learning_rate: 5e-05
train_batch_size: 16
eval_batch_size: 8
seed: 42
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_ratio: 0.1
num_epochs: 10
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Model Preparation Time	Wer	Cer
0.2867	0.4990	250	0.4180	0.021	22.3373	6.6987
0.1669	0.9980	500	0.2537	0.021	21.0459	6.0712
0.1596	1.4970	750	0.2277	0.021	19.3997	5.2772
0.1512	1.9960	1000	0.2166	0.021	18.7753	5.1203
0.1192	2.4950	1250	0.2125	0.021	19.6197	5.8610
0.1374	2.9940	1500	0.2064	0.021	18.5553	5.1774
0.1001	3.4930	1750	0.2124	0.021	19.0449	5.4769
0.1001	3.9920	2000	0.2081	0.021	18.5340	5.3495
0.0856	4.4910	2250	0.2193	0.021	20.1022	6.6464
0.085	4.9900	2500	0.2165	0.021	19.2223	5.8278
0.0596	5.4890	2750	0.2309	0.021	19.5274	6.0950
0.0756	5.9880	3000	0.2280	0.021	18.7966	5.2848
0.056	6.4870	3250	0.2427	0.021	21.0246	7.5117
0.0516	6.9860	3500	0.2451	0.021	18.7256	5.2858
0.0413	7.4850	3750	0.2577	0.021	20.3789	6.3678
0.0409	7.9840	4000	0.2629	0.021	21.2091	7.2445
0.0342	8.4830	4250	0.2694	0.021	20.2299	6.5342
0.0366	8.9820	4500	0.2729	0.021	20.6556	6.6759
0.0287	9.4810	4750	0.2830	0.021	20.1873	6.1130
0.033	9.9800	5000	0.2836	0.021	20.1448	6.2499

AleksanderObuchowski/whisper-large-v3-med-pl-muon-lora-decoder-only

whisper-large-v3-med-pl-muon-lora-decoder-only