justheuristic 5cc3cd99c3 wip: implement grad wrt logits 5 سال پیش
..
__init__.py c58d08cc06 remove run_and_await_k completely, rename gating_function to moe 5 سال پیش
expert.py 6fb99c8746 wip: parallel fault-tolerant moe backward pass 5 سال پیش
moe.py 5cc3cd99c3 wip: implement grad wrt logits 5 سال پیش