bert_uncased_L-4_H-512_A-8_massive

This model is a fine-tuned version of google/bert_uncased_L-4_H-512_A-8 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 0.5260
• Accuracy: 0.8844

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
2.7044	1.0	180	1.5553	0.6901
1.2648	2.0	360	0.9088	0.8082
0.7783	3.0	540	0.6655	0.8539
0.5308	4.0	720	0.5876	0.8578
0.3865	5.0	900	0.5480	0.8716
0.2889	6.0	1080	0.5289	0.8746
0.2207	7.0	1260	0.5367	0.8756
0.1701	8.0	1440	0.5260	0.8844
0.1389	9.0	1620	0.5364	0.8819
0.1076	10.0	1800	0.5423	0.8834
0.0898	11.0	1980	0.5524	0.8795
0.0763	12.0	2160	0.5524	0.8829
0.0633	13.0	2340	0.5643	0.8805
0.0573	14.0	2520	0.5642	0.8819
0.0519	15.0	2700	0.5634	0.8805

Framework versions

• Transformers 4.34.0
• Pytorch 1.14.0a0+410ce96
• Datasets 2.14.5
• Tokenizers 0.14.1