HBERTv1_48_L10_H512_A8_massive

This model is a fine-tuned version of gokuls/HBERTv1_48_L10_H512_A8 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 0.8593
• Accuracy: 0.8583

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
2.6536	1.0	180	1.3212	0.6542
1.0696	2.0	360	0.8384	0.7777
0.701	3.0	540	0.7514	0.8082
0.5181	4.0	720	0.7376	0.8096
0.3958	5.0	900	0.6764	0.8269
0.3031	6.0	1080	0.6955	0.8382
0.2378	7.0	1260	0.7173	0.8392
0.1719	8.0	1440	0.7289	0.8401
0.1294	9.0	1620	0.7609	0.8485
0.096	10.0	1800	0.7744	0.8465
0.0769	11.0	1980	0.8206	0.8490
0.05	12.0	2160	0.8085	0.8564
0.034	13.0	2340	0.8537	0.8534
0.0218	14.0	2520	0.8480	0.8564
0.0139	15.0	2700	0.8593	0.8583

Framework versions

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