Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/59466
Change saved parameter type from at::Tensor to at::IValue to support custom
class parameters, e.g. `__torch__.torch.classes.xnnpack.Conv2dOpContext`.
The NNC produced kernels won't deal with custom class parameters directly.
They simply pass through to the external operators that take these custom
class parameters, e.g. `prepacked::conv2d_clamp_run`.
It will reuse the `__getstate__` and `__setstate__` methods on the custom class
to persist and restore the state of the parameters.
When calling into the kernel, it will pass in the untyped raw pointer of the custom
class objects to the kernel as `void*`. It's similar to the regular tensor parameters,
for which it will pass in the raw data pointer of the tensor storage. The generated
kernel needs to hardcode the expected type for each parameter and cast before
calling the external ops.
ghstack-source-id: 131897904
Test Plan: - unit tests
Reviewed By: kimishpatel
Differential Revision: D28902496
fbshipit-source-id: 4b2c0895dd28f0b7d344aa08183d42ad6a355dae
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/56852
This is part of the changes to enable NNC AOT compilation for mobile.
It introduced a custom backend for NNC, which uses the components defined in the stacked PRs to load and execute a NNC-compiled model.
ghstack-source-id: 128285801
Test Plan:
- On X86 host:
```
buck build //xplat/caffe2/fb/lite_predictor:lite_predictor_nnc
buck-out/last/lite_predictor_nnc --model xplat/pytorch_models/build/pytorch_dev_linear/v1/nnc/compiled.pt --print_output true --input_dims '4,4' --input_type float
```
- On Android:
```
buck build fbsource//fbandroid/mode/gnustl //xplat/caffe2/fb/lite_predictor:lite_predictor_nncAndroid#android-armv7
adb push buck-out/last/lite_predictor_nncAndroid#android-armv7 /data/local/tmp
adb push xplat/pytorch_models/build/pytorch_dev_linear/v1/nnc/compiled.pt /data/local/tmp
adb shell 'cd /data/local/tmp; ./lite_predictor_nncAndroid\#android-armv7 --model compiled.pt --print_output true --input_dims "4,4" --input_type float'
```
Reviewed By: kimishpatel, raziel
Differential Revision: D27897153
fbshipit-source-id: 8e039089d1602782582747adfd75b31496b525ca
