Commits

Branch selector

 with_finetune

User selector

All users

Datepicker

All time

Commit History

Commits on Sep 26, 2017

• Merge branch 'master' into with_finetune

  Show description for 125645d
  ruotianluo

  committed
  125645d

  Copy full SHA for 125645d

• Make a common father class CaptionModel (the original CaptionModel is renamed as OldModel). The new CaptionModel is designed to reduce the code used for beam search.

  Show description for 35185f5
  ruotianluo

  committed
  35185f5

  Copy full SHA for 35185f5

• Stop expand finished beam when beam search.

  ruotianluo

  committed
  cca154d

  Copy full SHA for cca154d

Commits on Sep 5, 2017

• Fix the bug introduced by merging.

  ruotianluo

  committed
  3897855

  Copy full SHA for 3897855

• Merge branch 'master' into with_finetune

  Show description for 212bb6b
  ruotianluo

  committed
  212bb6b

  Copy full SHA for 212bb6b

• Fix the final bug in CaptionModel sample_beam.

  ruotianluo

  authored
  097ed64

  Copy full SHA for 097ed64

• Again further fix bugs in CaptionModel in sample_beam

  ruotianluo

  authored
  35719f3

  Copy full SHA for 35719f3

• Futher fix the sample_beam in CaptionModel.

  ruotianluo

  authored
  4880150

  Copy full SHA for 4880150

• Fix sample_beam in CaptionModel.

  ruotianluo

  authored
  c16f982

  Copy full SHA for c16f982

Commits on Sep 4, 2017

• Make sure to use vocab in infos when evaluating.

  ruotianluo

  committed
  0694263

  Copy full SHA for 0694263

• Fix the problem of duplicating name in dataloaderaw.

  ruotianluo

  committed
  a88a6d8

  Copy full SHA for a88a6d8

• Fix that the pretrained model doens't have use_att option.

  ruotianluo

  committed
  0aeb222

  Copy full SHA for 0aeb222

Commits on Aug 31, 2017

• Use torch.utils.data to load file.

  ruotianluo

  committed
  b58e41d

  Copy full SHA for b58e41d

Commits on Aug 12, 2017

• Merge branch 'master' into with_finetune

  Show description for ca0b24e
  ruotianluo

  committed
  ca0b24e

  Copy full SHA for ca0b24e

Commits on Aug 11, 2017

• Disable scheduled sampling during evaluation mode.

  ruotianluo

  committed
  0b8c057

  Copy full SHA for 0b8c057

• Fix a bug in resnet_utils.

  ruotianluo

  committed
  954b2ec

  Copy full SHA for 954b2ec

Commits on Aug 7, 2017

• Don't load attention features if the model is not attention model.

  ruotianluo

  committed
  0155760

  Copy full SHA for 0155760

Commits on Aug 5, 2017

• Remove coco/.

  ruotianluo

  committed
  b722828

  Copy full SHA for b722828

• Rewrite the readme.

  ruotianluo

  committed
  8024c33

  Copy full SHA for 8024c33

• Change the name prepro_feats_npy to prepro_feats

  ruotianluo

  committed
  fdf9fde

  Copy full SHA for fdf9fde

• Add option in eval.py to use different resnets. (Must match with what you use to preprocess.)

  ruotianluo

  committed
  062118e

  Copy full SHA for 062118e

Commits on Aug 2, 2017

• Specify the location of pretrained resnet model.

  ruotianluo

  committed
  879971b

  Copy full SHA for 879971b

• Merge branch 'master' into with_finetune

  Show description for 2de14d2
  ruotianluo

  committed
  2de14d2

  Copy full SHA for 2de14d2

• Save the histories of loss lr etc. in a histories pkl file to reduce the original infos file, because infos are needed for evaluation but histories are actually not necessary.

  ruotianluo

  committed
  2eb03b6

  Copy full SHA for 2eb03b6

• Allow both with tensorflow and without tensorflow to remove the redundant code (train_tb.py).

  ruotianluo

  committed
  e787220

  Copy full SHA for e787220

• Move all models from misc to models folder.

  ruotianluo

  committed
  e4d4811

  Copy full SHA for e4d4811

• 1. Add Top Down attention model, it achieves the best performance now.

  Show description for 7f4f268
  ruotianluo

  committed
  7f4f268

  Copy full SHA for 7f4f268

• Fix dropout problems:

  Show description for 4ec0819
  ruotianluo

  committed
  4ec0819

  Copy full SHA for 4ec0819

Commits on Aug 1, 2017

• Change mean because of 0.2 behaviour.

  ruotianluo

  committed
  93ddeb4

  Copy full SHA for 93ddeb4

Commits on Jul 19, 2017

• Change mean to fit pytorch 0.2 mean behavior.

  ruotianluo

  committed
  5845b61

  Copy full SHA for 5845b61

Commits on Jul 18, 2017

• Use inferred shape to expand. (related to last commit.)

  ruotianluo

  committed
  6456b1e

  Copy full SHA for 6456b1e

• Use inferred shape to expand. (related to last commit.)

  ruotianluo

  committed
  9d66010

  Copy full SHA for 9d66010

• Update AttModel.py

  

  raoyongming

  authored and

  ruotianluo

  committed
  bdc4a97

  Copy full SHA for bdc4a97

• Merge pull request #8 from rym9005023/with_finetune

  Show description for 39b68b1
  ruotianluo

  authored
  39b68b1

  Copy full SHA for 39b68b1

• Update AttModel.py

  raoyongming

  authored
  acc66e4

  Copy full SHA for acc66e4

Pagination

Previous

Next