OpenVINO backend: GPU compile-time memory optimizations#236

Draft

cavusmustafa wants to merge 5 commits into

ravi9:dev_backend_openvinofrom

cavusmustafa:ov-mem-optimizations

Commits on Jul 1, 2026

ggml-openvino: add GGML_OPENVINO_RELEASE_WEIGHTS to reclaim host weight RSS on GPU

Mustafa Cavus
and
claude
committed
ggml-openvino: stream weight requantization to cut the compile-time RSS peak

Mustafa Cavus
and
claude
committed
ggml-openvino: avoid redundant token_embd requantization at compile

Mustafa Cavus
and
claude
committed
ggml-openvino: gate compile-memory optimizations behind GGML_OPENVINO_REDUCE_COMPILE_MEM

Mustafa Cavus
and
claude
committed
ggml-openvino: add frontend model cache (GGML_OPENVINO_MODEL_CACHE_DIR)

Mustafa Cavus
and
claude
committed