onnxruntime
Keep loss_scale and Whole Loss Subgraph in FP32 during Mixed Precision Training
#4268
Merged
Overview
Commits
9
Changes
View On GitHub
Commits
  • Keep loss subgraph as FP32 when mixed-p training.
    Vincent Wang committed 5 years ago
  • Merge branch 'master' into weicwang/lossscale
    Vincent Wang committed 5 years ago
  • Fix case where there is no white-list loss op.
    Vincent Wang committed 5 years ago
  • Get nodes from loss_scale instead of whitelist.
    Vincent Wang committed 5 years ago
  • rename const variables.
    Vincent Wang committed 5 years ago
  • Merge branch 'master' into weicwang/lossscale
    Vincent Wang committed 5 years ago
  • Merge branch 'master' into weicwang/lossscale
    Vincent Wang committed 5 years ago
  • merge from master
    Vincent Wang committed 5 years ago
  • Merge branch 'master' into weicwang/lossscale
    Vincent Wang committed 5 years ago
FAQ Terms Privacy Refunds Impressum
Loading