Ansu/whisper-large-basque

composite_eu+faktoria+Dialektoak_54378

This model is a fine-tuned version of openai/whisper-large on the composite_eu, faktoria, Dialektoak dataset. It achieves the following results on the evaluation set:

• Loss: 0.2164
• Wer: 10.9775

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: 2e-05
• train_batch_size: 256
• eval_batch_size: 32
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• total_train_batch_size: 512
• total_eval_batch_size: 64
• 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
0.1619	0.42	500	0.2754	15.5930
0.1307	0.84	1000	0.2378	13.6187
0.089	1.26	1500	0.2186	12.2006
0.0867	1.68	2000	0.2166	11.5952
0.0512	2.1	2500	0.2170	11.0001
0.0519	2.52	3000	0.2164	10.9775
0.0511	2.94	3500	0.2175	11.0144
0.027	3.36	4000	0.2344	11.0719
0.0265	3.78	4500	0.2365	10.9713
0.0144	4.2	5000	0.2529	11.2525

Framework versions

• Transformers 4.38.0
• Pytorch 2.1.1+cu121
• Datasets 2.8.0
• Tokenizers 0.15.2