whisper-base-finetuned2222222222222222222222222222222

This model is a fine-tuned version of openai/whisper-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0018
• Wer: 0.125

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: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 5
• training_steps: 400
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
7.7056	0.8	20	6.4502	16.25
4.7836	1.6	40	2.9149	10.375
1.8399	2.4	60	0.8254	7.875
0.3132	3.2	80	0.0852	3.875
0.0335	4.0	100	0.0190	1.7500
0.0067	4.8	120	0.0080	1.0
0.0032	5.6	140	0.0050	0.375
0.0021	6.4	160	0.0039	0.125
0.0017	7.2	180	0.0034	0.125
0.0015	8.0	200	0.0030	0.125
0.0013	8.8	220	0.0027	0.125
0.0012	9.6	240	0.0025	0.125
0.0011	10.4	260	0.0023	0.125
0.001	11.2	280	0.0021	0.125
0.0009	12.0	300	0.0020	0.125
0.0009	12.8	320	0.0020	0.125
0.0009	13.6	340	0.0019	0.125
0.0008	14.4	360	0.0018	0.125
0.0009	15.2	380	0.0018	0.125
0.0008	16.0	400	0.0018	0.125

Framework versions

• Transformers 4.39.3
• Pytorch 2.2.2+cu121
• Datasets 2.14.5
• Tokenizers 0.15.2