Flatpak: Ollama doesn't use AMD plugin

[gk-abosserhoff]

Describe the bug

Ollama doesn't see the AMD plugin even though it is installed. The GPU will not be used.

Workaround

If I manually copy the folder:
/var/lib/flatpak/runtime/com.jeffser.Alpaca.Plugins.AMD/x86_64/stable/active/files/lib/ollama/rocm to
/var/lib/flatpak/runtime/com.jeffser.Alpaca.Plugins.Ollama/x86_64/stable/active/files/lib/ollama/ it finds the GPU again.

There also seems to be another issue with the ROCm libraries because it crashes with large models when the model should overflow into RAM:

rocblaslt error: Could not load /app/plugins/Ollama/lib/ollama/rocm/hipblaslt/library/TensileLibrary_lazy_gfx1201.dat
hipBLASLt error: Heuristic Fetch Failed!
This message will be only be displayed once, unless the ROCBLAS_VERBOSE_HIPBLASLT_ERROR environment variable is set.

rocBLAS warning: hipBlasLT failed, falling back to tensile. 
This message will be only be displayed once, unless the ROCBLAS_VERBOSE_TENSILE_ERROR environment variable is set.
graph_reserve: failed to allocate compute buffers
SIGSEGV: segmentation violation
PC=0x7f5813988faa m=9 sigcode=1 addr=0x7f5a02823ce8
signal arrived during cgo execution