machine-learningtorchluarocks

Torch with CUDA10


I am trying to install torch with cuda10 on my raspberry pi. I am following the following github thread ( answer of nagadomi: https://github.com/torch/cutorch/issues/834

Following all his steps still when I run ./install.sh I get the following error

make[2]: * No rule to make target 'exe/luajit-rocks/luajit-2.1/src/lib_ffi.c', needed by 'exe/luajit-rocks/luajit-2.1/vmdef.lua'. Stop. make[1]: * [CMakeFiles/Makefile2:212: exe/luajit-rocks/luajit-2.1/CMakeFiles/luajit-static.dir/all] Error 2 make: *** [Makefile:130: all] Error 2 ./install.sh: line 107: /home/pi/torch/install/bin/luarocks: No such file or directory Installing common Lua packages ./install.sh: line 111: /home/pi/torch/install/bin/luarocks: No such file or directory

I have now spent a week just trying to setup torch with CUDA. Basically I need cutorch library and it is not installing from `luarocks install cutorch

Can anyone help me please?


Solution

  • CUDA allows to write programs that can utilize the power of NVIDIA GPUs. It provides a framework to allows developers to write code that can be executed in parallel on the thousands of cores available in modern GPUs. On a RaspberryPI you may not have a NVIDIA GPU, so even if you install CUDA correctly, there is no hardware it can work with.

    You could add a Tensor Processing Unit (TPU) to a Raspberry Pi to accelerate machine learning workloads. Googles Coral Edge TPU is a popular option for this. The Coral USB Accelerator is a small hardware device that contains a TPU and can be connected to a Raspberry Pis USB port.