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 Loss	Epoch	Step	Validation Loss	Wer
0.7818	0.1678	25	0.9010	116.9199
0.6999	0.3356	50	0.6872	83.8398
0.5713	0.5034	75	0.6395	76.7265
0.6842	0.6711	100	0.6125	73.4807
0.5176	0.8389	125	0.5861	72.2376
0.3141	1.0067	150	0.5643	77.0028
0.3362	1.1745	175	0.5735	71.5470
0.3727	1.3423	200	0.5709	85.3591
0.3567	1.5101	225	0.5627	71.4088
0.378	1.6779	250	0.5513	72.0994
0.336	1.8456	275	0.5559	72.4448
0.3027	2.0134	300	0.5436	69.7514
0.1331	2.1812	325	0.5806	72.4448
0.3148	2.3490	350	0.5831	77.7624
0.2308	2.5168	375	0.5848	73.4807
0.287	2.6846	400	0.5694	68.5773
0.2077	2.8523	425	0.5771	69.6823
0.1206	3.0201	450	0.5789	70.7873
0.0931	3.1879	475	0.6320	72.0304
0.0908	3.3557	500	0.6382	70.0276
0.1106	3.5235	525	0.6384	69.5442
0.0975	3.6913	550	0.6312	70.5801
0.0745	3.8591	575	0.6349	70.7182

nkkbr/whisper-large-v3-zatoichi-ja-20251210_091216

Whisper Large v3 - Japanese Zatoichi ASR