All issues

Search results

Question about codebook collapse during training

Status: Open.

#28In ByteVisionLab/TokenFlow;

· iceporridge opened on Apr 26, 2026

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

Status: Open.

#24In ByteVisionLab/TokenFlow;

· tianzhangwu opened on Jan 16, 2025

Are you using --ema ?

Status: Open.

#23In ByteVisionLab/TokenFlow;

· tianzhangwu opened on Jan 15, 2025

The reconstruction loss increases when adding discriminator loss

Status: Open.

#22In ByteVisionLab/TokenFlow;

· tianzhangwu opened on Jan 15, 2025

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

Status: Open.

#17In ByteVisionLab/TokenFlow;

· tianzhangwu opened on Jan 8, 2025

The direct application of this method to VAE

Status: Open.

#15In ByteVisionLab/TokenFlow;

· lavinal712 opened on Jan 7, 2025

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

Status: Open.

#14In ByteVisionLab/TokenFlow;

· tianzhangwu opened on Jan 6, 2025

code for i2t inference

Status: Open.

#10In ByteVisionLab/TokenFlow;

· pkulium opened on Dec 19, 2024

related work suggestion

Status: Open.

#2In ByteVisionLab/TokenFlow;

· youngsheen opened on Dec 5, 2024