PSHuman

Running on Zero

App Files Files Community

PSHuman

15.2 MB

Ctrl+K

2 contributors

History: 172 commits

Daankular

Remove verbose= kwarg from ReduceLROnPlateau: removed in newer PyTorch (we are now on 2.8.0)

ab4dd9b verified 5 days ago

assets
Migrated from GitHub almost 2 years ago
blender
Migrated from GitHub almost 2 years ago
configs
Update configs/inference-768-6view.yaml almost 2 years ago
core
Migrated from GitHub almost 2 years ago
examples
Migrated from GitHub almost 2 years ago
lib
Fix device mismatch: maskrcnn detector lands on cuda under mmgps lingering default device, but torch.from_numpy() input stays on cpu 5 days ago
mvdiffusion
Implement attention slicing: chunk SDPA calls along the batch*heads dim since torch 2.8s fused kernels reject this shape on this GPU, forcing an O(seq_len^2 * batch) math fallback that OOMs even a 96GB card 5 days ago
node_config
Migrated from GitHub almost 2 years ago
scripts
Migrated from GitHub almost 2 years ago
utils
Migrated from GitHub almost 2 years ago
.gitattributes

1.6 kB
Migrated from GitHub almost 2 years ago
LICENSE.txt

1.11 kB
Migrated from GitHub almost 2 years ago
ORIGINAL_README.md

3.65 kB
Migrated from GitHub almost 2 years ago
README.md

325 Bytes
downgradio 5.8.0 about 1 year ago
app.py

8.68 kB
Bump GPU duration to 600s: memory issue is fixed (attention slicing), task now just needs more time for the full multiview diffusion + mesh reconstruction pipeline 5 days ago
econdataset.py

15.7 kB
Migrated from GitHub almost 2 years ago
inference.py

10.5 kB
Reset torch default device to cpu after offload.profile(): fixes device mismatches in econdataset/pymaf/smplx code that assumes CPU-default tensor creation, without affecting mmgps explicit hook-based device management 5 days ago
reconstruct.py

21.1 kB
Remove verbose= kwarg from ReduceLROnPlateau: removed in newer PyTorch (we are now on 2.8.0) 5 days ago
requirements.txt

4.5 kB
Bump diffusers to 0.29.0: mmgp (via optimum-quanto) needs PixArtTransformer2DModel, absent from the old 0.26.0 pin; verified all of PSHumans own diffusers imports (incl. LoRACompatibleConv/AdaLayerNorm) still resolve at 0.29.0 5 days ago
train_mvdiffusion_unit_unclip.py

58.2 kB
Migrated from GitHub almost 2 years ago