[release] Prepare dummy binary for 2.0 release (#1115)

[1.13] Tighten packaging versions (#1055)

Allow batch norm with all variations of batching when training=False (#…

…958)

* allow batch norm with all variations of batching when training=False

* make running mean/var always call contiguous

[release] fix cuda versions

[release] fix docs version

[release] binaries build