Distilbert Base Uncased Ag News

🚀 TextAttack Model Card

This model card introduces a fine - tuned distilbert - base - uncased model for sequence classification. It leverages TextAttack and the ag_news dataset loaded via the nlp library, offering high - performance classification capabilities.

🚀 Quick Start

This distilbert-base-uncased model was fine - tuned for sequence classification. Here are the key details of the fine - tuning process:

• Dataset: The ag_news dataset was loaded using the nlp library.
• Training Parameters:
  • Epochs: The model was fine - tuned for 5 epochs.
  • Batch Size: A batch size of 32 was used.
  • Learning Rate: The learning rate was set to 2e - 05.
  • Maximum Sequence Length: The maximum sequence length was 128.
• Loss Function: Since this was a classification task, the model was trained with a cross - entropy loss function.
• Performance: The best score the model achieved on this task was 0.9478947368421052, as measured by the eval set accuracy, which was found after 1 epoch.

For more information, check out TextAttack on Github.