[Test] Un-shadow duplicated test classes in test_lu_op.py and test_tensordot.py

[Anai-Guo]

PR Category

Execute Infrastructure

PR Types

Bug fixes

Description

Two unit-test files define a test class name twice at module level. Python keeps only the last binding, so unittest / pytest collect only the later class and the earlier case silently stops running.

1. test/legacy_test/test_lu_op.py

TestLUOp is defined at line 106 (# m < n, "case 1", x_shape = [3, 10, 12]) and again at line 387, where a zero-size case (x_shape = [2, 0, 12]) reuses the same name.

TestLUOp2 … TestLUOp5 still behave correctly — they subclass the first definition and captured the class object at class-creation time — but the m < n case itself is no longer collected, since the module attribute TestLUOp now points at the zero-size class.

Renamed the zero-size class to TestLUOpZeroSize, matching the TestLUAPIZeroSize naming already used in the same file. Both cases are then collected.

collected class	before	after
TestLUOp	L387 zero-size [2, 0, 12]	L106 m < n [3, 10, 12] (base of TestLUOp2–TestLUOp5)
TestLUOpZeroSize	(not collected)	L387 zero-size [2, 0, 12]

Class count is unchanged (9 before, 9 after); one case is recovered.

2. test/legacy_test/test_tensordot.py

TestTensordotAPIZeroSize is defined at line 365 and again at line 385. The second definition is byte-identical to the first except that it adds set_dtype → np.float64 — which is exactly what TestTensordotAPIFloat64ZeroSize (line 380, a subclass of the first definition) already provides.

So the duplicate buys no extra coverage and costs the default float32 zero-size case: the collected TestTensordotAPIZeroSize runs float64, and float64 is then tested twice.

Removed the second definition. TestTensordotAPIZeroSize goes back to the base set_dtype (np.float32) and TestTensordotAPIFloat64ZeroSize continues to cover float64. No class defined after line 400 subclasses TestTensordotAPIZeroSize, so nothing else changes binding.

collected class	before	after
TestTensordotAPIZeroSize	float64 (duplicate of TestTensordotAPIFloat64ZeroSize)	float32 (base dtype)
TestTensordotAPIFloat64ZeroSize	float64	float64 (unchanged)

Diff is +1 / −19, test-only.

是否引起精度变化

否

（Test-only. Two previously-uncollected cases start running again; no framework or kernel code is touched.）

🤖 Generated with Claude Code

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}