hBERTv1_new_pretrain_w_init_48_mnli

This model is a fine-tuned version of gokuls/bert_12_layer_model_v1_complete_training_new_wt_init_48 on the GLUE MNLI dataset. It achieves the following results on the evaluation set:

• Loss: 0.8954
• Accuracy: 0.5911

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
1.031	1.0	3068	0.9883	0.4718
0.9713	2.0	6136	0.9752	0.5146
0.9588	3.0	9204	0.9702	0.5167
0.953	4.0	12272	0.9590	0.5355
0.9323	5.0	15340	0.9371	0.5416
0.9068	6.0	18408	0.9213	0.5590
0.8891	7.0	21476	0.9168	0.5608
0.8741	8.0	24544	0.9472	0.5453
0.8589	9.0	27612	0.9185	0.5793
0.8432	10.0	30680	0.9134	0.5747
0.8289	11.0	33748	0.9139	0.5753
0.8137	12.0	36816	0.9113	0.5767
0.7988	13.0	39884	0.8925	0.5917
0.7828	14.0	42952	0.9037	0.5859
0.7705	15.0	46020	0.9129	0.5866
0.7576	16.0	49088	0.9237	0.5879
0.7463	17.0	52156	0.9212	0.5897
0.7341	18.0	55224	0.9226	0.5918

Framework versions

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