This model is a fine-tuned version of openai/whisper-base on the speech_impediment_audio dataset. It achieves the following results on the evaluation set:

Loss: 0.3906
Cer: 14.8734

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: 2000
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.0426	20.0	200	0.2896	15.5063
0.0016	40.0	400	0.3225	14.5570
0.0006	60.0	600	0.3447	13.9241
0.0003	80.0	800	0.3588	14.5570
0.0002	100.0	1000	0.3686	14.5570
0.0002	120.0	1200	0.3765	14.8734
0.0002	140.0	1400	0.3827	14.8734
0.0001	160.0	1600	0.3869	14.8734
0.0001	180.0	1800	0.3896	14.8734
0.0001	200.0	2000	0.3906	14.8734

Framework versions

Transformers 4.45.0.dev0
Pytorch 2.3.1+cu121
Datasets 2.20.0
Tokenizers 0.19.1