search

guoqincode / Open-AnimateAnyone

Unofficial Implementation of Animate Anyone
2.89k stars 233 forks source link

train stage 1 OOM #69

Closed LeonJoe13 closed 8 months ago

LeonJoe13 commented 8 months ago

Thanks for your great effort in this wonderful work! I have some questions about train stage 1, my env is V100 32G, torch 2.0.1, python 3.9.18 I set train_batch_size=1,
sample_size: 16 # for 40G 256 sample_stride: 4, sample_n_frames: 2 mixed_precision_training: true enable_xformers_memory_efficient_attention: True in train_stage_1.yaml, I still got OOM,could you help me?

guoqincode commented 8 months ago

I tried your configuration and it requires 36G VRAM.

guoqincode commented 8 months ago

Maybe you can start by not training unet and only train uent after some training steps (pose and appearance are good)

LeonJoe13 commented 8 months ago

Maybe you can start by not training unet and only train uent after some training steps (pose and appearance are good)

thank you! I frezze referencenet and it works!

wangxr1999 commented 8 months ago

How should it be modified？