nkkbr/whisper-large-v3-zatoichi-ja-JDG_ver_20260212_lr_8e-6

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: 0.4840
• Wer: 66.7281
• Cer: 20.3144

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: 8e-06
• 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	1.0001	98.2488	39.6164
0.9648	0.0492	3	0.9802	95.9447	37.9113
0.8521	0.0656	4	0.9561	94.6544	37.0854
0.8758	0.0820	5	0.9179	89.4931	35.3270
0.8352	0.0984	6	0.8658	84.7926	31.3441
0.857	0.1148	7	0.8329	86.7281	31.7570
0.8774	0.1311	8	0.7928	84.7005	28.8397
0.6609	0.1475	9	0.7602	90.5991	36.7524
0.611	0.1639	10	0.7362	94.1935	40.7886
0.7666	0.1803	11	0.7158	93.0876	35.6467
0.7385	0.1967	12	0.7011	106.9124	43.2530
0.7515	0.2131	13	0.6730	96.9585	35.2205
0.6566	0.2295	14	0.6519	106.3594	36.7390
0.5967	0.2459	15	0.6388	108.3871	36.2728
0.581	0.2623	16	0.6267	90.5991	29.3992
0.4996	0.2787	17	0.6219	75.0230	24.8435
0.4923	0.2951	18	0.6160	81.1982	27.1880
0.5767	0.3115	19	0.6069	84.7926	27.9073
0.4646	0.3279	20	0.6025	84.7005	27.7741
0.5232	0.3443	21	0.5975	85.0691	28.4801
0.5417	0.3607	22	0.5876	82.9493	27.1480
0.5517	0.3770	23	0.5863	81.7512	27.0281
0.5448	0.3934	24	0.5855	79.7235	27.9606
0.5594	0.4098	25	0.5812	81.9355	28.0538
0.474	0.4262	26	0.5736	80.0922	28.4268
0.4132	0.4426	27	0.5690	72.6267	25.0033
0.4865	0.4590	28	0.5642	72.9954	22.9786
0.4551	0.4754	29	0.5640	73.1797	23.2583
0.4381	0.4918	30	0.5587	77.9724	24.8568
0.5665	0.5082	31	0.5534	74.1014	23.3515
0.5163	0.5246	32	0.5512	77.5115	25.3763
0.4361	0.5410	33	0.5502	76.8664	25.1099
0.5875	0.5574	34	0.5465	78.2488	25.2431
0.4768	0.5738	35	0.5430	78.0645	25.8159
0.4679	0.5902	36	0.5388	77.1429	25.8825
0.4451	0.6066	37	0.5336	76.6820	24.8968
0.5174	0.6230	38	0.5318	77.9724	25.7759
0.482	0.6393	39	0.5300	82.0276	27.1746
0.5346	0.6557	40	0.5281	78.1567	26.3887
0.438	0.6721	41	0.5268	78.4332	25.7759
0.4897	0.6885	42	0.5236	74.1935	23.8844
0.4632	0.7049	43	0.5203	72.3502	22.4191
0.4604	0.7213	44	0.5170	71.4286	21.9795
0.4033	0.7377	45	0.5136	71.1521	22.6988
0.4472	0.7541	46	0.5114	71.1521	21.9662
0.5045	0.7705	47	0.5098	70.5991	21.9262
0.4855	0.7869	48	0.5077	69.6774	22.1393
0.42	0.8033	49	0.5056	70.1382	22.1793
0.5175	0.8197	50	0.5048	70.5991	22.2592
0.4794	0.8361	51	0.5041	70.5991	21.9928
0.4024	0.8525	52	0.5012	69.6774	21.8729
0.5713	0.8689	53	0.4982	69.7696	21.9662
0.3572	0.8852	54	0.4965	69.6774	21.8862
0.4756	0.9016	55	0.4952	70.1382	22.0061
0.449	0.9180	56	0.4952	70.2304	22.2725
0.3502	0.9344	57	0.4957	69.7696	22.4857
0.4227	0.9508	58	0.4946	69.6774	22.2592
0.4279	0.9672	59	0.4927	69.4009	22.3392
0.4525	0.9836	60	0.4919	69.4009	22.1926
0.3604	1.0	61	0.4931	69.6774	22.0328
0.3922	1.0164	62	0.4930	68.8479	21.6465
0.3027	1.0328	63	0.4921	69.0323	22.3658
0.2272	1.0492	64	0.4919	68.9401	22.5123
0.3115	1.0656	65	0.4929	68.0184	21.4200
0.2675	1.0820	66	0.4949	67.2811	21.1136
0.2775	1.0984	67	0.4981	67.4654	21.3001
0.3321	1.1148	68	0.4965	67.7419	21.3268
0.3657	1.1311	69	0.4957	68.0184	21.6465
0.3738	1.1475	70	0.4945	68.2028	21.4733
0.3074	1.1639	71	0.4937	68.0184	21.4200
0.248	1.1803	72	0.4937	67.8341	21.3268
0.3109	1.1967	73	0.4945	67.8341	21.2602
0.2877	1.2131	74	0.4947	67.9263	21.2335
0.2716	1.2295	75	0.4951	67.8341	21.1536
0.3274	1.2459	76	0.4950	68.3871	21.1536
0.3167	1.2623	77	0.4939	68.0184	21.1403
0.3253	1.2787	78	0.4932	67.9263	20.9405
0.2885	1.2951	79	0.4928	68.2028	21.0603
0.2749	1.3115	80	0.4926	68.6636	21.0204
0.2848	1.3279	81	0.4923	67.9263	20.8339
0.317	1.3443	82	0.4915	68.2028	20.7540
0.3156	1.3607	83	0.4910	72.6267	22.2193
0.301	1.3770	84	0.4897	72.8111	22.1660
0.2662	1.3934	85	0.4891	73.3641	22.2326
0.2843	1.4098	86	0.4879	73.1797	22.1260
0.2849	1.4262	87	0.4882	73.3641	22.2592
0.2793	1.4426	88	0.4880	72.9032	22.1127
0.3007	1.4590	89	0.4880	72.8111	22.2193
0.275	1.4754	90	0.4876	71.4286	22.0061
0.3088	1.4918	91	0.4877	72.0737	22.0594
0.274	1.5082	92	0.4876	72.3502	22.0727
0.1979	1.5246	93	0.4878	68.2949	20.5941
0.2966	1.5410	94	0.4880	67.9263	20.6474
0.2869	1.5574	95	0.4880	68.0184	20.8206
0.3406	1.5738	96	0.4876	68.2028	20.8339
0.3157	1.5902	97	0.4872	67.5576	20.6874
0.2947	1.6066	98	0.4864	67.4654	20.6474
0.2349	1.6230	99	0.4854	67.5576	20.5808
0.2613	1.6393	100	0.4848	67.6498	20.6874
0.2381	1.6557	101	0.4846	67.8341	20.5941
0.2716	1.6721	102	0.4844	68.2949	20.5941
0.2941	1.6885	103	0.4843	67.8341	20.5408
0.3053	1.7049	104	0.4841	67.8341	20.5275
0.3454	1.7213	105	0.4840	67.6498	20.4742
0.2696	1.7377	106	0.4838	67.9263	20.3810
0.2713	1.7541	107	0.4839	67.6498	20.4476
0.2466	1.7705	108	0.4839	67.7419	20.5408
0.26	1.7869	109	0.4840	67.0968	20.3277
0.286	1.8033	110	0.4840	67.0046	20.4609
0.325	1.8197	111	0.4845	67.5576	20.5275
0.2416	1.8361	112	0.4839	67.3733	20.6341
0.2441	1.8525	113	0.4841	67.4654	20.4875
0.25	1.8689	114	0.4840	67.2811	20.4875
0.3176	1.8852	115	0.4841	67.1889	20.5541
0.2874	1.9016	116	0.4839	66.9124	20.4742
0.2994	1.9180	117	0.4838	67.2811	20.4476
0.3202	1.9344	118	0.4838	67.1889	20.5009
0.2649	1.9508	119	0.4839	67.4654	20.4875
0.2685	1.9672	120	0.4839	67.1889	20.4476
0.2794	1.9836	121	0.4837	67.6498	20.5009
0.2618	2.0	122	0.4840	66.7281	20.3144

Framework versions

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