blangSDK
blangSDK copied to clipboard
Blang's software development kit
- one idea is to extend behaviour instead of overwriting, e.g. use both ``@Input`` for direct input plus some optional command line args for file (and same strategy for matrices,...
Latest failure example: ``` --model.plate.maxSize 100 \ --model.diagonal false \ --engine PT \ --engine.ladder.nChains 1 \ --engine.nThreads 1 \ --engine.random 1 \ --engine.nPassesPerScan 1 \ --samplers.useAnnotation false \ --samplers.additional blang.mcmc.BouncyParticleSampler...
E.g. ``` package assignments model Growth { param IntVar current ?: latentInt random IntVar next ?: latentInt laws { logf(current, next) { if (next == current - 1) return log(1.0/5.0)...
Provide an injection mechanism for detecting partial observability under a node. Detect if a sampler requested it. If not, and there is partial observability, throw an error.
E.g. take Dirichlet before commit 2509312366a2a99514c6a47092d3d0c9a9a35556, it was fine if concentration param is via a ConstantSupplier (I think!), but not if e.g. it's via ``Dirichlet(#[aVar, anotherVar])``. Basic solution would be...