Is tensor parallelism supported by neural speed?

[zhangnju]

an example of TP has been provided by Neural speed document: mpirun -np 2 -bind-to=socket ./build/bin/main_gptj -m ne-q4_0.bin --seed 1234 -t 56 -c 68 -n 32 -p "Once upon a time, there existed a little girl, who liked to have adventures. She wanted to go to places and meet new people, and have fun." --no_mmap

but I didn't find main_gptj existed in build/bin path, and also didn't find the option no_mmap

does Neural Speed still support TP feature?

[ClarkChin08]

Yes， it's still supported. Doc need to be updated, please wait the fix on how to use it.

[ClarkChin08]

https://github.com/intel/neural-speed/pull/224