learning_rate: 0.0001
train_batch_size: 256
eval_batch_size: 256
seed: 42
optimizer: Use adamw_torch_fused 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
num_epochs: 20
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss
5.7987	1.0	422	4.8848
4.7594	2.0	844	4.7385
4.6549	3.0	1266	4.6667
4.5905	4.0	1688	4.6235
4.543	5.0	2110	4.5943
4.5031	6.0	2532	4.5713
4.4668	7.0	2954	4.5578
4.4319	8.0	3376	4.5457
4.3979	9.0	3798	4.5386
4.3635	10.0	4220	4.5350
4.3284	11.0	4642	4.5385
4.2928	12.0	5064	4.5433
4.257	13.0	5486	4.5498
4.2213	14.0	5908	4.5586
4.1867	15.0	6330	4.5685
4.1538	16.0	6752	4.5806
4.1236	17.0	7174	4.5939
4.0969	18.0	7596	4.6058
4.0747	19.0	8018	4.6135
4.0583	20.0	8440	4.6171

fpadovani/candor_sh1_42

candor_sh1_42