Modelpublic

gokuls/HBERTv1_48_L6_H512_A8_massive

sourceHugging Faceupdated 3y agoView on Hugging Face

0likes7downloads

Model Card

HBERTv148L6H512A8_massive

This model is a fine-tuned version of gokuls/HBERTv1_48_L6_H512_A8 on the massive dataset. It achieves the following results on the evaluation set:

—Loss: 0.8008
—Accuracy: 0.8554

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
—trainbatchsize: 64
—evalbatchsize: 64
—seed: 33
—distributed_type: multi-GPU
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
2.2378	1.0	180	1.0048	0.7477
0.8333	2.0	360	0.7230	0.8091
0.5543	3.0	540	0.6624	0.8308
0.3904	4.0	720	0.6289	0.8485
0.2796	5.0	900	0.6249	0.8446
0.2095	6.0	1080	0.6813	0.8455
0.1532	7.0	1260	0.6995	0.8421
0.1142	8.0	1440	0.6917	0.8475
0.0826	9.0	1620	0.7257	0.8539
0.0617	10.0	1800	0.7759	0.8470
0.0415	11.0	1980	0.7927	0.8529
0.028	12.0	2160	0.7841	0.8539
0.0203	13.0	2340	0.8027	0.8510
0.0131	14.0	2520	0.8008	0.8554
0.0102	15.0	2700	0.8058	0.8515

Framework versions

—Transformers 4.34.0
—Pytorch 1.14.0a0+410ce96
—Datasets 2.14.5
—Tokenizers 0.14.0