du33169/roberta-large-finetuned-GLUE-STSB

STSB

This model is a fine-tuned version of FacebookAI/roberta-large on the GLUE STSB dataset. It achieves the following results on the evaluation set:

• Loss: 0.3741
• Pearson: 0.9204
• Spearmanr: 0.9179
• Combined Score: 0.9191

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: 64
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10.0

Training results

Training Loss	Epoch	Step	Combined Score	Validation Loss	Pearson	Spearmanr
No log	1.0	90	0.8853	0.4872	0.8861	0.8845
No log	2.0	180	0.9080	0.5326	0.9079	0.9081
No log	3.0	270	0.9154	0.3783	0.9163	0.9145
No log	4.0	360	0.9186	0.3566	0.9191	0.9180
No log	5.0	450	0.9189	0.3721	0.9195	0.9183
0.6287	6.0	540	0.9191	0.3741	0.9204	0.9179
0.6287	7.0	630	0.3786	0.9175	0.9169	0.9172
0.6287	8.0	720	0.4122	0.9200	0.9179	0.9189
0.6287	9.0	810	0.3854	0.9197	0.9185	0.9191
0.6287	10.0	900	0.3883	0.9196	0.9181	0.9188

Framework versions

• Transformers 4.43.3
• Pytorch 1.11.0+cu113
• Datasets 2.20.0
• Tokenizers 0.19.1