Add Early Stopping Mechanism

[pearguacamole]

Implement Early Stopping to Prevent Overfitting and Optimize Training

Closes #10

Description:

This pull request implements an EarlyStopping feature to halt training when there is no significant improvement in validation loss, helping to prevent overfitting and reduce unnecessary computation.

Implementation:

• Added an EarlyStopping class that monitors validation loss during training.
• Introduced a patience parameter to define how many epochs to wait after the last improvement before stopping training.
• Included feedback to the user about whether early stopping was triggered and at which epoch.

Key Changes:

1. New EarlyStopping class:

   • Parameters:
     • patience: Number of epochs to wait for validation loss improvement.
     • delta: Minimum change to qualify as an improvement.
   • Functionality: Stops training when no significant validation loss improvement is detected after the given patience period.

2. Training Loop Update:

   • Integrated the EarlyStopping callback into the training loop, checking for improvements in validation loss after each epoch.
   • Added logging to notify users when early stopping is activated.

3. Verbosity toggle:

   • Integrated the verbosity param into the training loop
   • It can toggle between showing progress bar, or printing verbose log of each epoch during training

---

Testing:

• Manually tested the early stopping functionality on a sample model with varying patience values.
• Verified that early stopping is triggered after the specified patience period without significant validation loss improvement.

---

Additional Notes:

• Can be further enhanced to work with other metrics beyond validation loss in future iterations.

---

[ravin-d-27]

Hi @pearguacamole ! Thank you! Will check and merge it soon.

[ravin-d-27]

Hi @pearguacamole ! Thank you for your contribution ! Please star the repository for better reach of this project !