HBERTv1_48_L8_H128_A2_massive

This model is a fine-tuned version of gokuls/HBERTv1_48_L8_H128_A2 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 0.9536
• Accuracy: 0.7732

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: 64
• eval_batch_size: 64
• seed: 33
• distributed_type: multi-GPU
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
3.6754	1.0	180	3.1304	0.2066
2.7764	2.0	360	2.3806	0.4191
2.1831	3.0	540	1.9090	0.5312
1.7788	4.0	720	1.5949	0.6016
1.4936	5.0	900	1.4032	0.6513
1.2858	6.0	1080	1.2747	0.6847
1.1232	7.0	1260	1.1651	0.7172
1.0058	8.0	1440	1.0993	0.7295
0.9035	9.0	1620	1.0379	0.7486
0.8245	10.0	1800	1.0164	0.7634
0.7634	11.0	1980	0.9935	0.7614
0.7116	12.0	2160	0.9657	0.7708
0.6763	13.0	2340	0.9624	0.7708
0.6477	14.0	2520	0.9536	0.7732
0.6258	15.0	2700	0.9487	0.7732

Framework versions

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