owensgroup/ml_perf_model

Issues

• Convolution perf model generalizatoin

  #10 opened 2 years ago by moderato
  1

• Study of overhead stats

  #8 opened 2 years ago by moderato
  1

• Performance model for variable embedding table size

  #4 opened 3 years ago by moderato
  3

• Multi-node multi-GPU support

  #6 opened 3 years ago by moderato
  2

• Migrate from nvprof to nsys/ncu

  #2 opened 3 years ago by moderato
  0

• PyTorch custom operator inputs fix (emb op)

  #12 opened 3 years ago by louisfeng
  0

• Any other models that have active time proportion (ATP) as low as DLRM?

  #9 opened 3 years ago by moderato
  1

• Get rid of profiler overheads in E2E

  #1 opened 3 years ago by moderato
  1

• Get input shapes of batched embedding lookup when extracting execution graph

  #7 opened 3 years ago by moderato
  1

• Code refactor

  #5 opened 3 years ago by moderato
  1

• Integrate with PARAM

  #3 opened 3 years ago by moderato
  0