mlx backend

[krzysztofrusek]

Are there any plans to include ml-explore/mlx as tfp backend?
It would be great for mac users running statistical computations locally.
If there are no such plans, I would be grateful for hints on how to add it, since the backend system has no public API.

[brianwa84]

There are not currently such plans. Testing might be quite challenging, unless MLX provides an emulator? The best documentation about multi-backend support is in here: https://github.com/tensorflow/probability/blob/main/SUBSTRATES.md You would need to work with bazel directly to compile and run some tests locally, since the backends are compiled by rewriting TFP code into new submodules. IIRC TFP's BUILD rules are incompatible with the latest bazel for some reason: https://github.com/tensorflow/probability/blob/main/testing/run_github_tests.sh#L27 Feel free to post back w/ questions if you start seriously pursuing. Brian Patton | Software Engineer | ***@***.***

…

On Fri, Mar 22, 2024 at 3:05 AM Krzysztof Rusek ***@***.***> wrote: Are there any plans to include ml-explore/mlx <https://github.com/ml-explore/mlx> as tfp backend? It would be great for mac users running statistical computations locally. If there are no such plans, I would be grateful for hints on how to add it, since the backend system has no public API. — Reply to this email directly, view it on GitHub <#1802>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFJFSI2ORSI6MT5HMCXUIYTYZPJ47AVCNFSM6AAAAABFCZMPG2VHI2DSMVQWIX3LMV43ASLTON2WKOZSGIYDCOBTHAZDAMI> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>