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
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.
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
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.