Recently, I was hoping to learn CUDA C++ but didn't have access to an NVIDIA GPU.
So I made this repo to solve this problem and thought others could find it useful too.
It allows you to access Google Colab GPUs in your terminal so you can easily use your typical devtools/IDEs (Neovim,Cursor,etc) while still having access to a GPU runtime.
`cgpu run nvcc...` is concise enough that coding agents probably can use it if that's your preference.
Ideally I would like to add more cloud options so if anyone has suggestions for free GPU resources please feel free to share!
One caveat is that Google Colab has usage limits so this is not intended for heavy workloads but it's useful if you just want to write CUDA programs and test/compile them.
0 comments
—
Show HN: Free GPUs in your terminal for learning CUDA · HackerTrans
It allows you to access Google Colab GPUs in your terminal so you can easily use your typical devtools/IDEs (Neovim,Cursor,etc) while still having access to a GPU runtime.
`cgpu run nvcc...` is concise enough that coding agents probably can use it if that's your preference.
Ideally I would like to add more cloud options so if anyone has suggestions for free GPU resources please feel free to share! One caveat is that Google Colab has usage limits so this is not intended for heavy workloads but it's useful if you just want to write CUDA programs and test/compile them.