Skip to content

Add masks to first iteration for iterative prompting #659

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

Draft
wants to merge 1 commit into
base: dev
Choose a base branch
from
Draft

Add masks to first iteration for iterative prompting #659

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
13 changes: 10 additions & 3 deletions micro_sam/training/sam_trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -203,11 +203,18 @@ def _compute_iterative_loss(self, batched_inputs, y_one_hot, num_subiter, multim
loss, mask_loss, iou_regression_loss, mean_model_iou = 0.0, 0.0, 0.0, 0.0

for i in range(0, num_subiter):
# Pass zero as "mask_inputs" for the first iteration when mask probability is 1
if i == 0 and self.mask_prob == 1:
for inp in batched_inputs:
inp["mask_inputs"] = torch.zeros((y_one_hot.shape[1], 1, 256, 256))

# We do multimasking only in the first sub-iteration as we then pass single prompt
# after the first sub-iteration, we don't do multimasking because we get multiple prompts.
batched_outputs = self.model(batched_inputs,
image_embeddings=image_embeddings,
multimask_output=multimask_output if i == 0 else False)
batched_outputs = self.model(
batched_inputs=batched_inputs,
image_embeddings=image_embeddings,
multimask_output=multimask_output if i == 0 else False
)

# Compute loss for tis sub-iteration.
net_loss, net_mask_loss, net_iou_regression_loss = self._compute_loss(batched_outputs, y_one_hot)
Expand Down
Loading