Update for compatibility with PyTorch 2.0.0 and latest GCC in multi-user environment

neurosim / DNN_NeuroSim_V2.1

Benchmark framework of compute-in-memory based accelerators for deep neural network (on-chip training chip focused)

98 stars 48 forks source link

Update for compatibility with PyTorch 2.0.0 and latest GCC in multi-user environment #16

Closed xfong closed 4 months ago

xfong commented 1 year ago

Test bed CPUs: AMD Epyc 7773X, Intel Xeon Gold 6252, Intel Xeon E5-2699A v4 and Intel Xeon E5-2630 v4

OS and Environment: Native Ubuntu 22.04 with gcc-11 and g++-11 Singularity containers running CentOS 7.9 and Ubuntu 22.04

Works with CUDA 11.7, 11.8 (tested on Tesla H100, P100, and Quadro P2000) Works with ROCm 5.4.3 (tested on MI210)

jaredblase commented 4 months ago

Very helpful changes to bump the dependencies up. I couldn't run the master branch code on my machine as it was giving out a vague CUDNN error. But when I pulled the changes from this pull request, I was able to run it easily. Thank you so much.