Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Oct 12, 2020

• bug fix: using lambda will make all colalte_fn to be collatefn(split='test'), because lambda is just a pointer to the local state, and the split will be the split at that local state, which is 'tes…

  Show description for c305eb0
  ruotianluo

  committed
  c305eb0

  Copy full SHA for c305eb0

• Use lmdbdict instead of original lmdb

  ruotianluo

  committed
  02509ec

  Copy full SHA for 02509ec

• update readme installing as a package

  ruotianluo

  committed
  5d1a632

  Copy full SHA for 5d1a632

Commits on Aug 18, 2020

• bug fix for dataloaderraw and masks.

  ruotianluo

  committed
  cd651fa

  Copy full SHA for cd651fa

Commits on Jul 31, 2020

• bug fix: fix self.parameters bug on data parallel introduced in pth1.5

  ruotianluo

  committed
  282e4e3

  Copy full SHA for 282e4e3

Commits on Jul 30, 2020

• bug fix: py3 for make_bu_data.

  ruotianluo

  committed
  cf554a5

  Copy full SHA for cf554a5

Commits on Jul 29, 2020

• Fix some bugs.

  ruotianluo

  committed
  d850d26

  Copy full SHA for d850d26

Commits on Jul 20, 2020

• fix overflow when mixed precision.

  ruotianluo

  committed
  adacc50

  Copy full SHA for adacc50

• Handle types better.(mainly for use amp.)

  ruotianluo

  committed
  fb74f67

  Copy full SHA for fb74f67

Commits on Jul 15, 2020

• fix train_pl.

  ruotianluo

  committed
  0028c3e

  Copy full SHA for 0028c3e

• cpu evaluation

  ruotianluo

  committed
  2253580

  Copy full SHA for 2253580

Commits on Jul 12, 2020

• update because pytorch-lightning 0.8.5.

  ruotianluo

  committed
  436d900

  Copy full SHA for 436d900

Commits on Jul 6, 2020

• fix ruotianluo/ImageCaptioning.pytorch#100 and ruotianluo/self-critical.pytorch#176

  ruotianluo

  committed
  b232d49

  Copy full SHA for b232d49

Commits on Jul 5, 2020

• update readme.

  ruotianluo

  committed
  4e4d5f2

  Copy full SHA for 4e4d5f2

• formatting and fix bug and support evaluation.

  ruotianluo

  committed
  1f720ee

  Copy full SHA for 1f720ee

• Add distributed training supported by pytorch lightning.

  ruotianluo

  committed
  ce83c5d

  Copy full SHA for ce83c5d

• Overwrite state_dict for Noamopt.

  ruotianluo

  committed
  4acf2c9

  Copy full SHA for 4acf2c9

Commits on Jun 29, 2020

• Remove redundant code.

  ruotianluo

  committed
  547a215

  Copy full SHA for 547a215

• Drop usage of loader in eval_split_n.

  ruotianluo

  committed
  fb0d9e0

  Copy full SHA for fb0d9e0

Commits on Jun 24, 2020

• update readme.

  ruotianluo

  committed
  c5cd43e

  Copy full SHA for c5cd43e

Commits on Jun 18, 2020

• Make sure mask is float (to avoid wierd behavior.)

  ruotianluo

  committed
  5fdf374

  Copy full SHA for 5fdf374

Commits on Jun 7, 2020

• Add transformer step config.

  ruotianluo

  committed
  edb932c

  Copy full SHA for edb932c

• fix model zoo link.

  ruotianluo

  committed
  5edfc00

  Copy full SHA for 5edfc00

Commits on Jun 6, 2020

• [dataloader] add data_in_memory option to allow save all the features in memory.

  ruotianluo

  committed
  18e0d4b

  Copy full SHA for 18e0d4b

• update model zoo

  ruotianluo

  committed
  506c409

  Copy full SHA for 506c409

• Add reduce_on_plataeu options.

  ruotianluo

  committed
  32bdff4

  Copy full SHA for 32bdff4

• Update configs.

  ruotianluo

  committed
  1db214b

  Copy full SHA for 1db214b

• update readme

  ruotianluo

  committed
  cb7e7be

  Copy full SHA for cb7e7be

• Add download link for vilbert features.

  ruotianluo

  committed
  49f96c8

  Copy full SHA for 49f96c8

• fix dump_to_lmdb for py3(also added saving all keys in the lmdb mimicing vilbert)

  ruotianluo

  committed
  0b3949d

  Copy full SHA for 0b3949d

• [optim] Allow warmup used in conjunction with pervious stepwise lr decay.

  ruotianluo

  committed
  6c0af5b

  Copy full SHA for 6c0af5b

• [transformer] add cached transformer, which implements the "past" function in gpt2 in huggingface transformers. This can make generation faster. At inference time I see huge speedup but not during …

  Show description for 3e4d52d
  ruotianluo

  committed
  3e4d52d

  Copy full SHA for 3e4d52d

Commits on Jun 2, 2020

• Add support for adamw optimizer.

  ruotianluo

  committed
  52f6b8d

  Copy full SHA for 52f6b8d

• [Minor]Fix legacy option help.

  ruotianluo

  committed
  c63cb57

  Copy full SHA for c63cb57

Commits on May 29, 2020

• MIP: step 7: move train, eval script to tools.

  ruotianluo

  committed
  b0de03e

  Copy full SHA for b0de03e

Pagination

Previous

Next