bert_uncased_L-4_H-128_A-2_massive

This model is a fine-tuned version of google/bert_uncased_L-4_H-128_A-2 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 1.5917
• Accuracy: 0.7122

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
3.8362	1.0	180	3.5577	0.2750
3.3785	2.0	360	3.1194	0.4215
3.0059	3.0	540	2.7843	0.4845
2.7219	4.0	720	2.5372	0.5273
2.4947	5.0	900	2.3286	0.5578
2.3072	6.0	1080	2.1582	0.5947
2.1494	7.0	1260	2.0276	0.6232
2.0206	8.0	1440	1.9108	0.6375
1.9207	9.0	1620	1.8206	0.6704
1.83	10.0	1800	1.7500	0.6891
1.7592	11.0	1980	1.6872	0.7004
1.7011	12.0	2160	1.6489	0.7019
1.6627	13.0	2340	1.6160	0.7093
1.6347	14.0	2520	1.5992	0.7118
1.6216	15.0	2700	1.5917	0.7122

Framework versions

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