Pretrain, finetune ANY AI model of ANY size on 1 or 10,000+ GPUs with zero code changes.
Fix ReduceLROnPlateau scheduler with check_val_every_n_epoch
- Only update plateau schedulers on epochs when validation runs - This prevents errors when monitored metrics are not available - Added test case for this scenario Co-authored-by: Borda <[email protected]>
C
copilot-swe-agent[bot] committed
817243398495a67ed90b460c1dd288be00c87af3
Parent: 531c1e9