whisper-tiny-fo-ravnursson-5k-v1

This model is a fine-tuned version of davidilag/whisper-tiny-no-is-5k-steps on Ravnursson_asr. It achieves the following results on the evaluation set:

Model description

More information needed

Intended for use in a university project

More information needed

The following hyperparameters were used during training:

learning_rate: 1e-05
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 16
optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 250
training_steps: 5000

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
4.0892	0.0610	250	3.7442	97.4754	44.3615
2.7318	0.1219	500	2.6703	94.0609	41.8589
2.4448	0.1829	750	2.4701	91.1839	39.7830
2.3914	0.2438	1000	2.3899	90.0912	39.0619
2.3689	0.3048	1250	2.3468	89.2277	38.4946
2.3692	0.3658	1500	2.3244	89.0206	38.3155
2.3428	0.4267	1750	2.3112	88.8135	38.1356
2.289	0.4877	2000	2.3054	88.6196	38.0820
2.3455	0.5486	2250	2.3010	88.5756	38.0607
2.3957	0.6096	2500	2.2961	88.6813	38.1688
2.2843	0.6706	2750	2.2949	88.7562	38.1033
2.2986	0.7315	3000	2.2952	88.6064	38.0567
2.3073	0.7925	3250	2.2942	88.5712	38.0694
2.275	0.8535	3500	2.2954	88.5844	38.0196
2.2693	0.9144	3750	2.2946	88.5756	38.0978
2.35	0.9754	4000	2.2952	88.5095	38.0465
2.3726	1.0363	4250	2.2951	88.5271	38.0717
2.3145	1.0973	4500	2.2959	88.6152	38.0267
2.268	1.1583	4750	2.2930	88.4170	38.0212
2.3001	1.2192	5000	2.2926	88.6064	38.0985