bert-base-parsbert-uncased-finetuned-khorshid

This model is a fine-tuned version of HooshvareLab/bert-base-parsbert-uncased on a sentiment analysis dataset of Hala-Khorshid T.V program dataset. It achieves the following results on the evaluation set:

• Loss: 4.1739

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: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss
4.7733	1.0	285	4.4369
4.2875	2.0	570	4.3070
4.132	3.0	855	4.1908

Framework versions

• Transformers 4.25.1
• Pytorch 1.13.1+cu116
• Datasets 2.8.0
• Tokenizers 0.13.2