Search results

Bug: MTP + Vision (multimodal) causes slot position corruption and OOM

Status: Open.

#22867 In ggml-org/llama.cpp;

· Scorp1o117 opened on May 9, 2026

Feature Request: Add an optimization mechanism for pass

enhancementNew feature or requestNew feature or request

Status: Open.

#22865 In ggml-org/llama.cpp;

· henry1833 opened on May 9, 2026

Feature Request: remove the build output of WebUI from repo

enhancementNew feature or requestNew feature or request

Status: Open.

#22860 In ggml-org/llama.cpp;

· foldl opened on May 9, 2026

Eval bug: srv operator(): http client error: Failed to read connection

bug-unconfirmed

Status: Open.

#22847 In ggml-org/llama.cpp;

· Mikec78660 opened on May 8, 2026

Compile bug: installed executables fail to load libllama-common.so.0 with custom CMAKE_INSTALL_PREFIX

bug-unconfirmed

Status: Open.

#22834 In ggml-org/llama.cpp;

· raphael10-collab opened on May 8, 2026

Eval bug: llama-server reproducibly enters degenerate per-slot generation state after contaminated prompt; returns 1-4 token completions until context reset

bug-unconfirmed

Status: Open.

#22828 In ggml-org/llama.cpp;

· papa923829 opened on May 8, 2026

Compile bug: warning #221-D: floating-point value does not fit in required floating-point type

bug-unconfirmed

CUDARelated to the CUDA backendRelated to the CUDA backend

windowsIssues specific to WindowsIssues specific to Windows

Status: Open.

#22824 In ggml-org/llama.cpp;

· jrxpana opened on May 7, 2026

Misc. bug: WebUI: Pressing Cancel/Escape after editing the System Message deletes it

bug-unconfirmed

Status: Open.

#22819 In ggml-org/llama.cpp;

· garak95 opened on May 7, 2026

Eval bug: [Vulkan] Tensor Split Assert: Split State Failed

bug-unconfirmed

tensor-parallelTensor parallelism feature with meta backend.Tensor parallelism feature with meta backend.

VulkanIssues specific to the Vulkan backendIssues specific to the Vulkan backend

Status: Open.

#22817 In ggml-org/llama.cpp;

· Stoney49th opened on May 7, 2026

Feature Request: Gracefully handle model crashes in llama-bench

enhancementNew feature or requestNew feature or request

Status: Open.

#22815 In ggml-org/llama.cpp;

· gdevenyi opened on May 7, 2026

Eval bug: tool call syntax leaking into Response content - Step-3.5-Flash-Q8

bug-unconfirmed

Status: Open.

#22814 In ggml-org/llama.cpp;

· MarkErik opened on May 7, 2026

Compile bug: ggml-virtgpu/ggml-backend-device.cpp fails to compile on Fedora44/g++ 16.1.1

bug-unconfirmed

Status: Open.

#22809 In ggml-org/llama.cpp;

· olliewalsh opened on May 7, 2026