mobilebert_sa_GLUE_Experiment_stsb_128

This model is a fine-tuned version of google/mobilebert-uncased on the GLUE STSB dataset. It achieves the following results on the evaluation set:

• Loss: 2.2718
• Pearson: 0.2225
• Spearmanr: 0.2230
• Combined Score: 0.2228

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: 128
• eval_batch_size: 128
• seed: 10
• distributed_type: multi-GPU
• 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	Pearson	Spearmanr	Combined Score
5.1866	1.0	45	2.5161	-0.0355	-0.0329	-0.0342
2.1786	2.0	90	2.3527	0.0407	0.0395	0.0401
2.1136	3.0	135	2.3101	0.0543	0.0555	0.0549
2.1	4.0	180	2.4469	0.0624	0.0656	0.0640
1.9564	5.0	225	2.8646	0.0815	0.0817	0.0816
1.8611	6.0	270	2.5597	0.1089	0.1054	0.1071
1.6702	7.0	315	2.7087	0.1644	0.1666	0.1655
1.385	8.0	360	2.2718	0.2225	0.2230	0.2228
1.2518	9.0	405	2.4105	0.2134	0.2022	0.2078
1.143	10.0	450	2.5834	0.1998	0.2083	0.2040
1.0191	11.0	495	2.6132	0.1856	0.1896	0.1876
0.9431	12.0	540	2.8187	0.1784	0.1872	0.1828
0.8725	13.0	585	2.8360	0.1815	0.1879	0.1847

Framework versions

• Transformers 4.26.0
• Pytorch 1.14.0a0+410ce96
• Datasets 2.8.0
• Tokenizers 0.13.2