CoreML argmin/argmax intermittently crashes process

[GregoryComer]

🐛 Describe the bug

The following repro intermittently crashes the process on my M1 Pro. Putting this in a notebook and running the cell twice appears to crash 100% of the time for me. I'm not 100% sure if it's a CoreML issue, but it does not crash when removing the CoreML partitioner and running on portable.

Repro:

import torch

from executorch.backends.apple.coreml.partition import CoreMLPartitioner
from executorch.exir import to_edge_transform_and_lower, EdgeCompileConfig, to_edge
from executorch.extension.pybindings.portable_lib import _load_for_executorch_from_buffer

class Model(torch.nn.Module):
    def __init__(self):
        super().__init__()
        
    def forward(self, x):
        return torch.argmax(x, dim=None, keepdim=False)

model = Model()
inputs = (
    torch.randn(10, 10),
)

eager_outputs = model(*inputs)
#print(f"Eager: {eager_outputs.shape} {eager_outputs}")

ep = torch.export.export(model.eval(), inputs)
print(ep)

print(f"EP: {ep.module()(*inputs)}")

lowered = to_edge_transform_and_lower(
    ep,
    partitioner=[CoreMLPartitioner()],
    compile_config=EdgeCompileConfig(_check_ir_validity=False)
).to_executorch()

print(lowered.exported_program())

et_model = _load_for_executorch_from_buffer(lowered.buffer)
et_outputs = et_model([*inputs])[0]

et_outputs - eager_outputs

This causes a segmentation fault.

Versions

coremltools version 8.3
executorch commit 67b6009 (Jun 14)

cc @kimishpatel @YifanShenSZ @cymbalrush @metascroy

[metascroy]

This should be fixed by #13018, but need to confirm