Expected training time for HistAug models #3
Description
Hi, and thanks for the great work on HistAug!
Before I start running experiments, I'd love to get a sense of the computational cost involved.
The README mentions hardware details (V100 32GB) for inference, but I couldn't find information regarding the training cost.
Could you share how long a full training run took in your own experimental environment? Specifically:
- How long did training take end-to-end (e.g., per epoch or total hours)?
- Did training time differ notably across the foundation models evaluated (e.g., CONCH, Virchow2, UNI)?
This would help me better understand the scale of compute required to reproduce your results.
Thank you!