no loss for training with small batch size,

When I train the Deepsurv model with small batch size (64), there is no loss and always early stop


However, if I train with larger batch size (1024), it almost always training properly

May I ask what might be the possible reason ? I am not familiar with the progress bar

Start by reproducing the Deepsurv training behavior with batch sizes 64 and 1024, comparing the reported loss, early stopping, and progress bar output. No files or tests are named, so trace the training entry point used by Deepsurv before determining whether the issue is in training or only progress reporting. Done means the cause is established and the affected behavior is corrected or clearly documented.

Written by the indexing model from the issue text.