Package Details: python-jaxlib-cuda 0.4.16-4

Git Clone URL: (read-only, click to copy)
Package Base: python-jaxlib-cuda
Description: XLA library for JAX
Upstream URL:
Keywords: deep-learning google jax maching-learning xla
Licenses: Apache
Conflicts: python-jaxlib
Provides: python-jaxlib
Submitter: daskol
Maintainer: daskol
Last Packager: daskol
Votes: 2
Popularity: 0.031910
First Submitted: 2023-02-12 23:18 (UTC)
Last Updated: 2023-09-22 10:43 (UTC)

Latest Comments

daskol commented on 2023-06-12 19:22 (UTC) (edited on 2023-06-12 19:25 (UTC) by daskol)

@actionless @petronny @mergen I am trying to fix the error. The issue is that Arch moved to gcc-13 but nvcc supports only gcc-12 as a host compiler. It is a bit strange since this kind of issue is solved for TensorFlow and there is no such an issue for example openxla/xla at all.

actionless commented on 2023-05-10 10:48 (UTC)

i'm having this error:

/usr/include/bits/mathcalls.h(62): error: identifier "_Float32" is undefined extern _Float32 cosf32 (_Float32 __x) noexcept (true); extern _Float32 __cosf32 (_Float32 __x) noexcept (true);

mb it doesn't like latest gcc

petronny commented on 2023-03-28 11:06 (UTC) (edited on 2023-03-28 11:06 (UTC) by petronny)

I'm getting the same error like @mergen does.

Full build log:

PS. It's built on a CPU-only machine.

mergen commented on 2023-03-26 08:57 (UTC) (edited on 2023-03-26 08:57 (UTC) by mergen)

My computer has cuda 12.1.0-3 installed from the community repo and I am able to use it with pytorch or pip installed jax.

However, when I try to install this package, I get the following error:

ERROR: Skipping '@org_tensorflow//tensorflow/compiler/xla/python:enable_gpu': no such package '@local_config_cuda//cuda': Repository command failed
Could not find any cuda.h matching version '' in any subdirectory:
WARNING: Target pattern parsing failed.