kennethge123/bigbench-entailed_polarity-bert-base-uncased-kd

Plainly Optimized Network

Dataset: BIGBENCH

Trainer Hyperparameters:

• lr = 5e-05
• per_device_batch_size = 8
• gradient_accumulation_steps = 2
• weight_decay = 0.0
• seed = 42

eval_loss	eval_accuracy	epoch
10.410	0.571	1.0
10.191	0.571	2.0
9.468	0.643	3.0
10.414	0.571	4.0
10.468	0.571	5.0
10.335	0.571	6.0
10.296	0.571	7.0
9.998	0.571	8.0
10.080	0.571	9.0
10.186	0.571	10.0
9.862	0.571	11.0
10.713	0.500	12.0
9.873	0.571	13.0
9.905	0.571	14.0
9.860	0.571	15.0
9.997	0.571	16.0
9.823	0.571	17.0
9.840	0.571	18.0
9.817	0.571	19.0