nkkbr/whisper-large-v3-zatoichi-ja-EX-10-combine

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: 0.7883
• Wer: 75.7101

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: 1e-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: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 4

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.9789	0.0382	5	1.0693	96.0870
1.1283	0.0763	10	0.9586	91.7101
0.7806	0.1145	15	0.8366	111.1884
0.8674	0.1527	20	0.7455	104.8696
0.7825	0.1908	25	0.6720	88.5797
0.6831	0.2290	30	0.6501	82.2609
0.6509	0.2672	35	0.6076	78.0580
0.5506	0.3053	40	0.6019	85.9420
0.4806	0.3435	45	0.5971	88.9855
0.5241	0.3817	50	0.5778	83.2174
0.5247	0.4198	55	0.5859	90.0290
0.4202	0.4580	60	0.5822	85.9420
0.4054	0.4962	65	0.5678	76.6667
0.4655	0.5344	70	0.5889	90.4928
0.3275	0.5725	75	0.5625	76.7536
0.3979	0.6107	80	0.5826	86.6377
0.3904	0.6489	85	0.5543	75.0145
0.3719	0.6870	90	0.5691	84.8696
0.4087	0.7252	95	0.5470	78.4638
0.3276	0.7634	100	0.5540	79.6522
0.2817	0.8015	105	0.5526	87.6522
0.3434	0.8397	110	0.5553	80.0870
0.3615	0.8779	115	0.5522	80.4058
0.3366	0.9160	120	0.5464	82.2609
0.4001	0.9542	125	0.5485	73.5652
0.3078	0.9924	130	0.5611	76.7246
0.1942	1.0305	135	0.5730	74.9275
0.2897	1.0687	140	0.6015	81.2464
0.1861	1.1069	145	0.5781	76.6377
0.1762	1.1450	150	0.5806	84.0290
0.1342	1.1832	155	0.5722	74.2319
0.169	1.2214	160	0.5795	76.5217
0.1453	1.2595	165	0.5867	77.4493
0.1495	1.2977	170	0.5946	77.5652
0.1588	1.3359	175	0.5988	75.4493
0.2011	1.3740	180	0.5979	80.8116
0.1579	1.4122	185	0.5853	80.0290
0.1805	1.4504	190	0.6044	80.4348
0.1865	1.4885	195	0.5941	77.4493
0.1783	1.5267	200	0.5965	79.7101
0.1348	1.5649	205	0.5967	78.8406
0.198	1.6031	210	0.5937	79.3333
0.126	1.6412	215	0.5998	81.0145
0.1151	1.6794	220	0.6167	81.3623
0.1682	1.7176	225	0.6072	80.7246
0.1168	1.7557	230	0.6008	75.1594
0.1568	1.7939	235	0.6289	82.9565
0.1075	1.8321	240	0.6328	78.6377
0.1994	1.8702	245	0.6366	77.7681
0.1558	1.9084	250	0.6303	75.1884
0.0819	1.9466	255	0.6234	78.4928
0.0948	1.9847	260	0.6336	77.8551
0.0589	2.0229	265	0.6449	78.6667
0.0614	2.0611	270	0.6595	78.4348
0.0549	2.0992	275	0.6809	78.8696
0.0441	2.1374	280	0.6811	76.3768
0.0813	2.1756	285	0.6882	79.1014
0.0568	2.2137	290	0.6929	77.8551
0.0394	2.2519	295	0.6864	76.2899
0.0579	2.2901	300	0.6814	75.6522
0.0476	2.3282	305	0.6900	77.2174
0.0449	2.3664	310	0.6853	75.4783
0.0506	2.4046	315	0.6965	76.3478
0.0486	2.4427	320	0.6974	76.1159
0.0347	2.4809	325	0.7053	77.5652
0.0437	2.5191	330	0.6910	77.1594
0.0458	2.5573	335	0.6903	76.0
0.0508	2.5954	340	0.7050	77.2464
0.0372	2.6336	345	0.6945	75.7681
0.0416	2.6718	350	0.6943	76.0870
0.0457	2.7099	355	0.7083	76.9565
0.0524	2.7481	360	0.7070	75.9420
0.0599	2.7863	365	0.7321	80.5217
0.0499	2.8244	370	0.7176	78.1159
0.0467	2.8626	375	0.6973	76.2029
0.0313	2.9008	380	0.7085	79.3043
0.0273	2.9389	385	0.7043	75.3333
0.0299	2.9771	390	0.7247	77.2174
0.0133	3.0153	395	0.7291	75.3333
0.0176	3.0534	400	0.7323	74.1159
0.0101	3.0916	405	0.7485	77.3043
0.0197	3.1298	410	0.7516	75.8841
0.015	3.1679	415	0.7549	75.4493
0.0241	3.2061	420	0.7676	77.4783
0.0258	3.2443	425	0.7565	76.7826
0.03	3.2824	430	0.7529	74.2029
0.0138	3.3206	435	0.7637	76.1449
0.0145	3.3588	440	0.7699	76.2899
0.0135	3.3969	445	0.7684	77.3333
0.0223	3.4351	450	0.7640	76.6667
0.0206	3.4733	455	0.7552	74.3478
0.0097	3.5115	460	0.7553	73.9130
0.0078	3.5496	465	0.7620	75.2174
0.0079	3.5878	470	0.7668	74.6667
0.0116	3.6260	475	0.7723	74.2319
0.0139	3.6641	480	0.7779	74.4348
0.0199	3.7023	485	0.7823	74.9275
0.017	3.7405	490	0.7832	75.3913
0.0131	3.7786	495	0.7872	75.9130
0.0066	3.8168	500	0.7886	76.3478
0.0105	3.8550	505	0.7885	75.6812
0.0193	3.8931	510	0.7881	75.7391
0.0131	3.9313	515	0.7880	75.5652
0.016	3.9695	520	0.7883	75.7101

Framework versions

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