Hello!
We, and our O(500) users really profit from triton to scale AI inference at the CMS experiment on the Large Hadron Collider at CERN.
We noticed in the following PR that for python 3.14 we need to compile wheels locally for grpio in python 3.14, which may not be possible in all use cases with the current upper bound pin in tritonclient.
Would it be possible to remove the current upper bound (grpcio<1.68,>=1.63.0) in the next release of tritonclient/triton (or, similarly, make py3.14 wheels for the older grpcio version)?
c.f. scikit-hep/coffea#1540
Hello!
We, and our O(500) users really profit from triton to scale AI inference at the CMS experiment on the Large Hadron Collider at CERN.
We noticed in the following PR that for python 3.14 we need to compile wheels locally for grpio in python 3.14, which may not be possible in all use cases with the current upper bound pin in tritonclient.
Would it be possible to remove the current upper bound (
grpcio<1.68,>=1.63.0) in the next release of tritonclient/triton (or, similarly, make py3.14 wheels for the older grpcio version)?c.f. scikit-hep/coffea#1540