add_BERT_48_rte

This model is a fine-tuned version of gokuls/add_bert_12_layer_model_complete_training_new_48 on the GLUE RTE dataset. It achieves the following results on the evaluation set:

• Loss: 0.6914
• Accuracy: 0.5343

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 4e-05
• train_batch_size: 128
• eval_batch_size: 128
• seed: 10
• distributed_type: multi-GPU
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.7366	1.0	20	0.7002	0.4693
0.7043	2.0	40	0.6989	0.4621
0.7196	3.0	60	0.6963	0.5307
0.7123	4.0	80	0.7180	0.4729
0.7159	5.0	100	0.6933	0.4982
0.7033	6.0	120	0.7186	0.4729
0.7041	7.0	140	0.7036	0.4729
0.6992	8.0	160	0.6952	0.4693
0.7056	9.0	180	0.6920	0.5271
0.6988	10.0	200	0.6978	0.4621
0.6988	11.0	220	0.6917	0.5271
0.7047	12.0	240	0.7013	0.4729
0.7036	13.0	260	0.7099	0.4729
0.7018	14.0	280	0.6914	0.5343
0.694	15.0	300	0.6924	0.5271
0.6942	16.0	320	0.6961	0.5271
0.6948	17.0	340	0.6988	0.5018
0.6973	18.0	360	0.6930	0.5271
0.6886	19.0	380	0.7256	0.4621

Framework versions

• Transformers 4.30.2
• Pytorch 1.14.0a0+410ce96
• Datasets 2.13.0
• Tokenizers 0.13.3