huggingface/pytorch-openai-transformer-lm

Issues

• Instructions on how to train a language model from scratch

  #29 opened 6 years ago by Froskekongen
  9

• Avoid model overfitting

  #31 opened 6 years ago by BangLiu
  9

• problems about DoubleHeadModel implementation

  #50 opened 6 years ago by eveliao
  1

• Instructions for encoding own sentences

  #58 opened 5 years ago by izaskr
  1

• Training from scratch: Repeated and mangled words

  #59 opened 5 years ago by maruker
  0

• Running on new dataset similar to rocstories

  #57 opened 6 years ago by priyanka-chaudhary
  0

• How to create transforms for entailment task?

  #40 opened 6 years ago by lordzuko
  12

• vocab = n_vocab + n_special + n_ctx means?

  #54 opened 6 years ago by JiahangOK
  1

• ConvAI

  #55 opened 6 years ago by bkj
  0

• Why do we need to apply mask while fine tuning?

  #53 opened 6 years ago by pranoy-k
  4

• Confused by multiply n to both clf_logits and clf_losses?

  #39 opened 6 years ago by Vimos
  3

• Implementation of Seq2Seq with Transformer

  #52 opened 6 years ago by bhedayat
  0

• why it didn't use softmax in computing multichoice loss

  #51 opened 6 years ago by eveliao
  1

• How should one modify the code to successfully run text classification?

  #43 opened 6 years ago by davidefiocco
  7

• Potentially incorrect regex in text_utils.py

  #48 opened 6 years ago by schmmd
  0

• Why is output vocab including positional embeddings?

  #47 opened 6 years ago by OanaMariaCamburu
  2

• Encoder paddings influence results?

  #45 opened 6 years ago by OanaMariaCamburu
  0

• Retrain the LM on new dataset?

  #46 opened 6 years ago by fabrahman
  0

• help to understand bpe logic

  #42 opened 6 years ago by BogdanDidenko
  2

• Results and questions on text generation experiments with pretrained LM model

  #36 opened 6 years ago by xiaoda99
  10

• How does position embedding implementation work?

  #44 opened 6 years ago by bcserna
  2

• In `transform_roc`, why do we need `xmb[:, :, :, 1] `?

  #12 opened 6 years ago by FrankWork
  3

• How is the file "cloze_test_test__spring2016 - cloze_test_ALL_test.csv" created?

  #41 opened 6 years ago by luffycodes
  5

• Vocabulary size code explanation and occasionally shape error

  #38 opened 6 years ago by Vimos
  2

• Redundant decoder

  #32 opened 6 years ago by joshim5
  1

• finetuning the model for NLI, but using sentence embeddings instead of word embeddings

  #30 opened 6 years ago by saurabhvyas
  2

• How to use Inference?

  #33 opened 6 years ago by masati91
  4

• So we can not change the word embedding with the pretrained LM?

  #34 opened 6 years ago by herbertchen1
  1

• How does Dropout2d help in cloze task?

  #11 opened 6 years ago by sai-prasanna
  12

• loading pretrained open ai model

  #28 opened 6 years ago by mehdimashayekhi
  3

• Can someone explain this line?

  #21 opened 6 years ago by teucer
  4

• Clarifying last step of the 'transform_roc' function

  #26 opened 6 years ago by sharpsy
  7

• Having various network heads

  #24 opened 6 years ago by rodgzilla
  1

• what is the use of dropout in the Transformer?

  #19 opened 6 years ago by teucer
  2

• Object is not specified

  #13 opened 6 years ago by Oktai15
  4

• DoubleHeadModel is not instanciated when n_gpu <= 1

  #14 opened 6 years ago by rodgzilla
  1

• Pre-trained LMHead

  #7 opened 6 years ago by rodgzilla
  3

• dimensioning bug?

  #2 opened 6 years ago by jtatusko
  4

• a git clone is large because history has the "add model" commit

  #3 opened 6 years ago by soumith
  1