Uploaded image for project: 'AI Platform Core Components'
  1. AI Platform Core Components
  2. AIPCC-7810

Build TensorFlow from source in RHOAI 3.4 EA-1

    • False
    • Hide

      None

      Show
      None
    • True
    • 67% To Do, 33% In Progress, 0% Done
    • Hide

      13/Jan/2026 Yellow

      While a solution exists, we are raising it during the AIPCC Tech call on 14/Jan for calibration and discussion

      Show
      13/Jan/2026 Yellow While a solution exists, we are raising it during the AIPCC Tech call on 14/Jan for calibration and discussion

      ADR: https://docs.google.com/document/d/1YeQHna4dj68GXlx8asPRi_jL05LtT6lmrF6s9moUjLU/edit?usp=sharing  

       

      Feature title:  Build TensorFlow from source

      Feature Overview: TensorFlow is a platform for machine learning initiated and largely maintained by Google. We are currently shipping it as pre built wheels from PyPI, but we'd like to build it from source like other packages. This effort is large as it involves different problems to solve. This initiative will track that effort.

      Product(s) associated:

      RHAIIS: No
      RHEL AI: No
      RHOAI: Yes (notebooks)

      Goals:

      Build accelerator optimized wheels for TensorFlow from source using AIPCC tooling and ecosystem.

      Requirements:

      Builds for CPU, CUDA and ROCm are needed.

      Done - Acceptance Criteria:

      Builds are published in our package index for these variants, and probe tested.

              spryor@redhat.com Sean Pryor
              emacchi@redhat.com Emilien Macchi
              Frank's Team
              Votes:
              0 Vote for this issue
              Watchers:
              6 Start watching this issue

                Created:
                Updated: