HBERTv1_48_L2_H64_A2_massive

This model is a fine-tuned version of gokuls/HBERTv1_48_L2_H64_A2 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 2.1655
• Accuracy: 0.4009

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
3.9873	1.0	180	3.7924	0.1190
3.6056	2.0	360	3.4147	0.1210
3.3391	3.0	540	3.2008	0.1422
3.1518	4.0	720	3.0285	0.1977
2.9855	5.0	900	2.8620	0.2356
2.8224	6.0	1080	2.7059	0.2671
2.6751	7.0	1260	2.5728	0.2986
2.5558	8.0	1440	2.4704	0.3399
2.4664	9.0	1620	2.3848	0.3566
2.3814	10.0	1800	2.3129	0.3719
2.3131	11.0	1980	2.2572	0.3792
2.2662	12.0	2160	2.2149	0.3920
2.2201	13.0	2340	2.1830	0.3935
2.1957	14.0	2520	2.1655	0.4009
2.1831	15.0	2700	2.1585	0.3994

Framework versions

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