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

[BE][2/6] fix typos in test/ (test/test_*.py) (#157636)

Browse Source 
Pull Request resolved: https://github.com/pytorch/pytorch/pull/157636
Approved by: https://github.com/yewentao256, https://github.com/mlazos
ghstack dependencies: #156311, #156609
This commit is contained in:
Xuehai Pan
2025-07-09 13:23:55 +08:00
committed by PyTorch MergeBot
parent ffe11b2bf2
commit fc0376e8b1
57 changed files with 194 additions and 195 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
test/test_reductions.py
View File

@ -735,7 +735,7 @@ class TestReductions(TestCase):
res2 = x1.sum(axis=(0, 2), keepdims=True)
self.assertEqual(res1, res2)
# TODO: kill this ane replace with common creation ops
# TODO: kill this and replace with common creation ops
def _make_tensors(self, shape, val_range=(-100, 100), use_floating=True, use_integral=True,
use_complex=False) -> dict[str, list[torch.Tensor]]:
float_types = [torch.double,
@ -1629,7 +1629,7 @@ class TestReductions(TestCase):
RuntimeError, "only when boundaries tensor dimension is 1"):
torch.searchsorted(boundaries, 1)
# incompatiable output tensor's dtype
# incompatible output tensor's dtype
def test_output_dtype(dtype, is_int32):
output = values_1d.to(dtype)
with self.assertRaisesRegex(
@ -2018,7 +2018,7 @@ class TestReductions(TestCase):
with self.assertRaisesRegex(RuntimeError, error_msg):
op(x, dim=dim)
# TODO: update this test to comapre against NumPy
# TODO: update this test to compare against NumPy
@onlyCUDA
def test_var(self, device):
cpu_tensor = torch.randn(2, 3, 3)
@ -2513,7 +2513,7 @@ class TestReductions(TestCase):
k = int((t.numel() - 1) / 2)
self.assertEqual(res, t.view(-1).sort()[0][k])
if t.numel() % 2 == 1:
# We can only test agains numpy for odd reductions because numpy
# We can only test against numpy for odd reductions because numpy
# returns the mean of the two medians and torch returns the lower
self.assertEqual(res.cpu().numpy(), np.median(t_numpy))
for dim in range(t.ndim):
@ -2524,7 +2524,7 @@ class TestReductions(TestCase):
self.assertEqual(res[0], (t.sort(dim)[0]).select(dim, k).unsqueeze_(dim))
self.assertEqual(res[0], t.gather(dim, res[1]))
if size % 2 == 1:
# We can only test agains numpy for odd reductions because numpy
# We can only test against numpy for odd reductions because numpy
# returns the mean of the two medians and torch returns the lower
self.assertEqual(res[0].cpu().numpy(), np.median(t_numpy, dim, keepdims=True), exact_dtype=False)
@ -2548,7 +2548,7 @@ class TestReductions(TestCase):
k = int((t.numel() - num_nan - 1) / 2)
self.assertEqual(res, t.view(-1).sort()[0][k])
if (t.numel() - num_nan) % 2 == 1:
# We can only test agains numpy for odd reductions because numpy
# We can only test against numpy for odd reductions because numpy
# returns the mean of the two medians and torch returns the lower
self.assertEqual(res.item(), numpy_op(t.cpu().numpy()))
for dim in range(t.ndim):
@ -2561,7 +2561,7 @@ class TestReductions(TestCase):
k = ((size - num_nan - 1) / 2).type(torch.long)
self.assertEqual(res[0], (t.sort(dim)[0]).gather(dim, k))
self.assertEqual(res[0], t.gather(dim, res[1]))
# We can only test agains numpy for odd reductions because numpy
# We can only test against numpy for odd reductions because numpy
# returns the mean of the two medians and torch returns the lower
mask = (size - num_nan) % 2 == 1
res = res[0].masked_select(mask).cpu()
@ -3526,7 +3526,7 @@ as the input tensor excluding its innermost dimension'):
# raises an error if no `dim` parameter is specified. This exists separately from tests in
# test_tensot_compare_ops_empty because not specifying a `dim` parameter in the former tests does
# not throw errors. Also, checking the return type of argmax requires supplying a different dtype
# argument than that for the input tensor. There is also variantion in numpy testing.
# argument than that for the input tensor. There is also variation in numpy testing.
def test_tensor_compare_ops_argmax_argmix_kthvalue_dim_empty(self, device):
shape = (2, 0, 4)
master_input = torch.randn(shape, device=device)