HBERTv1_48_L6_H128_A2_massive

This model is a fine-tuned version of gokuls/HBERTv1_48_L6_H128_A2 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 0.9439
• Accuracy: 0.7791

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
3.7427	1.0	180	3.3038	0.2110
3.0115	2.0	360	2.6361	0.3601
2.4129	3.0	540	2.1001	0.4988
1.946	4.0	720	1.7411	0.5617
1.6252	5.0	900	1.4891	0.6188
1.3825	6.0	1080	1.3293	0.6508
1.201	7.0	1260	1.2183	0.6990
1.074	8.0	1440	1.1335	0.7304
0.9698	9.0	1620	1.0699	0.7472
0.8878	10.0	1800	1.0251	0.7555
0.8286	11.0	1980	0.9941	0.7629
0.7817	12.0	2160	0.9766	0.7678
0.7388	13.0	2340	0.9558	0.7703
0.7157	14.0	2520	0.9489	0.7782
0.6877	15.0	2700	0.9439	0.7791

Framework versions

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