HBERTv1_48_L12_H128_A2_massive

This model is a fine-tuned version of gokuls/HBERTv1_48_L12_H128_A2 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 1.0042
• Accuracy: 0.7732

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
3.7792	1.0	180	3.3305	0.1913
3.0177	2.0	360	2.6809	0.2371
2.451	3.0	540	2.1770	0.4260
2.0079	4.0	720	1.8072	0.5780
1.681	5.0	900	1.5486	0.6350
1.4431	6.0	1080	1.3891	0.6695
1.2574	7.0	1260	1.2684	0.7049
1.1169	8.0	1440	1.1950	0.7152
1.0107	9.0	1620	1.1286	0.7418
0.9139	10.0	1800	1.0791	0.7590
0.8467	11.0	1980	1.0527	0.7580
0.7838	12.0	2160	1.0343	0.7639
0.7331	13.0	2340	1.0309	0.7673
0.7004	14.0	2520	1.0077	0.7713
0.6797	15.0	2700	1.0042	0.7732

Framework versions

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