CannyLab / tsne-cuda

GPU Accelerated t-SNE for CUDA with Python bindings
BSD 3-Clause "New" or "Revised" License
1.81k stars 130 forks source link

Binary installation for Windows10 not possible #131

Closed raulmendez closed 8 months ago

raulmendez commented 8 months ago

Conda installation conda install tsnecuda -c conda-forge DOES NOT find any files, whereas pip3 binary installation pip3 install tsnecuda==3.0.1+cu113 -f https://tsnecuda.isx.ai/tsnecuda_stable.html (downloading the corresponding linked wheel file locally), it does the installation but using Linux libraries, (but then it comes, for instance with libtsnecuda.so instead of libtsnecuda.dll). Any recommendations about how to install it on Windows 10?

DavidMChan commented 8 months ago

We don't officially support windows, however there's a PR #130 which allows you to build from source! We're looking to get it merged with the conda recipe eventually.

raulmendez commented 8 months ago

Hi David, The link you mentioned PR https://github.com/CannyLab/tsne-cuda/pull/130 does not provide the txt nor explicit instructions. I would need some more details to be able to installed it on my Windows (I tried the TSNE funciton at sklearn,.. but it is so slow).

Thanks!