All issues

Search results

CUDA: full test-backend-ops sweep segfaults in ggml_backend_cuda_graph_compute at ROLL on GB10 (passes with graphs disabled)

bugSomething isn't workingSomething isn't working

Status: Open.

#347 In TheTom/llama-cpp-turboquant;

· TheTom opened on Sep 3, 2026

Compile bug: Windows release archives link the runner's OpenSSL dynamically and never ship it (still open since #109) — one-line fix: -DLLAMA_BUILD_BORINGSSL=ON

Status: Open.

#346 In TheTom/llama-cpp-turboquant;

· breakerh opened on Sep 3, 2026

--gdn-replay makes qwen35 generate degenerate looping text

Status: Open.

#345 In TheTom/llama-cpp-turboquant;

· tmark00 opened on Sep 3, 2026

Proposal: TurboAnchorKV cold prompt KV experiment

Status: Open.

#332 In TheTom/llama-cpp-turboquant;

· TheTom opened on Aug 31, 2026

Vulkan: Polaris wave64 incompatible with hard-coded subgroup size 32

Status: Open.

#330 In TheTom/llama-cpp-turboquant;

· gaigher opened on Aug 31, 2026

Eval bug: llama-server OOM Over Time

Status: Open.

#325 In TheTom/llama-cpp-turboquant;

· shizhx opened on Aug 28, 2026

Eval bug: llama_model_load: error loading model: missing tensor 'blk.64.ssm_conv1d.weight'

Status: Open.

#323 In TheTom/llama-cpp-turboquant;

· JohnZhuYX opened on Aug 26, 2026

Misc. bug: failed to initialize router models: option 'reasoning-effort' not recognized

Status: Open.

#317 In TheTom/llama-cpp-turboquant;

· Jackal830 opened on Aug 25, 2026

SYCL reorder kernels undefined symbols

Status: Open.

#314 In TheTom/llama-cpp-turboquant;

· emcourtney1 opened on Aug 23, 2026

turbo4 cache output corruption introduced by PR #197 — bisected: last good b9907, first bad b9914

Status: Open.

#311 In TheTom/llama-cpp-turboquant;

· poshih opened on Aug 19, 2026

Misc. bug: more VRAM usage use compared to older builds

Status: Open.

#308 In TheTom/llama-cpp-turboquant;

· aminya opened on Aug 18, 2026

Misc. bug: Llama Web UI is broken

Status: Open.

#307 In TheTom/llama-cpp-turboquant;

· Almazick opened on Aug 18, 2026