search

pixeli99 / SVD_Xtend

Stable Video Diffusion Training Code and Extensions.
553 stars 51 forks source link

training gpu cost #4

Closed zhw-zhang closed 8 months ago

zhw-zhang commented 8 months ago

Hi, thank you for your open-source code. How much GPU consumption is required during training? Is it necessary to add additional deepspeed or checkpoint to save training memory consumption?

pixeli99 commented 8 months ago

This depends on your specific training setup, with my current configuration (512*320, 16 frames, with all parameters of unet trainable), it will use approximately 63GB of VRAM (I haven't enabled any optimizations, such as deepspeed and so on)

zhw-zhang commented 8 months ago

Thank you for your reply. So your batchsize is 1? At 512*320 resolution.

pixeli99 commented 8 months ago

Yes, batchsize=1