Whisper Small Persian

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

Loss: 0.0031
Wer: 13.3353

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: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 8
total_train_batch_size: 64
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
training_steps: 5000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0032	9.52	100	0.0031	13.1328
0.0027	19.05	200	0.0031	13.0171
0.002	28.57	300	0.0031	13.0171
0.0015	38.1	400	0.0031	13.0749
0.0005	47.62	500	0.0031	13.3642
0.0003	57.14	600	0.0031	13.3353
0.0002	66.67	700	0.0031	13.3353
0.0002	76.19	800	0.0031	13.3353
0.0001	85.71	900	0.0031	13.3063
0.0001	95.24	1000	0.0031	13.3063
0.0001	104.76	1100	0.0031	13.3063
0.0001	114.29	1200	0.0031	13.3931
0.0001	123.81	1300	0.0031	13.3931
0.0001	133.33	1400	0.0031	13.3642
0.0001	142.86	1500	0.0031	13.3931
0.0001	152.38	1600	0.0031	13.3931
0.0001	161.9	1700	0.0031	13.3642
0.0	171.43	1800	0.0031	13.3642
0.0001	180.95	1900	0.0031	13.3642
0.0	190.48	2000	0.0031	13.3642
0.0	200.0	2100	0.0031	13.3642
0.0	209.52	2200	0.0031	13.3642
0.0	219.05	2300	0.0031	13.3642
0.0	228.57	2400	0.0031	13.3642
0.0	238.1	2500	0.0031	13.3353
0.0	247.62	2600	0.0031	13.3353
0.0	257.14	2700	0.0031	13.3353
0.0	266.67	2800	0.0031	13.3353
0.0	276.19	2900	0.0031	13.3353
0.0	285.71	3000	0.0031	13.3353
0.0	295.24	3100	0.0031	13.3353
0.0	304.76	3200	0.0031	13.3353
0.0	314.29	3300	0.0031	13.3642
0.0	323.81	3400	0.0031	13.3642
0.0	333.33	3500	0.0031	13.3642
0.0	342.86	3600	0.0031	13.3642
0.0	352.38	3700	0.0031	13.3353
0.0	361.9	3800	0.0031	13.3353
0.0	371.43	3900	0.0031	13.3353
0.0	380.95	4000	0.0031	13.3353
0.0	390.48	4100	0.0031	13.3353
0.0	400.0	4200	0.0031	13.3353
0.0	409.52	4300	0.0031	13.3353
0.0	419.05	4400	0.0031	13.3353
0.0	428.57	4500	0.0031	13.3353
0.0	438.1	4600	0.0031	13.3353
0.0	447.62	4700	0.0031	13.3353
0.0	457.14	4800	0.0031	13.3353
0.0	466.67	4900	0.0031	13.3353
0.0	476.19	5000	0.0031	13.3353

Framework versions

Transformers 4.35.2
Pytorch 2.0.1+cu117
Datasets 2.15.0
Tokenizers 0.15.0