Sam3 GGUF

--- license: apache-2.0 tags: - sam3 - gguf - segmentation - grounding - vision library_name: ggml pipeline_tag: image-segmentation --- # SAM3-GGUF GGUF weights for **SAM 3** (Segment Anything Model 3), converted for use with [sam3.cpp](https://github.com/ropoctl/sam3cpp). ## Available Variants | File | Precision | Size | Inference (M4 Max) | |------|-----------|------|--------------------| | `sam3-image-f32.gguf` | F32 | 3.2 GB | 3.5s | | `sam3-image-f16.gguf` | F16 | 1.7 GB | 3.5s | | `sam3-image-q8_0.gguf` | Q8_0 | 1.0 GB | 3.2s | ## Usage ```python import sam3cpp model = sam3cpp.Sam3Model("hf://rob-laz/sam3-gguf/sam3-image-f16.gguf") pred = model.predict("photo.jpg", "person") print(f"{pred.count} detections") ``` See [sam3.cpp](https://github.com/ropoctl/sam3cpp) for build instructions and full documentation. ## Conversion Converted from the original [SAM 3](https://github.com/facebookresearch/sam3) MLX weights using `tools/convert_mlx_sam3_to_gguf.py`.