Oct 24 Yesterday 16:55 Add pinned NMT research artifacts 16:52 Add pinned NMT research artifacts Oct 21 4 days ago 03:04 Publish NCNN 2026.10.21.1 and Photos native/WASM media runtimes NCNN v7/json-v2 wrappers for Linux amd64/arm64 and macOS arm64; private Linux companion RPATH; exact source and SHA256 manifests. Publish existing FFmpeg 8.1.3-photos.1 and current scalar/SIMD128 panorama WASM artifacts under immutable runtime/releases paths. Existing llama runtime and model artifacts are preserved. Oct 1 24 days ago 06:30 Promote intel-llama 2026.10.01.11312 06:22 Publish intel-llama 2026.10.01.11312 runtime Sep 22 05:52 Publish Intel llama runtime 2026.09.22.11060 with Level Zero loader Sep 20 08:14 Publish versioned CUDA llama runtime 08:14 Publish versioned CPU llama runtime 07:04 publish intel llama runtime metadata 2026.09.20.11060 07:01 publish intel llama runtime 2026.09.20.11060 Sep 9 05:54 Repack Manticore and CUDA runtimes with gzip Sep 8 12:39 Publish Manticore 29.2.4 runtime modules Sep 7 11:08 Update checksums for CC-CEDICT query lexicon 11:08 Publish CC-CEDICT query lexicon Sep 4 15:20 Publish NCNN runtime v7 2026.09.04.1 04:02 Upload runtime/cpu-llama.meta.json with huggingface_hub 04:01 Upload runtime/cpu-llama.raw with huggingface_hub Sep 3 13:54 Publish NCNN runtime v6 2026.09.03.1 Publish Darwin arm64, Linux amd64, and Linux arm64 v6 runtime artifacts with unified release manifest and SHA256SUMS. Sep 1 05:24 publish ncnn runtime ABI v5 2026.09.01.1 05:11 publish ncnn runtime 2026.09.01.1 Aug 30 04:55 Publish NCNN runtime 2026.08.30.1 and SCRFD five-point detector Adds split face_detect/face_embed runtime support while retaining combined face compatibility. Replaces the bbox-only SCRFD artifact with the official five-point model. Aug 26 18:20 Update cuda-llama metadata to 2026.08.27.10615.1 18:20 Update cuda-llama runtime to 2026.08.27.10615.1 07:40 Publish CUDA llama runtime b10615 metadata 07:39 Upgrade CUDA llama runtime raw to b10615 Aug 6 04:48 Upload runtime/intel-llama.raw with huggingface_hub 04:10 Upload runtime/intel-llama.raw with huggingface_hub 04:05 Upload runtime/intel-llama.raw with huggingface_hub Aug 4 05:49 Delete intel-llama.raw 05:39 release: intel-llama.raw 2026.08.05 05:33 runtime: slim OpenCL mtmd Intel llama 2026.08.05 Aug 3 17:50 Refresh Intel llama SYCL runtime for Arc Pro B60 Version 2026.08.04. Includes oneAPI 2025.3.2 SYCL dependencies required for Battlemage detection. Jul 28 11:42 Add RapidOCR 512 readable-text persistent server 06:40 Fix full BGE native position-id runtime input Jul 27 11:12 Publish zvec v0.5.0 Linux amd64 runtime Jul 15 08:17 Publish metadata-aware NCNN runtime v4 Jun 30 11:05 Upload folder using huggingface_hub 09:43 Upload folder using huggingface_hub Jun 18 10:04 Refresh photo-search-adapter for bge-m3-0.5B-ncnn-int8 Jun 17 10:59 Upload folder using huggingface_hub 10:12 Remove fp32 debug BGE-M3 NCNN smoke bundle 09:55 Add fp32 debug full BGE-M3 NCNN bundle Jun 5 08:20 Upload folder using huggingface_hub 08:15 Upload folder using huggingface_hub Jun 1 10:13 Add llama runtime raw artifacts Add runtime/intel-llama.raw and runtime/cuda-llama.raw; remove temporary upload probe. 07:52 Probe runtime upload path Temporary probe before runtime raw upload May 26 07:26 Add staged RapidOCR sidecar linux amd64 07:26 Add NCNN runtime v4 linux amd64 07:26 Add NCNN runtime v4 header linux amd64 May 21 07:29 Upload linux/amd64/RapidOcrNcnn with huggingface_hub 
