Skip to content

model: LFM2-VL fixes #17577

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 10 commits into
base: master
Choose a base branch
from
+165 −12
Open

model: LFM2-VL fixes #17577

Changes from 1 commit
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
baacb35
Adjust to pytorch
tdakhran Nov 25, 2025
bfdf624
Add antialiasing upscale
tdakhran Nov 27, 2025
c236bee
Increase number of patches to 1024
tdakhran Nov 27, 2025
1fea2d1
Handle default marker insertion for LFM2
tdakhran Nov 27, 2025
1f4f2bd
Switch to flag
tdakhran Nov 27, 2025
b1ba04d
Reformat
tdakhran Nov 28, 2025
7c27ef9
Cuda implementation of antialias kernel
tdakhran Nov 28, 2025
c03f031
Change placement in ops.cpp
tdakhran Nov 28, 2025
215388f
consistent float literals
tdakhran Nov 28, 2025
50ba22e
Pad only for LFM2
tdakhran Nov 28, 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
Prev Previous commit
Next Next commit
Change placement in ops.cpp
  • Loading branch information
@tdakhran
tdakhran committed Nov 28, 2025
commit c03f03147ea2f7efd8a2f663beb3ecbe2611440a
42 changes: 21 additions & 21 deletions ggml/src/ggml-cpu/ops.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7402,9 +7402,27 @@ static void ggml_compute_forward_upscale_f32(
sf1 = ne1 > 1 && ne01 > 1 ? (float)(ne1 - 1) / (ne01 - 1) : sf1;
}

// Similar to F.interpolate(..., mode="bilinear", align_corners=False, antialias=True)
// https://github.com/pytorch/pytorch/blob/8871ff29b743948d1225389d5b7068f37b22750b/aten/src/ATen/native/cpu/UpSampleKernel.cpp
if (mode == GGML_SCALE_MODE_BILINEAR && (mode_flags & GGML_SCALE_FLAG_ANTIALIAS)) {
if (mode == GGML_SCALE_MODE_NEAREST) {
for (int64_t i3 = 0; i3 < ne3; i3++) {
const int64_t i03 = i3 / sf3;
for (int64_t i2 = ith; i2 < ne2; i2 += nth) {
const int64_t i02 = i2 / sf2;
for (int64_t i1 = 0; i1 < ne1; i1++) {
const int64_t i01 = i1 / sf1;
for (int64_t i0 = 0; i0 < ne0; i0++) {
const int64_t i00 = i0 / sf0;

const float * x = (float *)((char *) src0->data + i00*nb00 + i01*nb01 + i02*nb02 + i03*nb03);
float * y = (float *)((char *) dst->data + i0*nb0 + i1*nb1 + i2*nb2 + i3*nb3);

*y = *x;
}
}
}
}
} else if (mode == GGML_SCALE_MODE_BILINEAR && (mode_flags & GGML_SCALE_FLAG_ANTIALIAS)) {
Copy link
Contributor Author

@tdakhran tdakhran Nov 28, 2025 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I was thinking whether to introduce a new scaling mode GGML_SCALE_MODE_BILINEAR_ANTIALIAS or a flag. I would like to hear your feedback on this.
Another question is where to place correctness tests?

Copy link
Collaborator

@ngxson ngxson Nov 28, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think it's fine to introduce this GGML_SCALE_FLAG_ANTIALIAS as a flag. Btw, I think we would need to explicitly GGML_ASSERT that the antialias is only supported by bilinear (maybe add a TODO to implement it in other modes)

// Similar to F.interpolate(..., mode="bilinear", align_corners=False, antialias=True)
// https://github.com/pytorch/pytorch/blob/8871ff29b743948d1225389d5b7068f37b22750b/aten/src/ATen/native/cpu/UpSampleKernel.cpp
auto triangle_filter = [](float x) -> float {
return std::max(1.0f - fabsf(x), 0.f);
};
Expand Down Expand Up @@ -7461,24 +7479,6 @@ static void ggml_compute_forward_upscale_f32(
}
}
}
} else if (mode == GGML_SCALE_MODE_NEAREST) {
for (int64_t i3 = 0; i3 < ne3; i3++) {
const int64_t i03 = i3 / sf3;
for (int64_t i2 = ith; i2 < ne2; i2 += nth) {
const int64_t i02 = i2 / sf2;
for (int64_t i1 = 0; i1 < ne1; i1++) {
const int64_t i01 = i1 / sf1;
for (int64_t i0 = 0; i0 < ne0; i0++) {
const int64_t i00 = i0 / sf0;

const float * x = (float *)((char *) src0->data + i00*nb00 + i01*nb01 + i02*nb02 + i03*nb03);
float * y = (float *)((char *) dst->data + i0*nb0 + i1*nb1 + i2*nb2 + i3*nb3);

*y = *x;
}
}
}
}
} else if (mode == GGML_SCALE_MODE_BILINEAR) {
for (int64_t i3 = 0; i3 < ne3; i3++) {
const int64_t i03 = i3 / sf3;
Expand Down