Task doesn't seem to converge: Summurization, Translation

[tchaton]

🐛 Bug

To Reproduce

On a machine with 1 GPU.

FLASH_TESTING=1 python flash_examples/finetuning/summarization.py

FLASH_TESTING=1 python flash_examples/finetuning/translation.py

Steps to reproduce the behavior:

1. Go to '...'
2. Run '....'
3. Scroll down to '....'
4. See error

Code sample

Expected behavior

Environment

• PyTorch Version (e.g., 1.0):
• OS (e.g., Linux):
• How you installed PyTorch (conda, pip, source):
• Build command you used (if compiling from source):
• Python version:
• CUDA/cuDNN version:
• GPU models and configuration:
• Any other relevant information:

Additional context