apaszke

PyTorch was never a Chainer fork. The whole codebase are C libs from Lua Torch, and a bunch of Python code that was written entirely for this project. Chainer was an inspiration, but no code was ever shared between those two projects.

PyTorch uses NVIDIA NCCL for multi-GPU communication (under BSD license). Gloo is only one of the three backends that can be used for distributed communication out of the box.

It's not controlled by Facebook in any way. It's true that a large part of the core team works there, but development is public and guided by community needs first.

It's not like you have to give up a lot - the graphs are simple data structures and creating them is not the expensive part of the training. The computation has to be re-done at every step in a static framework too, and this is the part that matters.

gloo is only one of the three currently supported backends. One can easily switch to MPI, and pick an implementation that comes with a license you want.

It supports numpy, and conversion from PyTorch tensors to numpy arrays is a matter of calling the numpy() method.

To not have the kernel launch overhead you'd need to stop launching GPU kernels but that's now how things work in any framework ;)

These ops are just not needed in PyTorch. while is just a Python while loop. Scan is a for loop, map is a list comprehension that applies modules. No need for anything fancy.

Yes! There's an issue on that, where we'll be coordinating the work: https://github.com/pytorch/pytorch/issues/494

You save the parameters and the code of the model definition

torch?

But it is possible to write your model in purely functional style. Check out the PR to examples repo with functional ResNets https://github.com/pytorch/examples/pull/22.

torch.nn offers a very similar interface to Keras (e.g. see Alexnet definition at https://github.com/pytorch/vision/blob/master/torchvision/mo...).

No! Python 0 based indexing everywhere.

The core libraries are the same as in Lua torch, but the interface is redesigned and new.

I am aware of 'ul' conflict, but I haven't heard of it before, so I just hoped that it's not popular :x If you've found a bug in installation script then please post an issue at GitHub :)

I am aware that there are lots of universal window managers, but I just wanted something simple that would have a terminal API. This is why I've created termtile

I know that there are loads of universal window managers, but I wanted a one with a terminal API ;)

Thanks! It's nice to hear that!

What is the bug you're talking about?

I didn't include it, because OS X can already make the window full screen. Do you think it would be useful just to resize the window without creating a new desktop?

Hey guys, I'm a creator of termitile. Many of you have pointed out that there are lots of window managers for OS X and my intention wasn't to create just another one. What I really needed was a bunch of script that I could alias and use directly from the terminal without cluttering my global keyboard shortcuts. I couldn't find any other programs that had a terminal API, so I've just hacked this one :)

So just to be clear, it's not intended to be a replacement for Amethyst, Divvy, or any other universal manager, but just a quick solutions just for the terminal, because you can control it with regular commands!

I hope you like it!

I know about it, but I didn't have an 'll' alias so i picked it just to be consistent with 'rr'. If you wish you can configure the aliases in the install.sh file (if you decline to overwrite an existing alias it asks you to type another one) :)