[Feature]: Restructure the code to build a wheel and use importlib to embed non-python files

[bioinfornatics]

Suggestion Description

Dear Tensile Team,

I am reaching out to advocate for the structuring of code in the RocM ecosystem in alignment with Linux packaging conventions. This initiative is not merely a step towards enhancing the usability of RocM but is also a strategic move to significantly elevate its prominence and accessibility in the development community.

The RocM ecosystem, despite its powerful capabilities, is often perceived as challenging to deploy, particularly when adhering to Linux standards. This perception acts as a barrier, deterring potential adopters and limiting the ecosystem's reach and impact. By restructuring the code to facilitate packaging (creating wheel distributions, for instance), we can dismantle these barriers, simplifying the installation and deployment process.

Benefits of Structuring Code for Packaging:

• Ease of Deployment: Structured and packaged code can be seamlessly integrated and deployed, making the RocM ecosystem more approachable for new and existing users.
• Standard Compliance: Aligning with Linux packaging conventions enhances compatibility and reliability, fostering trust and wider acceptance within the Linux user base.
• Community Engagement: A more accessible and user-friendly structure encourages community engagement, leading to more contributions, enhanced features, and robust support.
• Enhanced Collaboration: Packaged code eases the process of collaboration, allowing developers to focus on innovation rather than grappling with deployment intricacies.

The pull request at https://github.com/ROCm/Tensile/pull/1870 serves as a testament to the community's acknowledgment and the initial steps towards this enhancement. It is an embodiment of the potential improvements that structured and packaged code can bring to the RocM ecosystem.

In conclusion, restructuring the code to support standard packaging conventions is not just an upgrade; it's a transformative step towards making the RocM ecosystem more robust, accessible, and future-proof. I am looking forward to your thoughts on this proposal and am eager to contribute to this positive change.

Best regards,

Operating System

No response

GPU

No response

ROCm Component

No response

[bstefanuk]

@bioinfornatics Thank you for submitting this issue, we appreciate your time and commitment to supporting the project and completely agree that improving the packaging of Tensile will help improve adoption of the ROCm ecosystem.

I've reviewed your associated PR where you mention you "would like to package ROCm software to industrialize [your] ROCm DeepLearning environment". We would be glad to support your adoption of ROCm!

Could you please clarify the exact nature of the change you would like to see? How are you integrating Tensile into your project? How are you packaging ROCm?