polixonrio/whisper-small-fy-NL

Whisper Small Western Frisian (Netherlands)

This model is a fine-tuned version of openai/whisper-small on the mozilla-foundation/common_voice_11_0 fy-NL dataset. This is an attempt for cross lingual transfer from Dutch to Frisian, since Whisper doesn't support Frisian.

It achieves the following results on the evaluation set:

• Loss: 0.5443
• Wer: 22.2969

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: 32
• 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
0.0067	10.01	1000	0.4810	23.0115
0.0008	21.0	2000	0.5200	22.3576
0.0004	31.01	3000	0.5443	22.2969
0.0003	42.0	4000	0.5610	22.3719
0.0002	52.01	5000	0.5674	22.3898

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.13.0+cu117
• Datasets 2.7.1.dev0
• Tokenizers 0.13.2