bert_uncased_L-6_H-128_A-2_massive

This model is a fine-tuned version of google/bert_uncased_L-6_H-128_A-2 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 1.5439
• Accuracy: 0.7314

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
3.8348	1.0	180	3.5466	0.3192
3.3738	2.0	360	3.1057	0.4407
2.9882	3.0	540	2.7502	0.5140
2.6778	4.0	720	2.4684	0.5789
2.4276	5.0	900	2.2516	0.6158
2.2242	6.0	1080	2.0764	0.6419
2.0619	7.0	1260	1.9388	0.6827
1.932	8.0	1440	1.8294	0.6827
1.8283	9.0	1620	1.7395	0.6975
1.7411	10.0	1800	1.6747	0.7118
1.6698	11.0	1980	1.6142	0.7137
1.6176	12.0	2160	1.5737	0.7231
1.5796	13.0	2340	1.5439	0.7314
1.5509	14.0	2520	1.5284	0.7300
1.5409	15.0	2700	1.5226	0.7285

Framework versions

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