Consistently apply MIT OR Apache-2.0 license and clarify licence on test files

[pixelspark]

#149

@haixuanTao , @SludgePhD , @CameronTofer , @zimond , @aliemjay , @philpax , @maekawatoshiki please confirm you agree to dual-licensing your contributions as "MIT OR Apache-2.0" as clarified in this pull request.

[philpax]

I license past and future contributions under the dual MIT/Apache-2.0 license, allowing licensees to choose either at their option.

[maekawatoshiki]

I agree to dual-license my past & future contributions as MIT OR Apache-2.0.

[haixuanTao]

OK, for both licenses MIT or APACHE as well for both past and future.

[SludgePhD]

I agree to dual-license my past & future contributions as MIT OR Apache-2.0.

[aliemjay]

I agree to license my work under MIT OR Apache-2.0

[zimond]

I agree to license my work under MIT OR Apache-2.0

[pixelspark]

OK, so that leaves @CameronTofer . Can you agree to dual-license your contribution (#139, see your commits )?

[pixelspark]

We are now licensed "MIT OR Apache-2.0" ! Thanks all for the replies.

@CameronTofer unfortunately I had to remove your contribution for now. Creating certainty about the licensing situation is in the best interest of the project. If you agree to relicense your contributions we can simply merge it back (I prepared a PR: #152).

@antimora curious to see what you will be using in burn-rs!

[antimora]

@pixelspark Awesome! Thank you for the clarification.

We would like to use your code in the Burn project as a burn-wgpu backend. Currently, Burn has two backends (torch and ndarray). The burn-wgpu backend will allow accessing GPU on my many platforms without the need to install torch. This is great for training and inference. Also, burn-wgpu will allow acceleration on the web. Just recently we published an example running MNIST inference on the web using wasm. You can try the live demo. So if we have a burn-wgpu backend, it will be easy just to swap the burn-ndarray backend with burn-wgpu, to have GPU acceleration on the web.

(I am tagging @nathanielsimard here so he's aware of this ticket. He's the creator and developer of Burn. He was also excited about the possibility to run train/inference on AMD GPUs).