A lightweight wrapper for PyTorch that provides a simple declarative API for context switching between devices, distributed modes, mixed-precision, and PyTorch extensions.
Updated launcher docs to reflect new torchrun launcher
Handle DistributedSampler check when in distributed mode
Updated README docs to reflect correct use of enum options
Updated Dockerfiles to correctly install Horovod -- missing [pytorch] dep
Moved to **kwargs style of passing keyword args to the sub-classed PyTorch dataloader -- multiprocessing might be handling the args to some fork/spawn ops in an obfuscated way that only **kwargs will work (see how Lightning handles this)
**kwargs
style of passing keyword args to the sub-classed PyTorch dataloader -- multiprocessing might be handling the args to some fork/spawn ops in an obfuscated way that only**kwargs
will work (see how Lightning handles this)