Skip to content

Correct numerical regression in vision embeddings #41374

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
wants to merge 3 commits into
base: main
Choose a base branch
from
+36 −12
Open

Correct numerical regression in vision embeddings #41374

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
86eef88
created modeling file
i3hz Oct 6, 2025
84c0576
XMerge remote-tracking branch 'upstream/main'
i3hz Oct 6, 2025
0dac06b
Merge remote-tracking branch 'upstream/main'
i3hz Oct 6, 2025
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
16 changes: 12 additions & 4 deletions src/transformers/models/idefics2/modeling_idefics2.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -151,11 +151,19 @@ def forward(self, pixel_values: torch.FloatTensor, patch_attention_mask: torch.B
nb_patches_h = p_attn_mask[:, 0].sum()
nb_patches_w = p_attn_mask[0].sum()

h_indices = torch.arange(nb_patches_h, device=position_ids.device, dtype=pixel_values.dtype)
w_indices = torch.arange(nb_patches_w, device=position_ids.device, dtype=pixel_values.dtype)
step_h = 1.0 / nb_patches_h
step_w = 1.0 / nb_patches_w

fractional_coords_h = h_indices / nb_patches_h * (1 - 1e-6)
fractional_coords_w = w_indices / nb_patches_w * (1 - 1e-6)
h_indices = torch.arange(nb_patches_h, device=position_ids.device, dtype=torch.float32)
w_indices = torch.arange(nb_patches_w, device=position_ids.device, dtype=torch.float32)
fractional_coords_h = h_indices * step_h
fractional_coords_w = w_indices * step_w

fractional_coords_h = torch.clamp(fractional_coords_h, max=(1.0 - 1e-6))
fractional_coords_w = torch.clamp(fractional_coords_w, max=(1.0 - 1e-6))

fractional_coords_h = fractional_coords_h.to(pixel_values.dtype)
fractional_coords_w = fractional_coords_w.to(pixel_values.dtype)

bucket_coords_h = torch.bucketize(fractional_coords_h, boundaries, right=True)
bucket_coords_w = torch.bucketize(fractional_coords_w, boundaries, right=True)
Expand Down
16 changes: 12 additions & 4 deletions src/transformers/models/idefics3/modeling_idefics3.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -149,11 +149,19 @@ def forward(self, pixel_values: torch.FloatTensor, patch_attention_mask: torch.B
nb_patches_h = p_attn_mask[:, 0].sum()
nb_patches_w = p_attn_mask[0].sum()

h_indices = torch.arange(nb_patches_h, device=position_ids.device, dtype=pixel_values.dtype)
w_indices = torch.arange(nb_patches_w, device=position_ids.device, dtype=pixel_values.dtype)
step_h = 1.0 / nb_patches_h
step_w = 1.0 / nb_patches_w

fractional_coords_h = h_indices / nb_patches_h * (1 - 1e-6)
fractional_coords_w = w_indices / nb_patches_w * (1 - 1e-6)
h_indices = torch.arange(nb_patches_h, device=position_ids.device, dtype=torch.float32)
w_indices = torch.arange(nb_patches_w, device=position_ids.device, dtype=torch.float32)
fractional_coords_h = h_indices * step_h
fractional_coords_w = w_indices * step_w

fractional_coords_h = torch.clamp(fractional_coords_h, max=(1.0 - 1e-6))
fractional_coords_w = torch.clamp(fractional_coords_w, max=(1.0 - 1e-6))

fractional_coords_h = fractional_coords_h.to(pixel_values.dtype)
fractional_coords_w = fractional_coords_w.to(pixel_values.dtype)

bucket_coords_h = torch.bucketize(fractional_coords_h, boundaries, right=True)
bucket_coords_w = torch.bucketize(fractional_coords_w, boundaries, right=True)
Expand Down
16 changes: 12 additions & 4 deletions src/transformers/models/smolvlm/modeling_smolvlm.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -147,11 +147,19 @@ def forward(self, pixel_values: torch.FloatTensor, patch_attention_mask: torch.B
nb_patches_h = p_attn_mask[:, 0].sum()
nb_patches_w = p_attn_mask[0].sum()

h_indices = torch.arange(nb_patches_h, device=position_ids.device, dtype=pixel_values.dtype)
w_indices = torch.arange(nb_patches_w, device=position_ids.device, dtype=pixel_values.dtype)
step_h = 1.0 / nb_patches_h
step_w = 1.0 / nb_patches_w

fractional_coords_h = h_indices / nb_patches_h * (1 - 1e-6)
fractional_coords_w = w_indices / nb_patches_w * (1 - 1e-6)
h_indices = torch.arange(nb_patches_h, device=position_ids.device, dtype=torch.float32)
w_indices = torch.arange(nb_patches_w, device=position_ids.device, dtype=torch.float32)
fractional_coords_h = h_indices * step_h
fractional_coords_w = w_indices * step_w

fractional_coords_h = torch.clamp(fractional_coords_h, max=(1.0 - 1e-6))
fractional_coords_w = torch.clamp(fractional_coords_w, max=(1.0 - 1e-6))

fractional_coords_h = fractional_coords_h.to(pixel_values.dtype)
fractional_coords_w = fractional_coords_w.to(pixel_values.dtype)

bucket_coords_h = torch.bucketize(fractional_coords_h, boundaries, right=True)
bucket_coords_w = torch.bucketize(fractional_coords_w, boundaries, right=True)
Expand Down