frozenleaves/pytorch
1
0
Fork 0
mirror of https://github.com/pytorch/pytorch.git synced 2025-10-20 21:14:14 +08:00
Code Issues Packages Projects Releases Wiki Activity

Revert "Fix truediv numerics between eager and compile (#164144)"

Browse Source 
This reverts commit 68913d8f2a953bdbada4033101b04f6e8d49dabe.

Reverted https://github.com/pytorch/pytorch/pull/164144 on behalf of https://github.com/malfet due to It breaks CI again, why was it landed for 3 times in a row without any changes? ([comment](https://github.com/pytorch/pytorch/pull/164144#issuecomment-3390973016))
This commit is contained in:
PyTorch MergeBot
2025-10-10 16:10:25 +00:00
parent b57ab9a3f2
commit abb2f7179e
2 changed files with 1 additions and 34 deletions
Download Patch File Download Diff File Expand all files Collapse all files

25
test/inductor/test_cuda_repro.py
View File

@ -2580,31 +2580,6 @@ def triton_poi_fused_add_reflection_pad2d_0(in_ptr0, in_ptr1, out_ptr0, xnumel,
actual = compiled(*example_inputs) actual = compiled(*example_inputs)
self.assertEqual(actual, correct) self.assertEqual(actual, correct)
def test_truediv_numerics_with_eager(self):
from decimal import Decimal
y, x = 7.0, 11.0
@torch.compile
def compiled_divide(x, y):
return x / y
for y_dtype in [torch.float16, torch.bfloat16, torch.float32, torch.float64]:
for x_dtype in [
torch.float16,
torch.bfloat16,
torch.float32,
torch.float64,
]:
y_ten = torch.tensor([y], dtype=y_dtype, device="cuda")
x_ten = torch.tensor([x], dtype=x_dtype, device="cuda")
torch._dynamo.reset()
compiled_div = Decimal(compiled_divide(x, y_ten).item())
eager_div = Decimal((x / y_ten).item())
self.assertEqual(eager_div, compiled_div)
if __name__ == "__main__": if __name__ == "__main__":
from torch._inductor.test_case import run_tests from torch._inductor.test_case import run_tests

10
torch/_inductor/codegen/triton.py
View File

@ -1075,15 +1075,7 @@ class TritonOverrides(OpOverrides):
@staticmethod @staticmethod
def truediv(x, y): def truediv(x, y):
x_dtype = getattr(x, "dtype", None) out = f"({x} / {y})"
y_dtype = getattr(y, "dtype", None)
if x_dtype == torch.float32 and y_dtype == torch.float32:
# x / y in Triton is lowered to div.full which is approx
# we want div_rn to adhere with eager
out = f"triton.language.div_rn({x}, {y})"
else:
out = f"({x} / {y})"
if low_precision_fp_var(x) or low_precision_fp_var(y): if low_precision_fp_var(x) or low_precision_fp_var(y):
out_dtype = get_dtype_handler().truediv(x, y) out_dtype = get_dtype_handler().truediv(x, y)
if out_dtype in (torch.float16, torch.float32): if out_dtype in (torch.float16, torch.float32):