Loading…
Loading…
depth-anything/Depth-Anything-V2-Small-hfA GGUF conversion of Depth Anything V2 Small for PIE4K's ggml depth runtime (libpie4k_depthany), which
the PIE4K area editor uses for Height from image. The file uses the
vision.cpp depth-anything GGUF layout (default WHCN layout) and is
kept at fp32.
| File | Precision | Size (bytes) | sha256 |
|---|---|---|---|
depth-anything-v2-small-F32.gguf |
fp32 | 99165120 | a0d0fbdeea10123fe2697b13b89633daf81fff47a0857efa3decde14944cf87a |
5426e4f0f36572d16453bbda7a8389317b1bef99, model.safetensors
(sha256 3152477ce0d8d6978d76b995120de97cb5b928701fd0f817769f59e249a16b70)tools/depth/setup_depth.sh in the PIE4K repository downloads model.safetensors and runs vision.cpp's
scripts/convert.py at commit 26a752912d49f6c4ff4545b35a1bdf7400d349ed:
uv run --script scripts/convert.py depth-anything depth-anything-v2-small.safetensors \
--output depth-anything-v2-small-F32.gguf
<models>/depth-anything-v2-small/depth-anything-v2-small-F32.gguf, where <models> is
$PIE4K_HOME/models or ~/.local/share/pie4k/models.
The upstream Hugging Face page
(depth-anything/Depth-Anything-V2-Small-hf)
lists apache-2.0. See that page for the terms.