---
library_name: transformers
license: apache-2.0
base_model: Samuael/asr-amharic-phoneme-based-37-6
tags:
- generated_from_trainer
datasets:
- alffa_amharic
metrics:
- wer
model-index:
- name: tigrinya-asr-characters
  results:
  - task:
      name: Automatic Speech Recognition
      type: automatic-speech-recognition
    dataset:
      name: alffa_amharic
      type: alffa_amharic
      config: clean
      split: None
      args: clean
    metrics:
    - name: Wer
      type: wer
      value: 0.5721231766612642
---

<!-- 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. -->

# tigrinya-asr-characters

This model is a fine-tuned version of [Samuael/asr-amharic-phoneme-based-37-6](https://huggingface.co/Samuael/asr-amharic-phoneme-based-37-6) on the alffa_amharic dataset.
It achieves the following results on the evaluation set:
- Loss: 1.3509
- Wer: 0.5721
- Phoneme Cer: 0.2155

## 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: 0.0001
- train_batch_size: 32
- eval_batch_size: 8
- seed: 42
- 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_steps: 100
- num_epochs: 5

### Training results

| Training Loss | Epoch  | Step | Validation Loss | Wer    | Phoneme Cer |
|:-------------:|:------:|:----:|:---------------:|:------:|:-----------:|
| 3.8729        | 0.2312 | 200  | 3.8491          | 0.9962 | 0.6422      |
| 3.2897        | 0.4624 | 400  | 3.3571          | 0.9941 | 0.5266      |
| 2.9586        | 0.6936 | 600  | 2.9319          | 0.9433 | 0.4522      |
| 2.5878        | 0.9249 | 800  | 2.5198          | 0.8909 | 0.3965      |
| 2.272         | 1.1561 | 1000 | 2.2232          | 0.8314 | 0.3587      |
| 1.8661        | 1.3873 | 1200 | 2.0279          | 0.7801 | 0.3200      |
| 1.9839        | 1.6185 | 1400 | 1.8774          | 0.7202 | 0.2930      |
| 1.5144        | 1.8497 | 1600 | 1.7565          | 0.6910 | 0.2715      |
| 1.5869        | 2.0809 | 1800 | 1.6566          | 0.6694 | 0.2601      |
| 1.3379        | 2.3121 | 2000 | 1.5864          | 0.6413 | 0.2469      |
| 1.4446        | 2.5434 | 2200 | 1.5409          | 0.6272 | 0.2387      |
| 1.3749        | 2.7746 | 2400 | 1.5214          | 0.6207 | 0.2381      |
| 1.7298        | 3.0058 | 2600 | 1.4709          | 0.6121 | 0.2333      |
| 1.3708        | 3.2370 | 2800 | 1.4397          | 0.5991 | 0.2277      |
| 1.3355        | 3.4682 | 3000 | 1.4168          | 0.5894 | 0.2252      |
| 1.227         | 3.6994 | 3200 | 1.3986          | 0.5900 | 0.2229      |
| 1.4346        | 3.9306 | 3400 | 1.3821          | 0.5862 | 0.2223      |
| 1.0866        | 4.1618 | 3600 | 1.3747          | 0.5829 | 0.2208      |
| 1.1822        | 4.3931 | 3800 | 1.3586          | 0.5764 | 0.2164      |
| 0.927         | 4.6243 | 4000 | 1.3532          | 0.5748 | 0.2164      |
| 1.0948        | 4.8555 | 4200 | 1.3509          | 0.5721 | 0.2155      |


### Framework versions

- Transformers 4.46.3
- Pytorch 2.5.1+cu121
- Datasets 3.1.0
- Tokenizers 0.20.3