Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Apr 1, 2020

• Support for Python 3, PyTorch 1.3+, CPU, ship weights

  lantiga

  committed
  a73d9c3

  Copy full SHA for a73d9c3

Commits on Mar 9, 2019

• Merge pull request deep-learning-with-pytorch#1 from ruotianluo/master

  Show description for 8c5862e
  elistevens

  authored
  8c5862e

  Copy full SHA for 8c5862e

Commits on Oct 7, 2018

• Fix eval script with python3 and pytorch0.4 (ruotianluo#72)

  Show description for a1445e4
  

  dmitriy-serdyuk

  authored and

  ruotianluo

  committed
  a1445e4

  Copy full SHA for a1445e4

Commits on Sep 28, 2018

• Pytorch 4 (ruotianluo#71)

  Show description for 7ad027f
  

  dmitriy-serdyuk

  authored and

  ruotianluo

  committed
  7ad027f

  Copy full SHA for 7ad027f

• Bump pytorch to 4.1 (ruotianluo#70)

  Show description for 0ff90bd
  

  dmitriy-serdyuk

  authored and

  ruotianluo

  committed
  0ff90bd

  Copy full SHA for 0ff90bd

Commits on Sep 20, 2018

• Add model weights

  lantiga

  committed
  49e8189

  Copy full SHA for 49e8189

Commits on Mar 8, 2018

• Update Readme for legacy conversion.

  ruotianluo

  committed
  622b6a5

  Copy full SHA for 622b6a5

• Change convert_old.py(no need to specify the feature size in advance, we can directly use the loaded array size to create dataset);

  Show description for 99de7af
  ruotianluo

  committed
  99de7af

  Copy full SHA for 99de7af

• saving to and loading from single h5 (ruotianluo#38)

  Show description for 6b8a740
  

  ReyhaneAskari

  authored and

  ruotianluo

  committed
  6b8a740

  Copy full SHA for 6b8a740

Commits on Dec 11, 2017

• Update README.md

  Show description for 9c1dd52
  ruotianluo

  authored
  9c1dd52

  Copy full SHA for 9c1dd52

Commits on Oct 3, 2017

• Suppress unk when beam search.

  ruotianluo

  committed
  13516df

  Copy full SHA for 13516df

Commits on Sep 26, 2017

• 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 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 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

• 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

Pagination

Previous

Next