BCINexus · Publishing

Publishing models

A published model is weights plus the graph that produced them. Weights alone are a number soup nobody can evaluate, so the registry does not accept them on their own.

The two file formats

FormatContainsUsed for
.bltwTrained weights onlyLocal reuse inside BCILattice — reload a checkpoint into the graph you already have.
.bltmGraph definition plus weightsPublishing. Self-contained, so someone else can load it and reproduce the architecture without your project.

Export targets

Raw PyTorch checkpoints, ONNX export, and edge deployment targets are paid-plan features. Training locally and reusing weights inside the app is not gated.

What to include

  • The .bltm bundle, so the architecture travels with the weights.
  • The dataset the model was trained on, linked rather than described in prose.
  • Headline metrics with the split policy that produced them — a number without a split is not comparable to anything.
  • The reproducibility stamp from training: seed, library versions, artifact hashes.
  • A licence, so reuse terms are unambiguous.

Models published as part of a study inherit that study's review outcome. A model published without any reproducibility manifest triggers a readiness warning, which reviewers will usually ask about before approving.

Reusing someone else's model

  • Browse the model registry and filter by modality, task, and architecture.
  • Load a published model in BCILattice and evaluate it against your own recordings before trusting the headline number.
  • Report what you find as a reproduction attempt — successful or not. Failed reproductions are as useful to the record as successful ones.
  • Submit results to a benchmark to compare on a shared task rather than on differently-cut data.

Versions

Model versions are snapshots with their own metrics and artifacts. Retraining produces a new version rather than overwriting the old one, so a citation and a benchmark entry keep referring to the weights that actually produced the reported result.

Related

Model catalog

The BCI model classes bundled with the desktop app.

Using the ML Suite

Building and training the graph in the first place.

Community & benchmarks

Leaderboards, reproductions, and comparison.

Publish readiness

The reproducibility checks that apply to trained models.

Something missing or out of date? Email support or request a doc page. Include the page name and what you expected to find.