lucidrains/transfusion-pytorch

Issues

• todo

  #29 opened 20 days ago by lucidrains
  0

• Image Generation

  #22 opened a month ago by taeinkwon
  75

• issue with checkpoint saving

  #30 opened 19 days ago by mijanur132
  1

• Generate text description of image prompt

  #27 opened 20 days ago by mijanur132
  4

• Unsure of the functionality of one line

  #28 opened 21 days ago by Ivan-Zhong
  1

• Any idea about image caption implementation.

  #26 opened a month ago by Ma-Weijian
  3

• Example of using VAE for image.

  #20 opened a month ago by dingkwang
  3

• Overfit example seems to modify the text_and_images in place

  #24 opened a month ago by robertmash2
  2

• Default times are multiplied by batch size

  #23 opened a month ago by RefractAI
  4

• Issue with "decoding text" not finishing

  #21 opened 2 months ago by siyuan5
  3

• Why are you eliminating the influence of positional information on modalities using RoPE?

  #19 opened 2 months ago by shin-wn
  15

• modality_length_to_times_fn always default

  #18 opened 2 months ago by RefractAI
  1

• [BUG] Fail to run the test samples

  #17 opened 3 months ago by Masaaki-75
  4

• Offset setting miss due to addition of meta_id and modality_meta_info?

  #16 opened 3 months ago by shin-wn
  1

• omnigen plz

  #15 opened 3 months ago by af-74413592
  0

• Could you prepare a reasoning demo?

  #13 opened 3 months ago by win10ogod
  1

• Got error when running example

  #12 opened 3 months ago by Bing1002
  6

• text layernorm

  #11 opened 3 months ago by cliangyu
  2

• Question about Diffusion Loss

  #8 opened 4 months ago by JJJYmmm
  1

• Bug: modality_token_transform is empty?

  #6 opened 4 months ago by GindaChen
  3

• Is there any pretrained model weight?

  #5 opened 4 months ago by chenfengshijie
  1

• Question

  #1 opened 4 months ago by zhang-haojie
  1

• Batch size >1 not working, and loss queries

  #3 opened 4 months ago by RefractAI
  1

• Question about num_text_tokens setting in Transfusion

  #4 opened 4 months ago by wdlctc
  4