nkkbr/whisper-large-v3-zatoichi-ja-JDG_ver_20260212_lr_1.6e-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: 0.4725
• Wer: 65.3456
• Cer: 20.0480

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: 1.6e-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	0.9978	100.8295	39.7762
0.9628	0.0492	3	0.9683	94.1935	37.2452
0.8419	0.0656	4	0.9179	88.8479	34.9940
0.8414	0.0820	5	0.8498	87.3733	32.4897
0.7754	0.0984	6	0.7923	90.4147	31.4640
0.7996	0.1148	7	0.7558	94.1014	38.7372
0.8085	0.1311	8	0.7284	88.8479	34.0749
0.5933	0.1475	9	0.7010	103.9631	38.3509
0.5553	0.1639	10	0.6865	101.8433	41.0550
0.7173	0.1803	11	0.6567	99.6313	33.1158
0.6767	0.1967	12	0.6382	94.3779	30.9045
0.6969	0.2131	13	0.6288	83.3180	27.4144
0.6143	0.2295	14	0.6184	89.8618	30.2917
0.5587	0.2459	15	0.6087	76.5899	24.4305
0.5589	0.2623	16	0.6075	84.7926	27.9472
0.4769	0.2787	17	0.5991	85.0691	28.8930
0.4749	0.2951	18	0.5933	80.6452	29.3859
0.5503	0.3115	19	0.5862	79.4470	29.7456
0.4433	0.3279	20	0.5826	80.8295	30.4649
0.5179	0.3443	21	0.5764	78.7097	27.5743
0.5142	0.3607	22	0.5709	74.7465	25.6694
0.5343	0.3770	23	0.5798	78.1567	26.7350
0.521	0.3934	24	0.5758	75.8525	25.6694
0.5391	0.4098	25	0.5662	71.9816	23.4981
0.4464	0.4262	26	0.5606	73.7327	24.6836
0.3951	0.4426	27	0.5586	74.2857	24.3906
0.4829	0.4590	28	0.5555	79.1705	25.8958
0.4526	0.4754	29	0.5603	72.2581	23.2050
0.435	0.4918	30	0.5533	72.4424	23.3782
0.5434	0.5082	31	0.5514	71.8894	23.5114
0.5069	0.5246	32	0.5529	73.0876	23.3116
0.4182	0.5410	33	0.5488	72.9032	23.0452
0.5657	0.5574	34	0.5420	74.3779	23.4981
0.4704	0.5738	35	0.5382	75.4839	23.4448
0.4541	0.5902	36	0.5338	80.2765	25.5362
0.4476	0.6066	37	0.5269	79.2627	25.4962
0.5102	0.6230	38	0.5214	72.8111	23.0185
0.4658	0.6393	39	0.5182	74.0092	25.1099
0.5239	0.6557	40	0.5191	73.4562	24.8435
0.4436	0.6721	41	0.5147	73.3641	24.5904
0.4811	0.6885	42	0.5100	74.4700	25.4562
0.4516	0.7049	43	0.5058	75.1152	25.4696
0.442	0.7213	44	0.5015	70.9677	22.0594
0.3938	0.7377	45	0.4989	71.7972	22.0461
0.4374	0.7541	46	0.4975	70.8756	21.9262
0.481	0.7705	47	0.4982	68.6636	21.7530
0.4753	0.7869	48	0.4994	69.5853	22.2592
0.3993	0.8033	49	0.4992	69.4931	21.9662
0.5051	0.8197	50	0.4995	69.9539	22.0194
0.4733	0.8361	51	0.4979	68.5714	21.9528
0.3906	0.8525	52	0.4960	68.9401	21.8196
0.5722	0.8689	53	0.4957	73.5484	23.3649
0.3469	0.8852	54	0.4932	72.8111	23.2183
0.4614	0.9016	55	0.4907	73.9171	23.3915
0.4409	0.9180	56	0.4925	72.6267	23.5114
0.3628	0.9344	57	0.4960	72.3502	23.4714
0.4173	0.9508	58	0.4947	71.8894	23.2183
0.4377	0.9672	59	0.4898	71.3364	23.0452
0.4469	0.9836	60	0.4859	71.6129	23.8844
0.3472	1.0	61	0.4851	73.7327	24.3906
0.3362	1.0164	62	0.4855	73.4562	23.1384
0.2497	1.0328	63	0.4854	68.4793	21.4067
0.1866	1.0492	64	0.4885	68.5714	22.5390
0.2637	1.0656	65	0.4899	67.1889	21.9928
0.2228	1.0820	66	0.4903	69.1244	22.1260
0.2375	1.0984	67	0.4897	65.0691	20.6874
0.275	1.1148	68	0.4884	66.8203	20.7939
0.3114	1.1311	69	0.4880	66.1751	20.6341
0.3197	1.1475	70	0.4862	66.5438	20.4742
0.2613	1.1639	71	0.4859	67.9263	20.7673
0.2022	1.1803	72	0.4865	67.5576	20.6474
0.2445	1.1967	73	0.4875	68.0184	20.7273
0.2407	1.2131	74	0.4878	67.5576	20.7273
0.2119	1.2295	75	0.4869	67.8341	20.6474
0.2705	1.2459	76	0.4865	67.7419	20.5275
0.2523	1.2623	77	0.4855	71.8894	21.7930
0.2707	1.2787	78	0.4841	72.7189	21.9262
0.2363	1.2951	79	0.4834	72.0737	21.7930
0.2173	1.3115	80	0.4827	71.7051	21.6598
0.2328	1.3279	81	0.4821	71.6129	21.6065
0.2602	1.3443	82	0.4818	71.6129	21.5665
0.2539	1.3607	83	0.4808	70.8756	21.3934
0.2401	1.3770	84	0.4802	71.3364	21.5399
0.2215	1.3934	85	0.4795	71.6129	21.4866
0.2261	1.4098	86	0.4794	72.6267	21.7530
0.2372	1.4262	87	0.4790	72.1659	21.6731
0.2301	1.4426	88	0.4787	71.7972	21.5266
0.2527	1.4590	89	0.4783	70.8756	21.3800
0.2241	1.4754	90	0.4778	65.9908	19.4219
0.265	1.4918	91	0.4771	66.1751	19.6084
0.2274	1.5082	92	0.4759	65.8065	19.5418
0.1545	1.5246	93	0.4748	66.3594	19.8348
0.2438	1.5410	94	0.4744	65.9908	19.6616
0.2323	1.5574	95	0.4735	66.5438	19.8481
0.276	1.5738	96	0.4727	66.1751	19.7682
0.2563	1.5902	97	0.4721	65.8065	19.9680
0.2298	1.6066	98	0.4715	65.8065	19.9414
0.1907	1.6230	99	0.4714	66.2673	20.0346
0.2068	1.6393	100	0.4711	66.4516	20.0080
0.1927	1.6557	101	0.4713	67.2811	20.1012
0.2202	1.6721	102	0.4714	67.0046	20.2211
0.2302	1.6885	103	0.4713	66.9124	20.2211
0.2505	1.7049	104	0.4717	66.7281	20.1146
0.2915	1.7213	105	0.4716	66.4516	20.1545
0.2092	1.7377	106	0.4721	66.2673	20.1279
0.2201	1.7541	107	0.4722	65.9908	20.0480
0.1894	1.7705	108	0.4725	65.9908	20.1812
0.2025	1.7869	109	0.4725	66.0829	20.1812
0.2206	1.8033	110	0.4732	65.5300	20.1945
0.2585	1.8197	111	0.4731	65.3456	19.9814
0.1979	1.8361	112	0.4730	65.3456	20.0346
0.1952	1.8525	113	0.4733	65.5300	20.1678
0.1948	1.8689	114	0.4729	65.3456	20.1012
0.2638	1.8852	115	0.4729	65.8986	20.0613
0.2348	1.9016	116	0.4731	65.7143	20.3011
0.2567	1.9180	117	0.4729	65.6221	20.0613
0.2572	1.9344	118	0.4729	65.5300	20.0879
0.2162	1.9508	119	0.4729	65.6221	20.1012
0.213	1.9672	120	0.4730	65.6221	20.1012
0.2259	1.9836	121	0.4727	65.7143	20.1678
0.2158	2.0	122	0.4725	65.3456	20.0480

Framework versions

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