Hi maintainers,
The current PyPI release of jwave (0.2.1, from September 2024) depends on jaxdf <0.3.0, which in turn caps jax <0.5.0. This prevents installing jwave from PyPI in environments running modern JAX (0.9.x).
However, jaxdf 0.3.0 — which supports jax 0.9.x — is already available on PyPI, and the jwave main branch already depends on jaxdf >=0.3.0 and works with jax 0.9.x. So it seems like the only missing piece is a new PyPI release of jwave itself.
Would it be possible to cut a new release (e.g. 0.3.0) to PyPI from the current main branch?
Our use case: we maintain openlifu-python, a library for transcranial focused ultrasound simulation that depends on jwave. We need jax 0.9.x because it is the earliest version providing CUDA support on aarch64 (NVIDIA GH/GB-series platforms), and the jax <0.5.0 cap from the current PyPI release blocks us from using jwave in that environment.
Thank you for your work on jwave — happy to help test a pre-release if that would be useful.
Hi maintainers,
The current PyPI release of jwave (0.2.1, from September 2024) depends on jaxdf <0.3.0, which in turn caps jax <0.5.0. This prevents installing jwave from PyPI in environments running modern JAX (0.9.x).
However, jaxdf 0.3.0 — which supports jax 0.9.x — is already available on PyPI, and the jwave main branch already depends on jaxdf >=0.3.0 and works with jax 0.9.x. So it seems like the only missing piece is a new PyPI release of jwave itself.
Would it be possible to cut a new release (e.g. 0.3.0) to PyPI from the current main branch?
Our use case: we maintain openlifu-python, a library for transcranial focused ultrasound simulation that depends on jwave. We need jax 0.9.x because it is the earliest version providing CUDA support on aarch64 (NVIDIA GH/GB-series platforms), and the jax <0.5.0 cap from the current PyPI release blocks us from using jwave in that environment.
Thank you for your work on jwave — happy to help test a pre-release if that would be useful.