nkkbr/whisper-large-v3-zatoichi-ja-JDG_ver_20260212_lr_2.0e-5

Whisper Large v3 - Japanese Zatoichi ASR

This model is a fine-tuned version of openai/whisper-large-v3 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4721
• Wer: 66.6359
• Cer: 20.0746

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: 2e-05
• train_batch_size: 32
• eval_batch_size: 8
• seed: 42
• optimizer: Use adamw_torch_fused with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
1.0467	0.0164	1	1.2545	98.1567	40.8419
1.1366	0.0328	2	0.9974	100.4608	39.9760
0.9623	0.0492	3	0.9468	92.9954	36.7524
0.8227	0.0656	4	0.8737	90.4147	33.9417
0.7999	0.0820	5	0.8299	86.7281	31.3041
0.7585	0.0984	6	0.7691	94.7465	38.0445
0.7779	0.1148	7	0.7384	91.7972	32.0767
0.787	0.1311	8	0.7099	101.1060	36.4859
0.5704	0.1475	9	0.6876	101.2903	38.5107
0.5429	0.1639	10	0.6649	96.1290	36.1396
0.6962	0.1803	11	0.6433	93.2719	31.6105
0.6661	0.1967	12	0.6266	94.1014	31.6238
0.6868	0.2131	13	0.6249	84.7005	28.0805
0.6043	0.2295	14	0.6109	79.3548	25.8292
0.5467	0.2459	15	0.6145	84.5161	28.6799
0.5758	0.2623	16	0.6086	79.0783	27.3212
0.4723	0.2787	17	0.5990	82.2120	34.0749
0.476	0.2951	18	0.5992	86.5438	41.0151
0.5545	0.3115	19	0.5883	85.4378	32.9692
0.4452	0.3279	20	0.5830	75.9447	26.7084
0.5191	0.3443	21	0.5775	76.0369	25.3763
0.5186	0.3607	22	0.5731	74.3779	25.7759
0.5398	0.3770	23	0.5813	74.0092	23.9776
0.5192	0.3934	24	0.5793	73.1797	24.3639
0.5424	0.4098	25	0.5664	72.9032	23.8977
0.4434	0.4262	26	0.5615	75.2995	23.6579
0.4001	0.4426	27	0.5614	75.4839	23.7778
0.4888	0.4590	28	0.5571	71.4286	23.1917
0.4519	0.4754	29	0.5642	70.8756	23.5247
0.4317	0.4918	30	0.5631	71.4286	23.6579
0.5509	0.5082	31	0.5596	71.8894	23.5913
0.5109	0.5246	32	0.5555	72.9954	23.4048
0.4176	0.5410	33	0.5491	72.8111	23.2716
0.5682	0.5574	34	0.5445	76.6820	23.6846
0.4805	0.5738	35	0.5391	75.7604	23.5913
0.4523	0.5902	36	0.5344	72.4424	22.8320
0.4466	0.6066	37	0.5338	70.9677	23.0052
0.5174	0.6230	38	0.5314	71.2442	22.8453
0.463	0.6393	39	0.5275	71.0599	22.4724
0.5266	0.6557	40	0.5233	68.9401	22.1527
0.4467	0.6721	41	0.5153	69.3088	22.0994
0.4784	0.6885	42	0.5122	70.7834	22.1393
0.4456	0.7049	43	0.5103	71.3364	21.9262
0.4364	0.7213	44	0.5081	71.5207	22.0194
0.3935	0.7377	45	0.5084	70.6912	22.1260
0.4374	0.7541	46	0.5074	70.3226	22.1127
0.4791	0.7705	47	0.5079	69.4931	22.5523
0.4756	0.7869	48	0.5087	69.4009	22.5523
0.3973	0.8033	49	0.5082	68.3871	22.0461
0.5047	0.8197	50	0.5066	69.4009	23.4714
0.4794	0.8361	51	0.5038	69.4009	23.2050
0.3947	0.8525	52	0.5013	68.2949	21.8862
0.5726	0.8689	53	0.5004	69.2166	22.3392
0.3449	0.8852	54	0.5001	69.4931	22.2326
0.4603	0.9016	55	0.5002	70.0461	22.5390
0.4423	0.9180	56	0.5014	74.1014	23.9510
0.3694	0.9344	57	0.5012	75.2074	24.6969
0.4133	0.9508	58	0.4972	74.1935	24.5105
0.4305	0.9672	59	0.4930	75.1152	24.9367
0.4464	0.9836	60	0.4914	71.4286	24.2973
0.3402	1.0	61	0.4947	75.5760	23.8444
0.3283	1.0164	62	0.4945	77.2350	24.8701
0.2436	1.0328	63	0.4921	72.9032	24.4705
0.1821	1.0492	64	0.4932	67.8341	22.1393
0.2508	1.0656	65	0.4963	67.3733	21.4333
0.2097	1.0820	66	0.4989	66.9124	21.1936
0.2275	1.0984	67	0.5002	67.0968	21.5799
0.2842	1.1148	68	0.4968	65.4378	20.6341
0.2982	1.1311	69	0.4932	66.5438	20.4609
0.3103	1.1475	70	0.4900	70.5991	21.8063
0.2519	1.1639	71	0.4897	67.7419	21.0470
0.1926	1.1803	72	0.4921	67.1889	20.9538
0.2351	1.1967	73	0.4948	67.5576	21.0071
0.2368	1.2131	74	0.4945	67.0968	20.9405
0.2021	1.2295	75	0.4935	71.9816	22.6189
0.2649	1.2459	76	0.4914	68.2028	21.1403
0.2413	1.2623	77	0.4887	71.6129	22.3258
0.2593	1.2787	78	0.4866	72.2581	22.2059
0.2202	1.2951	79	0.4857	69.1244	20.6341
0.2029	1.3115	80	0.4857	69.7696	20.6208
0.227	1.3279	81	0.4853	69.7696	20.5541
0.2448	1.3443	82	0.4845	69.4009	20.5142
0.2366	1.3607	83	0.4832	68.8479	20.4209
0.2253	1.3770	84	0.4822	69.5853	20.7140
0.2132	1.3934	85	0.4809	68.9401	20.2611
0.2218	1.4098	86	0.4795	73.4562	22.0727
0.2257	1.4262	87	0.4785	73.4562	21.9928
0.2143	1.4426	88	0.4777	67.5576	20.1012
0.2412	1.4590	89	0.4775	68.2949	20.4343
0.2081	1.4754	90	0.4762	67.2811	20.1146
0.2528	1.4918	91	0.4748	67.2811	20.0879
0.2184	1.5082	92	0.4740	67.2811	20.1012
0.1486	1.5246	93	0.4727	67.3733	20.2744
0.2317	1.5410	94	0.4724	66.5438	19.9814
0.2182	1.5574	95	0.4712	66.6359	20.1279
0.2537	1.5738	96	0.4705	66.8203	19.9414
0.2453	1.5902	97	0.4703	66.6359	20.0480
0.2181	1.6066	98	0.4697	67.1889	20.0613
0.1844	1.6230	99	0.4697	67.1889	19.9680
0.192	1.6393	100	0.4698	67.0968	19.9281
0.1787	1.6557	101	0.4697	66.9124	19.8748
0.2061	1.6721	102	0.4700	66.6359	19.9147
0.2218	1.6885	103	0.4703	66.6359	19.9814
0.2354	1.7049	104	0.4705	67.1889	20.1146
0.2686	1.7213	105	0.4706	66.9124	20.0480
0.1979	1.7377	106	0.4710	66.8203	20.0480
0.2075	1.7541	107	0.4712	66.9124	20.1146
0.1811	1.7705	108	0.4716	66.6359	20.1012
0.1903	1.7869	109	0.4722	66.4516	20.0480
0.2148	1.8033	110	0.4723	66.3594	20.0213
0.2501	1.8197	111	0.4722	66.4516	20.0480
0.185	1.8361	112	0.4723	66.2673	20.1012
0.1836	1.8525	113	0.4722	66.3594	20.0346
0.187	1.8689	114	0.4722	66.8203	20.2211
0.2457	1.8852	115	0.4721	66.4516	20.0613
0.2309	1.9016	116	0.4720	66.5438	20.1678
0.2469	1.9180	117	0.4717	66.2673	20.0746
0.2438	1.9344	118	0.4718	66.5438	20.0746
0.2073	1.9508	119	0.4720	66.8203	20.0746
0.1977	1.9672	120	0.4718	66.6359	20.0746
0.2149	1.9836	121	0.4717	66.6359	20.1279
0.2062	2.0	122	0.4721	66.6359	20.0746

Framework versions

• Transformers 4.57.3
• Pytorch 2.9.1+cu128
• Datasets 4.4.1
• Tokenizers 0.22.1