p2p inferencing not work

[mintyleaf]

LocalAI version:
7adbc16 (7adbc16)

Environment, CPU architecture, OS, and Version:
some vast.ai nvidia gpu instance

Describe the bug

To Reproduce
just try to start p2p worker node in the working p2p environment

Additional context
using latest master docker image with my p2p fixes makes everything work in federated mode and connect two nodes with third just head server cpu node (+ using own DHT for faster discovery), yet, that only works for federated mode, worker is stuck in the loop here:

LocalAI/core/cli/worker/worker_p2p.go

Lines 70 to 76 in eaf0e30

	go func() {
	for {
	log.Info().Msgf("Starting llama-cpp-rpc-server on '%s:%d'", address, port)
	grpcProcess := assets.ResolvePath(
	r.BackendAssetsPath,
	"util",

i'll try to investigate that further, but maybe you can get the idea of what is happening faster

[mudler]

Seems there was a regression: should be fixed by #4220

[pratikbin]

#3968