Benson Ma

Hi @cyyever could you rebase this onto latest main? We have fixed a bunch of CI errors in the last week

Hi @tiankongdeguiji we are currently working on releasing the builds on PyPI, as it is being blocked by storage limit issues on PyPI side. However, the wheels should be available...

> > thx! What's the difference between 1.4.0 and 1.4.1? > > hi, [@q10](https://github.com/q10) , could you take a look? Hi @tiankongdeguiji, there were minor fixes for torch version compatibility...

Hi @foreverYoungGitHub we're currently working on this in https://github.com/pytorch/FBGEMM/pull/5143.

The missing symbol `_ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE15_M_replace_coldEPcmPKcmm` appears to indicate that the glibcxx that is installed on the system is old. Could you try this on a more recent OS version, i.e. one...

> we investigated this a bit more in preparation for the torchao v0.13.0 release, and here is what we observe: > > * PyTorch 2.7.1 and 2.8 + fbgemm_gpu stable...

What is the directory you are currently in? If you're in the project root directory, it may cause Python import confusion

> According to https://docs.pytorch.org/FBGEMM/general/Releases.html I use pytorch 2.9.1, python 3.13.7, cuda 13.0.2 and fbgemm-gpu-genai 1.4.1, however it still > > ``` > In [1]: from fbgemm_gpu.experimental.gen_ai.quantize import int4_row_quantize_zp, pack_int4 >...

@ezyang I cannot speak for non-Conda environments, but we have been building and installing fbgemm in Conda environments where the CUDA installations are located in the Conda environment's path. @isuruf...

Hi @isuruf thanks for reporting this. We will work to remove this rpath from the binary.