learning_rate: 0.0001
train_batch_size: 8
eval_batch_size: 8
seed: 3407
gradient_accumulation_steps: 4
total_train_batch_size: 32
optimizer: Use OptimizerNames.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_steps: 2000
training_steps: 20000
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss
0.107	90.9091	1000	0.0932
0.0895	181.8182	2000	0.0909
0.0718	272.7273	3000	0.0931
0.067	363.6364	4000	0.0952
0.0614	454.5455	5000	0.1001
0.0574	545.4545	6000	0.0994
0.0539	636.3636	7000	0.1040
0.0485	727.2727	8000	0.1060
0.0502	818.1818	9000	0.1092
0.0454	909.0909	10000	0.1113
0.044	1000.0	11000	0.1130
0.0454	1090.9091	12000	0.1130
0.0414	1181.8182	13000	0.1172
0.0422	1272.7273	14000	0.1157
0.0395	1363.6364	15000	0.1202
0.0411	1454.5455	16000	0.1173
0.0401	1545.4545	17000	0.1179
0.0484	1636.3636	18000	0.1207
0.0405	1727.2727	19000	0.1178
0.0422	1818.1818	20000	0.1180

sil-ai/senga-LUK-20k-quiet-speecht5

senga-LUK-20k-quiet-speecht5