HBERTv1_48_L4_H256_A4_massive

This model is a fine-tuned version of gokuls/HBERTv1_48_L4_H256_A4 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 0.7151
• Accuracy: 0.8401

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
3.206	1.0	180	2.2341	0.4579
1.8169	2.0	360	1.3642	0.6749
1.1697	3.0	540	1.0289	0.7423
0.8587	4.0	720	0.8583	0.7821
0.6723	5.0	900	0.8038	0.8008
0.5497	6.0	1080	0.7459	0.8101
0.4755	7.0	1260	0.7419	0.8146
0.3992	8.0	1440	0.7095	0.8278
0.336	9.0	1620	0.7278	0.8288
0.2956	10.0	1800	0.7151	0.8401
0.2593	11.0	1980	0.7130	0.8347
0.2301	12.0	2160	0.7215	0.8367
0.2038	13.0	2340	0.7191	0.8372
0.1881	14.0	2520	0.7273	0.8362
0.1766	15.0	2700	0.7256	0.8401

Framework versions

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