Rapidsai

----
- Publish branch-0.14 to conda (262)
- Local gpuCI build script (285)

----
- Use RMM's `DeviceBuffer` directly (235) `John Kirkham`_
- Add RMM pool support from dask-cuda-worker/LocalCUDACluster (236) `Peter Andreas Entschev`_
- Restrict CuPy to <7.2 (239) `Benjamin Zaitlen`_
- Fix UCX configurations (246) `Peter Andreas Entschev`_
- Respect `temporary-directory` config for spilling (247) `John Kirkham`_
- Relax CuPy pin (248) `John Kirkham`_
- Added `ignore_index` argument to `partition_by_hash()` (253) `Mads R. B. Kristensen`_
- Use `"dask"` serialization to move to/from host (256) `John Kirkham`_
- Drop Numba `DeviceNDArray` code for `sizeof` (257) `John Kirkham`_
- Support spilling of device objects in dictionaries (260) `Mads R. B. Kristensen`_

----

- Add ucx-py dependency to CI (212) `Ray Douglass`_
- Follow-up revision of local_cudf_merge benchmark (213) `Richard (Rick) Zamora`_
- Add codeowners file (217) `Ray Douglass`_
- Add pypi upload script (218) `Ray Douglass`_
- Skip existing on PyPi uploads (219) `Ray Douglass`_
- Make benchmarks use rmm_cupy_allocator (220) `Mads R. B. Kristensen`_
- cudf-merge-benchmark now reports throughput (222) `Mads R. B. Kristensen`_
- Fix dask-cuda-worker --interface/--net-devices docs (223) `Peter Andreas Entschev`_
- Use RMM for serialization when available (227) `Peter Andreas Entschev`_

----

- Use UCX-Py initialization API (152) `Peter Andreas Entschev`_
- Remove all CUDA labels (160) `Mike Wendt`_
- Setting UCX options through dask global config (168) `Mads R. B. Kristensen`_
- Make test_cudf_device_spill xfail (170) `Peter Andreas Entschev`_
- Updated CI, cleanup tests and reformat Python files (171) `Mads R. B. Kristensen`_
- Fix GPU dependency versions (173) `Dillon Cullinan`_
- Set LocalCUDACluster n_workers equal to the length of CUDA_VISIBLE_DEVICES (174) `Matthew Rocklin`_
- Simplify dask-cuda code (175) `Mads R. B. Kristensen`_
- DGX inherit from LocalCUDACluster (177) `Mads R. B. Kristensen`_
- Single-node CUDA benchmarks (179) `Mads R. B. Kristensen`_
- Set TCP for UCX tests (180) `Peter Andreas Entschev`_
- Single-node cuDF merge benchmarks (183) `Mads R. B. Kristensen`_
- Add black and isort checks in CI (185) `Peter Andreas Entschev`_
- Remove outdated xfail/importorskip test entries (188) `Peter Andreas Entschev`_
- Use UCX-Py's TopologicalDistance to determine IB interfaces in DGX (189) `Peter Andreas Entschev`_
- Dask Performance Report (192) `Mads R. B. Kristensen`_
- Allow test_cupy_device_spill to xfail (195) `Peter Andreas Entschev`_
- Use ucx-py from rapidsai-nightly in CI (196) `Peter Andreas Entschev`_
- LocalCUDACluster sets closest network device (200) `Mads R. B. Kristensen`_
- Expand cudf-merge benchmark (201) `Richard (Rick) Zamora`_
- Added --runs to merge benchmark (202) `Mads R. B. Kristensen`_
- Move UCX code to LocalCUDACluster and deprecate DGX (205) `Peter Andreas Entschev`_
- Add markdown output option to cuDF merge benchmark (208) `Benjamin Zaitlen`_

----

- Change the updated new_worker_spec API for upstream (128) `Matthew Rocklin`_
- Update TOTAL_MEMORY to match new distributed MEMORY_LIMIT (131) `Peter Andreas Entschev`_
- Bum Dask requirement to 2.4 (133) `Matthew Rocklin`_
- Use YYMMDD tag in nightly build (134) `Markku Luukkainen`_
- Automatically determine CPU affinity (138) `Peter Andreas Entschev`_
- Fix full memory use check testcase (139) `Sangeeth Keeriyadath`_
- Use pynvml to get memory info without creating CUDA context (140) `Peter Andreas Entschev`_
- Pass missing local_directory to Nanny from dask-cuda-worker (141) `Peter Andreas Entschev`_
- New worker_spec function for worker recipes (147) `Peter Andreas Entschev`_
- Add new Scheduler class supporting environment variables (149) `Peter Andreas Entschev`_
- Support for TCP over UCX (152) `Peter Andreas Entschev`_


.. _`Matthew Rocklin`: https://github.com/mrocklin
.. _`Peter Andreas Entschev`: https://github.com/pentschev
.. _`Markku Luukkainen`: https://github.com/mluukkainen
.. _`Sangeeth Keeriyadath`: https://github.com/ksangeek

---

- Fix serialization of collections and bump dask to 2.3.0 (118) `Peter Andreas Entschev`_
- Add versioneer (88) `Matthieu Bulte`_
- Python CodeCov Integration (91) `Dillon Cullinan`_
- Update cudf, dask, dask-cudf, distributed version requirements (97) `Peter Andreas Entschev`_
- Improve device memory spilling performance (98) `Peter Andreas Entschev`_
- Update dask-cuda for dask 2.2 (101) `Matthew Rocklin`_
- Streamline CUDA_REL environment variable (102) `Olli Koskinen`_
- Replace ncores= parameter with nthreads= (101) `Matthew Rocklin`_
- Fix remove CodeCov upload from build script (115) `Dillon Cullinan`_
- Remove CodeCov upload (116) `Dillon Cullinan`_

.. _`Matthieu Bulte`: https://github.com/matthieubulte
.. _`Dillon Cullinan`: https://github.com/dillon-cullinan
.. _`Peter Andreas Entschev`: https://github.com/pentschev
.. _`Matthew Rocklin`: https://github.com/mrocklin
.. _`Olli Koskinen`: https://github.com/okoskinen