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

add logsigmoid op #1520

Merged
merged 10 commits into from
Dec 14, 2024
Merged

add logsigmoid op #1520

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
9c84e23
add logsigmoid op
beverlylytle Dec 5, 2024
bc9dda3
new logsigmoid impl and custom backward
beverlylytle Dec 11, 2024
96486e8
remove print statements
beverlylytle Dec 11, 2024
0d31b7c
use the buffer
beverlylytle Dec 12, 2024
6f14d8a
Merge branch 'main' into add_logsigmoid
beverlylytle Dec 12, 2024
0f3cbec
use smaller tensor
beverlylytle Dec 13, 2024
b2a7ef1
clean up _inplace_to_out_of_place
beverlylytle Dec 13, 2024
d9c8766
use empty instead of zeroes for placeholder
beverlylytle Dec 13, 2024
f4cacf8
move to grad_transform
beverlylytle Dec 13, 2024
6bdcc03
revert move to grad_transform
beverlylytle Dec 13, 2024
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
17 changes: 17 additions & 0 deletions thunder/tests/opinfos.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1678,6 +1678,23 @@ def gen(op, device, dtype, requires_grad):
elementwise_unary_ops.append(leaky_relu_opinfo)


logsigmoid_opinfo = OpInfo(
ltorch.logsigmoid,
dtypes=(datatypes.floating,),
sample_input_generator=elementwise_unary_generator,
torch_reference=torch.nn.functional.logsigmoid,
test_directives=(
# test tols are too tight for these half precision tests
beverlylytle marked this conversation as resolved.
Show resolved Hide resolved
DecorateInfo(
pytest.mark.skip,
"test_core_vs_torch_consistency",
dtypes=(datatypes.float16, datatypes.bfloat16),
),
),
)
elementwise_unary_ops.append(logsigmoid_opinfo)


relu_opinfo = OpInfo(
ltorch.relu,
sample_input_generator=elementwise_unary_generator,
Expand Down
8 changes: 8 additions & 0 deletions thunder/torch/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1812,6 +1812,14 @@ def leaky_relu(a: TensorProxy, /, negative_slope: float = 0.01, inplace: bool =
_inplace_to_out_of_place[leaky_relu] = leaky_relu, 2


@torchsymbol(torch.nn.functional.logsigmoid, is_method=False)
def logsigmoid(a: TensorProxy, /):
return log(sigmoid(a))
beverlylytle marked this conversation as resolved.
Show resolved Hide resolved


_inplace_to_out_of_place[logsigmoid] = logsigmoid, -1
beverlylytle marked this conversation as resolved.
Show resolved Hide resolved


# TODO Should this use clamp? -- Would that propagate NaNs properly?
@torchsymbol(torch.relu, torch.nn.functional.relu, id="torch.relu", is_method=True)
def relu(a: TensorLike, /, inplace: bool = False) -> TensorLike:
Expand Down
1 change: 0 additions & 1 deletion thunder/torch/default_torch_ops.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -354,7 +354,6 @@
torch.nn.functional.kl_div,
torch.nn.functional.l1_loss,
torch.nn.functional.local_response_norm,
torch.nn.functional.logsigmoid,
torch.nn.functional.lp_pool1d,
torch.nn.functional.lp_pool2d,
torch.nn.functional.lp_pool3d,
Expand Down
Loading