whisper-small-sorani-v1

This model is a fine-tuned version of openai/whisper-small on the None dataset. It achieves the following results on the evaluation set:

Model description

More information needed

More information needed

More information needed

The following hyperparameters were used during training:

learning_rate: 3e-05
train_batch_size: 16
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_steps: 1250
num_epochs: 15
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer
0.3599	0.3365	500	0.3697	52.7891
0.2722	0.6729	1000	0.2894	41.9277
0.2465	1.0094	1500	0.2676	40.5501
0.1811	1.3459	2000	0.2399	37.7610
0.1754	1.6824	2500	0.2218	34.5224
0.0896	2.0188	3000	0.2165	32.1297
0.1024	2.3553	3500	0.2097	31.3563
0.0998	2.6918	4000	0.2038	30.7038
0.0606	3.0283	4500	0.2031	29.3745
0.052	3.3647	5000	0.2153	29.5292
0.0633	3.7012	5500	0.2065	28.5818
0.0305	4.0377	6000	0.2216	28.3014
0.0323	4.3742	6500	0.2163	27.3395
0.0347	4.7106	7000	0.2180	27.9099
0.0215	5.0471	7500	0.2291	27.5232
0.0237	5.3836	8000	0.2320	27.0978
0.0236	5.7201	8500	0.2318	27.0253
0.0126	6.0565	9000	0.2399	26.6193
0.0141	6.3930	9500	0.2468	26.3486
0.0123	6.7295	10000	0.2493	25.7299
0.009	7.0659	10500	0.2435	26.5371
0.0082	7.4024	11000	0.2500	25.9039
0.0128	7.7389	11500	0.2517	26.3679
0.0042	8.0754	12000	0.2594	26.0731
0.003	8.4118	12500	0.2690	26.1698
0.0056	8.7483	13000	0.2602	25.4592
0.0057	9.0848	13500	0.2617	24.4103
0.0046	9.4213	14000	0.2621	25.2078
0.0046	9.7577	14500	0.2672	24.9468
0.0016	10.0942	15000	0.2693	24.4538
0.0019	10.4307	15500	0.2692	24.0719
0.0014	10.7672	16000	0.2674	24.5795
0.0008	11.1036	16500	0.2768	24.5601
0.0007	11.4401	17000	0.2743	23.8786
0.0007	11.7766	17500	0.2751	24.2411
0.0003	12.1131	18000	0.2808	24.1638
0.0002	12.4495	18500	0.2815	24.1589
0.0006	12.7860	19000	0.2827	23.5934
0.0002	13.1225	19500	0.2847	23.3759
0.0005	13.4590	20000	0.2881	23.6466
0.0001	13.7954	20500	0.2900	23.3227
0.0	14.1319	21000	0.2903	23.3372
0.0	14.4684	21500	0.2914	23.2744
0.0	14.8048	22000	0.2918	23.4484