Skip to content

v2.0.2
Compare
Choose a tag to compare
@milakov milakov released this 19 Dec 20:53
· 119 commits to master since this release
v2.0.2
96d0f20
  • Gradient modifier layer added
  • Structured_data_constant_reader added
  • Error functions accept the 3rd optional input layer - mask
  • ADAM training algo implemented, use "--momentum_type adam", rate should generally be much smaller than for other methods
  • Changed default value for cuda_fixed_working_buffers_ratio to 0.4
Assets 2
Loading