mirror of https://github.com/llvm/torch-mlir
Make numerical stability test more perverse
To test the summation stability of `torch.aten.var`, add a large constant to it, which increases the effective precision requirements.pull/1156/head
parent
c94431f71c
commit
6484776a25
|
@ -688,4 +688,4 @@ class VarCorrectionLargeInputModule(torch.nn.Module):
|
|||
|
||||
@register_test_case(module_factory=lambda: VarCorrectionLargeInputModule())
|
||||
def VarCorrectionLargeInputModule_basic(module, tu: TestUtils):
|
||||
module.forward(tu.rand(3, 4, 1024, 8192))
|
||||
module.forward(100 + tu.rand(3, 4, 1024, 8192))
|
||||
|
|
Loading…
Reference in New Issue