ggml · Issues· 363 open

Why doesn't ggml distrbute libggml-cpu-{arch}.so got all CPU architectures?

ggml-cuda: missing __syncthreads() in ssm-scan.cu

[Feature request] Use a stable GGML CPU thread pool instead of re-creating it for each request

Upstream the mmap and splits loader from llama.cpp

ggml-cuda: missing <cuda/iterator>/<cuda/execution>/<cuda/stream_ref> includes break build with CCCL 3.1+ (argsort.cu, top-k.cu)

MUSA: flash-attn TILE GQA path (ncols2>1) emits all-NaN logits when attention window length is a multiple of 256

vulkan: vkGetDeviceQueue2 returns VK_NULL_HANDLE on older AMD drivers, first vkQueueSubmit aborts

Convert ggml to Rust

cpu: support GGML_CPU_ALL_VARIANTS with static linking

ggml-metal: a transient command-buffer failure (e.g. iOS background-GPU rejection) permanently latches the whole backend into an error state

libggml 0.13.1 fails in MUL_MAT_ID() on i915 (Intel comet lake) on freebsd-16

libggml isn't throwing an error if intel gpu fence timeout happens but DOES if GGML_VK_PERF_LOGGER=1

CUDA backend crashes with "an illegal memory access" on SM 8.9 (RTX 4060 Laptop) when processing ViT-style small matmul shapes

Question about requirements.txt dependencies

Windows build fails due to MLX backend (undefined symbols, no Windows build tags)