metadata
language:
- fa
license: apache-2.0
base_model: openai/whisper-tiny
tags:
- fa-asr-leaderboard
- generated_from_trainer
datasets:
- mozilla-foundation/common_voice_16_0
metrics:
- wer
model-index:
- name: Whisper Tiny Fa - Javad Razavian
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice 16.0
type: mozilla-foundation/common_voice_16_0
config: fa
split: test
args: 'config: fa, split: test'
metrics:
- name: Wer
type: wer
value: 94.28095502498613
Whisper Tiny Fa - Javad Razavian
This model is a fine-tuned version of openai/whisper-tiny on the Common Voice 16.0 dataset. It achieves the following results on the evaluation set:
- Loss: 0.9459
- Wer: 94.2810
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-06
- train_batch_size: 16
- eval_batch_size: 256
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 5000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 4.6309 | 0.08 | 100 | 4.1290 | 140.4220 |
| 2.5371 | 0.16 | 200 | 2.5264 | 128.3176 |
| 1.5224 | 0.24 | 300 | 1.7147 | 120.6830 |
| 1.2351 | 0.33 | 400 | 1.4970 | 112.3542 |
| 1.073 | 0.41 | 500 | 1.3917 | 103.7479 |
| 1.0077 | 0.49 | 600 | 1.3232 | 104.2199 |
| 0.9541 | 0.57 | 700 | 1.2781 | 99.6669 |
| 0.8933 | 0.65 | 800 | 1.2369 | 99.8612 |
| 0.8746 | 0.73 | 900 | 1.2076 | 99.5003 |
| 0.8306 | 0.81 | 1000 | 1.1809 | 99.8890 |
| 0.8309 | 0.89 | 1100 | 1.1583 | 96.5297 |
| 0.7982 | 0.98 | 1200 | 1.1370 | 94.2254 |
| 0.7719 | 1.06 | 1300 | 1.1243 | 96.8351 |
| 0.7799 | 1.14 | 1400 | 1.1065 | 92.6707 |
| 0.7512 | 1.22 | 1500 | 1.0941 | 93.1427 |
| 0.7212 | 1.3 | 1600 | 1.0838 | 94.6696 |
| 0.7315 | 1.38 | 1700 | 1.0709 | 96.0855 |
| 0.7002 | 1.46 | 1800 | 1.0595 | 96.0022 |
| 0.719 | 1.54 | 1900 | 1.0517 | 94.7807 |
| 0.7157 | 1.63 | 2000 | 1.0420 | 95.5303 |
| 0.7004 | 1.71 | 2100 | 1.0337 | 94.2810 |
| 0.6792 | 1.79 | 2200 | 1.0278 | 96.7518 |
| 0.6933 | 1.87 | 2300 | 1.0196 | 95.7801 |
| 0.669 | 1.95 | 2400 | 1.0113 | 98.0566 |
| 0.6627 | 2.03 | 2500 | 1.0063 | 96.8351 |
| 0.655 | 2.11 | 2600 | 1.0006 | 96.0577 |
| 0.6511 | 2.2 | 2700 | 0.9939 | 97.0572 |
| 0.6352 | 2.28 | 2800 | 0.9899 | 95.4470 |
| 0.6339 | 2.36 | 2900 | 0.9874 | 97.2238 |
| 0.6354 | 2.44 | 3000 | 0.9820 | 96.8351 |
| 0.611 | 2.52 | 3100 | 0.9777 | 94.5308 |
| 0.6143 | 2.6 | 3200 | 0.9752 | 99.0006 |
| 0.6242 | 2.68 | 3300 | 0.9729 | 98.7229 |
| 0.6324 | 2.76 | 3400 | 0.9681 | 99.1394 |
| 0.6237 | 2.85 | 3500 | 0.9646 | 96.8906 |
| 0.6285 | 2.93 | 3600 | 0.9621 | 96.1410 |
| 0.5934 | 3.01 | 3700 | 0.9601 | 97.4736 |
| 0.6129 | 3.09 | 3800 | 0.9575 | 92.9761 |
| 0.6154 | 3.17 | 3900 | 0.9575 | 97.5847 |
| 0.6334 | 3.25 | 4000 | 0.9555 | 101.0827 |
| 0.5956 | 3.33 | 4100 | 0.9536 | 94.7529 |
| 0.5956 | 3.41 | 4200 | 0.9507 | 100.3054 |
| 0.6053 | 3.5 | 4300 | 0.9504 | 94.5308 |
| 0.6199 | 3.58 | 4400 | 0.9491 | 95.0861 |
| 0.6064 | 3.66 | 4500 | 0.9482 | 91.8656 |
| 0.6154 | 3.74 | 4600 | 0.9478 | 94.1144 |
| 0.5909 | 3.82 | 4700 | 0.9466 | 91.5047 |
| 0.584 | 3.9 | 4800 | 0.9459 | 94.1144 |
| 0.5935 | 3.98 | 4900 | 0.9459 | 94.0589 |
| 0.5939 | 4.07 | 5000 | 0.9459 | 94.2810 |
Framework versions
- Transformers 4.37.0.dev0
- Pytorch 2.1.0+cu121
- Datasets 2.16.1
- Tokenizers 0.15.0