Today
-
Remove test qualifiers from coverage
-
Remove throughput figures: limited test hardware
-
Drop Vulkan throughput claim: smoke-tested only
-
State runtime coverage explicitly
-
Update iMatrix filename
-
Correct platform note, Windows compatibility and usage, archive names
-
Update metadata for renamed artifacts and Windows universal build
-
Update metadata for renamed artifacts and Windows universal build
-
Update metadata for renamed artifacts and Windows universal build
-
Update metadata for renamed artifacts and Windows universal build
-
Rename precompiled artifacts by platform, arch and backend
-
Add Windows x86_64 universal build: CUDA 12.8, Vulkan, CPU
Yesterday
-
Card: note Linux-only builds and universal builds in progress
-
Card: list under quantizations (base_model_relation)
-
Add config.json build descriptor (enables Hub download tracking)
-
Card: align metadata with the Clover quantization family
-
Link to Clover, add compatibility, Windows notes and Ko-fi block
-
Add Ko-fi support link at the top
-
Upload README.md with huggingface_hub
-
Upload SHA256SUMS with huggingface_hub
-
Upload BUILD_INFO.json with huggingface_hub
-
Upload llama-clover-b4838ffb-converter.tar.gz with huggingface_hub
-
Upload llama-clover-b4838ffb-linux-x64-cuda12.tar.gz with huggingface_hub
-
Upload llama-clover-b4838ffb-linux-x64-cpu.tar.gz with huggingface_hub
-
initial commit
