z

The switch to the new back-end is coming and with it the doors will be open to really nailing down a productive workflow. Below are my thoughts on what are...

This P.R. starts the process of adding a `SlurmDistributor` strategy to handle distributing work across a SLURM cluster.

Support `DDP(static_graph=True)` and gradient accumulation

3

I got `SystemError: returned NULL without setting an error` when setting **accumulate_grad_batches = 2**. But I see nothing helpful in the log. Error gone when changing `DDPStrategy(static_graph=False,)`, or `accumulate_grad_batches` back...

I have this on my local clone, just a setup.py to make this repo "```pip install -e .```"-able. I can work on packaging it for PyPi if that's of interest...