Michael Ekstrand
Michael Ekstrand
Should add (and test) support for running the k-NN recommenders on GPUs. Related: - #374 - #381 - #377
Right now, the new item KNN code is using brute force for neighborhood limiting. We should be able to implement the "kvp heap" logic with bulk heaps to do this...
It would be useful to be able to load and configure algorithms from serialized files (e.g. TOML). This will enable general command-line tools.
After #374, I need to review the documentation to make sure that PyTorch usage is properly documented and Numba is gone.
Instead of #305, it may be a good idea to just use `ipyparallel` for batch evaluation, so we can reuse their management code instead of continuing to do our own...
Checklist - [x] Title of this PR is meaningful: e.g. "Adding my_nifty_package", not "updated meta.yaml". - [x] License file is packaged (see [here](https://github.com/conda-forge/staged-recipes/blob/5eddbd7fc9d1502169089da06c3688d9759be978/recipes/example/meta.yaml#L64-L73) for an example). - [x] Source is...
I realize I'm probably operating in very unsupported territory here, but I was doing some experimenting with Sharp on WASM and wanted to report what I found in case the...
Warning during tests: ``` /Users/mde48/LensKit/lkpy/lenskit/lenskit/crossfold.py:282: DeprecationWarning: DataFrameGroupBy.apply operated on the grouping columns. This behavior is deprecated, and in a future version of pandas the grouping columns will be excluded from...
I'm seeing tests that use parallel processing lockup on my MacBook Pro. Have not isolated cause (or other variables).
The [structlog](https://www.structlog.org/en/stable/) package provides *structured logging* support for Python. I would like to evaluate whether to use structured logging for LensKit.