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
1.407	0.0333	1	1.2545	98.1567
1.5379	0.0667	2	0.9910	96.2212
1.0993	0.1	3	0.9364	92.4424
1.136	0.1333	4	0.8575	90.6912
0.9113	0.1667	5	0.7996	94.6544
0.794	0.2	6	0.7502	101.5668
0.9065	0.2333	7	0.7192	101.0138
0.7117	0.2667	8	0.6781	116.9585
0.7077	0.3	9	0.6509	115.5760
0.6987	0.3333	10	0.6295	98.7097
0.5863	0.3667	11	0.6173	78.4332
0.6197	0.4	12	0.6097	76.1290
0.6296	0.4333	13	0.6059	77.1429
0.6102	0.4667	14	0.6010	83.6866
0.7319	0.5	15	0.5982	85.5300
0.6282	0.5333	16	0.5882	87.0968
0.6061	0.5667	17	0.5820	86.4516
0.614	0.6	18	0.5686	83.1336
0.6684	0.6333	19	0.5615	82.1198
0.5062	0.6667	20	0.5567	77.0507
0.5313	0.7	21	0.5523	82.7650
0.5547	0.7333	22	0.5533	84.6083
0.5068	0.7667	23	0.5521	82.3963
0.5531	0.8	24	0.5517	83.4101
0.5318	0.8333	25	0.5497	83.5945
0.6925	0.8667	26	0.5468	86.6359
0.5107	0.9	27	0.5428	85.7143
0.5536	0.9333	28	0.5375	83.6866
0.4961	0.9667	29	0.5340	81.2903
0.52	1.0	30	0.5328	79.4470
0.3528	1.0333	31	0.5324	84.6083
0.4116	1.0667	32	0.5295	83.8710
0.3783	1.1	33	0.5289	79.8157
0.3432	1.1333	34	0.5293	78.8018
0.3321	1.1667	35	0.5297	77.6959
0.3924	1.2	36	0.5289	80.9217
0.3327	1.2333	37	0.5267	80.5530
0.3641	1.2667	38	0.5251	77.1429
0.393	1.3	39	0.5253	77.6959
0.3907	1.3333	40	0.5255	78.4332
0.3524	1.3667	41	0.5262	79.2627
0.362	1.4	42	0.5265	80.1843
0.3334	1.4333	43	0.5259	80.1843
0.4142	1.4667	44	0.5250	79.4470
0.3404	1.5	45	0.5235	79.4470
0.3449	1.5333	46	0.5213	78.3410
0.3388	1.5667	47	0.5197	77.0507
0.4168	1.6	48	0.5192	76.2212
0.307	1.6333	49	0.5185	76.4977
0.3574	1.6667	50	0.5182	75.4839
0.362	1.7	51	0.5172	75.2074
0.3535	1.7333	52	0.5168	75.0230
0.3706	1.7667	53	0.5161	75.0230
0.3499	1.8	54	0.5161	75.0230
0.4151	1.8333	55	0.5156	74.8387
0.3503	1.8667	56	0.5153	75.3917
0.3406	1.9	57	0.5153	75.0230
0.3507	1.9333	58	0.5151	75.0230
0.3703	1.9667	59	0.5150	75.0230
0.3701	2.0	60	0.5152	75.4839

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-6-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-6-TRAIN_2_TO_10_EVAL_1_COSINE