Code

Add quantized training (CPU part)#5800

Code

Merged

shiyu1994 merged 31 commits intomastermicrosoft/LightGBM:masterfrom quantized-trainingmicrosoft/LightGBM:quantized-trainingCopy head branch name to clipboardMay 5, 2023

+2,920-290

Commits

Commits on Dec 22, 2022

• add quantized training (first stage)

  shiyu1994

  committed
  8187759

  Copy full SHA for 8187759

Commits on Mar 23, 2023

• Merge remote-tracking branch 'origin/master' into quantized-training

  shiyu1994

  committed
  8480873

  Copy full SHA for 8480873

• add histogram construction functions for integer gradients

  shiyu1994

  committed
  9e5d46b

  Copy full SHA for 9e5d46b

• add stochastic rounding

  shiyu1994

  committed
  dd2a3b4

  Copy full SHA for dd2a3b4

• update docs

  shiyu1994

  committed
  41c6c79

  Copy full SHA for 41c6c79

• fix compilation errors by adding template instantiations

  shiyu1994

  committed
  dfb5bc4

  Copy full SHA for dfb5bc4

• update files for compilation

  shiyu1994

  committed
  d830128

  Copy full SHA for d830128

• fix compilation of gpu version

  shiyu1994

  committed
  e82675b

  Copy full SHA for e82675b

Commits on Mar 30, 2023

• initialize gradient discretizer before share states

  shiyu1994

  committed
  1d68e97

  Copy full SHA for 1d68e97

• Merge remote-tracking branch 'origin/master' into quantized-training

  shiyu1994

  committed
  4ccdf34

  Copy full SHA for 4ccdf34

Commits on Apr 5, 2023

• add a test case for quantized training

  shiyu1994

  committed
  27dbf8c

  Copy full SHA for 27dbf8c

• add quantized training for data distributed training

  shiyu1994

  committed
  5c8aac1

  Copy full SHA for 5c8aac1

Commits on Apr 6, 2023

• Delete origin.pred

  shiyu1994

  authored
  1fd115a

  Copy full SHA for 1fd115a

• Delete ifelse.pred

  shiyu1994

  authored
  197b394

  Copy full SHA for 197b394

• Delete LightGBM_model.txt

  shiyu1994

  authored
  7140bb8

  Copy full SHA for 7140bb8

• remove useless changes

  shiyu1994

  authored
  1f142d5

  Copy full SHA for 1f142d5

Commits on Apr 19, 2023

• Merge remote-tracking branch 'origin/master' into quantized-training

  shiyu1994

  committed
  22a98b7

  Copy full SHA for 22a98b7

• Merge branch 'quantized-training' of https://github.com/Microsoft/LightGBM into quantized-training

  shiyu1994

  committed
  bc848a0

  Copy full SHA for bc848a0

• fix lint error

  shiyu1994

  committed
  d5fc93d

  Copy full SHA for d5fc93d

• remove debug loggings

  shiyu1994

  committed
  ed066d0

  Copy full SHA for ed066d0

Commits on Apr 24, 2023

• fix mismatch of vector and allocator types

  shiyu1994

  committed
  06826f0

  Copy full SHA for 06826f0

Commits on Apr 25, 2023

• remove changes in main.cpp

  shiyu1994

  committed
  025ad39

  Copy full SHA for 025ad39

• fix bugs with uninitialized gradient discretizer

  shiyu1994

  committed
  baef468

  Copy full SHA for baef468

• initialize ordered gradients in gradient discretizer

  shiyu1994

  committed
  ce93015

  Copy full SHA for ce93015

• disable quantized training with gpu and cuda

  Show description for 2b1118c
  shiyu1994

  committed
  2b1118c

  Copy full SHA for 2b1118c

Commits on Apr 26, 2023

• fix bug in data parallel tree learner

  shiyu1994

  committed
  487f2c4

  Copy full SHA for 487f2c4

• make quantized training test deterministic

  shiyu1994

  committed
  8c0e67b

  Copy full SHA for 8c0e67b

• make quantized training in test case more accurate

  shiyu1994

  committed
  6a76fde

  Copy full SHA for 6a76fde

• refactor test_quantized_training

  shiyu1994

  committed
  0812403

  Copy full SHA for 0812403

Commits on May 4, 2023

• fix leaf splits initialization with quantized training

  shiyu1994

  committed
  9c8894b

  Copy full SHA for 9c8894b

Commits on May 5, 2023

• check distributed quantized training result

  shiyu1994

  committed
  788e1aa

  Copy full SHA for 788e1aa