Issues

• Question about codebook collapse during training

  ByteVisionLab/TokenFlow#28 · Filter by author iceporridge opened on Apr 26, 2026
  4

• CLIP ViT-B/14 in the paper, but ViT-B/16 in the code.

  ByteVisionLab/TokenFlow#24 · Filter by author tianzhangwu opened on Jan 16, 2025
  1

• Are you using --ema ?

  ByteVisionLab/TokenFlow#23 · Filter by author tianzhangwu opened on Jan 15, 2025
  1

• The reconstruction loss increases when adding discriminator loss

  ByteVisionLab/TokenFlow#22 · Filter by author tianzhangwu opened on Jan 15, 2025
  3

• Is there a timetable for single-scale version of TokenFlow?

  ByteVisionLab/TokenFlow#17 · Filter by author tianzhangwu opened on Jan 8, 2025

• The direct application of this method to VAE

  ByteVisionLab/TokenFlow#15 · Filter by author lavinal712 opened on Jan 7, 2025
  1

• The final input to the LLM is still a continuous feature representation for i2t

  ByteVisionLab/TokenFlow#14 · Filter by author tianzhangwu opened on Jan 6, 2025
  5

• code for i2t inference

  ByteVisionLab/TokenFlow#10 · Filter by author pkulium opened on Dec 19, 2024
  3

• related work suggestion

  ByteVisionLab/TokenFlow#2 · Filter by author youngsheen opened on Dec 5, 2024