nkkbr/whisper-large-v3-zatoichi-ja-zatoichi-TEST-2-EX-6-TRAIN_2_TO_26_EVAL_1_COSINE

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: 0.4506
• Wer: 75.0681

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: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.3854	0.0127	1	1.2957	98.1381
1.4152	0.0253	2	1.0247	96.0945
1.0787	0.0380	3	1.0070	94.7321
1.0403	0.0506	4	0.9820	92.6431
1.0091	0.0633	5	0.9406	87.5568
0.9546	0.0759	6	0.8870	89.7820
0.9311	0.0886	7	0.8534	89.0554
0.9116	0.1013	8	0.8080	100.3179
0.8877	0.1139	9	0.7776	96.0945
0.6861	0.1266	10	0.7560	102.9973
0.8515	0.1392	11	0.7217	103.1335
0.6287	0.1519	12	0.6890	111.6712
0.7156	0.1646	13	0.6621	110.6721
0.6571	0.1772	14	0.6429	100.5450
0.6851	0.1899	15	0.6359	88.4196
0.6575	0.2025	16	0.6190	87.6476
0.7203	0.2152	17	0.6126	77.1571
0.7072	0.2278	18	0.6051	77.2934
0.6095	0.2405	19	0.5961	82.1526
0.7896	0.2532	20	0.5892	87.5568
0.6042	0.2658	21	0.5866	85.3769
0.7063	0.2785	22	0.5875	90.3270
0.5943	0.2911	23	0.5786	82.1072
0.6516	0.3038	24	0.5712	82.5159
0.5662	0.3165	25	0.5678	84.4687
0.591	0.3291	26	0.5656	87.9655
0.628	0.3418	27	0.5639	84.3324
0.5737	0.3544	28	0.5576	83.6058
0.5231	0.3671	29	0.5518	82.8338
0.6269	0.3797	30	0.5497	83.6058
0.4757	0.3924	31	0.5469	85.5586
0.5437	0.4051	32	0.5441	84.1962
0.5336	0.4177	33	0.5473	82.1072
0.5153	0.4304	34	0.5403	82.5159
0.5292	0.4430	35	0.5359	84.8774
0.4669	0.4557	36	0.5298	85.5132
0.566	0.4684	37	0.5217	81.8347
0.5262	0.4810	38	0.5173	81.9709
0.404	0.4937	39	0.5156	74.9773
0.5276	0.5063	40	0.5115	75.7493
0.4519	0.5190	41	0.5084	77.2934
0.4074	0.5316	42	0.5089	80.1544
0.5793	0.5443	43	0.5084	80.7902
0.4602	0.5570	44	0.5027	79.0645
0.4422	0.5696	45	0.4975	75.6131
0.4064	0.5823	46	0.4945	71.7984
0.4939	0.5949	47	0.4954	70.9809
0.592	0.6076	48	0.4966	70.4360
0.5544	0.6203	49	0.4974	72.2979
0.5218	0.6329	50	0.4986	72.8883
0.5341	0.6456	51	0.4996	75.5677
0.5519	0.6582	52	0.5025	78.3833
0.5196	0.6709	53	0.5043	79.0191
0.5251	0.6835	54	0.5022	77.8383
0.3997	0.6962	55	0.4982	76.2943
0.4862	0.7089	56	0.4941	74.3869
0.522	0.7215	57	0.4918	72.2525
0.458	0.7342	58	0.4884	73.1608
0.4555	0.7468	59	0.4878	77.6567
0.5084	0.7595	60	0.4883	74.3869
0.4656	0.7722	61	0.4878	77.5658
0.4471	0.7848	62	0.4868	79.6549
0.4575	0.7975	63	0.4834	77.9746
0.4096	0.8101	64	0.4824	79.6094
0.4917	0.8228	65	0.4837	79.8365
0.4396	0.8354	66	0.4830	81.5168
0.4612	0.8481	67	0.4820	80.8356
0.6037	0.8608	68	0.4809	76.3397
0.6233	0.8734	69	0.4804	76.4759
0.4937	0.8861	70	0.4796	76.7030
0.4161	0.8987	71	0.4792	76.4759
0.4997	0.9114	72	0.4772	74.3869
0.4594	0.9241	73	0.4746	75.1589
0.5552	0.9367	74	0.4717	72.3433
0.5164	0.9494	75	0.4702	70.8901
0.511	0.9620	76	0.4696	70.4814
0.4388	0.9747	77	0.4698	70.9355
0.4342	0.9873	78	0.4691	70.2543
0.4286	1.0	79	0.4699	71.1172
0.328	1.0127	80	0.4691	72.0708
0.2731	1.0253	81	0.4676	73.2970
0.3507	1.0380	82	0.4663	73.4332
0.3272	1.0506	83	0.4659	72.9337
0.3078	1.0633	84	0.4665	73.0245
0.3021	1.0759	85	0.4674	71.8438
0.3319	1.0886	86	0.4668	70.5268
0.2959	1.1013	87	0.4674	71.0263
0.2539	1.1139	88	0.4694	71.6621
0.297	1.1266	89	0.4708	72.2525
0.2806	1.1392	90	0.4727	73.4332
0.2894	1.1519	91	0.4731	74.4777
0.3233	1.1646	92	0.4723	74.4323
0.2903	1.1772	93	0.4699	74.2961
0.3894	1.1899	94	0.4674	72.7975
0.3188	1.2025	95	0.4648	71.3896
0.274	1.2152	96	0.4639	70.7993
0.3164	1.2278	97	0.4640	70.3451
0.3096	1.2405	98	0.4647	70.2997
0.314	1.2532	99	0.4640	71.5259
0.342	1.2658	100	0.4638	71.4805
0.3044	1.2785	101	0.4633	71.4351
0.3617	1.2911	102	0.4625	70.4814
0.3293	1.3038	103	0.4615	71.3442
0.2864	1.3165	104	0.4617	71.7984
0.3038	1.3291	105	0.4615	72.8883
0.3027	1.3418	106	0.4618	73.3878
0.2246	1.3544	107	0.4627	74.2507
0.3087	1.3671	108	0.4621	73.8420
0.3413	1.3797	109	0.4602	73.1608
0.3352	1.3924	110	0.4594	72.2979
0.3036	1.4051	111	0.4584	71.7075
0.335	1.4177	112	0.4572	71.7984
0.3337	1.4304	113	0.4564	71.0718
0.2833	1.4430	114	0.4562	70.5722
0.2723	1.4557	115	0.4558	70.3906
0.3181	1.4684	116	0.4559	70.2997
0.3266	1.4810	117	0.4558	70.7084
0.3424	1.4937	118	0.4557	71.0718
0.2744	1.5063	119	0.4555	70.7993
0.2523	1.5190	120	0.4552	70.3906
0.2371	1.5316	121	0.4550	69.9364
0.2985	1.5443	122	0.4547	70.2089
0.2731	1.5570	123	0.4549	70.7993
0.29	1.5696	124	0.4549	70.6630
0.3358	1.5823	125	0.4548	70.7539
0.2824	1.5949	126	0.4545	71.0718
0.2507	1.6076	127	0.4545	71.4351
0.3095	1.6203	128	0.4544	71.9346
0.4167	1.6329	129	0.4544	72.1617
0.3025	1.6456	130	0.4548	74.2961
0.387	1.6582	131	0.4541	75.6131
0.2967	1.6709	132	0.4540	75.9764
0.3051	1.6835	133	0.4538	75.7493
0.2935	1.6962	134	0.4535	76.0672
0.3093	1.7089	135	0.4537	76.1126
0.3067	1.7215	136	0.4534	76.5213
0.294	1.7342	137	0.4533	76.9755
0.3125	1.7468	138	0.4528	76.8392
0.3131	1.7595	139	0.4526	76.3851
0.3453	1.7722	140	0.4524	76.3851
0.2877	1.7848	141	0.4520	76.1126
0.3226	1.7975	142	0.4521	76.1126
0.3423	1.8101	143	0.4517	75.5677
0.286	1.8228	144	0.4515	75.5677
0.3066	1.8354	145	0.4513	75.6131
0.272	1.8481	146	0.4510	75.5223
0.2917	1.8608	147	0.4510	75.3406
0.3386	1.8734	148	0.4509	75.2952
0.2745	1.8861	149	0.4509	75.0681
0.3083	1.8987	150	0.4509	75.0681
0.3454	1.9114	151	0.4509	75.2044
0.3412	1.9241	152	0.4510	74.9773
0.3342	1.9367	153	0.4510	75.0227
0.2733	1.9494	154	0.4509	74.6140
0.2795	1.9620	155	0.4507	75.2498
0.3091	1.9747	156	0.4509	74.9773
0.3062	1.9873	157	0.4505	74.7048
0.2928	2.0	158	0.4506	75.0681

Framework versions

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