Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix rocm hip bugs #40971

Closed
wants to merge 1 commit into from
Closed

Fix rocm hip bugs #40971

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion python/paddle/fluid/tests/unittests/op_test.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1775,7 +1775,7 @@ def check_grad(self,
no_grad_set=None,
numeric_grad_delta=0.005,
in_place=False,
max_relative_error=0.005,
max_relative_error=0.008,
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

op_test.py是所有单测算子的基础类，不允许随意对其的阈值进行修改，请修改对应的算子单测文件test_row_conv_op的阈值。

Copy link
Author

@lql341 lql341 Apr 10, 2022 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

op_test.py是所有单测算子的基础类，不允许随意对其的阈值进行修改，请修改对应的算子单测文件test_row_conv_op的阈值。

目前这个算子加入最大相对误差输入，通过了测试

159     def test_check_grad_ignore_x(self):
160         self.check_grad(
161             ['Filter'], 'Out', max_relative_error=0.06, no_grad_set=set('X'), check_dygraph=False)

user_defined_grads=None,
user_defined_grad_outputs=None,
check_dygraph=True,
Expand Down
2 changes: 1 addition & 1 deletion python/paddle/fluid/tests/unittests/test_poisson_op.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -56,7 +56,7 @@ def verify_output(self, outs):
hist, prob = output_hist(np.array(outs[0]), self.lam, self.a, self.b)
self.assertTrue(
np.allclose(
hist, prob, rtol=0.01),
hist, prob, rtol=0.05),
"actual: {}, expected: {}".format(hist, prob))

def test_check_output(self):
Expand Down
2 changes: 1 addition & 1 deletion python/paddle/fluid/tests/unittests/test_strided_slice_op.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -759,7 +759,7 @@ class Net01(ArrayLayer):
def array_slice(self, tensors):
return tensors[::-1]

self.create_case(Net01(array_size=10))
self.create_case(Net01(input_size=112, array_size=10))

class Net02(ArrayLayer):
def array_slice(self, tensors):
Expand Down
2 changes: 1 addition & 1 deletion python/paddle/fluid/tests/unittests/test_transfer_dtype_op.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -70,7 +70,7 @@ def test_check_output(self):

class TestTransferDtypeOpBf16ToFp32(OpTest):
def setUp(self):
ipt = np.array(np.random.randint(10, size=[10, 10])).astype('uint16')
ipt = np.array(np.random.randint(1, size=[10, 10])).astype('uint16')
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

这个单测在CUDA下是可以正常运行的，CUDA与ROCm处理结果应该保持一致，且单测修复原则是避免修改单测本身的输入。

self.inputs = {'X': ipt}
self.outputs = {'Out': convert_uint16_to_float(ipt)}
self.attrs = {
Expand Down