lorenzoncina/whisper-small-en-no-8bitoptm

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: 0.3266
• Wer: 13.0386

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: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1529	0.1	1000	0.4381	17.7766
0.2372	0.2	2000	0.3988	15.9201
0.1706	0.3	3000	0.3841	15.5069
0.2781	0.4	4000	0.3697	14.8122
0.2167	0.5	5000	0.3576	14.2563
0.3609	0.6	6000	0.4041	18.0670
0.2455	0.7	7000	0.3372	13.4813
0.2502	0.8	8000	0.3393	13.5810
0.2564	0.9	9000	0.3303	13.1041
0.2394	1.0	10000	0.3266	13.0386

Framework versions

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