learning_rate: 1e-05
train_batch_size: 1
eval_batch_size: 8
seed: 42
distributed_type: multi-GPU
num_devices: 4
total_train_batch_size: 4
total_eval_batch_size: 32
optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
training_steps: 5000
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer
3.9496	0.3251	500	3.8825	99.9935
3.0031	0.6502	1000	3.0268	94.9919
2.8602	0.9753	1500	2.8218	118.7130
2.808	1.3004	2000	2.7146	97.5562
2.6178	1.6255	2500	2.6478	89.8751
2.6159	1.9506	3000	2.5992	88.6473
2.5032	2.2757	3500	2.5672	90.1134
2.5137	2.6008	4000	2.5444	91.5938
2.4869	2.9259	4500	2.5319	91.5938
2.5279	3.2510	5000	2.5273	93.3256

Paulia12138/whisper-large-v3-caption-en

whisper-large-v3-caption-en