lorenzoncina/whisper-small-en-4-epochs

Whisper Small English

This model is a fine-tuned version of openai/whisper-small on the mozilla-foundation/common_voice_11_0 en dataset. It achieves the following results on the evaluation set:

• Loss: 0.3107
• Wer: 12.0213

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: 32
• eval_batch_size: 16
• 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: 40000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1577	0.06	2500	0.4077	16.2349
0.2244	0.12	5000	0.3698	14.7325
0.3231	0.19	7500	0.3434	13.7448
0.2536	0.25	10000	0.3406	13.4981
0.2234	0.31	12500	0.3510	14.1304
0.1989	0.38	15000	0.3388	13.6394
0.2449	0.44	17500	0.3394	13.4293
0.2302	0.5	20000	0.3198	12.5020
0.213	0.56	22500	0.3167	12.4904
0.2395	0.62	25000	0.3145	12.7533
0.1152	0.69	27500	0.3181	12.6087
0.0901	1.01	30000	0.3134	12.3240
0.1595	1.07	32500	0.3107	12.0213
0.1249	1.13	35000	0.3131	12.0869
0.1404	1.2	37500	0.3117	12.4635
0.1812	1.26	40000	0.3104	12.1415

Framework versions

• Transformers 4.28.0.dev0
• Pytorch 2.0.0+cu117
• Datasets 2.11.1.dev0
• Tokenizers 0.13.2