54-Balanced-Nepali dataset. It achieves the following results on the evaluation set:
- Loss: 0.5896
- Wer: 0.4541
- Cer: 0.1135
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: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 2
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 0.1466 | 0.24 | 300 | 0.5865 | 0.4662 | 0.1138 |
| 0.2372 | 0.48 | 600 | 0.6488 | 0.5092 | 0.1314 |
| 0.2822 | 0.72 | 900 | 0.5943 | 0.4713 | 0.1179 |
| 0.2517 | 0.96 | 1200 | 0.5902 | 0.4876 | 0.1252 |
| 0.1943 | 1.2 | 1500 | 0.6223 | 0.4814 | 0.1228 |
| 0.1776 | 1.44 | 1800 | 0.5888 | 0.4566 | 0.1155 |
| 0.1567 | 1.6800 | 2100 | 0.6007 | 0.4612 | 0.1154 |
| 0.1495 | 1.92 | 2400 | 0.5896 | 0.4541 | 0.1135 |
Framework versions
- Transformers 4.44.2
- Pytorch 2.4.1+cu121
- Datasets 3.0.1
- Tokenizers 0.19.1