hBERTv2_new_pretrain_wnli

This model is a fine-tuned version of gokuls/bert_12_layer_model_v2_complete_training_new on the GLUE WNLI dataset. It achieves the following results on the evaluation set:

• Loss: 0.6857
• Accuracy: 0.5634

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
0.8646	1.0	5	0.7422	0.4366
0.7094	2.0	10	0.7290	0.4366
0.7047	3.0	15	0.7053	0.5634
0.7203	4.0	20	0.7022	0.4366
0.7	5.0	25	0.6977	0.4366
0.7098	6.0	30	0.6885	0.5634
0.695	7.0	35	0.7045	0.4366
0.7053	8.0	40	0.6858	0.5634
0.7095	9.0	45	0.7070	0.4366
0.7012	10.0	50	0.6857	0.5634
0.6995	11.0	55	0.6969	0.4507
0.6913	12.0	60	0.6875	0.5634
0.6963	13.0	65	0.6959	0.4789
0.6996	14.0	70	0.7190	0.4366
0.6957	15.0	75	0.6963	0.5634

Framework versions

• Transformers 4.29.2
• Pytorch 1.14.0a0+410ce96
• Datasets 2.12.0
• Tokenizers 0.13.3