I have rewrited tvm.exec.rpc_tracker and tvm.exec.rpc_server, but I can’t modify the environment CUDA_VISIBLE_DEVICES after defining first rpc server. Are there any advices to use multiple gpus in python script, no in the terminal?
I have rewrited tvm.exec.rpc_tracker and tvm.exec.rpc_server, but I can’t modify the environment CUDA_VISIBLE_DEVICES after defining first rpc server. Are there any advices to use multiple gpus in python script, no in the terminal?