nkkbr/whisper-large-v3-zatoichi-ja-JDG_ver_20260212_lr_6e-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: 0.4924
• Wer: 68.4793
• Cer: 20.8072

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: 6e-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	1.0002	96.7742	38.6573
0.9654	0.0492	3	0.9812	97.6037	39.1901
0.8532	0.0656	4	0.9597	95.3917	37.6315
0.879	0.0820	5	0.9297	92.2581	36.1796
0.8458	0.0984	6	0.9093	89.0323	35.0606
0.8937	0.1148	7	0.8571	84.2396	31.2775
0.8999	0.1311	8	0.8307	82.5806	29.9853
0.6964	0.1475	9	0.7972	85.4378	29.8388
0.6418	0.1639	10	0.7632	84.9770	29.7322
0.7961	0.1803	11	0.7405	93.9171	38.2310
0.7758	0.1967	12	0.7234	94.7465	36.0996
0.781	0.2131	13	0.7070	97.5115	39.6563
0.684	0.2295	14	0.6849	102.3963	40.4955
0.6273	0.2459	15	0.6644	92.3502	30.6114
0.6014	0.2623	16	0.6511	99.9078	32.5962
0.5211	0.2787	17	0.6413	105.2535	35.7533
0.5164	0.2951	18	0.6318	94.1014	31.7837
0.584	0.3115	19	0.6205	94.4700	31.9702
0.4791	0.3279	20	0.6125	100.9217	35.2071
0.5292	0.3443	21	0.6065	88.9401	28.3069
0.546	0.3607	22	0.6022	82.8571	27.1214
0.5594	0.3770	23	0.5996	80.8295	26.1889
0.5623	0.3934	24	0.5950	73.7327	24.1108
0.5725	0.4098	25	0.5912	78.4332	25.7360
0.4956	0.4262	26	0.5853	77.8802	26.3487
0.4311	0.4426	27	0.5810	78.5253	26.3088
0.4967	0.4590	28	0.5774	83.9631	29.6390
0.4686	0.4754	29	0.5751	84.3318	29.8921
0.4491	0.4918	30	0.5700	82.6728	28.1737
0.5749	0.5082	31	0.5648	77.9724	26.0557
0.5251	0.5246	32	0.5624	77.9724	26.0024
0.4511	0.5410	33	0.5608	77.3272	25.1632
0.6026	0.5574	34	0.5584	77.2350	25.1765
0.4902	0.5738	35	0.5560	77.3272	25.6161
0.4794	0.5902	36	0.5519	78.1567	26.7883
0.4549	0.6066	37	0.5468	71.9816	23.2716
0.5304	0.6230	38	0.5454	71.9816	23.3515
0.4953	0.6393	39	0.5442	77.9724	26.6551
0.5459	0.6557	40	0.5413	78.3410	26.2155
0.4451	0.6721	41	0.5382	78.5253	24.8834
0.4988	0.6885	42	0.5350	77.8802	24.8302
0.4762	0.7049	43	0.5321	78.1567	24.8302
0.4704	0.7213	44	0.5290	77.8802	24.8435
0.4159	0.7377	45	0.5258	72.0737	22.5789
0.4568	0.7541	46	0.5233	70.8756	22.4724
0.5148	0.7705	47	0.5230	70.3226	22.0328
0.4968	0.7869	48	0.5211	70.8756	22.2059
0.4344	0.8033	49	0.5193	70.3226	22.3924
0.5264	0.8197	50	0.5186	70.5069	23.2450
0.4916	0.8361	51	0.5175	70.5069	22.2992
0.4111	0.8525	52	0.5143	70.5069	22.1393
0.576	0.8689	53	0.5112	69.4931	21.8596
0.3659	0.8852	54	0.5090	70.4147	21.9129
0.484	0.9016	55	0.5070	71.4286	23.0452
0.4585	0.9180	56	0.5071	71.2442	23.1784
0.3552	0.9344	57	0.5066	69.8618	23.0318
0.4321	0.9508	58	0.5058	69.9539	22.8720
0.4303	0.9672	59	0.5040	70.1382	23.0185
0.4616	0.9836	60	0.5029	70.7834	23.0318
0.3698	1.0	61	0.5025	70.5069	22.7787
0.426	1.0164	62	0.5022	70.5991	23.6979
0.3282	1.0328	63	0.5018	70.0461	23.4981
0.2497	1.0492	64	0.5018	69.8618	23.7245
0.3411	1.0656	65	0.5026	68.5714	23.2583
0.2908	1.0820	66	0.5035	68.6636	22.5123
0.301	1.0984	67	0.5061	68.1106	22.2992
0.3594	1.1148	68	0.5063	68.4793	22.2326
0.3926	1.1311	69	0.5060	68.7558	22.0727
0.3992	1.1475	70	0.5047	68.4793	21.9395
0.3345	1.1639	71	0.5031	67.7419	22.7787
0.2743	1.1803	72	0.5019	67.5576	21.2069
0.3456	1.1967	73	0.5015	66.8203	20.9538
0.3097	1.2131	74	0.5008	67.3733	21.0204
0.299	1.2295	75	0.5012	67.8341	21.0603
0.3532	1.2459	76	0.5005	68.2028	21.0870
0.3423	1.2623	77	0.5000	67.8341	21.0737
0.3539	1.2787	78	0.4994	68.1106	21.1403
0.3155	1.2951	79	0.4997	68.4793	21.1536
0.3026	1.3115	80	0.5005	68.5714	21.2335
0.3068	1.3279	81	0.5000	69.0323	21.3401
0.3462	1.3443	82	0.5001	68.9401	21.3800
0.3479	1.3607	83	0.4992	68.6636	21.0737
0.327	1.3770	84	0.4984	68.8479	21.1536
0.2866	1.3934	85	0.4980	69.9539	21.2069
0.3091	1.4098	86	0.4972	69.8618	21.3268
0.3094	1.4262	87	0.4968	69.8618	21.0737
0.3051	1.4426	88	0.4961	69.5853	21.0870
0.3243	1.4590	89	0.4960	69.5853	21.1403
0.2997	1.4754	90	0.4956	69.2166	20.9405
0.3317	1.4918	91	0.4957	69.4931	21.0071
0.2967	1.5082	92	0.4958	69.4931	20.9538
0.2181	1.5246	93	0.4963	73.1797	22.6855
0.3186	1.5410	94	0.4963	72.8111	22.5789
0.3121	1.5574	95	0.4963	69.0323	21.1003
0.372	1.5738	96	0.4962	68.5714	21.0204
0.3427	1.5902	97	0.4959	68.3871	21.0470
0.3271	1.6066	98	0.4953	68.5714	20.9804
0.2564	1.6230	99	0.4949	68.7558	20.9671
0.2894	1.6393	100	0.4937	68.6636	20.8605
0.2655	1.6557	101	0.4935	68.8479	20.9005
0.2964	1.6721	102	0.4932	68.7558	20.7406
0.3223	1.6885	103	0.4930	69.1244	20.8339
0.3305	1.7049	104	0.4929	68.9401	20.8872
0.3763	1.7213	105	0.4929	68.6636	20.8339
0.3013	1.7377	106	0.4927	68.5714	20.7273
0.2986	1.7541	107	0.4927	68.8479	20.8072
0.2743	1.7705	108	0.4926	69.1244	20.9538
0.2885	1.7869	109	0.4924	68.4793	20.9405
0.3161	1.8033	110	0.4929	68.2949	20.8072
0.3606	1.8197	111	0.4926	67.9263	20.8472
0.2606	1.8361	112	0.4930	67.7419	20.7007
0.2654	1.8525	113	0.4923	68.2949	20.8206
0.2771	1.8689	114	0.4926	68.1106	20.7939
0.3455	1.8852	115	0.4929	68.3871	20.8339
0.3141	1.9016	116	0.4927	68.0184	20.7406
0.3246	1.9180	117	0.4926	68.1106	20.7540
0.3521	1.9344	118	0.4922	68.2028	20.8206
0.2872	1.9508	119	0.4925	68.0184	20.7140
0.2952	1.9672	120	0.4923	67.9263	20.6874
0.3049	1.9836	121	0.4925	68.2949	20.8072
0.2862	2.0	122	0.4924	68.4793	20.8072

Framework versions

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