I have RTX3060 in my desktop, so which cuda toolkit should i install for machine learning/deep learning?
The CUDA 12.1 SDK is the current toolkit for ML and DL whether you are Windows or LINUX OS this webpage (https://developer.nvidia.com/cuda-downloads) has an option for you. Just choose which OS and I’ll recommend the network installer it easier and quicker. We have options for installing CUDA under WSL2 on Win11 under the green LINUX button if WSL is your route. NVIDIA RAPIDS our ML/DL CUDAX-AI accelerated algorithms is OpenSource software (https://rapids.ai/start.html) that runs on in LINUX(WSL also) that makes processing datasets a snap. Preprocessing data is transformed in seconds instead of minutes. This can all be done RAPIDS using familiar code constructs from Pandas, numpy and Scikit - learn often only having to change module imports from pandas to cudf (
import pandas as pdf , becomes
import cudf as pdf) and your off and running with RAPIDS. I am a Ubuntu user on both Window WSL and LINUX what OS are you using?
Installing CUDA on Windows is only 2 commands and on LINUX just 4 commands, the CUDA SDK will also update the display driver to be compatible with the CUDA toolkit version.