Open SWAPv1 opened 1 week ago
Thanks. Sure. You can modify the temporal_length
and video_length
to 32 or other lengths in the config.yaml and train the model.
great thank you! Do you know how much more vram I would need to train a 32 frame length video model using the 512 or 1024 model?
I am not sure. I know that 25 frames (1024 x 576) may be the max. length for training on A100 (40G) GPUs.
Hello. I'm amazed by your model. However the 16 frame is very limiting. Is there a way to train the model so it can be extended to 32 frames or more?