Chenyu Jiang
Chenyu Jiang
Thanks for the fast reply! I tried to set `FMOE_FASTER_GROUP_SIZE=4`, but still not seeing any overlap:
Hi @zms1999, extremely sorry for the (very) delayed response.. After the fix, now I can see overlapping in the example program. Thanks a lot for the fix! It is tremendously...
Sorry for bothering again, but I am still running into problems when running the above example code with SwitchGate (i.e., add `gate=SwitchGate` when initializing `FMoETransformerMLP`. The error message is: ```...