First of all, thank you very much for creating and maintaining the excellent pytorch_dlprim project! It provides a very valuable opportunity for us to run PyTorch on embedded devices without CUDA.
When trying to deploy the project to devices based on the ARM64 (aarch64) architecture, such as the RK3588 with Mali GPU, I noticed that the currently released pre-built wheel packages (in Releases or on PyPI) only support linux_x86_64 and win_amd64.
In order to lower the barrier to entry for using it on ARM devices and allow more developers and researchers to quickly test and utilize your work, would it be possible to consider providing pre-built wheel packages for linux_aarch64 or manylinux2014_aarch64 in the future?
We understand that local compilation on ARM devices is feasible, but the process can be time-consuming and requires a more advanced setup from the user. A pre-built package would significantly improve the initial user experience and promote the project's adoption in edge computing and embedded AI fields.
Thank you again for your contribution and consideration!
First of all, thank you very much for creating and maintaining the excellent pytorch_dlprim project! It provides a very valuable opportunity for us to run PyTorch on embedded devices without CUDA.
When trying to deploy the project to devices based on the ARM64 (aarch64) architecture, such as the RK3588 with Mali GPU, I noticed that the currently released pre-built wheel packages (in Releases or on PyPI) only support linux_x86_64 and win_amd64.
In order to lower the barrier to entry for using it on ARM devices and allow more developers and researchers to quickly test and utilize your work, would it be possible to consider providing pre-built wheel packages for
linux_aarch64ormanylinux2014_aarch64in the future?We understand that local compilation on ARM devices is feasible, but the process can be time-consuming and requires a more advanced setup from the user. A pre-built package would significantly improve the initial user experience and promote the project's adoption in edge computing and embedded AI fields.
Thank you again for your contribution and consideration!