Disable ProfilingGraphExecutorImpl for mobile (#30067)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/30067
### Summary
The mobile build has been broken since last week due to a runtime error caused by a missing operator in JIT:
```shell
libc++abi.dylib: terminating with uncaught exception of type torch::jit::script::ErrorReport:
Unknown builtin op: aten::_adaptive_avg_pool2d_backward.
Could not find any similar ops to aten::_adaptive_avg_pool2d_backward. This op may not exist or may not be currently supported in TorchScript.
:
at <string>:9:28
grad_self = grad.expand(self.size()) / (self_size[-1] * self_size[-2])
else:
grad_self = torch._adaptive_avg_pool2d_backward(grad, self)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ <--- HERE
return grad_self
```
### How this happens
Since we've disabled the autograd for the opensourced version, the `backward` ops won't get registered by JIT.
When `forward` runs, a `GraphExecutor` will be created according to the value of `executor_mode`. In the mobile case , this one was set to true, which gives us the `ProfilingGraphExecutorImpl` object. Seems like this executor will eventually try to emit IR for autograd schemas? which causes the error.
### Fix
There are two ways to fix it.
1. Add a macro to disable `profiling_mode` as well as `executor_mode` on mobile. Like what `FBCODE_CAFFE2` does [here](https://github.com/pytorch/pytorch/blob/master/torch/csrc/jit/profiling_graph_executor_impl.cpp#L22).
2. Disable the two modes in runtime, by calling ` torch::jit::getExecutorMode() = false;` before calling forward.
(IMO, The second fix is sort of a workaround as it doesn't make sense from a user perspective (Why I need to do this). But the up side is that we don't have to introduce yet another macro )
Feel free to drop comments, if there is a better way to fix it.
### How this was not detected by our mobile CI
We're working on adding runtime tests to our mobile build to prevent similar issues like this.
### Test Plan
- The error above disappears
- Don't break CI
cc AshkanAliabadi
Test Plan: Imported from OSS
Differential Revision: D18605998
Pulled By: xta0
fbshipit-source-id: 11fa85c2b44d54bc28a9c45731af0f5d17d5804c
