whisper-tiny-basque-lr1e-5-freezeFalse

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

• Loss: 0.3068
• Wer: 15.2741

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: 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: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.3108	0.66	1000	0.4898	27.4135
0.2097	1.32	2000	0.3735	20.8880
0.179	1.98	3000	0.3456	18.7898
0.1488	2.64	4000	0.3320	17.3161
0.1348	3.3	5000	0.3175	16.7728
0.1271	3.96	6000	0.3110	16.1421
0.1176	4.62	7000	0.3099	15.7050
0.1095	5.28	8000	0.3078	15.2866
0.11	5.94	9000	0.3059	15.2804
0.1048	6.61	10000	0.3068	15.2741

Framework versions

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