bert_uncased_L-8_H-512_A-8_massive

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

• Loss: 0.6061
• Accuracy: 0.8839

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
2.4813	1.0	180	1.3295	0.7506
1.0577	2.0	360	0.7897	0.8318
0.6168	3.0	540	0.6254	0.8623
0.404	4.0	720	0.5679	0.8672
0.2787	5.0	900	0.5594	0.8647
0.1993	6.0	1080	0.5499	0.8746
0.1406	7.0	1260	0.5674	0.8706
0.1044	8.0	1440	0.5678	0.8770
0.0775	9.0	1620	0.5668	0.8829
0.0565	10.0	1800	0.5830	0.8829
0.044	11.0	1980	0.6016	0.8785
0.0342	12.0	2160	0.6061	0.8839
0.0298	13.0	2340	0.6151	0.8815
0.0247	14.0	2520	0.6126	0.8810
0.0208	15.0	2700	0.6156	0.8829

Framework versions

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