Closed zhw-zhang closed 1 year ago
I trained maximum 300 steps on 5 videos-prompts that tooks 10min on colab t4.
Thanks for your reply, what setting do you fine-tune animatediff, why only 5 video-text pairs are needed?
Still experimenting. I'm resolving my own problem of video stylizing. Small step count (even 1 step) enough for me that model learn motion that i need. (I guess base model already knows all that motions).
I see, thank you very much
Hello, thank you very much for releasing your code, because my GPU is not very sufficient at the moment; so I would like to ask the cost of training on webvid first, how long did you train on which device?