whisper-tiny-minds-v1

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

• Loss: 0.7887
• Wer Ortho: 0.4046
• Wer: 0.3804

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

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
1.7167	3.57	100	1.3603	0.5324	0.4132
0.3753	7.14	200	0.5665	0.4695	0.3894
0.1274	10.71	300	0.5589	0.4626	0.3912
0.0207	14.29	400	0.6216	0.4327	0.3834
0.0045	17.86	500	0.6684	0.4121	0.3697
0.0017	21.43	600	0.7018	0.4171	0.3792
0.0009	25.0	700	0.7218	0.4239	0.3876
0.0007	28.57	800	0.7272	0.4102	0.3781
0.0005	32.14	900	0.7427	0.4077	0.3787
0.0004	35.71	1000	0.7512	0.4077	0.3787
0.0004	39.29	1100	0.7573	0.4034	0.3757
0.0003	42.86	1200	0.7650	0.4027	0.3751
0.0003	46.43	1300	0.7714	0.4059	0.3769
0.0002	50.0	1400	0.7759	0.4052	0.3775
0.0002	53.57	1500	0.7796	0.4077	0.3798
0.0002	57.14	1600	0.7831	0.4046	0.3798
0.0002	60.71	1700	0.7858	0.4040	0.3792
0.0002	64.29	1800	0.7873	0.4040	0.3792
0.0002	67.86	1900	0.7883	0.4034	0.3792
0.0002	71.43	2000	0.7887	0.4046	0.3804

Framework versions

• Transformers 4.30.2
• Pytorch 2.0.1+cu118
• Datasets 2.13.1
• Tokenizers 0.13.3