learning_rate: 0.0001
train_batch_size: 16
eval_batch_size: 16
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 32
optimizer: Use OptimizerNames.PAGED_ADAMW_8BIT 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: 30
training_steps: 50
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss
9.0286	0.0909	2	6.8793
9.1448	0.1818	4	6.8793
8.4749	0.2727	6	4.8854
4.4436	0.3636	8	4.2949
2.8813	0.4545	10	4.2554
1.8565	0.5455	12	4.6461
1.2493	0.6364	14	4.3707
1.1609	0.7273	16	4.8081
1.2741	0.8182	18	4.5106
1.2503	0.9091	20	4.9242
1.5987	1.0	22	4.4682
1.3656	1.0909	24	5.4809
1.4069	1.1818	26	4.9377
1.3247	1.2727	28	4.6579
1.3614	1.3636	30	4.0041
1.4241	1.4545	32	4.2715
1.5301	1.5455	34	4.3900
1.4486	1.6364	36	4.5534
1.3155	1.7273	38	5.2379
1.2132	1.8182	40	5.6748
1.2602	1.9091	42	5.3015
1.0974	2.0	44	4.9654
1.1286	2.0909	46	4.7520
1.0629	2.1818	48	4.8735
1.033	2.2727	50	5.0132

snaoi-csl/whisper-ft-large-1000-distli-param-2

whisper-ft-large-1000-distli-param-2