bert_uncased_L-2_H-512_A-8_massive

This model is a fine-tuned version of google/bert_uncased_L-2_H-512_A-8 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 0.5162
• Accuracy: 0.8741

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
2.9188	1.0	180	1.8187	0.6085
1.5237	2.0	360	1.0972	0.7585
0.9995	3.0	540	0.8257	0.7978
0.7371	4.0	720	0.6883	0.8318
0.5734	5.0	900	0.6198	0.8455
0.4599	6.0	1080	0.5870	0.8569
0.3835	7.0	1260	0.5654	0.8574
0.3262	8.0	1440	0.5414	0.8652
0.279	9.0	1620	0.5282	0.8697
0.2435	10.0	1800	0.5281	0.8677
0.2199	11.0	1980	0.5156	0.8697
0.1999	12.0	2160	0.5162	0.8741
0.1824	13.0	2340	0.5224	0.8736
0.1712	14.0	2520	0.5223	0.8731
0.1627	15.0	2700	0.5206	0.8716

Framework versions

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