whisper-base-basque

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: 0.2498
• Wer: 62.1644

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
0.4078	0.25	500	0.5613	136.6055
0.2533	0.5	1000	0.3973	105.6700
0.1994	0.74	1500	0.3350	73.1485
0.1723	0.99	2000	0.3101	55.7387
0.1403	1.24	2500	0.2895	49.8689
0.1318	1.49	3000	0.2800	70.4321
0.1279	1.73	3500	0.2711	80.4296
0.1192	1.98	4000	0.2667	60.5533
0.104	2.23	4500	0.2605	54.0402
0.0986	2.48	5000	0.2601	53.4158
0.0929	2.73	5500	0.2539	61.0653
0.0971	2.97	6000	0.2521	45.2104
0.0806	3.22	6500	0.2526	51.5736
0.0812	3.47	7000	0.2509	53.9903
0.0817	3.72	7500	0.2498	56.8440
0.0799	3.96	8000	0.2511	65.1305
0.0723	4.21	8500	0.2500	55.6326
0.0724	4.46	9000	0.2498	60.2910
0.0707	4.71	9500	0.2501	59.5854
0.0685	4.96	10000	0.2498	62.1644

Framework versions

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