RFC-0045-decoupling-cuda-code.md [commenting] #82
Conversation
highlights manuscript
file decoupling and directory restructuring
Prior Art
Add next steps
Update RFC-0039-decouple-cuda-codes.md
Add next steps
Check word usage
|
Hi @bjtuwjx! Thank you for your pull request and welcome to our community. Action RequiredIn order to merge any pull request (code, docs, etc.), we require contributors to sign our Contributor License Agreement, and we don't seem to have one on file for you. ProcessIn order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA. Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with If you have received this in error or have any questions, please contact us at [email protected]. Thanks! |
|
Thank you for signing our Contributor License Agreement. We can now accept your code for this (and any) Meta Open Source project. Thanks! |
| ## **Motivation** | ||
|
|
||
| For a long time, NVIDIA GPUs and the CUDA architecture have dominated the PyTorch ecosystem. However, as an increasing number of vendors introduce their high-performance AI chips, current ecosystem is revealing the following key issues: | ||
| - *Code coupling*. The CUDA code is too tightly coupled with the PyTorch codebase, resulting in poor modularity and high maintenance costs. |
There was a problem hiding this comment.
Can you give some details on this? What is the more concrete impact on modularity? And how this would reduce maintenance cost?
|
|
||
| For a long time, NVIDIA GPUs and the CUDA architecture have dominated the PyTorch ecosystem. However, as an increasing number of vendors introduce their high-performance AI chips, current ecosystem is revealing the following key issues: | ||
| - *Code coupling*. The CUDA code is too tightly coupled with the PyTorch codebase, resulting in poor modularity and high maintenance costs. | ||
| - *Integration effort*. Currently, different hardware backends may adopt varying integration methods into PyTorch. The integration approaches and code lack standardization and consistency, leading to a significant amount of repetitive code and substantial integration effort. |
There was a problem hiding this comment.
We have done a lot of work with the PrivateUse1 backend extension point. In particular, added all the asked-for extension points, building OpenReg as a in-tree testing backend for this extension, added autoload, actively working on more documentation and updated many tools to use the accelerator API to enable smooth transition for the end user.
What would the proposed RFC provide on top of this? And why would we prefer investing in this refactor rather than continue improving the PrivateUse1-related extension points?
| For a long time, NVIDIA GPUs and the CUDA architecture have dominated the PyTorch ecosystem. However, as an increasing number of vendors introduce their high-performance AI chips, current ecosystem is revealing the following key issues: | ||
| - *Code coupling*. The CUDA code is too tightly coupled with the PyTorch codebase, resulting in poor modularity and high maintenance costs. | ||
| - *Integration effort*. Currently, different hardware backends may adopt varying integration methods into PyTorch. The integration approaches and code lack standardization and consistency, leading to a significant amount of repetitive code and substantial integration effort. | ||
| - *Code migration*. Due to the lack of integration code specification, different hardware backends provide APIs with varying names and styles, resulting in high code migration costs for PyTorch users. |
There was a problem hiding this comment.
We are building the accelerator API to address this particular points (relatively independent on how the backend itself is implemented (in/out of tree)).
How would the proposal here help compared to continuing to extend the accelerator API?
This RFC proposal aims to decouple the CUDA-related code from the PyTorch main codebase and refactor it into an independent and modularized directory hierarchy with the help of a build optimization toolkit. Specifically, the proposal covers the following work:
Click here for a preview of this RFC.