I’m publishing ornith15 (registered Aug 29), an inference snap in the style of Canonical’s inference-snaps serving Ornith 1.5 9B on CPU, Intel CPU, and Intel GPU engines. It uses snap components to package the model weights (GGUF + OpenVINO conversions) and engine backends (llama.cpp, OpenVINO Model Server) separately from the application snap.
I am looking to be granted component upload rights for:
- ornith15
Components:
- ornith15+llamacpp (19 MB — llama.cpp runtime)
- ornith15+openvino-model-server (411 MB — OpenVINO Model Server runtime)
- ornith15+model-9b-q4-k-m-gguf (5.4 GB — GGUF model weights)
- ornith15+mmproj-9b-bf16-gguf (818 MB — multimodal projector)
- ornith15+model-9b-ov-1-of-2 (4.5 GB — OpenVINO model, part 1 of 2)
- ornith15+model-9b-ov-2-of-2 (1.4 GB — OpenVINO model, part 2 of 2)
The source can be found here.