Add logs for gpt2_large and update loss tracking logic

[Usman-Rafique]

Add GPT2-Large Logs and Update Loss Tracking

Changes

1. Added logs for GPT2-Large model

   • Includes performance metrics and training progress for the new model

2. Updated loss tracking logic

   • Removed validation loss tracking
   • Now only tracking training loss due to the large dataset size
   • Implemented more efficient logging of training loss

Rationale

• GPT2-Large logs provide insights into the performance of this larger model variant
• Updated loss tracking reflects the reality of our training process, where we typically
  process less than 1% of the total dataset
• Focusing on training loss provides a more relevant metric for our current use case

Testing

• Verified correct logging of GPT2
• Training is running fine
• All tests are passing