Skip to content

make loss.backward() work #84

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

Open
allenhung1025 wants to merge 1 commit into
base: main
Choose a base branch
from
Open

make loss.backward() work #84

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
36 changes: 18 additions & 18 deletions box_embeddings/modules/intersection/gumbel_intersection.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,17 +11,17 @@
def _compute_logaddexp_with_clipping_and_separate_forward(
t1: TBoxTensor, t2: TBoxTensor, intersection_temperature: float
) -> Tuple[torch.Tensor, torch.Tensor]:
t1_data = torch.stack((t1.z, -t1.Z), -2)
t2_data = torch.stack((t2.z, -t2.Z), -2)
t1_data = torch.stack((t1.z.clone(), -t1.Z.clone()), -2)
t2_data = torch.stack((t2.z.clone(), -t2.Z.clone()), -2)
lse = torch.logaddexp(
t1_data / intersection_temperature, t2_data / intersection_temperature
)

z = intersection_temperature * lse[..., 0, :]
Z = -intersection_temperature * lse[..., 1, :]
z = intersection_temperature * lse[..., 0, :].clone()
Z = -intersection_temperature * lse[..., 1, :].clone()

z_value = torch.max(z, torch.max(t1.z, t2.z)) # type: ignore
Z_value = torch.min(Z, torch.min(t1.Z, t2.Z))
z_value = torch.max(z, torch.max(t1.z.clone(), t2.z.clone())) # type: ignore
Z_value = torch.min(Z, torch.min(t1.Z.clone(), t2.Z.clone()))

z_final = (z - z.detach()) + z_value.detach()
Z_final = (Z - Z.detach()) + Z_value.detach()
Expand All @@ -32,32 +32,32 @@ def _compute_logaddexp_with_clipping_and_separate_forward(
def _compute_logaddexp_with_clipping(
t1: TBoxTensor, t2: TBoxTensor, intersection_temperature: float
) -> Tuple[torch.Tensor, torch.Tensor]:
t1_data = torch.stack((t1.z, -t1.Z), -2)
t2_data = torch.stack((t2.z, -t2.Z), -2)
t1_data = torch.stack((t1.z.clone(), -t1.Z.clone()), -2)
t2_data = torch.stack((t2.z.clone(), -t2.Z.clone()), -2)
lse = torch.logaddexp(
t1_data / intersection_temperature, t2_data / intersection_temperature
)

z = intersection_temperature * lse[..., 0, :]
Z = -intersection_temperature * lse[..., 1, :]
z = intersection_temperature * lse[..., 0, :].clone()
Z = -intersection_temperature * lse[..., 1, :].clone()

z_value = torch.max(z, torch.max(t1.z, t2.z)) # type: ignore
Z_value = torch.min(Z, torch.min(t1.Z, t2.Z))
z_value = torch.max(z, torch.max(t1.z.clone(), t2.z.clone())) # type: ignore
Z_value = torch.min(Z, torch.min(t1.Z.clone(), t2.Z.clone()))

return z_value, Z_value


def _compute_logaddexp(
t1: TBoxTensor, t2: TBoxTensor, intersection_temperature: float
) -> Tuple[torch.Tensor, torch.Tensor]:
t1_data = torch.stack((t1.z, -t1.Z), -2)
t2_data = torch.stack((t2.z, -t2.Z), -2)
t1_data = torch.stack((t1.z.clone(), -t1.Z.clone()), -2)
t2_data = torch.stack((t2.z.clone(), -t2.Z.clone()), -2)
lse = torch.logaddexp(
t1_data / intersection_temperature, t2_data / intersection_temperature
)

z = intersection_temperature * lse[..., 0, :]
Z = -intersection_temperature * lse[..., 1, :]
z = intersection_temperature * lse[..., 0, :].clone()
Z = -intersection_temperature * lse[..., 1, :].clone()

return z, Z

Expand Down Expand Up @@ -116,10 +116,10 @@ def gumbel_intersection(
if box_debug_level > 0:
with torch.no_grad(): # type:ignore
assert (
torch.max(t1.z, t2.z) < z
torch.max(t1.z.clone(), t2.z.clone()) < z
), "max(a,b) < beta*log(exp(a/beta) + exp(b/beta)) not holding"
assert (
torch.min(t1.z, t2.z) > Z
torch.min(t1.z.clone(), t2.z.clone()) > Z
), "min(a,b) > -beta*log(exp(-a/beta) + exp(-b/beta)) not holding"

return left.from_zZ(z, Z)
Expand Down