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
1.3173	0.0303	1	1.3310	98.1200
1.4269	0.0606	2	1.0497	95.8818
1.1875	0.0909	3	1.0013	91.9427
0.911	0.1212	4	0.9277	86.2131
0.9791	0.1515	5	0.8755	87.8245
0.9068	0.1818	6	0.8177	107.2516
0.8325	0.2121	7	0.7731	105.5506
0.8549	0.2424	8	0.7312	103.1334
0.7945	0.2727	9	0.7003	98.2990
0.6762	0.3030	10	0.6662	99.6419
0.642	0.3333	11	0.6423	96.4190
0.7638	0.3636	12	0.6280	82.6321
0.6288	0.3939	13	0.6193	76.4548
0.5627	0.4242	14	0.6232	77.1710
0.5446	0.4545	15	0.6161	78.8720
0.6459	0.4848	16	0.6100	80.4834
0.6125	0.5152	17	0.6112	84.0645
0.571	0.5455	18	0.6102	85.8550
0.5716	0.5758	19	0.6028	86.0340
0.5593	0.6061	20	0.5941	85.5864
0.5686	0.6364	21	0.5839	79.4987
0.5995	0.6667	22	0.5774	78.1558
0.625	0.6970	23	0.5712	76.8129
0.5875	0.7273	24	0.5647	77.0815
0.6402	0.7576	25	0.5604	76.1862
0.4817	0.7879	26	0.5591	75.1119
0.5744	0.8182	27	0.5584	75.5595
0.6008	0.8485	28	0.5595	77.5291
0.5015	0.8788	29	0.5623	79.3196
0.6453	0.9091	30	0.5628	80.6625
0.549	0.9394	31	0.5597	81.6473
0.6709	0.9697	32	0.5561	88.4512
0.7261	1.0	33	0.5525	86.7502
0.398	1.0303	34	0.5486	85.9445
0.4016	1.0606	35	0.5450	84.6016
0.424	1.0909	36	0.5423	85.4073
0.3512	1.1212	37	0.5403	83.8854
0.4034	1.1515	38	0.5397	75.2910
0.4133	1.1818	39	0.5403	75.7386
0.3322	1.2121	40	0.5423	76.3653
0.3711	1.2424	41	0.5451	79.2301
0.4051	1.2727	42	0.5455	79.5882
0.2811	1.3030	43	0.5452	79.6777
0.3401	1.3333	44	0.5441	80.3939
0.3766	1.3636	45	0.5427	79.9463
0.3366	1.3939	46	0.5417	80.2149
0.3978	1.4242	47	0.5407	79.5882
0.3663	1.4545	48	0.5405	78.4244
0.3142	1.4848	49	0.5410	78.8720
0.3226	1.5152	50	0.5414	77.7977
0.4009	1.5455	51	0.5405	77.5291
0.3005	1.5758	52	0.5401	77.4396
0.3417	1.6061	53	0.5404	77.4396
0.3489	1.6364	54	0.5400	77.6186
0.3349	1.6667	55	0.5393	77.9767
0.3161	1.6970	56	0.5385	78.5139
0.3953	1.7273	57	0.5382	78.8720
0.4236	1.7576	58	0.5376	78.9615
0.3445	1.7879	59	0.5371	78.9615
0.2952	1.8182	60	0.5366	79.3196
0.3413	1.8485	61	0.5367	79.1406
0.3672	1.8788	62	0.5365	79.3196
0.3108	1.9091	63	0.5361	79.1406
0.3339	1.9394	64	0.5362	79.4987
0.3165	1.9697	65	0.5362	79.1406
0.4807	2.0	66	0.5360	79.3196

Framework versions

Transformers 4.57.3
Pytorch 2.9.1+cu128
Datasets 4.4.1
Tokenizers 0.22.1

nkkbr/whisper-large-v3-zatoichi-ja-EX-5-TRAIN_2_TO_10_EVAL_1_COSINE

Whisper Large v3 - Japanese Zatoichi ASR

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

nkkbr/
whisper-large-v3-zatoichi-ja-EX-5-TRAIN_2_TO_10_EVAL_1_COSINE