deepdml/whisper-small-mix-pt

Whisper Small Mixed-Portuguese

This model is a fine-tuned version of openai/whisper-small on the pt datasets:

• mozilla-foundation/common_voice_17_0
• google/fleurs
• facebook/multilingual_librispeech

It achieves the following results on the evaluation set:

• Loss: 0.2000
• Wer: 10.6349

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: 64
• 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: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1575	0.2	1000	0.2125	12.1855
0.1986	0.4	2000	0.2062	11.5701
0.0942	1.131	3000	0.1979	11.0154
0.0577	1.331	4000	0.2000	10.6349
0.0516	2.062	5000	0.2007	10.6701

Framework versions

• Transformers 4.42.0.dev0
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1