老董

> I laid some foundations in the `variable-rework` branch. As of now I'm quite busy with other duties, I'll be working full time on the project in a couple of...

@drallensmith ,@Ns85, @evolvingfridge , I am still at a loss--- whether the implemented CTRNN is an extension of RNN? I mean rnn structure (usually lstm) would sometimes be used to...

@Osrip ，check this: [pytorch_NEAT](https://github.com/uber-research/PyTorch-NEAT)

@zhreshold, Sorry, I didn't make it clear, the "image" I said refers to picture.for example, if I need to detect objects in a game, it is inefficient to capture it...

> json文件还是不能有注释 确实，即使目前，json里还是不能有注释

> This looks very nice, thanks for sharing! Once this has been polished/tested a bit, we could probably mention it in the `tch-rs` main readme file for better discoverability if...

> If it's just to avoid having an additional dependency for crates that would want to use this, I would lean more towards keeping an external crate, and in general...

@LaurentMazare , thanks for your suggestion. But here with tch v0.5 on win10, I wonder why only this is workable `let filename = std::env::temp_dir().join(format!("tch2-{}", std::process::id()));`, and this `let filename =...

@LaurentMazare, the vgg code I've checked before, because I want to know how to set train/eval mode for a torch model. And what I want here is a network that...

> @dbsxdbsx Had the same issue. After some searching on the repo, realised this is a duplicate of #291. The solution, as per [#291 (comment)](https://github.com/LaurentMazare/tch-rs/issues/291#issuecomment-890404057), is to add > >...