PixelFace

Dataset

MMCelebA-HQ: image, pickle, json

Pretraiend Models

1. Text Encoder
2. Image Encoder
3. CheckPoint

Train on Multi-Modal CelebA-HQ

CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.launch --nproc_per_node NPROC_PER_NODE --master_port MASTER_PORT main.py --cfg cfg/mmceleba.yml

or

CUDA_VISIBLE_DEVICES=0,1 torchrun --nproc_per_node NPROC_PER_NODE --master_port MASTER_PORT main.py --cfg cfg/mmceleba.yml

Test

CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.launch --nproc_per_node NPROC_PER_NODE --master_port MASTER_PORT main.py --cfg cfg/eval_mmceleba.yml

or

CUDA_VISIBLE_DEVICES=0,1 torchrun --nproc_per_node NPROC_PER_NODE --master_port MASTER_PORT main.py --cfg cfg/eval_mmceleba.yml