Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Jun 5, 2020

• Adjusting the learning rate to freeze at last, when passed the decay interval.

  mnaumovfb

  committed
  4d4371d

  Copy full SHA for 4d4371d

Commits on May 31, 2020

• Adjusting ONNX calls to work with large models (more than 2GB in size).

  mnaumovfb

  committed
  3df2e30

  Copy full SHA for 3df2e30

Commits on May 28, 2020

• added more visualization options (#76)

  Show description for 74c5bb0
  dkorchevgithub

  authored
  74c5bb0

  Copy full SHA for 74c5bb0

Commits on May 23, 2020

• Enable LR warmup and decay policy (#73)

  Show description for fcd0969
  rachithayp

  authored
  fcd0969

  Copy full SHA for fcd0969

Commits on May 11, 2020

• added visualization of DLRM embeddings (#72)

  Show description for 6b029a5
  dkorchevgithub

  authored
  6b029a5

  Copy full SHA for 6b029a5

Commits on Mar 27, 2020

• Update README.md

  mnaumovfb

  authored
  4705ea1

  Copy full SHA for 4705ea1

Commits on Mar 25, 2020

• Update README.md

  mnaumovfb

  authored
  3a4f50f

  Copy full SHA for 3a4f50f

• Update README.md

  mnaumovfb

  authored
  f8ec4bf

  Copy full SHA for f8ec4bf

Commits on Mar 4, 2020

• Adjusting restart from saved model during training. Need to skip early batches when enumerate is used.

  Show description for 7e6dce5
  mnaumovfb

  committed
  7e6dce5

  Copy full SHA for 7e6dce5

Commits on Feb 27, 2020

• Switch the binary dataloader to int32 datatype (#60)

  tgrel

  authored
  03a1534

  Copy full SHA for 03a1534

Commits on Feb 20, 2020

• Adjusting the use of --max-ind-range post processing as discussed in https://github.com/facebookresearch/dlrm/issues/53

  mnaumovfb

  committed
  924032c

  Copy full SHA for 924032c

• Adding flexibility in saving/loading model to/from different devices. Also, enforcing single copy of embeddings across devices on multiple GPUs.

  mnaumovfb

  committed
  5d31ee1

  Copy full SHA for 5d31ee1

Commits on Feb 15, 2020

• adding back end of epoch check for now.

  mnaumovfb

  committed
  b345969

  Copy full SHA for b345969

• Update README.md

  mnaumovfb

  authored
  5f00309

  Copy full SHA for 5f00309

• Update README.md

  mnaumovfb

  authored
  3331553

  Copy full SHA for 3331553

• Tgrel/minor mlperf fixes (#54)

  Show description for a5d707f
  tgrel

  authored
  a5d707f

  Copy full SHA for a5d707f

Commits on Feb 9, 2020

• Fix latent bug in caffe2 version when --max-ind-range is used

  mnaumovfb

  committed
  1768658

  Copy full SHA for 1768658

Commits on Feb 5, 2020

• Adding support for testing and mlperf flags to caffe2 version.

  mnaumovfb

  committed
  58c2806

  Copy full SHA for 58c2806

Commits on Jan 26, 2020

• Adding support for binary loader proposed in pull request https://github.com/facebookresearch/dlrm/pull/49 Also, adding an option for early stopping based on AUC. Finally, adding exact model config…

  Show description for bdba86d
  mnaumovfb

  committed
  bdba86d

  Copy full SHA for bdba86d

Commits on Jan 22, 2020

• Fixing a bug in dataset loader. When multiple passes over the data are done you must reset/rotate the day being used at the end.

  Show description for 7c55962
  mnaumovfb

  committed
  7c55962

  Copy full SHA for 7c55962

Commits on Jan 12, 2020

• Adding support for weigthed BCEloss (wbce). Also, adjusting Caffe2 version reporting of accuracy and loss to match that of PyTorch.

  Show description for 3acf18f
  mnaumovfb

  committed
  3acf18f

  Copy full SHA for 3acf18f

Commits on Dec 31, 2019

• Slight changes to data loaders for consistency.

  Show description for 0fb34f3
  mnaumovfb

  committed
  0fb34f3

  Copy full SHA for 0fb34f3

Commits on Dec 30, 2019

• Adjusting metrics to be consistent with and without --mlperf-logging flag.

  Show description for 4545b83
  mnaumovfb

  committed
  4545b83

  Copy full SHA for 4545b83

Commits on Dec 19, 2019

• Adding --mlperf-threshold command line parameter that allows early stopping when specified accuracy is reached.

  Show description for ce8557b
  mnaumovfb

  committed
  ce8557b

  Copy full SHA for ce8557b

Commits on Dec 17, 2019

• Minor adjustments.

  Show description for 9e4c0d1
  mnaumovfb

  committed
  9e4c0d1

  Copy full SHA for 9e4c0d1

• Merge branch 'tgrel/dlrm' of https://github.com/tgrel/dlrm into tgrel-tgrel/dlrm

  mnaumovfb

  committed
  dc30563

  Copy full SHA for dc30563

Commits on Dec 16, 2019

• Adjusting test data loading to work with --test-num-workers.

  Show description for 08fdeaf
  mnaumovfb

  committed
  08fdeaf

  Copy full SHA for 08fdeaf

Commits on Dec 13, 2019

• Change license to MIT

  tgrel

  committed
  66086e6

  Copy full SHA for 66086e6

Commits on Dec 11, 2019

• Adding quotient-remainder and mixed-dimensions embedding tricks.

  Show description for cac6723
  mnaumovfb

  committed
  cac6723

  Copy full SHA for cac6723

Commits on Dec 9, 2019

• Add batch-size-independent way of specifying evaluation frequency. Compute AUC and AP metrics accurately

  tgrel

  committed
  d5cc170

  Copy full SHA for d5cc170

• Usability improvement

  tgrel

  committed
  6d6b9a7

  Copy full SHA for 6d6b9a7

Commits on Dec 7, 2019

• Update requirements with necessary caffe2 dependencies. (#45)

  

  tayo

  authored and

  mnaumovfb

  committed
  e257b6e

  Copy full SHA for e257b6e

Commits on Dec 3, 2019

• Dataloading improvements for the terabyte dataset

  tgrel

  committed
  587d4cc

  Copy full SHA for 587d4cc

• Add ROC AUC and average precision scores, include dataloading time in profiling measurements

  tgrel

  committed
  d4f97cb

  Copy full SHA for d4f97cb

Commits on Nov 27, 2019

• Add dockerfile and run_and_time.sh

  tgrel

  committed
  f155e7c

  Copy full SHA for f155e7c

Pagination

Previous

Next