learning_rate: 0.0001
train_batch_size: 32
eval_batch_size: 16
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
0.1664	0.4996	602	0.1688	0.022	15.9176	5.4619
0.1296	0.9992	1204	0.1216	0.022	11.5699	4.7418
0.082	1.4988	1806	0.0956	0.022	11.4329	5.6251
0.0863	1.9983	2408	0.0833	0.022	9.2359	4.2807
0.0543	2.4979	3010	0.0779	0.022	9.8054	5.3263
0.0505	2.9975	3612	0.0712	0.022	7.3651	3.1850
0.0371	3.4971	4214	0.0725	0.022	9.3228	5.1141
0.0334	3.9967	4816	0.0704	0.022	8.2300	4.2879
0.0204	4.4963	5418	0.0707	0.022	7.1219	3.1809
0.0213	4.9959	6020	0.0700	0.022	6.8613	3.1437
0.0151	5.4954	6622	0.0716	0.022	7.2146	3.3620
0.0106	5.9950	7224	0.0701	0.022	6.9385	3.2018
0.0082	6.4946	7826	0.0730	0.022	6.7416	3.1243
0.009	6.9942	8428	0.0726	0.022	6.9771	3.3038
0.0062	7.4938	9030	0.0741	0.022	6.8941	3.2291
0.0052	7.9934	9632	0.0737	0.022	6.9578	3.3553
0.0045	8.4929	10234	0.0751	0.022	6.6721	3.0858
0.0038	8.9925	10836	0.0756	0.022	7.1953	3.2857
0.0038	9.4921	11438	0.0766	0.022	7.0775	3.2391
0.0033	9.9917	12040	0.0768	0.022	6.9829	3.1870

AleksanderObuchowski/whisper-large-v3-med-pl-lora

whisper-large-v3-med-pl-lora