DavidHuji/CapDec

Issues

• Why noise is added to “prefix”（visual embeding by clip） instead of token（text embeding by tokenizer）in the code？

  #19 opened 9 months ago by Panlizhi
  1

• Error(s) in loading state_dict for ClipCaptionModel

  #15 opened a year ago by sszbuu
  0

• modality_offset

  #14 opened a year ago by Flaick
  1

• how to download / create single_caption_per_sample_val.json file

  #7 opened 2 years ago by BoiAkay
  6

• how to get train.json?

  #13 opened a year ago by qq123aa456
  1

• About FlickrStyle10K

  #12 opened a year ago by LiLiParty
  1

• About using the Sentcap dataset and setting parameters for train.py

  #10 opened 2 years ago by Rouchestand
  5

• training parameters on Flickr30k dataset

  #11 opened 2 years ago by mzhaoshuai
  2

• about the performance of the Zerocap demonstrated in this paper

  #9 opened 2 years ago by wanboyang
  1

• about the parameters of train.py

  #8 opened 2 years ago by tianguang2525
  1

• Different results at CiDER metric in paper

  #6 opened 2 years ago by taewhankim
  1

• About the type of gpt2

  #4 opened 2 years ago by zjr2000
  1

• annotations files of coco during evaluation

  #5 opened 2 years ago by mzhaoshuai
  1

• About the variance of 0.016

  #3 opened 2 years ago by 232525
  3

• Requirements File

  #2 opened 2 years ago by Adesoji1
  3