Andrey Shedko
Results
2
issues of
Andrey Shedko
## What? It would be nice to issue a warning when a weight (this includes any learned parameters, such as bias values, mean and variance in batch norm, etc.) is...
area/quant
type/suggestion
### Bug description When running a `torch.compile`-d model with ModelParallelStrategy, saving a non-distributed checkpoint fails. That is due to a mismatch in fqn paths between `get_optimizer_state_dict` and `rekey_optim_state_dict`. get_optimizer_state_dict replaces...
bug
checkpointing
needs triage
ver: 2.5.x