metadata

language:
  - fa
license: apache-2.0
base_model: makhataei/Whisper-Small-Common-Voice
tags:
  - fa-asr
  - generated_from_trainer
datasets:
  - mozilla-foundation/common_voice_16_0
metrics:
  - wer
model-index:
  - name: Whisper Small Persian
    results: []

Whisper Small Persian

This model is a fine-tuned version of makhataei/Whisper-Small-Common-Voice on the Common Voice 16.0 dataset. It achieves the following results on the evaluation set:

Loss: 0.7543
Wer: 46.0283

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: 10
eval_batch_size: 10
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 40
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 50
training_steps: 2000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0137	0.16	100	0.7336	48.8685
0.0153	0.31	200	0.7163	45.0962
0.016	0.47	300	0.7235	45.9726
0.0133	0.62	400	0.7356	45.7130
0.0186	0.78	500	0.6912	44.0760
0.012	0.93	600	0.6939	44.9664
0.0092	1.09	700	0.7184	44.8180
0.0071	1.25	800	0.7321	44.0691
0.0072	1.4	900	0.7485	46.3714
0.0061	1.56	1000	0.7180	44.9502
0.0063	1.71	1100	0.7036	46.2741
0.004	1.87	1200	0.7296	45.0684
0.0036	2.03	1300	0.7275	46.3158
0.002	2.18	1400	0.7432	45.9332
0.0014	2.34	1500	0.7472	45.5692
0.0012	2.49	1600	0.7408	44.7902
0.0201	2.65	1700	0.7566	45.7593
0.0011	2.8	1800	0.7551	45.3327
0.0011	2.96	1900	0.7550	46.1720
0.0004	3.12	2000	0.7543	46.0283

Framework versions

Transformers 4.35.2
Pytorch 2.0.1+cu117
Datasets 2.15.0
Tokenizers 0.15.0