Commits

Branch selector

 master

User selector

All users

All time

Commit History

Commits on Feb 20, 2019

• Merge pull request #233 from haolibai/master

  Show description for c989447
  jiaxiang-wu

  authored
  c989447

  Copy full SHA for c989447

Commits on Feb 19, 2019

• bugfix: remove 0s in __build_(bucket)_norm_quant_point()

  haolibai

  committed
  2b70cd5

  Copy full SHA for 2b70cd5

Commits on Feb 18, 2019

• Merge pull request #228 from jiaxiang-wu/ssd-chn-prune

  Show description for 3589a0b
  haolibai

  authored
  3589a0b

  Copy full SHA for 3589a0b

Commits on Feb 17, 2019

• remove commented-out code

  jiaxiang-wu

  committed
  e337255

  Copy full SHA for e337255

• use secondary sampling to speed-up LASSO

  jiaxiang-wu

  committed
  2045700

  Copy full SHA for 2045700

Commits on Feb 14, 2019

• remove redundant logs

  jiaxiang-wu

  committed
  300ad3c

  Copy full SHA for 300ad3c

• bugfix: some variables not initialized before broadcasting

  jiaxiang-wu

  committed
  a6a62ff

  Copy full SHA for a6a62ff

• bugfix: avoid infinite while loop

  jiaxiang-wu

  committed
  a93617e

  Copy full SHA for a93617e

• allow to skip certain Conv2D operations

  jiaxiang-wu

  committed
  7f43850

  Copy full SHA for 7f43850

• bugfix: incorrect padding size in Conv2D

  jiaxiang-wu

  committed
  c301ff1

  Copy full SHA for c301ff1

Commits on Feb 13, 2019

• add FLAGS for warm start to skip channel selection

  jiaxiang-wu

  committed
  7805d57

  Copy full SHA for 7805d57

• simplify the array concatenation

  jiaxiang-wu

  committed
  cb92289

  Copy full SHA for cb92289

• use hand-coded Adam for least-square regression

  jiaxiang-wu

  committed
  428c1d5

  Copy full SHA for 428c1d5

• bugfix: gamma's upper bound is not correctly set

  jiaxiang-wu

  committed
  1a4b1b3

  Copy full SHA for 1a4b1b3

• stop zero-padding to save time (complicates implementation)

  jiaxiang-wu

  committed
  8a6b11a

  Copy full SHA for 8a6b11a

• sort idle GPUs in the ascending order of mem usage

  jiaxiang-wu

  committed
  a417f34

  Copy full SHA for a417f34

• add logging outputs during channel pruning

  jiaxiang-wu

  committed
  14ce512

  Copy full SHA for 14ce512

Commits on Feb 12, 2019

• minor updates

  jiaxiang-wu

  committed
  0b78c0c

  Copy full SHA for 0b78c0c

Commits on Jan 31, 2019

• Merge pull request #220 from jiaxiang-wu/chn-prune

  Show description for 6576705
  jiaxiang-wu

  authored
  6576705

  Copy full SHA for 6576705

• fall-back to the original name for ChannelPrunedLearner

  jiaxiang-wu

  committed
  21b2be0

  Copy full SHA for 21b2be0

• rename channel pruning learners

  jiaxiang-wu

  committed
  405c5dc

  Copy full SHA for 405c5dc

• use floor() to round # of channels to integer

  jiaxiang-wu

  committed
  3e675b8

  Copy full SHA for 3e675b8

• bugfix: incorrect # of channels to be preserved

  jiaxiang-wu

  committed
  0d7bca2

  Copy full SHA for 0d7bca2

• switch between per-layer and final evaluation for pruning

  jiaxiang-wu

  committed
  a39ffcb

  Copy full SHA for a39ffcb

• bugfix: target value is not integer during binary search

  jiaxiang-wu

  committed
  246c651

  Copy full SHA for 246c651

• do not skip pruning for the last layer by default

  jiaxiang-wu

  committed
  5c17aa0

  Copy full SHA for 5c17aa0

• measure the time consumption for channel selection

  jiaxiang-wu

  committed
  9c7d6f1

  Copy full SHA for 9c7d6f1

Commits on Jan 30, 2019

• post-pruning fine-tuning run passed

  jiaxiang-wu

  committed
  d811614

  Copy full SHA for d811614

• remove unused FLAGS

  jiaxiang-wu

  committed
  96fb433

  Copy full SHA for 96fb433

• construct the meta least-square optimization

  jiaxiang-wu

  committed
  30992c1

  Copy full SHA for 30992c1

• construct the meta LASSO problem

  jiaxiang-wu

  committed
  0f26fb8

  Copy full SHA for 0f26fb8

• directly compute the weighing matrix based on least-square regression

  jiaxiang-wu

  committed
  98b5d89

  Copy full SHA for 98b5d89

• bugfix: incorrect dimension layout in the feature matrix

  jiaxiang-wu

  committed
  897bf1a

  Copy full SHA for 897bf1a

Commits on Jan 29, 2019

• move __smpl_inputs_n_outputs() to the front

  jiaxiang-wu

  committed
  5a35707

  Copy full SHA for 5a35707

• remove debug-purpose log outputs

  jiaxiang-wu

  committed
  79651f3

  Copy full SHA for 79651f3