kennethge123/superglue-rte-t5-base-kd

Plainly Optimized Network

Dataset: SUPERGLUE

Trainer Hyperparameters:

• lr = 5e-05
• per_device_batch_size = 8
• gradient_accumulation_steps = 2
• weight_decay = 1e-09
• seed = 42

eval_loss	eval_accuracy	epoch
20.646	0.587	1.0
20.427	0.609	2.0
20.235	0.580	3.0
19.325	0.623	4.0
19.596	0.623	5.0
18.737	0.717	6.0
18.698	0.717	7.0
18.311	0.725	8.0
18.347	0.739	9.0
18.437	0.710	10.0
17.924	0.717	11.0
18.021	0.732	12.0
17.819	0.739	13.0
18.181	0.725	14.0
18.062	0.732	15.0
18.068	0.725	16.0
18.177	0.717	17.0
17.865	0.732	18.0
17.977	0.732	19.0