bert-small-finetuned-glue-rte

This model is a fine-tuned version of google/bert_uncased_L-4_H-512_A-8 on the glue dataset. It achieves the following results on the evaluation set:

• Loss: 2.8715
• Accuracy: 0.6318

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: 2e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	2.62	50	1.8285	0.6318
No log	5.26	100	2.0806	0.6462
No log	7.87	150	2.1598	0.6282
No log	10.51	200	2.2774	0.6318
No log	13.15	250	2.3676	0.6245
No log	15.77	300	2.4581	0.6462
No log	18.41	350	2.6175	0.6354
No log	21.05	400	2.6697	0.6354
No log	23.67	450	2.7717	0.6354
0.0101	26.31	500	2.7975	0.6462
0.0101	28.92	550	2.8532	0.6390
0.0101	31.56	600	2.9054	0.6209
0.0101	34.21	650	2.8715	0.6318

Framework versions

• Transformers 4.21.2
• Pytorch 1.12.1+cu113
• Datasets 2.4.0
• Tokenizers 0.12.1