📚 The doc issue
We need to improve the documentation for our delegates. Right now, information is sparse and it's difficult to find answers to common questions. We need to create a standard template for delegate documentation and then ensure that each delegate has a page in the docs under "Backend Delegates" based on this template.
Key information to cover includes:
- System requirements, both for development and deployment - what hardware and software is supported? That may include iOS/Android versions, supported NPU hardware, microcontroller families, toolchains, and any other relevant requirements.
- At a high level, what operators are well supported?
- A code snippet / sample for export, including the relevant includes and usage of the partitioner. What partitioner options are there? For example, AOT compile for Core ML, or per-op mode for XNNPACK.
- Runtime targets - what CMake target needs to be linked?
- What quantization schemes and operators are supported?
Suggest a potential alternative/fix
No response
cc @mergennachin@byjlw
📚 The doc issue
We need to improve the documentation for our delegates. Right now, information is sparse and it's difficult to find answers to common questions. We need to create a standard template for delegate documentation and then ensure that each delegate has a page in the docs under "Backend Delegates" based on this template.
Key information to cover includes:
Suggest a potential alternative/fix
No response
cc @mergennachin@byjlw