Skip to content

[Accuracy diff No.12] Fix accuracy diff for paddle.concat API #329

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

Closed
wants to merge 1 commit into from
Closed

[Accuracy diff No.12] Fix accuracy diff for paddle.concat API #329

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
45 changes: 45 additions & 0 deletions tester/base.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -487,6 +487,32 @@ def gen_paddle_input(self):
elif "driver" in self.paddle_kwargs:
self.paddle_kwargs["driver"] = "gels"

if self.api_config.api_name == "paddle.concat":
# handle type promotion here
input_list = self.paddle_kwargs.get("x", None)
is_x_kwargs = input_list is not None
input_list = self.paddle_args[0] if not is_x_kwargs else input_list
is_tuple = isinstance(input_list, tuple)
if is_tuple:
input_list = list(input_list)
promoted_type = numpy.promote_types(str(input_list[0].dtype).split('.')[-1], str(input_list[1].dtype).split('.')[-1])
num_inputs = len(input_list)
if num_inputs > 2:
for i in range(2, num_inputs):
promoted_type = numpy.promote_types(str(input_list[i].dtype).split('.')[-1], promoted_type)
for i in range(num_inputs):
input_list[i] = input_list[i].astype(str(promoted_type))
if is_tuple:
if is_x_kwargs:
self.paddle_kwargs["x"] = tuple(input_list)
else:
self.paddle_args[0] = tuple(input_list)
else:
if is_x_kwargs:
self.paddle_kwargs["x"] = input_list
else:
self.paddle_args[0] = input_list

if self.need_check_grad():
if (self.api_config.api_name[-1] == "_" and self.api_config.api_name[-2:] != "__") or self.api_config.api_name == "paddle.Tensor.__setitem__":
self.paddle_args, self.paddle_kwargs = self.copy_paddle_input()
Expand Down Expand Up @@ -842,6 +868,25 @@ def gen_torch_input(self):
else:
self.torch_kwargs[key] = arg_config

if self.api_config.api_name == "paddle.concat":
# handle type promotion here
input_list = self.torch_kwargs["x"]
is_tuple = isinstance(input_list, tuple)
if is_tuple:
input_list = list(input_list)
input_list = list(input_list)
promoted_type = numpy.promote_types(str(input_list[0].dtype).split('.')[-1], str(input_list[1].dtype).split('.')[-1])
num_inputs = len(input_list)
if num_inputs > 2:
for i in range(2, num_inputs):
promoted_type = numpy.promote_types(str(input_list[i].dtype).split('.')[-1], promoted_type)
for i in range(num_inputs):
input_list[i] = input_list[i].type(self.convert_dtype_to_torch_type(promoted_type))
if is_tuple:
self.torch_kwargs["x"] = tuple(input_list)
else:
self.torch_kwargs["x"] = input_list

if self.need_check_grad():
if (self.api_config.api_name[-1] == "_" and self.api_config.api_name[-2:] != "__") or self.api_config.api_name == "paddle.Tensor.__setitem__":
self.torch_args, self.torch_kwargs = self.copy_torch_input()
Expand Down