HBERTv1_48_L12_H256_A4_massive

This model is a fine-tuned version of gokuls/HBERTv1_48_L12_H256_A4 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 1.1699
• Accuracy: 0.7339

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
3.7238	1.0	180	3.4052	0.1382
3.1325	2.0	360	2.8875	0.2022
2.7162	3.0	540	2.5311	0.3030
2.4123	4.0	720	2.3315	0.3576
2.1258	5.0	900	2.0547	0.4186
1.8697	6.0	1080	1.8215	0.4889
1.6446	7.0	1260	1.6681	0.5421
1.4509	8.0	1440	1.5200	0.5853
1.2995	9.0	1620	1.4177	0.6188
1.1585	10.0	1800	1.3337	0.6557
1.0714	11.0	1980	1.2620	0.7059
0.9816	12.0	2160	1.2374	0.7147
0.9053	13.0	2340	1.1849	0.7290
0.8582	14.0	2520	1.1721	0.7324
0.8253	15.0	2700	1.1699	0.7339

Framework versions

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