bert_uncased_L-4_H-768_A-12_massive

This model is a fine-tuned version of google/bert_uncased_L-4_H-768_A-12 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 0.5747
• Accuracy: 0.8898

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
2.1636	1.0	180	0.9552	0.7983
0.7696	2.0	360	0.5934	0.8539
0.4303	3.0	540	0.4996	0.8721
0.2708	4.0	720	0.4900	0.8780
0.1756	5.0	900	0.4886	0.8780
0.113	6.0	1080	0.5020	0.8829
0.076	7.0	1260	0.5211	0.8810
0.0517	8.0	1440	0.5452	0.8864
0.035	9.0	1620	0.5516	0.8883
0.026	10.0	1800	0.5652	0.8864
0.0193	11.0	1980	0.5696	0.8869
0.0156	12.0	2160	0.5592	0.8888
0.0134	13.0	2340	0.5762	0.8893
0.0103	14.0	2520	0.5726	0.8883
0.0107	15.0	2700	0.5747	0.8898

Framework versions

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