llama.cpp

f221d562 - ggml : alloc ggml_contexts on the heap (whisper/2525)