Modelpublic

RemySkye/rwkv7-g1i-1.5B-i1-GGUF

sourceHugging Faceapache-2.0updated 1mo agoView on Hugging Face

0likes351downloads

Model Card

rwkv7-g1i-1.5B-i1-GGUF

GGUF quantizations for the main source model `BlinkDL/rwkv7-g1`.

—Source file: rwkv7-g1i-1.5b-20260805-ctx16384.pth
—Calibration dataset: lemon07r/bartowski-imatrix-v5-semantic
—Imatrix context: 512 tokens
—llama.cpp revision: c92e806d1c81091c9035edce99c35374da1b465e
—RWKV converter revision: ebfb744281c31a07aad5606ec7473f79f837e92a
—RWKV-aware custom tensor maps are used for Q3KL/M/S, Q4KM, and Q5KM/S.

The BF16 master and imatrix are uploaded before the quantized files. Large GGUFs, if any, are stored as llama.cpp split shards in their own repository subfolder; imatrix shards are stored under imatrix/.