diff --git a/exp/moe/moe_356m_8gpu.sh b/exp/moe/moe_356m_8gpu.sh index 3ec4e78b..be071e6a 100644 --- a/exp/moe/moe_356m_8gpu.sh +++ b/exp/moe/moe_356m_8gpu.sh @@ -16,7 +16,7 @@ if [ -n "${3}" ]; then NUM_EXPERTS=$3; fi -CAPACITY_FACTOR=16384 +CAPACITY_FACTOR=1 if [ -n "${4}" ]; then CAPACITY_FACTOR=$4; fi diff --git a/exp/moe/moe_46m_8gpu.sh b/exp/moe/moe_46m_8gpu.sh index 69637724..7d7d32ce 100644 --- a/exp/moe/moe_46m_8gpu.sh +++ b/exp/moe/moe_46m_8gpu.sh @@ -16,7 +16,7 @@ if [ -n "${3}" ]; then NUM_EXPERTS=$3; fi -CAPACITY_FACTOR=16384 +CAPACITY_FACTOR=2 if [ -n "${4}" ]; then CAPACITY_FACTOR=$4; fi