nkkbr/whisper-large-v3-zatoichi-ja-JDG_ver_20260212_lr_1.2e-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: 0.4759
• Wer: 66.5438
• Cer: 19.8615

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: 1.2e-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	0.9999	101.0138	40.1226
0.9643	0.0492	3	0.9715	94.5622	37.7781
0.8445	0.0656	4	0.9281	91.5207	35.9798
0.8506	0.0820	5	0.8691	85.6221	31.7437
0.7927	0.0984	6	0.8330	86.5438	31.6904
0.8313	0.1148	7	0.7765	88.8479	31.3974
0.8246	0.1311	8	0.7492	94.0092	38.7905
0.6164	0.1475	9	0.7237	89.1244	33.8351
0.5764	0.1639	10	0.7022	89.1244	34.2081
0.7295	0.1803	11	0.6738	96.6820	34.8075
0.6899	0.1967	12	0.6534	96.4977	32.9293
0.7089	0.2131	13	0.6403	93.1797	31.9302
0.6281	0.2295	14	0.6259	95.1152	32.7428
0.5662	0.2459	15	0.6129	83.5945	27.6675
0.5531	0.2623	16	0.6100	78.5253	25.8958
0.4798	0.2787	17	0.6074	82.5806	27.3611
0.4798	0.2951	18	0.5998	82.0276	26.9215
0.5592	0.3115	19	0.5916	79.2627	25.6427
0.4481	0.3279	20	0.5890	83.2258	26.7883
0.518	0.3443	21	0.5819	80.2765	27.4011
0.5234	0.3607	22	0.5754	78.3410	28.5334
0.5424	0.3770	23	0.5791	77.3272	27.7341
0.5292	0.3934	24	0.5755	75.5760	26.6152
0.5394	0.4098	25	0.5709	72.6267	24.3240
0.4536	0.4262	26	0.5625	72.2581	23.0318
0.3991	0.4426	27	0.5578	72.5346	22.9786
0.4798	0.4590	28	0.5537	73.8249	23.5114
0.4492	0.4754	29	0.5590	74.7465	24.6437
0.4362	0.4918	30	0.5530	73.6406	23.9243
0.5548	0.5082	31	0.5452	72.9032	24.1508
0.5036	0.5246	32	0.5448	72.5346	23.2450
0.422	0.5410	33	0.5448	72.1659	22.9386
0.5673	0.5574	34	0.5414	72.9032	22.9652
0.4702	0.5738	35	0.5368	72.9954	22.8054
0.4552	0.5902	36	0.5319	73.6406	22.8720
0.4454	0.6066	37	0.5271	79.9078	25.5628
0.5087	0.6230	38	0.5225	73.8249	23.0052
0.4716	0.6393	39	0.5179	83.4101	28.9064
0.528	0.6557	40	0.5163	70.5069	22.4724
0.4392	0.6721	41	0.5143	70.7834	21.8862
0.4823	0.6885	42	0.5112	70.1382	21.8063
0.4568	0.7049	43	0.5063	70.2304	21.7397
0.445	0.7213	44	0.5029	69.5853	21.8596
0.3939	0.7377	45	0.4996	70.6912	21.8996
0.4384	0.7541	46	0.4983	70.9677	21.7264
0.4906	0.7705	47	0.4973	70.8756	21.9528
0.4769	0.7869	48	0.4970	70.5069	21.8330
0.4062	0.8033	49	0.4967	70.9677	22.9519
0.5141	0.8197	50	0.4969	70.5069	23.1384
0.475	0.8361	51	0.4956	70.5069	22.8453
0.3942	0.8525	52	0.4928	70.5069	22.1393
0.5694	0.8689	53	0.4911	70.4147	22.1260
0.35	0.8852	54	0.4890	74.1014	23.7378
0.4661	0.9016	55	0.4875	73.5484	23.5780
0.4384	0.9180	56	0.4881	73.5484	23.6046
0.3531	0.9344	57	0.4907	72.9032	23.5780
0.4088	0.9508	58	0.4907	73.4562	24.4971
0.4314	0.9672	59	0.4877	72.9954	24.5771
0.4452	0.9836	60	0.4860	74.1935	24.9900
0.3466	1.0	61	0.4880	73.1797	23.3649
0.3563	1.0164	62	0.4887	73.7327	23.3515
0.2687	1.0328	63	0.4880	72.6267	23.1917
0.2011	1.0492	64	0.4885	72.3502	23.1650
0.2743	1.0656	65	0.4903	71.3364	22.9519
0.2373	1.0820	66	0.4922	71.5207	22.7255
0.2512	1.0984	67	0.4944	72.4424	23.0718
0.2986	1.1148	68	0.4927	67.4654	21.1269
0.3317	1.1311	69	0.4911	68.1106	21.5799
0.3394	1.1475	70	0.4880	67.5576	21.2868
0.2788	1.1639	71	0.4875	67.6498	21.1403
0.2189	1.1803	72	0.4883	68.1106	20.9538
0.266	1.1967	73	0.4898	68.4793	21.0737
0.2578	1.2131	74	0.4907	68.7558	21.1136
0.2328	1.2295	75	0.4914	73.4562	22.6589
0.2969	1.2459	76	0.4906	72.9954	22.8320
0.2743	1.2623	77	0.4892	72.5346	22.6988
0.2908	1.2787	78	0.4880	67.6498	20.7007
0.2577	1.2951	79	0.4873	68.2949	20.7540
0.2409	1.3115	80	0.4871	67.7419	20.6074
0.2543	1.3279	81	0.4862	67.9263	20.4343
0.2809	1.3443	82	0.4851	67.9263	20.4209
0.2794	1.3607	83	0.4840	67.7419	20.3277
0.2603	1.3770	84	0.4838	68.8479	20.4875
0.2374	1.3934	85	0.4826	69.1244	20.3943
0.2465	1.4098	86	0.4821	68.7558	20.2078
0.2538	1.4262	87	0.4822	69.4931	20.2211
0.2485	1.4426	88	0.4816	73.2719	21.7397
0.2711	1.4590	89	0.4817	72.5346	21.5399
0.2416	1.4754	90	0.4818	72.7189	21.6198
0.2819	1.4918	91	0.4810	68.7558	20.1012
0.2419	1.5082	92	0.4811	67.8341	19.9281
0.1723	1.5246	93	0.4804	67.7419	20.0080
0.2656	1.5410	94	0.4797	67.6498	19.9281
0.253	1.5574	95	0.4790	67.7419	20.0080
0.3	1.5738	96	0.4780	67.3733	20.0613
0.2811	1.5902	97	0.4777	67.3733	20.0746
0.2541	1.6066	98	0.4764	67.5576	20.0879
0.2056	1.6230	99	0.4758	67.3733	19.9814
0.2256	1.6393	100	0.4754	67.0046	20.2078
0.2091	1.6557	101	0.4749	67.0968	20.2078
0.24	1.6721	102	0.4752	67.0046	20.1279
0.2539	1.6885	103	0.4748	67.0046	20.0746
0.2731	1.7049	104	0.4750	67.2811	20.1545
0.3121	1.7213	105	0.4751	66.8203	20.0613
0.2305	1.7377	106	0.4750	66.4516	20.0080
0.2393	1.7541	107	0.4752	66.9124	20.1279
0.2114	1.7705	108	0.4753	67.0968	20.2478
0.223	1.7869	109	0.4758	67.0046	20.0879
0.2457	1.8033	110	0.4760	67.1889	20.1412
0.2815	1.8197	111	0.4757	66.8203	20.0480
0.2156	1.8361	112	0.4754	66.5438	19.9547
0.2149	1.8525	113	0.4758	66.9124	19.9680
0.2151	1.8689	114	0.4758	66.6359	19.8748
0.283	1.8852	115	0.4758	66.6359	19.8748
0.2471	1.9016	116	0.4757	67.0968	19.9414
0.2707	1.9180	117	0.4761	66.4516	19.8481
0.2785	1.9344	118	0.4759	66.9124	19.9414
0.2363	1.9508	119	0.4757	66.5438	19.8215
0.2343	1.9672	120	0.4758	66.7281	19.8615
0.246	1.9836	121	0.4754	67.0046	19.9814
0.2285	2.0	122	0.4759	66.5438	19.8615

Framework versions

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