backend libraries not found but present, fallback used instead of grpc but grpc called from backend

[JLT032]

LocalAI version:
localAI 4.x

Environment, CPU architecture, OS, and Version:
amd64, VM

Describe the bug
backend libraries not found

To Reproduce

cd backends/rocm-llama-cpp
./run.sh

ldd llama-cpp-grpc
linux-vdso.so.1 (0x00007f647be10000)
libomp.so => not found
libhipblas.so.2 => not found
librocblas.so.4 => not found
libamdhip64.so.6 => not found
libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007f647bcfc000)
libstdc++.so.6 => /lib/x86_64-linux-gnu/libstdc++.so.6 (0x00007f647ba00000)
libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1 (0x00007f647bccf000)
libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f647b80c000)
/lib64/ld-linux-x86-64.so.2 (0x00007f647be12000)

ldd llama-cpp-fallback
linux-vdso.so.1 (0x00007fc0a17c4000)
libomp.so => not found
libhipblas.so.2 => not found
librocblas.so.4 => not found
libamdhip64.so.6 => not found
libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007fc0a16b0000)
libstdc++.so.6 => /lib/x86_64-linux-gnu/libstdc++.so.6 (0x00007fc0a1400000)
libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1 (0x00007fc0a1683000)
libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007fc0a120c000)
/lib64/ld-linux-x86-64.so.2 (0x00007fc0a17c6000)

ldd llama-cpp-rpc-server
linux-vdso.so.1 (0x00007f4359dbd000)
libomp.so => not found
libhipblas.so.2 => not found
librocblas.so.4 => not found
libamdhip64.so.6 => not found
libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007f4359ca9000)
libstdc++.so.6 => /lib/x86_64-linux-gnu/libstdc++.so.6 (0x00007f4359a00000)
libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1 (0x00007f4359c7c000)
libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f435980c000)
/lib64/ld-linux-x86-64.so.2 (0x00007f4359dbf000)

Expected behavior
backend libraries found

Logs

Additional context
manually setting
LD_LIBRARY_PATH=/usr/local/rocm/lib
or
LD_LIBRARY_PATH=/home/myuser/localai/backends/rocm-llama-cpp/lib
both fail because libc6.so is shipped separately yet the required libraries are not found

ls
ls: symbol lookup error: libc.so.6: undefined symbol: __nptl_change_stack_perm, version GLIBC_PRIVATE

see also output for run.sh is
.....

• echo 'Using binary: llama-cpp-fallback'
  Using binary: llama-cpp-fallback
• exec /localai/backends/rocm-llama-cpp/lib/ld.so /localai/backends/rocm-llama-cpp/llama-cpp-fallback
  /localai/backends/rocm-llama-cpp/llama-cpp-fallback: error while loading shared libraries: libhipblaslt.so.0: cannot open shared object file: No such file or directory

[JLT032]

update: creating a symlink in /localai/backends/rocm-llama-cpp/lib to libhipblaslt.so in /usr/local/rocm/lib to point to libhipblaslt.so.0 under /localai/backends/rocm-llama-cpp/lib deblocks the issue of loading but now no ROCM device is found

stderr+ exec /localai/backends/rocm-llama-cpp/lib/ld.so /localai/backends/rocm-llama-cpp/llama-cpp-fallback --addr 127.0.0.1:39173
10:29:50 PM.613stderrWARNING: All log messages before absl::InitializeLog() is called are written to STDERR
10:29:50 PM.613stderrI0000 00:00:1775507390.612985 44209 config.cc:230] gRPC experiments enabled: call_status_override_on_cancellation, event_engine_dns, event_engine_listener, http2_stats_fix, monitoring_experiment, pick_first_new, trace_record_callops, work_serializer_clears_time_cache, work_serializer_dispatch
10:29:50 PM.613stderrI0000 00:00:1775507390.613221 44209 ev_epoll1_linux.cc:125] grpc epoll fd: 3
10:29:50 PM.613stderrI0000 00:00:1775507390.613356 44209 server_builder.cc:392] Synchronous server. Num CQs: 1, Min pollers: 1, Max Pollers: 2, CQ timeout (msec): 10000
10:29:50 PM.614stderrI0000 00:00:1775507390.613976 44209 ev_epoll1_linux.cc:359] grpc epoll fd: 5
10:29:50 PM.614stderrI0000 00:00:1775507390.614303 44209 tcp_socket_utils.cc:634] TCP_USER_TIMEOUT is available. TCP_USER_TIMEOUT will be used thereafter
10:29:50 PM.614stdoutServer listening on 127.0.0.1:39173
10:29:50 PM.614stderrstart_llama_server: starting llama server
10:29:50 PM.614stderrstart_llama_server: waiting for model to be loaded
10:29:52 PM.574stderrsystem info: n_threads = 4, n_threads_batch = -1, total_threads = 4
10:29:52 PM.574stderr
10:29:52 PM.575stderrggml_cuda_init: failed to initialize ROCm: no ROCm-capable device is detected
....