HBERTv1_48_L2_H256_A4_massive

This model is a fine-tuned version of gokuls/HBERTv1_48_L2_H256_A4 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 0.6816
• Accuracy: 0.8362

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
3.0609	1.0	180	2.0828	0.5416
1.7322	2.0	360	1.2770	0.6995
1.1671	3.0	540	0.9723	0.7521
0.8923	4.0	720	0.8411	0.7723
0.7282	5.0	900	0.7748	0.7890
0.6334	6.0	1080	0.7396	0.8042
0.5372	7.0	1260	0.7192	0.8146
0.4854	8.0	1440	0.7008	0.8229
0.4327	9.0	1620	0.6971	0.8224
0.3975	10.0	1800	0.6840	0.8323
0.3598	11.0	1980	0.6947	0.8313
0.334	12.0	2160	0.6791	0.8337
0.317	13.0	2340	0.6859	0.8323
0.2969	14.0	2520	0.6816	0.8347
0.2918	15.0	2700	0.6816	0.8362

Framework versions

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