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
0.1457	90.9091	1000	0.1290
0.1273	181.8182	2000	0.1296
0.1134	272.7273	3000	0.1334
0.1069	363.6364	4000	0.1320
0.1018	454.5455	5000	0.1401
0.0972	545.4545	6000	0.1381
0.0926	636.3636	7000	0.1432
0.0891	727.2727	8000	0.1485
0.0884	818.1818	9000	0.1464
0.0837	909.0909	10000	0.1501
0.0817	1000.0	11000	0.1501
0.0837	1090.9091	12000	0.1509
0.0789	1181.8182	13000	0.1548
0.08	1272.7273	14000	0.1534
0.0767	1363.6364	15000	0.1551
0.0785	1454.5455	16000	0.1559
0.0775	1545.4545	17000	0.1542
0.0838	1636.3636	18000	0.1572
0.0777	1727.2727	19000	0.1557
0.0802	1818.1818	20000	0.1543

sil-ai/senga-LUK-20k-l1-speecht5

senga-LUK-20k-l1-speecht5