issues
search
Oneflow-Inc
/
OneFlow-Benchmark
OneFlow models for benchmarking.
104
stars
31
forks
source link
Optimize bert attention mask calculation
#157
Closed
ShawnXuan
closed
3 years ago
ShawnXuan
commented
3 years ago
move calculation of
addr
out of layer
fix loss print bug in fp32
regression test result:
addr
out of layerregression test result: