Skip to content
GitLab
Projects Groups Topics Snippets
  • /
  • Help
    • Help
    • Support
    • Community forum
    • Submit feedback
    • Contribute to GitLab
  • Register
  • Sign in
  • Project Juggler Project Juggler
  • Project information
    • Project information
    • Activity
    • Labels
    • Members
  • Repository
    • Repository
    • Files
    • Commits
    • Branches
    • Tags
    • Contributor statistics
    • Graph
    • Compare revisions
  • Issues 33
    • Issues 33
    • List
    • Boards
    • Service Desk
    • Milestones
  • Merge requests 26
    • Merge requests 26
  • CI/CD
    • CI/CD
    • Pipelines
    • Jobs
    • Schedules
  • Deployments
    • Deployments
    • Environments
    • Releases
  • Packages and registries
    • Packages and registries
    • Container Registry
  • Monitor
    • Monitor
    • Metrics
    • Incidents
  • Analytics
    • Analytics
    • Value stream
    • CI/CD
    • Repository
  • Wiki
    • Wiki
  • Snippets
    • Snippets
  • Activity
  • Graph
  • Create a new issue
  • Jobs
  • Commits
  • Issue Boards
Collapse sidebar
  • EIC
  • Project JugglerProject Juggler
  • Merge requests
  • !370

Proof of concept algorithm that uses tensorflow-lite for ML on clusters

  • Review changes

  • Download
  • Patches
  • Plain diff
Open Wouter Deconinck requested to merge tflite-proof-of-concept into master Feb 08, 2022
  • Overview 3
  • Commits 1
  • Pipelines 0
  • Changes 3

This demonstrates how to include a tensorflow-lite model into a gaudi algorithm, compile, and link.

There are number of outstanding issues...

What is the best way to pass info from the options file to the algorithm? Should we include the expected input to the tensor as options? Right now we have an envisioned calling syntax somewhat like this:

cluster_identification = ClusterIdentification("cluster_identification",
        inputClusterCollection = "EcalBarrelImagingHitsML",
        outputParticles = "ReconstructedParticles",
        modelTFLiteFile = "calibrations/cluster_identification.tflite",
)

but we could add

        inputFeatures = ['energy', 'eta', 'phi']

which would use those features with the model. Ideally we would hardcode the features as metadata along with the tflite file to prevent incorrect use.

Assignee
Assign to
Reviewers
Request review from
Time tracking
Source branch: tflite-proof-of-concept