nkkbr/whisper-large-v3-zatoichi-ja-JDG_ver_20260212_lr_1.8e-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: 0.4699
• Wer: 66.4516
• Cer: 20.0080

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: 1.8e-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	0.9975	98.9862	38.6040
0.9623	0.0492	3	0.9529	93.7327	36.9522
0.8283	0.0656	4	0.8875	89.9539	34.8475
0.8133	0.0820	5	0.8385	89.1244	32.8627
0.7654	0.0984	6	0.7780	90.3226	31.5439
0.7871	0.1148	7	0.7467	90.4147	32.1300
0.7974	0.1311	8	0.7176	91.8894	35.9931
0.5786	0.1475	9	0.7020	102.7650	38.8970
0.5542	0.1639	10	0.6792	104.0553	40.2424
0.7073	0.1803	11	0.6534	96.0369	31.8636
0.6749	0.1967	12	0.6350	102.8571	34.3546
0.698	0.2131	13	0.6320	90.2304	30.9311
0.6138	0.2295	14	0.6138	76.5899	25.0566
0.5518	0.2459	15	0.6133	77.2350	25.4163
0.5693	0.2623	16	0.6070	77.5115	25.1365
0.4768	0.2787	17	0.5961	83.2258	28.7731
0.4734	0.2951	18	0.5965	80.1843	39.4831
0.5552	0.3115	19	0.5867	82.9493	32.4231
0.4442	0.3279	20	0.5810	82.4885	28.9730
0.5199	0.3443	21	0.5754	75.0230	23.9510
0.5181	0.3607	22	0.5692	74.5622	25.4962
0.533	0.3770	23	0.5814	73.6406	27.0814
0.5214	0.3934	24	0.5788	74.6544	27.5609
0.5405	0.4098	25	0.5651	73.3641	24.1508
0.445	0.4262	26	0.5583	75.1152	23.9243
0.3923	0.4426	27	0.5602	75.0230	23.5380
0.4868	0.4590	28	0.5534	77.2350	24.5904
0.4493	0.4754	29	0.5584	71.1521	23.1784
0.4254	0.4918	30	0.5595	71.7051	23.4581
0.5473	0.5082	31	0.5574	71.4286	23.3649
0.5057	0.5246	32	0.5543	72.4424	23.0984
0.415	0.5410	33	0.5503	73.9171	23.3249
0.5682	0.5574	34	0.5454	75.2995	23.4981
0.4804	0.5738	35	0.5396	76.2212	23.5247
0.4547	0.5902	36	0.5336	71.9816	22.8853
0.4501	0.6066	37	0.5299	78.6175	25.5228
0.5138	0.6230	38	0.5276	76.9585	25.0433
0.4667	0.6393	39	0.5244	71.3364	22.6056
0.523	0.6557	40	0.5210	73.7327	25.0300
0.4442	0.6721	41	0.5141	73.5484	24.9900
0.4771	0.6885	42	0.5111	71.6129	22.5390
0.4482	0.7049	43	0.5083	72.5346	22.5390
0.4369	0.7213	44	0.5044	71.4286	22.4191
0.3939	0.7377	45	0.5027	71.0599	22.3525
0.4351	0.7541	46	0.5024	70.5069	22.0061
0.4768	0.7705	47	0.5037	68.6636	22.0994
0.4756	0.7869	48	0.5045	69.2166	22.5390
0.3988	0.8033	49	0.5049	69.0323	22.2992
0.5043	0.8197	50	0.5037	69.1244	22.0061
0.4753	0.8361	51	0.5019	68.0184	21.8729
0.393	0.8525	52	0.4997	68.2028	21.6331
0.5722	0.8689	53	0.4982	69.0323	21.8862
0.3456	0.8852	54	0.4955	68.3871	21.9129
0.4647	0.9016	55	0.4938	72.7189	23.7512
0.4431	0.9180	56	0.4946	73.7327	23.4847
0.3653	0.9344	57	0.4964	72.7189	23.3249
0.4116	0.9508	58	0.4937	71.7972	23.2183
0.4366	0.9672	59	0.4886	72.8111	23.2583
0.4465	0.9836	60	0.4842	72.3502	24.3906
0.3432	1.0	61	0.4849	73.2719	24.3906
0.3308	1.0164	62	0.4853	74.3779	23.4581
0.2445	1.0328	63	0.4850	68.6636	21.6331
0.1849	1.0492	64	0.4877	68.8479	22.2326
0.2543	1.0656	65	0.4912	68.4793	22.0328
0.2168	1.0820	66	0.4925	67.9263	21.4866
0.2329	1.0984	67	0.4915	66.5438	21.1802
0.2677	1.1148	68	0.4898	70.5991	22.3392
0.3012	1.1311	69	0.4879	65.4378	20.5009
0.3084	1.1475	70	0.4864	66.6359	20.3810
0.2551	1.1639	71	0.4862	66.5438	20.6874
0.1999	1.1803	72	0.4880	66.9124	20.5941
0.2429	1.1967	73	0.4898	66.9124	20.9271
0.2409	1.2131	74	0.4893	67.0046	20.6740
0.2062	1.2295	75	0.4879	71.2442	22.2859
0.2622	1.2459	76	0.4873	70.7834	22.0594
0.2466	1.2623	77	0.4851	71.2442	21.8862
0.2619	1.2787	78	0.4835	71.5207	22.0328
0.2295	1.2951	79	0.4821	71.7051	21.9928
0.2092	1.3115	80	0.4818	72.5346	22.1260
0.23	1.3279	81	0.4814	69.2166	20.8072
0.2476	1.3443	82	0.4809	68.8479	20.7007
0.2425	1.3607	83	0.4799	68.2028	20.5675
0.2293	1.3770	84	0.4792	68.8479	20.7140
0.2174	1.3934	85	0.4779	72.9032	21.9528
0.2208	1.4098	86	0.4781	72.9954	22.0727
0.232	1.4262	87	0.4769	68.5714	20.3943
0.2224	1.4426	88	0.4759	68.2028	20.3410
0.2432	1.4590	89	0.4756	68.3871	20.3277
0.2128	1.4754	90	0.4752	68.1106	20.4476
0.2592	1.4918	91	0.4744	67.2811	20.1812
0.2232	1.5082	92	0.4735	67.1889	20.1279
0.1532	1.5246	93	0.4727	66.3594	19.9680
0.2344	1.5410	94	0.4715	66.6359	20.0746
0.2243	1.5574	95	0.4708	67.0968	20.3144
0.2635	1.5738	96	0.4699	66.2673	20.2211
0.2494	1.5902	97	0.4695	66.0829	20.1012
0.2258	1.6066	98	0.4694	65.9908	20.2211
0.1882	1.6230	99	0.4688	66.5438	20.1945
0.2084	1.6393	100	0.4691	66.8203	20.1279
0.1919	1.6557	101	0.4688	67.2811	20.2478
0.2097	1.6721	102	0.4689	66.7281	20.1678
0.2217	1.6885	103	0.4690	66.4516	20.0746
0.2423	1.7049	104	0.4693	66.3594	20.1279
0.2825	1.7213	105	0.4692	66.1751	20.0480
0.2083	1.7377	106	0.4694	66.0829	20.1279
0.2145	1.7541	107	0.4696	65.8065	20.0879
0.1843	1.7705	108	0.4700	66.3594	20.1012
0.2057	1.7869	109	0.4703	65.9908	20.0613
0.2141	1.8033	110	0.4704	66.3594	20.0346
0.2492	1.8197	111	0.4706	66.4516	20.0213
0.1897	1.8361	112	0.4707	66.5438	20.0613
0.1877	1.8525	113	0.4705	67.2811	20.2344
0.1913	1.8689	114	0.4702	67.0046	20.2211
0.2543	1.8852	115	0.4702	66.4516	20.0613
0.2303	1.9016	116	0.4701	66.6359	20.0080
0.2524	1.9180	117	0.4699	66.2673	19.9414
0.2503	1.9344	118	0.4699	66.4516	20.1412
0.2075	1.9508	119	0.4698	66.3594	20.0613
0.2046	1.9672	120	0.4698	66.4516	20.0613
0.2199	1.9836	121	0.4698	66.1751	20.0213
0.2092	2.0	122	0.4699	66.4516	20.0080

Framework versions

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