bert_uncased_L-2_H-768_A-12_massive

This model is a fine-tuned version of google/bert_uncased_L-2_H-768_A-12 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 0.5434
• Accuracy: 0.8746

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
2.5143	1.0	180	1.2564	0.7024
1.0135	2.0	360	0.7279	0.8205
0.6173	3.0	540	0.5817	0.8559
0.433	4.0	720	0.5234	0.8598
0.312	5.0	900	0.5019	0.8657
0.23	6.0	1080	0.5028	0.8711
0.1742	7.0	1260	0.5037	0.8682
0.1314	8.0	1440	0.5018	0.8692
0.1031	9.0	1620	0.5188	0.8731
0.081	10.0	1800	0.5231	0.8711
0.0671	11.0	1980	0.5407	0.8716
0.0569	12.0	2160	0.5309	0.8721
0.0466	13.0	2340	0.5463	0.8711
0.0414	14.0	2520	0.5434	0.8746
0.039	15.0	2700	0.5464	0.8721

Framework versions

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