search

showlab / MotionDirector

[ECCV 2024 Oral] MotionDirector: Motion Customization of Text-to-Video Diffusion Models.
https://showlab.github.io/MotionDirector/
Apache License 2.0
850 stars 54 forks source link

earlier checkpoints doing better than later checkpoints #34

Open sam-motamed opened 7 months ago

sam-motamed commented 7 months ago

Hi,

In my experiments, earlier checkpoints (~ 200 - 400) for both temporal and spatial tuning do much better than later checkpoints. To the point that later checkpoints seem like they generate pure noise videos! what could be the cause of this?

LiuHuijie6410 commented 6 months ago

Hello, could you please share the early results you observed?

ruizhaocv commented 3 months ago

I think it is abnormal to produce pure noise videos. Can you provide more details, such as the package versions and the configs you used?