[TESTS] Hardcode target and device when getting tvm output

[CaptainDuke]

tvm/tests/python/frontend/onnx/test_forward.py

Lines 81 to 84 in 3445532

	"""Generic function to execute and get tvm output"""
	# TODO: Resolve the issues and remove the following lines
	target = "llvm"
	device = tvm.cpu(0)

I comment the code but test case test_gemm failed. It seems onnx_runtime and tvm differ on the precision.

[masahi]

cc @mbrookhart @jwfromm I remember we found this problem before and fixed it then...

[electriclilies]

Yeah, this hardcode shouldn't be there. @mbrookhart and I definitely removed it (or tried to)..

[mbrookhart]

:/ I don't see the removal of the hard-coded target in the PR we did. https://github.com/apache/tvm/pull/7438/files

But I could have sworn it was in there. Perhaps we messed up a rebase?

[masahi]

Looks like it was fixed in #8621