InstructSAM gets a C++/GGUF port for local multimodal segmentation with Qwen3-VL-2B
remyxai · reddit · 2026-07-25
InstructSAM is being ported to C++/GGUF for faster local multimodal inference
The author describes a multimodal synthesis pipeline that uses VLMs for captioning and segmentation models for localization, then found InstructSAM — a repo that augments Qwen3-VL-2B with a SAM3-based segmentation head.
By using that architecture, they can collapse a two-stage pipeline into two passes through a single smaller model. The catch is practical: their current setup runs on 4-bit quantized weights in llama.cpp, and adding a segmentation head means the model must emit images, which is a substantial change to the project.
To make it work, they forked a related SAM3-compatible repo, added llama.cpp dependencies, and ported InstructSAM to C++ with GGUF conversion. They also included a Dockerfile to simplify build and conversion while respecting licensing constraints.
Repo: smellslikeml/instructSAM.cpp
More from coding & agent
- Shopify design lead on AI-era workflow: designers are becoming full-stack editors — KlausCodes · 2026-10-26
- Ultrametric open-sources Open Startup Repo to let AI agents run startup operations — garrytan · 2026-10-26
- MCP fault injection study: agent task scores drop 60% under realistic tool failures, scale doesn't help — dair_ai · 2026-10-26
- General agents burn ~140M tokens/month on LinkedIn outreach; scripts do it 60x cheaper — ChrisHarpon2 · 2026-10-26
- OpenAI upgrades Decisions API: 400 questions per call, native voice and image support — OpenAIDevs · 2026-10-26
- NLI alone rejects 5/6 new agent memories; a 0.45 similarity gate fixes it all — Alternative_Pin9598 · 2026-10-26