Bert_Classifier

This model is a fine-tuned version of bert-base-cased on the yelp_review_full dataset. It achieves the following results on the evaluation set:

• Loss: 2.8684
• Accuracy: 0.5133

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: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	1.0	188	2.3115	0.5267
No log	2.0	376	2.5268	0.5467
0.2313	3.0	564	2.8684	0.5133

Framework versions

• Transformers 4.21.3
• Pytorch 1.12.1+cu113
• Datasets 2.4.0
• Tokenizers 0.12.1