Skip to content

Gemma 4 26B A4B it QAT failing to load images with bundled mmproj in LM Studio #2134

Description

@jdar-gems

Which version of LM Studio?
0.4.18

Which operating system?
Win 11

What is the bug?
Model opens/runs 95tps, Only will not allow "any" small image, without crash. All updates, running CUDA 12gguf. Does same w/Vulcan llama.ccp V2.23.1

See: Issue #21402 is in ggml-org/llama.cpp, titled "Gemma 4 mmproj crashes on CUDA: SIGABRT in clip_model_loader::load_tensors."

Screenshots

Image Image Image Image Image Image Image

Logs
Failed to generate AI response
The model has crashed without additional information. (Exit code: 18446744072635812000)

To Reproduce
Load model, ask for text reply, no issues. Code, no issue.
Load image (have tried as small as 30kb)
starts to process, kicks model out with exit code shown.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions