whisper-tiny-minds14

This model is a fine-tuned version of openai/whisper-tiny on the PolyAI/minds14 dataset. It achieves the following results on the evaluation set:

• Loss: 0.7473
• Wer Ortho: 0.2788
• Wer: 0.2834

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: 3.220378398329722e-05
• train_batch_size: 16
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9850038588304092,0.9902432649395926) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 205
• num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
No log	1.0	28	2.9124	0.5305	0.3991
No log	2.0	56	1.3721	0.4516	0.4091
No log	3.0	84	0.6397	0.3856	0.3872
No log	4.0	112	0.5424	0.3849	0.3819
No log	5.0	140	0.5124	0.4460	0.4410
No log	6.0	168	0.5153	0.3479	0.3477
No log	7.0	196	0.5565	0.3418	0.3424
No log	8.0	224	0.5882	0.3208	0.3229
No log	9.0	252	0.6248	0.3356	0.3371
No log	10.0	280	0.6545	0.3282	0.3300
No log	11.0	308	0.7122	0.3060	0.3093
No log	12.0	336	0.7473	0.2788	0.2834
No log	13.0	364	0.7717	0.3072	0.3093
No log	14.0	392	0.7852	0.3424	0.3447
No log	15.0	420	0.8127	0.3307	0.3318
No log	16.0	448	0.8471	0.3300	0.3294
No log	17.0	476	0.8614	0.3405	0.3406
0.4338	18.0	504	0.8992	0.3627	0.3630
0.4338	19.0	532	0.9157	0.3640	0.3648
0.4338	20.0	560	0.9274	0.3578	0.3589
0.4338	21.0	588	0.9275	0.3387	0.3377
0.4338	22.0	616	0.9371	0.3381	0.3371

Framework versions

• Transformers 4.32.0.dev0
• Pytorch 2.0.1
• Datasets 2.13.2.dev1
• Tokenizers 0.13.3