bert_uncased_L-10_H-128_A-2_massive

This model is a fine-tuned version of google/bert_uncased_L-10_H-128_A-2 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 1.4064
• Accuracy: 0.7467

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
3.8032	1.0	180	3.4795	0.3296
3.2716	2.0	360	2.9915	0.4491
2.8593	3.0	540	2.6360	0.5145
2.5442	4.0	720	2.3533	0.5765
2.296	5.0	900	2.1403	0.6006
2.0936	6.0	1080	1.9655	0.6463
1.9277	7.0	1260	1.8291	0.6719
1.7937	8.0	1440	1.7114	0.6911
1.6829	9.0	1620	1.6267	0.7088
1.5946	10.0	1800	1.5575	0.7231
1.5258	11.0	1980	1.4976	0.7354
1.4663	12.0	2160	1.4616	0.7364
1.4256	13.0	2340	1.4296	0.7437
1.3984	14.0	2520	1.4126	0.7442
1.3824	15.0	2700	1.4064	0.7467

Framework versions

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