Thank you for your excellent work, and congratulations on having your work accepted by ICLR.
I followed the default configuration in your scripts/iterative_fl.sh for debugging. I'm currently using a single A100 (80GB) GPU with train_batch_size=4 and gradient_accumulation_steps=4, and I encountered an OOM error.
Additionally, when I changed the settings to train_batch_size=2, the memory usage was 72GB.
When I further changed it to train_batch_size=1, the memory usage increased to 79GB.
I was wondering if you encountered this issue in your experiments. Could you kindly share the specific memory usage in your setup?
Thank you for your excellent work, and congratulations on having your work accepted by ICLR.
I followed the default configuration in your scripts/iterative_fl.sh for debugging. I'm currently using a single A100 (80GB) GPU with train_batch_size=4 and gradient_accumulation_steps=4, and I encountered an OOM error.
Additionally, when I changed the settings to train_batch_size=2, the memory usage was 72GB.
When I further changed it to train_batch_size=1, the memory usage increased to 79GB.
I was wondering if you encountered this issue in your experiments. Could you kindly share the specific memory usage in your setup?