nkkbr/whisper-large-v3-zatoichi-ja-JDG_ver_20260212_lr_2.4e-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: 0.4770
• Wer: 65.9908
• Cer: 20.2478

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: 2.4e-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	0.9971	96.1290	38.2709
0.9617	0.0492	3	0.9421	91.7051	36.3261
0.819	0.0656	4	0.8619	88.5714	33.0225
0.7894	0.0820	5	0.7940	88.9401	31.2642
0.7276	0.0984	6	0.7509	98.1567	35.1272
0.7551	0.1148	7	0.7176	91.8894	36.0730
0.7604	0.1311	8	0.7011	107.6498	43.8524
0.5633	0.1475	9	0.6788	91.8894	33.2889
0.5365	0.1639	10	0.6578	101.8433	34.6876
0.6899	0.1803	11	0.6396	79.9078	28.6266
0.6611	0.1967	12	0.6289	92.1659	31.2642
0.6896	0.2131	13	0.6173	79.3548	25.5894
0.6012	0.2295	14	0.6132	77.8802	25.2831
0.545	0.2459	15	0.6145	83.5023	29.5058
0.5696	0.2623	16	0.6085	88.8479	38.4175
0.464	0.2787	17	0.6038	88.8479	46.2102
0.4828	0.2951	18	0.5961	87.0046	35.5535
0.5386	0.3115	19	0.5978	78.4332	24.9234
0.449	0.3279	20	0.5932	81.6590	26.5619
0.5242	0.3443	21	0.5830	82.3963	27.3878
0.5174	0.3607	22	0.5830	77.1429	27.1880
0.5472	0.3770	23	0.5928	74.3779	24.6969
0.5343	0.3934	24	0.5828	73.2719	23.7911
0.5413	0.4098	25	0.5718	78.7097	27.2013
0.4468	0.4262	26	0.5677	79.8157	27.3745
0.4047	0.4426	27	0.5691	71.7972	23.5913
0.4946	0.4590	28	0.5616	71.9816	24.8435
0.4561	0.4754	29	0.5658	72.2581	23.7778
0.4334	0.4918	30	0.5666	71.7051	23.7645
0.5598	0.5082	31	0.5551	71.8894	23.3649
0.502	0.5246	32	0.5475	72.1659	23.0984
0.4187	0.5410	33	0.5467	73.0876	23.3515
0.5715	0.5574	34	0.5423	74.1014	23.0052
0.4789	0.5738	35	0.5384	73.6406	23.1650
0.4525	0.5902	36	0.5350	72.0737	22.7255
0.4557	0.6066	37	0.5297	72.0737	22.8320
0.5211	0.6230	38	0.5256	71.0599	22.6855
0.4644	0.6393	39	0.5207	69.2166	22.2992
0.5284	0.6557	40	0.5190	69.5853	22.2193
0.4439	0.6721	41	0.5163	69.6774	22.2992
0.4751	0.6885	42	0.5151	69.4009	21.9795
0.446	0.7049	43	0.5147	68.9401	22.0061
0.4395	0.7213	44	0.5123	69.6774	22.0594
0.3977	0.7377	45	0.5082	69.9539	22.2725
0.4395	0.7541	46	0.5067	68.9401	22.4324
0.4774	0.7705	47	0.5078	69.6774	22.4724
0.4766	0.7869	48	0.5108	69.4931	22.7654
0.4034	0.8033	49	0.5104	68.3871	22.6589
0.4986	0.8197	50	0.5096	71.5207	23.5913
0.4802	0.8361	51	0.5070	73.0876	24.8435
0.405	0.8525	52	0.5035	69.1244	22.3658
0.5756	0.8689	53	0.5016	69.4009	22.2326
0.3444	0.8852	54	0.4990	70.0461	22.2592
0.4663	0.9016	55	0.4974	75.2074	24.0975
0.4439	0.9180	56	0.4990	73.6406	23.9377
0.3734	0.9344	57	0.5000	75.8525	25.1898
0.4051	0.9508	58	0.4959	75.3917	25.0433
0.4348	0.9672	59	0.4924	73.7327	24.6303
0.4501	0.9836	60	0.4922	74.0092	23.8311
0.3364	1.0	61	0.4957	73.2719	22.9919
0.3188	1.0164	62	0.4951	77.7880	24.9500
0.2346	1.0328	63	0.4938	70.7834	23.2716
0.1708	1.0492	64	0.5012	73.9171	24.3506
0.2457	1.0656	65	0.5077	69.2166	22.0328
0.2044	1.0820	66	0.5078	72.1659	23.1118
0.2232	1.0984	67	0.5047	71.6129	22.7654
0.263	1.1148	68	0.5030	67.5576	20.8872
0.2868	1.1311	69	0.5017	71.2442	21.8196
0.2978	1.1475	70	0.4991	72.5346	22.6589
0.2437	1.1639	71	0.4988	67.1889	21.1269
0.1885	1.1803	72	0.5017	71.5207	22.7388
0.2228	1.1967	73	0.5043	71.2442	22.5789
0.228	1.2131	74	0.5029	71.2442	22.6855
0.1948	1.2295	75	0.4996	71.0599	22.2859
0.2618	1.2459	76	0.4972	71.3364	22.0194
0.2326	1.2623	77	0.4949	68.1106	20.5675
0.2485	1.2787	78	0.4919	68.3871	20.4343
0.2135	1.2951	79	0.4907	67.0968	20.2344
0.196	1.3115	80	0.4896	67.7419	20.4476
0.2207	1.3279	81	0.4893	67.8341	20.4076
0.235	1.3443	82	0.4887	68.3871	20.4076
0.2279	1.3607	83	0.4872	68.3871	20.3543
0.2252	1.3770	84	0.4849	68.8479	20.4476
0.2054	1.3934	85	0.4829	68.2028	20.5142
0.2042	1.4098	86	0.4817	68.9401	20.5009
0.2148	1.4262	87	0.4804	73.4562	22.2725
0.2079	1.4426	88	0.4798	72.4424	22.3258
0.2288	1.4590	89	0.4796	67.6498	20.8472
0.1996	1.4754	90	0.4792	68.3871	20.7806
0.2454	1.4918	91	0.4781	68.3871	20.8605
0.2094	1.5082	92	0.4774	68.1106	20.7406
0.1352	1.5246	93	0.4765	67.3733	20.7806
0.2189	1.5410	94	0.4763	67.1889	20.8072
0.2046	1.5574	95	0.4759	66.8203	20.5941
0.2421	1.5738	96	0.4757	67.5576	20.7806
0.2359	1.5902	97	0.4751	67.3733	20.6208
0.2085	1.6066	98	0.4752	67.0046	20.3677
0.1719	1.6230	99	0.4753	66.8203	20.2478
0.1851	1.6393	100	0.4753	67.0968	20.3410
0.1732	1.6557	101	0.4751	67.0968	20.5009
0.1959	1.6721	102	0.4757	67.2811	20.7007
0.2079	1.6885	103	0.4757	66.9124	20.6074
0.2229	1.7049	104	0.4762	66.6359	20.2877
0.26	1.7213	105	0.4767	66.6359	20.3543
0.1832	1.7377	106	0.4767	66.6359	20.3943
0.1988	1.7541	107	0.4773	65.8065	20.2744
0.1756	1.7705	108	0.4773	66.3594	20.3543
0.1849	1.7869	109	0.4777	66.1751	20.1945
0.1949	1.8033	110	0.4783	65.8986	20.2478
0.2346	1.8197	111	0.4785	65.6221	20.1678
0.1759	1.8361	112	0.4782	65.5300	20.1545
0.1743	1.8525	113	0.4777	65.8065	20.2478
0.1757	1.8689	114	0.4777	65.9908	20.2478
0.232	1.8852	115	0.4772	65.9908	20.2478
0.2329	1.9016	116	0.4774	66.4516	20.3543
0.2334	1.9180	117	0.4770	66.2673	20.3410
0.2299	1.9344	118	0.4770	66.3594	20.3410
0.2007	1.9508	119	0.4771	65.8986	20.2744
0.1906	1.9672	120	0.4770	66.1751	20.3011
0.2036	1.9836	121	0.4770	66.3594	20.4209
0.1983	2.0	122	0.4770	65.9908	20.2478

Framework versions

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