Open wangxihao opened 11 months ago
Hello. Thank you. Motion module trained with position encoding 24. This multiplier stack original module side by side by multiplying by factor 2. If you set it to 2, then position encoding will have capacity of 48 frames, 3 - 72, 4 - 96 and so on.
ср, 18 окт. 2023 г., 09:49 Tho @.***>:
[image: image] https://user-images.githubusercontent.com/26833394/276069121-6e1d64c5-a39f-47a7-88fb-3d2a70f2a50a.png Nice work! When I train on sample_n_frames=32 , what should I set the value of the motion_module_pe_multiplier to? Is it okay to keep the value as 1? What is the basis of the principle here? Thank you!
— Reply to this email directly, view it on GitHub https://github.com/tumurzakov/AnimateDiff/issues/14, or unsubscribe https://github.com/notifications/unsubscribe-auth/AADLQLGISAEWKPXXK5G6NPDX75GWBAVCNFSM6AAAAAA6E2LZ5CVHI2DSMVQWIX3LMV43ASLTON2WKOZRHE2DQNZUG43DMOI . You are receiving this because you are subscribed to this thread.Message ID: @.***>
Nice work! When I train on sample_n_frames=32 , what should I set the value of the motion_module_pe_multiplier to? Is it okay to keep the value as 1? What is the basis of the principle here? Thank you!