HBERTv1_48_L12_H768_A12_massive_data_augmented

This model is a fine-tuned version of gokuls/HBERTv1_48_L12_H768_A12 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.6902
• Accuracy: 0.8377

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
1.0914	1.0	4000	0.6902	0.8377
0.6178	2.0	8000	0.6890	0.8283
0.463	3.0	12000	0.7721	0.8254
0.3683	4.0	16000	0.7829	0.8288
0.3005	5.0	20000	0.8556	0.8303
0.2502	6.0	24000	0.9171	0.8214
0.2106	7.0	28000	1.0074	0.8160
0.1779	8.0	32000	1.0923	0.8239
0.1506	9.0	36000	1.1525	0.8254
0.1259	10.0	40000	1.2103	0.8249
0.1059	11.0	44000	1.3093	0.8269
0.0894	12.0	48000	1.4000	0.8288
0.0745	13.0	52000	1.5050	0.8313
0.0624	14.0	56000	1.5424	0.8288
0.0521	15.0	60000	1.5973	0.8308

Framework versions

• Transformers 4.34.1
• Pytorch 1.14.0a0+410ce96
• Datasets 2.14.6
• Tokenizers 0.14.1