|
|
--- |
|
|
library_name: transformers |
|
|
language: |
|
|
- te |
|
|
base_model: kattojuprashanth238/whisper-small-te-v5 |
|
|
tags: |
|
|
- generated_from_trainer |
|
|
datasets: |
|
|
- None |
|
|
metrics: |
|
|
- wer |
|
|
model-index: |
|
|
- name: Whisper Small Te - Prashanth Kattoju |
|
|
results: |
|
|
- task: |
|
|
name: Automatic Speech Recognition |
|
|
type: automatic-speech-recognition |
|
|
dataset: |
|
|
name: HCU ASR Telugu Corpus |
|
|
type: None |
|
|
config: "te" |
|
|
split: None |
|
|
metrics: |
|
|
- name: Wer |
|
|
type: wer |
|
|
value: 15.932914046121594 |
|
|
--- |
|
|
|
|
|
<!-- 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. --> |
|
|
|
|
|
# Whisper Small Te - Prashanth Kattoju |
|
|
|
|
|
This model is a fine-tuned version of [kattojuprashanth238/whisper-small-te-v5](https://huggingface.co/kattojuprashanth238/whisper-small-te-v5) on the HCU ASR Telugu Corpus. |
|
|
It achieves the following results on the evaluation set: |
|
|
- Loss: 0.2009 |
|
|
- Wer Ortho: 51.4451 |
|
|
- Wer: 15.9329 |
|
|
|
|
|
## 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: 1e-05 |
|
|
- train_batch_size: 4 |
|
|
- eval_batch_size: 8 |
|
|
- seed: 42 |
|
|
- gradient_accumulation_steps: 4 |
|
|
- total_train_batch_size: 16 |
|
|
- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments |
|
|
- lr_scheduler_type: constant_with_warmup |
|
|
- lr_scheduler_warmup_steps: 50 |
|
|
- training_steps: 1500 |
|
|
|
|
|
### Training results |
|
|
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Wer Ortho | Wer | |
|
|
|:-------------:|:-------:|:----:|:---------------:|:---------:|:-------:| |
|
|
| 0.086 | 2.5063 | 50 | 0.1613 | 65.3179 | 19.2872 | |
|
|
| 0.0096 | 5.0 | 100 | 0.1647 | 57.8035 | 17.8197 | |
|
|
| 0.006 | 7.5063 | 150 | 0.1880 | 55.4913 | 14.6751 | |
|
|
| 0.0032 | 10.0 | 200 | 0.1541 | 46.8208 | 12.3690 | |
|
|
| 0.0027 | 12.5063 | 250 | 0.1631 | 48.5549 | 12.1593 | |
|
|
| 0.0002 | 15.0 | 300 | 0.1617 | 44.5087 | 11.5304 | |
|
|
| 0.0005 | 17.5063 | 350 | 0.1965 | 57.8035 | 14.6751 | |
|
|
| 0.0034 | 20.0 | 400 | 0.1589 | 49.7110 | 12.9979 | |
|
|
| 0.0036 | 22.5063 | 450 | 0.1761 | 49.7110 | 12.3690 | |
|
|
| 0.0106 | 25.0 | 500 | 0.1630 | 62.4277 | 19.9161 | |
|
|
| 0.005 | 27.5063 | 550 | 0.1809 | 54.3353 | 14.0461 | |
|
|
| 0.0021 | 30.0 | 600 | 0.1801 | 52.6012 | 11.5304 | |
|
|
| 0.0021 | 32.5063 | 650 | 0.1895 | 53.7572 | 14.4654 | |
|
|
| 0.0027 | 35.0 | 700 | 0.1576 | 51.4451 | 13.6268 | |
|
|
| 0.0033 | 37.5063 | 750 | 0.2080 | 58.9595 | 27.0440 | |
|
|
| 0.0019 | 40.0 | 800 | 0.1958 | 49.7110 | 15.7233 | |
|
|
| 0.0001 | 42.5063 | 850 | 0.1851 | 46.8208 | 12.9979 | |
|
|
| 0.0 | 45.0 | 900 | 0.1887 | 47.9769 | 13.4172 | |
|
|
| 0.0 | 47.5063 | 950 | 0.1904 | 47.9769 | 13.4172 | |
|
|
| 0.0 | 50.0 | 1000 | 0.1916 | 49.1329 | 12.9979 | |
|
|
| 0.0 | 52.5063 | 1050 | 0.1929 | 51.4451 | 14.8847 | |
|
|
| 0.0 | 55.0 | 1100 | 0.1940 | 50.8671 | 14.8847 | |
|
|
| 0.0 | 57.5063 | 1150 | 0.1951 | 50.8671 | 15.9329 | |
|
|
| 0.0 | 60.0 | 1200 | 0.1961 | 50.8671 | 15.9329 | |
|
|
| 0.0 | 62.5063 | 1250 | 0.1969 | 51.4451 | 15.9329 | |
|
|
| 0.0 | 65.0 | 1300 | 0.1978 | 51.4451 | 15.9329 | |
|
|
| 0.0 | 67.5063 | 1350 | 0.1986 | 51.4451 | 15.9329 | |
|
|
| 0.0 | 70.0 | 1400 | 0.1994 | 51.4451 | 15.9329 | |
|
|
| 0.0 | 72.5063 | 1450 | 0.2002 | 51.4451 | 15.9329 | |
|
|
| 0.0 | 75.0 | 1500 | 0.2009 | 51.4451 | 15.9329 | |
|
|
|
|
|
|
|
|
### Framework versions |
|
|
|
|
|
- Transformers 4.48.2 |
|
|
- Pytorch 2.6.0 |
|
|
- Datasets 3.2.0 |
|
|
- Tokenizers 0.21.0 |
|
|
|
