[Diff since v0.11.4](https://github.com/FluxML/Flux.jl/compare/v0.11.4...v0.11.5)


**Closed issues:**
- Huge performance difference between sparse and dense representation on GPU (#189)
- onecold is very slow (#556)
- onecold does not work on CuMatrix (#864)
- Multi-dimensional onehot (#1229)

**Merged pull requests:**
- Add CTC loss to new Losses module (#1287) (@maetshju)
- remove implicit conversions (#1393) (@CarloLucibello)
- Add Parallel layer (#1462) (@darsnack)
- Improve docs for `crossentropy` & friends (#1463) (@mcabbott)
- One-arg unsqueeze method (#1469) (@mcabbott)
- remove dataset tests (#1470) (@CarloLucibello)
- Fix RNN tests on GPU (#1473) (@jeremiedb)