Training script#51

Open

nviolante25 wants to merge 23 commits intoStability-AI:mainfrom

nviolante25:training

Commits on Apr 4, 2025

[ADD] sgm training code, loading code for DL2DV and initial config

Nicolas Violante
authored and
Nicolas Violante
committed

Commits on Apr 7, 2025

[ADD] SevaAutoenoderKL (wrapper to handle frames with image autoencoder)
Nicolas Violante
committed
[ADD] SevaWrapper
Nicolas Violante
committed
[ADD] correct naming and remove print
nviolante25
committed

Commits on Apr 8, 2025

[FIX] correct output in dataset
Nicolas Violante
committed
Merge branch 'training' of github.com:nviolante25/stable-virtual-camera into training
Nicolas Violante
committed
[ADD] conditioning in plucker and binary mask, mask out target latents
nviolante25
committed
[FIX] remove input frames from return dict
Nicolas Violante
committed

Commits on Apr 9, 2025

[FIX] correct shapes in wrapper, and dataset (needs to be checked) pass logging of images, add mask and plucer to VanillaCFG
Nicolas Violante
committed

Commits on Apr 10, 2025

Commits on Apr 11, 2025

[ADD] better names in loss weight
nviolante25
committed

Commits on Apr 15, 2025

[ADD] load SEVA checkpoint
nviolante25
committed
[FIX] upper idx when sampling adjacent frames
Nicolas Violante
committed

Commits on Apr 17, 2025

[FIX] verify that colmap poses correspond to the images, there couldbe more images than poses
Nicolas Violante
committed