RPC Failed To Connect

[xuanzhec]

Hi all,
I've tried to implement the rpc feature of llama.cpp between my devices including Macs and windows pc using CUDA backend via wifi network. When all devices completed to implement, the main host's terminal indicated that

...
main: llama backend init
main: load the model and apply lora adapter, if any
llama_model_load_from_file_impl: using device RPC(RPC[[ip_address]:50052]) - 0 MiB free 
llama_model_load_from_file_impl: using device RPC(RPC[[ip_address]][:50052]) - 0 MiB free
llama_model_load_from_file_impl: using device RPC(RPC[[ip_address]][:50052]) - 0 MiB free
llama_model_load_from_file_impl: using device Metal (Apple M4 Pro) - 19659 MiB free # the main host
...

Then after loading tensors, it has been displaying that

Failed to connect to :50052
Failed to connect to :50052
Failed to connect to :50052

It seems that the main host running the llama-cli is not connected with the rpc servers as well. Moreover, all devices shall be using the same wifi network and being able to find all under the network with each other I believe, please let me know how. Thank you so much.

[abc-nix]

It says 0 MiB free, so it cannot allocate any layers to the RPC devices. Either the RPC servers are unreachable (maybe firewall is blocking the connection), or they have problems being recognized (maybe different versions of llama.cpp on the RPC and main host).

[xuanzhec]

Thank you. Actually, it shall be the client issue caused by the network setup, solved it now.