README.md

---
language:
- dv
base_model: alakxender/w2v-bert-2.0-dhivehi-cv
tags:
- generated_from_trainer
datasets:
- mozilla-foundation/common_voice_17_0
metrics:
- wer
model-index:
- name: w2v Bert 2.0 Dv
  results:
  - task:
      name: Automatic Speech Recognition
      type: automatic-speech-recognition
    dataset:
      name: Common Voice 17.0
      type: mozilla-foundation/common_voice_17_0
      config: dv
      split: test
      args: 'config: dv, split: test'
    metrics:
    - name: Wer
      type: wer
      value: 0.45908364040881594
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

# w2v Bert 2.0 Dv - alakxender

This model is a fine-tuned version of [alakxender/w2v-bert-2.0-dhivehi-cv](https://huggingface.co/alakxender/w2v-bert-2.0-dhivehi-cv) on the Common Voice 17.0 dataset.
It achieves the following results on the evaluation set:
- Loss: 0.3580
- Wer: 0.4591

## 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: 32
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 64
- 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    |
|:-------------:|:------:|:----:|:---------------:|:------:|
| 1.9272        | 3.8961 | 300  | 0.3712          | 0.5096 |
| 0.1846        | 7.7922 | 600  | 0.3580          | 0.4591 |


### Framework versions

- Transformers 4.41.0.dev0
- Pytorch 2.3.0+cu121
- Datasets 2.19.0
- Tokenizers 0.19.1