zuazo/whisper-tiny-gl-train

Whisper Tiny Galician

This model is a fine-tuned version of openai/whisper-tiny on the mozilla-foundation/common_voice_13_0 gl dataset. It achieves the following results on the evaluation set:

• Loss: 0.6343
• Wer: 27.0570

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: 3.75e-05
• train_batch_size: 256
• eval_batch_size: 128
• 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
0.3105	23.26	1000	0.5440	30.7025
0.0797	46.51	2000	0.5683	28.1984
0.0326	69.77	3000	0.6091	27.9701
0.0209	93.02	4000	0.6289	27.7176
0.0172	116.28	5000	0.6343	27.0570

Framework versions

• Transformers 4.37.2
• Pytorch 2.2.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.1