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: 0.7215
Wer: 43.6054

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-07
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: 3000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0006	0.16	100	0.7099	42.5759
0.0016	0.31	200	0.7139	42.6710
0.0051	0.47	300	0.7114	42.6803
0.004	0.62	400	0.7109	42.6130
0.0047	0.78	500	0.7101	42.3139
0.0029	0.93	600	0.7105	42.4113
0.0016	1.09	700	0.7101	42.4067
0.0016	1.25	800	0.7129	42.3997
0.0028	1.4	900	0.7145	42.3951
0.0019	1.56	1000	0.7135	42.3557
0.0019	1.71	1100	0.7132	42.3534
0.0021	1.87	1200	0.7128	42.4739
0.0023	2.03	1300	0.7144	42.9330
0.0021	2.18	1400	0.7151	42.8403
0.0019	2.34	1500	0.7158	42.8820
0.0014	2.49	1600	0.7166	42.9121
0.0262	2.65	1700	0.7171	42.9933
0.0022	2.8	1800	0.7183	42.9608
0.0012	2.96	1900	0.7185	43.3921
0.0018	3.12	2000	0.7187	43.3295
0.0015	3.27	2100	0.7191	43.3480
0.0013	3.43	2200	0.7198	43.3898
0.0019	3.58	2300	0.7200	43.4106
0.0017	3.74	2400	0.7206	43.4315
0.0015	3.9	2500	0.7209	43.4268
0.0016	4.05	2600	0.7210	43.6031
0.0267	4.21	2700	0.7212	43.5891
0.001	4.36	2800	0.7213	43.5938
0.0015	4.52	2900	0.7215	43.6031
0.0019	4.67	3000	0.7215	43.6054

Framework versions

Transformers 4.35.2
Pytorch 2.0.1+cu117
Datasets 2.15.0
Tokenizers 0.15.0