learning_rate: 2e-05
train_batch_size: 8
eval_batch_size: 8
seed: 42
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_ratio: 0.1
num_epochs: 20
mixed_precision_training: Native AMP
label_smoothing_factor: 0.15

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
0.7046	1.0	21	0.6945	0.5175	0.3530	0.2679	0.5175
0.6918	2.0	42	0.6841	0.5175	0.3530	0.2679	0.5175
0.6247	3.0	63	0.5395	0.7018	0.6799	0.7978	0.7018
0.405	4.0	84	0.3227	0.8947	0.8945	0.9031	0.8947
0.2622	5.0	105	0.5418	0.8596	0.8586	0.8772	0.8596
0.185	6.0	126	0.3322	0.9211	0.9210	0.9245	0.9211
0.1968	7.0	147	0.2518	0.9386	0.9386	0.9399	0.9386
0.2289	8.0	168	0.2705	0.9386	0.9386	0.9399	0.9386
0.1514	9.0	189	0.1881	0.9561	0.9562	0.9575	0.9561
0.072	10.0	210	0.1173	0.9649	0.9649	0.9655	0.9649
0.066	11.0	231	0.1214	0.9737	0.9737	0.9750	0.9737
0.0299	12.0	252	0.1070	0.9737	0.9737	0.9738	0.9737
0.0651	13.0	273	0.1556	0.9649	0.9649	0.9655	0.9649
0.0889	14.0	294	0.1653	0.9737	0.9737	0.9750	0.9737

filip-taschner/young_old_classificator

young_old_classificator