nkkbr/whisper-large-v3-zatoichi-ja-JDG_ver_20260212_lr_2.6e-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: 0.4773
• 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: 2.6e-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	0.9922	95.5760	38.0045
0.9571	0.0492	3	0.9386	91.1521	36.0464
0.816	0.0656	4	0.8551	88.3871	32.6229
0.7837	0.0820	5	0.7866	93.4562	33.7552
0.7219	0.0984	6	0.7451	89.6774	31.9702
0.75	0.1148	7	0.7113	96.4055	34.9540
0.7531	0.1311	8	0.7032	102.6728	42.7201
0.5658	0.1475	9	0.6799	91.6129	30.3850
0.5375	0.1639	10	0.6590	101.3825	34.3546
0.6928	0.1803	11	0.6400	82.4885	30.4516
0.6639	0.1967	12	0.6328	86.3594	29.6923
0.6957	0.2131	13	0.6191	87.1889	29.3060
0.6033	0.2295	14	0.6152	86.5438	30.2384
0.546	0.2459	15	0.6178	79.4470	26.9881
0.5733	0.2623	16	0.6120	84.6083	35.9398
0.4664	0.2787	17	0.6069	87.5576	48.6879
0.486	0.2951	18	0.6005	89.4931	40.6687
0.5403	0.3115	19	0.6013	81.2903	29.3060
0.4509	0.3279	20	0.5989	81.3825	28.2803
0.5292	0.3443	21	0.5897	80.5530	28.5067
0.521	0.3607	22	0.5869	76.1290	26.5486
0.5514	0.3770	23	0.5952	74.6544	24.5771
0.536	0.3934	24	0.5856	73.3641	23.7512
0.5473	0.4098	25	0.5772	74.9309	24.0176
0.4493	0.4262	26	0.5712	73.3641	23.7378
0.4029	0.4426	27	0.5729	76.0369	25.5495
0.4914	0.4590	28	0.5688	72.9032	25.2964
0.4604	0.4754	29	0.5688	72.4424	24.0176
0.4342	0.4918	30	0.5661	73.3641	23.8444
0.5586	0.5082	31	0.5571	74.1014	23.9110
0.5055	0.5246	32	0.5507	73.0876	23.4448
0.4227	0.5410	33	0.5495	74.1935	23.3382
0.5744	0.5574	34	0.5451	74.6544	23.8311
0.4773	0.5738	35	0.5430	74.0092	23.6446
0.4552	0.5902	36	0.5383	73.4562	23.7112
0.4579	0.6066	37	0.5311	71.4286	22.8187
0.5264	0.6230	38	0.5274	70.6912	22.4990
0.464	0.6393	39	0.5249	69.8618	22.4857
0.5331	0.6557	40	0.5264	70.1382	22.2992
0.4498	0.6721	41	0.5247	69.7696	22.2326
0.4764	0.6885	42	0.5230	69.3088	22.0994
0.4523	0.7049	43	0.5228	69.4931	22.2326
0.443	0.7213	44	0.5207	70.5069	22.7388
0.4035	0.7377	45	0.5161	70.9677	22.8054
0.4433	0.7541	46	0.5129	68.7558	22.4990
0.4827	0.7705	47	0.5135	69.6774	22.9786
0.4771	0.7869	48	0.5165	73.7327	24.2840
0.3967	0.8033	49	0.5161	68.6636	22.3392
0.4987	0.8197	50	0.5138	69.2166	22.4590
0.4819	0.8361	51	0.5112	71.9816	23.5514
0.4081	0.8525	52	0.5073	71.1521	23.8577
0.5707	0.8689	53	0.5054	70.5991	22.2326
0.3475	0.8852	54	0.5042	75.1152	24.2707
0.4659	0.9016	55	0.5039	74.2857	24.0842
0.4509	0.9180	56	0.5058	73.9171	24.1108
0.3794	0.9344	57	0.5061	75.6682	25.2298
0.4051	0.9508	58	0.5029	75.9447	25.2031
0.4361	0.9672	59	0.4998	74.6544	25.6427
0.4501	0.9836	60	0.5010	70.5069	24.2041
0.33	1.0	61	0.5060	81.9355	28.5334
0.3195	1.0164	62	0.5050	79.7235	26.5352
0.2317	1.0328	63	0.5024	70.4147	23.0052
0.1693	1.0492	64	0.5108	69.4009	21.9928
0.2429	1.0656	65	0.5185	70.4147	21.9395
0.2035	1.0820	66	0.5189	70.4147	22.0194
0.2233	1.0984	67	0.5178	68.9401	21.7530
0.2641	1.1148	68	0.5137	69.9539	22.1260
0.2858	1.1311	69	0.5117	71.0599	22.2193
0.2929	1.1475	70	0.5077	71.5207	22.0461
0.2392	1.1639	71	0.5077	66.8203	21.1003
0.1883	1.1803	72	0.5115	68.2028	21.5932
0.2268	1.1967	73	0.5142	71.6129	22.9919
0.2293	1.2131	74	0.5103	71.4286	22.8986
0.1939	1.2295	75	0.5047	66.5438	20.7673
0.2568	1.2459	76	0.5011	67.9263	20.7406
0.23	1.2623	77	0.4984	68.2949	20.7007
0.2496	1.2787	78	0.4963	68.6636	20.7273
0.2092	1.2951	79	0.4956	68.0184	20.5808
0.1916	1.3115	80	0.4954	68.4793	20.8339
0.2179	1.3279	81	0.4948	68.8479	20.7939
0.2318	1.3443	82	0.4940	69.1244	20.5941
0.2306	1.3607	83	0.4918	68.2028	20.4609
0.2227	1.3770	84	0.4894	69.1244	20.7273
0.2002	1.3934	85	0.4873	68.3871	20.5541
0.2032	1.4098	86	0.4861	69.1244	20.6607
0.2127	1.4262	87	0.4846	70.1382	21.0204
0.2047	1.4426	88	0.4838	68.4793	21.1003
0.2278	1.4590	89	0.4835	68.0184	20.6740
0.1966	1.4754	90	0.4831	68.1106	20.6474
0.2435	1.4918	91	0.4817	67.8341	20.5408
0.2068	1.5082	92	0.4807	68.1106	20.7007
0.1343	1.5246	93	0.4803	67.5576	20.6740
0.2137	1.5410	94	0.4798	67.2811	20.6874
0.2063	1.5574	95	0.4791	67.2811	20.4609
0.2417	1.5738	96	0.4785	66.8203	20.2877
0.2355	1.5902	97	0.4780	66.6359	20.4742
0.2041	1.6066	98	0.4778	66.6359	20.3277
0.1724	1.6230	99	0.4771	66.1751	20.2078
0.1839	1.6393	100	0.4769	66.9124	20.3011
0.1705	1.6557	101	0.4772	66.6359	20.1012
0.1946	1.6721	102	0.4769	66.6359	20.1812
0.2056	1.6885	103	0.4766	66.9124	20.2611
0.221	1.7049	104	0.4767	65.5300	20.1146
0.2623	1.7213	105	0.4772	65.2535	20.1279
0.1801	1.7377	106	0.4775	65.4378	20.2078
0.1978	1.7541	107	0.4781	65.7143	20.2877
0.1754	1.7705	108	0.4782	66.2673	20.4476
0.1869	1.7869	109	0.4786	66.3594	20.4609
0.1933	1.8033	110	0.4786	65.9908	20.2877
0.2287	1.8197	111	0.4786	66.1751	20.3943
0.173	1.8361	112	0.4784	66.1751	20.2744
0.1718	1.8525	113	0.4780	66.0829	20.3810
0.1739	1.8689	114	0.4778	66.2673	20.3277
0.2266	1.8852	115	0.4781	65.9908	20.3410
0.229	1.9016	116	0.4775	65.5300	20.1945
0.23	1.9180	117	0.4776	65.9908	20.3410
0.2291	1.9344	118	0.4775	66.0829	20.3277
0.1967	1.9508	119	0.4774	65.8065	20.3144
0.1857	1.9672	120	0.4774	65.9908	20.2744
0.1971	1.9836	121	0.4773	65.5300	20.3011
0.1941	2.0	122	0.4773	65.9908	20.2478

Framework versions

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