bert_uncased_L-4_H-128_A-2_rte

This model is a fine-tuned version of google/bert_uncased_L-4_H-128_A-2 on the GLUE RTE dataset. It achieves the following results on the evaluation set:

• Loss: 0.6659
• Accuracy: 0.6173

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: 5e-05
• train_batch_size: 256
• eval_batch_size: 256
• seed: 10
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.6967	1.0	10	0.6907	0.5523
0.689	2.0	20	0.6875	0.5740
0.6832	3.0	30	0.6854	0.5379
0.6776	4.0	40	0.6809	0.5632
0.6694	5.0	50	0.6780	0.5812
0.6599	6.0	60	0.6749	0.5740
0.6469	7.0	70	0.6726	0.5993
0.6392	8.0	80	0.6712	0.5776
0.6221	9.0	90	0.6682	0.5884
0.6034	10.0	100	0.6684	0.5957
0.5867	11.0	110	0.6717	0.5993
0.5686	12.0	120	0.6690	0.6065
0.5596	13.0	130	0.6659	0.6173
0.5377	14.0	140	0.6720	0.6101
0.5249	15.0	150	0.6820	0.6029
0.5019	16.0	160	0.6896	0.6065
0.492	17.0	170	0.6977	0.6029
0.4786	18.0	180	0.7027	0.6173

Framework versions

• Transformers 4.46.3
• Pytorch 2.2.1+cu118
• Datasets 2.17.0
• Tokenizers 0.20.3