HBERTv1_48_L10_H768_A12_massive

This model is a fine-tuned version of gokuls/HBERTv1_48_L10_H768_A12 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 0.8249
• Accuracy: 0.8623

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
1.923	1.0	180	0.8820	0.7595
0.7644	2.0	360	0.7177	0.8087
0.5434	3.0	540	0.6450	0.8352
0.392	4.0	720	0.6084	0.8515
0.2895	5.0	900	0.6436	0.8441
0.2245	6.0	1080	0.6745	0.8510
0.1599	7.0	1260	0.7248	0.8465
0.1185	8.0	1440	0.7497	0.8490
0.0914	9.0	1620	0.7286	0.8564
0.0638	10.0	1800	0.7846	0.8583
0.0468	11.0	1980	0.7941	0.8569
0.0284	12.0	2160	0.7986	0.8569
0.0139	13.0	2340	0.8076	0.8588
0.0083	14.0	2520	0.8281	0.8598
0.005	15.0	2700	0.8249	0.8623

Framework versions

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