Fix non-fp tensor bugs of contiguous activation checkpointing (#1376)
* Update checkpointing.py
* Fix formatting
* Add flexibility of pipeline module and engine
* Separate PRs
* Separate PRs
* Update checkpointing.py
* Update checkpointing.py
* Reflect code review for contiguous activation checkpointing
* remove useless condition
Co-authored-by: Olatunji Ruwase <olruwase@microsoft.com>