search

neeek2303 / EMOPortraits

Official implementation of EMOPortraits: Emotion-enhanced Multimodal One-shot Head Avatars
311 stars 17 forks source link

cleanup #23

Closed johndpope closed 2 weeks ago

johndpope commented 2 weeks ago

this code is riddled with notebook pytorch checkpoints. it's complicated enough - makes searching / navigating codebase hard as if i search for unet3d - it's multiple hits.

rm -rf ./losses/.ipynb_checkpoints
rm -rf ./losses/rt_gene_loss/.ipynb_checkpoints
rm -rf  ./experiments/.ipynb_checkpoints
rm -rf  ./datasets/.ipynb_checkpoints
rm -rf  ./utils/.ipynb_checkpoints
rm -rf  ./models/stage_2/base/.ipynb_checkpoints
rm -rf  ./models/stage_1/volumetric_avatar/.ipynb_checkpoints
rm -rf  ./models/.ipynb_checkpoints
rm -rf  ./networks/common/.ipynb_checkpoints
rm -rf  ./networks/.ipynb_checkpoints
rm -rf  ./networks/volumetric_avatar/.ipynb_checkpoints
rm -rf  ./notebooks/.ipynb_checkpoints

https://gist.github.com/johndpope/981bfc11d31977629cb12bb082031acf

example - https://github.com/search?q=repo%3Aneeek2303%2FEMOPortraits%20idt_output_channels&type=code

johndpope commented 2 weeks ago

this actual code related to contrastive loss doesn't exist in repo - its just in the jupyter notebook

data_dict['idt_embed_face_target'] = idt_embed_cycle_face

i can see calculations done on them - https://github.com/search?q=idt_embed_face_target&type=code but it's a riddle how you assigned - i have to dig further into code.