pujiang2018
pujiang2018
@abenmao let's take time for possible fix.
@bin1guo do we still need to benchmark OPT model? Suggest to run the llama model.
could you pls give more details with examples?
@marvin-Yu How about this issue, any finding?
Let's close since 2 weeks passed. @zhm-algo pls reopen it if the issue is still there.
To solve this, need to move the kernels into pre-built library. Before that, pls use new GCC.
we now upgraded to PyTorch 2.3, which needs a newer GCC to build cpp_extensions.
@xiangzez could you pls to take a look?
My concern is that the packages in requirements.txt may trigger some issues during security checking, let's target for next version.
new quantization mechanism is under design, need some time to make the potential fix.