Daniel Falbel

This allows converting between torch tensors and numpy arrays with no copy: ``` library(torch) x

Continuous benchmarking

1

Explore https://github.com/r-lib/bench#continuous-benchmarking to avoid performance regressions

Implement Iterable Dataset class and Data loaders for it

Consider preserving precision when casting doubles to char

2

`flatten_chr` and `map_chr` uses a maximum 6 digits precision when casting doubles to char. ```r x [1] "0.000000" "0.000000" sapply(x, as.character) #> [1] "1e-07" "1e-08" ```

Rtools 42 adds MinGW to the path

2

It seems that when installing rtools 42 the mingw compiler is added to the `PATH` even if setting the `windows-path-include-mingw` variable. This contradicts what's documented [here](https://github.com/r-lib/actions/blob/v2-branch/setup-r/action.yml#L35): > This argument is...

Prepare for release: * [x] `git pull` * [x] Check [current CRAN check results](https://cran.rstudio.org/web/checks/check_results_torch.html) * [x] Bump required R version in DESCRIPTION to 3.6 * [x] [Polish NEWS](https://style.tidyverse.org/news.html#news-release) * [x]...

I think `names(jit_ops)` should work and return the list of possible operators.