metadata
language:
- te
license: apache-2.0
base_model: openai/whisper-medium
tags:
- hf-asr-leaderboard
- generated_from_trainer
datasets:
- Sonal0205/telugu_asr_recent
metrics:
- wer
model-index:
- name: Whisper Medium te - telugu new
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: Telugu ASR
type: Sonal0205/telugu_asr_recent
args: 'config: te, split: test'
metrics:
- name: Wer
type: wer
value: 23.513674197384066
Whisper Medium te - telugu new
This model is a fine-tuned version of openai/whisper-medium on the Telugu ASR dataset. It achieves the following results on the evaluation set:
- Loss: 0.0925
- Wer: 23.5137
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: 16
- 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
- training_steps: 4000
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
0.0452 | 1.0582 | 1000 | 0.0992 | 42.1522 |
0.0148 | 2.1164 | 2000 | 0.0884 | 27.2889 |
0.0065 | 3.1746 | 3000 | 0.0924 | 24.7919 |
0.0013 | 4.2328 | 4000 | 0.0925 | 23.5137 |
Framework versions
- Transformers 4.42.0.dev0
- Pytorch 2.4.0.dev20240606+cu118
- Datasets 2.19.1
- Tokenizers 0.19.1