bert_uncased_L-6_H-768_A-12_massive

This model is a fine-tuned version of google/bert_uncased_L-6_H-768_A-12 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 0.5459
• Accuracy: 0.8888

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
2.0422	1.0	180	0.8433	0.8160
0.6599	2.0	360	0.5532	0.8618
0.3564	3.0	540	0.5180	0.8701
0.212	4.0	720	0.4955	0.8805
0.1358	5.0	900	0.5076	0.8844
0.0859	6.0	1080	0.5193	0.8864
0.059	7.0	1260	0.5459	0.8888
0.038	8.0	1440	0.5811	0.8834
0.0255	9.0	1620	0.5875	0.8849
0.0171	10.0	1800	0.5881	0.8834
0.0122	11.0	1980	0.6051	0.8829
0.0086	12.0	2160	0.6117	0.8879
0.007	13.0	2340	0.6032	0.8864
0.006	14.0	2520	0.6112	0.8824
0.0055	15.0	2700	0.6130	0.8844

Framework versions

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