Joqsan
Joqsan
Hi, I'd like to take on the EDICT implementation, if someone hasn't started it.
I'll give it a try.
Hi @hemangjoshi37a As far I can see the only difference would be with full reductions. For instance, instead of having to do `y.sum().numpy()[0]` we'll just need `y.sum().numpy()`.
> Very cool! A few comments, and a bunch of tests failing. Yeah. Looking into it.
I added a (skipped) test for tensors with zeros in shape. It looks like I was wrong and the problem doesn't lie directly in backward itself but in doing binary...
> You interested in a job at the tiny corp by chance? Both this and your last PR were quite high quality. Yes, I think it'd be pretty cool, since...
> Full time jobs are in person in San Diego, or we have remote task based contract jobs. Which are you interested in? > > See #788 for the hiring...
> Cool, finish this up and then think of which contract you want to tackle. There's the big 10k ones that should take 1-6 weeks, the 1k ones which are...
> A bunch of the failing onnx tests depend on this. > > https://gist.github.com/geohot/546d7fded8df870f6e246b857091e331 Working on this.
Tests are passing (: Some comments: - Refactored `__getitem__` and added support for strides >= 1. Also left the comments for this function as they were, since they are still...