Oct 20 7 days ago

1. 09:33

   Avoid torchvision in the main process (compressai shim, no lightning/lpips) Running

   multimodalart HF Staff
   Verified signature

   4622bdb
2. 09:25

   Finer bisect

   multimodalart HF Staff
   Verified signature

   5c3db0d
3. 09:19

   Staged bisect

   multimodalart HF Staff
   Verified signature

   31b2855
4. 09:18

   Bisect switch

   multimodalart HF Staff
   Verified signature

   e05ff03
5. 09:13

   Library-load diagnostics

   multimodalart HF Staff
   Verified signature

   3986911
6. 09:10

   GPU probe

   multimodalart HF Staff
   Verified signature

   a473d77
7. 08:58

   Pin torch 2.11, drop libcudart preload, CUDA diagnostics

   multimodalart HF Staff
   Verified signature

   80e1690
8. 08:52

   Cap COLMAP threads, single-threaded OpenBLAS

   multimodalart HF Staff
   Verified signature

   7767094
9. 08:52

   CodecSplat demo

   multimodalart HF Staff
   Verified signature

   c9ad6b5
10. 08:33

    In-process COLMAP (thread-capped); CUDA-driver diagnostics

    multimodalart HF Staff
    Verified signature

    ef7ba95
11. 08:25

    Use prebuilt diff-gaussian-rasterization wheel (splatting_cuda decoder); drop gsplat JIT

    multimodalart HF Staff
    Verified signature

    7dfaaee
12. 08:22

    Keep pycolmap out of the main process; CUDA diagnostics

    multimodalart HF Staff
    Verified signature

    a5c56a7
13. 07:56

    Run COLMAP in an isolated subprocess

    multimodalart HF Staff
    Verified signature

    4400661
14. 07:51

    Single-threaded OpenBLAS for pycolmap

    multimodalart HF Staff
    Verified signature

    ab7fe1e
15. 07:43

    CodecSplat demo

    multimodalart HF Staff
    Verified signature

    7405e85