ggml : add ggml_interpolate_sf() with explicit scale factors

[ManaEstras]

Overview

Add ggml_interpolate_sf() API that accepts explicit scale factors for interpolation operations.

Motivation

The existing ggml_interpolate() computes scale factors internally as ne_dst / ne_src. However, some vision-language models require interpolation with scale factors that cannot be expressed as simple dimension ratios.

For example, when interpolating position embeddings, a model may use:

scale_factor = (target_dim + offset) / base_grid_size

where offset is a learned or configured value (e.g., 0.1). This cannot be achieved by manipulating input/output tensor dimensions alone.

This PR introduces ggml_interpolate_sf() which allows the caller to specify sf0 and sf1 directly, enabling exact control over the interpolation coordinate mapping.

Changes

API:

• Add ggml_interpolate_sf(ctx, a, ne0, ne1, ne2, ne3, mode, sf0, sf1)
• Add GGML_SCALE_FLAG_CUSTOM_SF flag

Backend support:

• CPU, CUDA, Metal, SYCL, Vulkan, OpenCL

Bug fix:

• Fix out-of-bounds access in nearest interpolation when custom scale factors cause source index overflow

Tests:

• Add 38 test cases covering upscale, downscale, non-square inputs, and various interpolation modes

Testing

• 38 interpolate_sf tests pass on CPU
• 38 interpolate_sf tests pass on Metal
• ctest -L main passes

Requirements

• I have read and agree with the contributing guidelines
• AI usage disclosure: YES - AI assisted with code review and documentation drafting. All code was human-authored and manually reviewed.

[ggml-gh-bot]

Hi @ManaEstras, thanks for your contribution!

Per our contribution guidelines, the automated PR checker found the following issue(s) that need your attention:

• Multiple open PRs from a new contributor: We limit new contributors (those without a previously merged PR) to 1 open PR at a time. You currently have 3 open PRs.

• Multiple backend changes in one PR: When adding support for a new model or feature, focus on CPU support only in the initial PR. Add support for other backends like CUDA in follow-up PRs. If you have a good reason to modify multiple backends in one PR, please explain it.

---

Please note that maintainers reserve the right to make final decisions on PRs. If you believe there is a mistake, please comment below.