whisper-base-basque-lr1e-5-freezeFalse

This model is a fine-tuned version of openai/whisper-base on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2464
• Wer: 12.0644

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: 192
• eval_batch_size: 96
• 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
0.2484	0.5	1000	0.4151	22.1306
0.1688	0.99	2000	0.3039	16.3482
0.1328	1.49	3000	0.2742	14.3499
0.1187	1.98	4000	0.2630	13.8379
0.1002	2.48	5000	0.2546	14.3499
0.0927	2.97	6000	0.2480	12.5453
0.0818	3.47	7000	0.2485	12.3829
0.0794	3.96	8000	0.2452	12.2330
0.0702	4.46	9000	0.2469	12.0520
0.0734	4.96	10000	0.2464	12.0644

Framework versions

• Transformers 4.25.1
• Pytorch 2.5.1+cu121
• Datasets 2.8.0
• Tokenizers 0.13.3