bert_uncased_L-6_H-256_A-4_massive

This model is a fine-tuned version of google/bert_uncased_L-6_H-256_A-4 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 0.6810
• Accuracy: 0.8559

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
3.414	1.0	180	2.7276	0.4629
2.4079	2.0	360	1.9650	0.6390
1.8021	3.0	540	1.5333	0.7221
1.4214	4.0	720	1.2586	0.7634
1.1564	5.0	900	1.0714	0.7900
0.966	6.0	1080	0.9539	0.8023
0.8253	7.0	1260	0.8706	0.8165
0.7209	8.0	1440	0.8109	0.8303
0.6401	9.0	1620	0.7647	0.8377
0.5755	10.0	1800	0.7404	0.8411
0.5227	11.0	1980	0.7118	0.8431
0.4854	12.0	2160	0.6995	0.8515
0.4528	13.0	2340	0.6902	0.8544
0.4399	14.0	2520	0.6810	0.8559
0.4285	15.0	2700	0.6820	0.8515

Framework versions

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