Commit History
28 days ago
-
Update app.py Running
Verified signature91ab218
-
Remove verbose= kwarg from ReduceLROnPlateau: removed in newer PyTorch (we are now on 2.8.0)
Verified signatureab4dd9b -
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
Verified signature181be20 -
Fix device mismatch: maskrcnn detector lands on cuda under mmgps lingering default device, but torch.from_numpy() input stays on cpu
Verified signaturecf35ba7 -
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
Verified signaturecb5743a -
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
Verified signature63d710a -
Diagnose why SDPA auto-select avoids flash/efficient backends: try FLASH_ATTENTION explicitly and log the actual error, plus dtype/contiguity/mask info
Verified signature4df1266 -
Add memory diagnostics at each attention call to see if VRAM usage is growing unbounded or just spiking once
Verified signature4ba1e6a -
Add GPU memory diagnostic print to confirm whether size=xlarge is actually being granted
Verified signature33ea232 -
Request ZeroGPU xlarge (full 96GB Blackwell card) instead of default large (48GB half-card): the multiview attention activation memory OOMs on the half-card even with mmgp weight offloading
Verified signature8526e94 -
Try mmgp VerylowRAM_LowVRAM profile: LowRAM_LowVRAM still OOMs during attention
Verified signature432c498 -
Fix remaining comment typo
Verified signature3a0919b -
Fix comment typo (apostrophe escaping) from previous commit
Verified signature97e5396 -
Exclude image_normalizer from mmgp offload management (its .scale() method bypasses mmgps forward-hook, stranding buffers on CPU); pin it on GPU directly instead
Verified signaturea143759 -
Fix create_mean_pose() to build its numpy array from CPU tensors explicitly, instead of masking it with a global torch default-device reset that broke mmgp's own device handling
Verified signature7f3d882 -
Reset torch default device to cpu after offload.profile(): mmgp leaves it as cuda, which broke SMPLDataset tensor creation later in the script
Verified signaturefc017ad -
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
Verified signature6ce5fd0 -
Use MMGP budget-based offloading/quantization instead of a blanket .to('cuda'): the pipeline genuinely OOMs on this ZeroGPU MIG slice's effective VRAM
Verified signature9eacf98 -
Revert forced EFFICIENT_ATTENTION backend (unsupported for these tensor shapes on this GPU); back to auto-selected SDPA
Verified signature42c2660 -
Force SDPA's memory-efficient backend for PSHuman's multiview attention: rule out a math-fallback OOM on Blackwell
Verified signatureff5e65e
-
Disable expandable_segments CUDA allocator: it hits an NVML assertion on MIG-sliced Blackwell GPUs
Verified signaturebb80551 -
Keep PSHuman's custom multiview attention processors but swap their inner xformers.ops.memory_efficient_attention call for torch's native scaled_dot_product_attention (xformers' Hopper kernel crashes on Blackwell/sm_120)
Verified signature2080275 -
Drop xformers: its Hopper-specific flash-attention kernel crashes on Blackwell (sm_120) with 'CUDA error: invalid argument'. torch 2.8's native SDPA (used automatically by diffusers) replaces it.
Verified signature4725c84 -
Bump warp-lang to 1.17.0: kaolin 0.18.0 physics subsystem imports warp.fem.linalg.inverse_qr, missing from the old 1.4.2 pin
Verified signaturee4d389f -
Bump sympy to 1.13.3 and triton to 3.4.0: both hard-required by torch 2.8.0 (sympy>=1.13.3, triton==3.4.0 exactly on linux x86_64)
Verified signature2cc7752 -
Bump typing_extensions to 4.12.2: torch 2.8.0 requires >=4.10.0, conflicting with the old 4.9.0 pin
Verified signaturee7397e9 -
Upgrade to torch 2.8.0+cu128 (adds Blackwell/sm_12x kernels) with matching kaolin 0.18.0, pytorch3d 0.7.8+pt2.8.0cu128, torch_scatter 2.1.2+pt28cu128, xformers 0.0.32.post2
Verified signaturec2448e9 -
Add nvidia-cuda-runtime + LD_LIBRARY_PATH shim so nvdiffrast's JIT-compiled CUDA extension can find libcudart.so.13 on Blackwell/CUDA-13 ZeroGPU workers
Verified signaturef5c6af6 -
Fix torch pin to 2.1.2 (xformers==0.0.23.post1 requires torch==2.1.2 exactly, conflicting with the 2.1.0 I pinned previously)
Verified signatureb908182 -
Pin torch/torchvision/torchaudio to 2.1.0+cu121: unpinned torch was drifting to latest PyPI (now 2.14.0/cu13x), breaking ABI with the hard-pinned kaolin/pytorch3d/torch_scatter cu121 wheels (libcudart.so.13 not found)
Verified signaturef75e574 -
Pin setuptools<82: pkg_resources was fully removed in setuptools 82, breaking nvdiffrast/torch.utils.cpp_extension import
Verified signature606c14a -
Add setuptools so pkg_resources is available for nvdiffrast/torch.utils.cpp_extension import
Verified signaturef2c23c5
-
Update requirements.txt
Verified signature3e0afe2 -
Update requirements.txt
Verified signature8561d1a
-
increase ZeroGPU needed time for inference
Verified signatureafb1563 -
includes ZeroGPU deco for main function
Verified signatureeb0fbff -
import spaces for ZeroGPU
Verified signaturebd444a3
-
set xformers 0.0.23.post1
Verified signaturef79ea6f -
Update requirements.txt
Verified signature729a66a -
Update requirements.txt
Verified signature62b7383 -
Update requirements.txt
Verified signature51191bd -
downgradio 5.8.0
Verified signature67e2a93 -
Update requirements.txt
Verified signatureea3c3d6 -
back to working properly config, psguman currently not compatible with Zero
Verified signaturee1eb50a -
Update requirements.txt
Verified signature9f8cd50 -
Update requirements.txt
Verified signature934dbb6 -
Update requirements.txt
Verified signature0121406 -
upgrade xformers==0.0.29.post1
Verified signature55f4540 -
back to torch modules 2.1.2 v
Verified signature0dcb8cc -
look for the right putorch3d
Verified signature3a9651e