HBERTv1_48_L12_H64_A2_massive

This model is a fine-tuned version of gokuls/HBERTv1_48_L12_H64_A2 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 1.8009
• Accuracy: 0.5553

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
3.9467	1.0	180	3.7242	0.1121
3.5338	2.0	360	3.3402	0.1121
3.2555	3.0	540	3.1034	0.1820
3.0004	4.0	720	2.8411	0.3074
2.7522	5.0	900	2.6134	0.3384
2.5415	6.0	1080	2.4233	0.3856
2.366	7.0	1260	2.2615	0.4112
2.2167	8.0	1440	2.1359	0.4609
2.1008	9.0	1620	2.0361	0.4835
2.0016	10.0	1800	1.9606	0.5061
1.9204	11.0	1980	1.8984	0.5298
1.8517	12.0	2160	1.8549	0.5352
1.8078	13.0	2340	1.8158	0.5499
1.78	14.0	2520	1.8009	0.5553
1.7531	15.0	2700	1.7919	0.5548

Framework versions

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