HBERTv1_48_L4_H768_A12_massive

This model is a fine-tuned version of gokuls/HBERTv1_48_L4_H768_A12 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 0.7904
• Accuracy: 0.8726

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
1.714	1.0	180	0.7757	0.7826
0.6529	2.0	360	0.6221	0.8328
0.4238	3.0	540	0.5757	0.8544
0.2832	4.0	720	0.5940	0.8544
0.2056	5.0	900	0.6066	0.8495
0.1417	6.0	1080	0.6677	0.8559
0.0983	7.0	1260	0.6791	0.8519
0.0741	8.0	1440	0.7092	0.8495
0.0495	9.0	1620	0.7061	0.8687
0.0356	10.0	1800	0.7682	0.8633
0.0243	11.0	1980	0.7785	0.8623
0.0144	12.0	2160	0.7833	0.8677
0.0099	13.0	2340	0.7941	0.8711
0.0063	14.0	2520	0.7904	0.8726
0.0037	15.0	2700	0.8014	0.8677

Framework versions

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