v0.24.0

What's Changed

• v0.22.0 — graph-ML + neural-process substrate (P1–P4, S9–S18, R1/R2) by @thankrandomness in #98
• P5 — parallel non-text training loop + differentiable segment-aggregate by @thankrandomness in #99
• S19 PR0 — extract S18 distributional losses + unify the std-floor by @thankrandomness in #100
• S19 PR1 — AnyContextPredictor family (CNP / AttnCNP / TNP) by @thankrandomness in #101
• S19 — amortized in-context predictor (episodic training + serving + conformal) by @thankrandomness in #102
• release: v0.23.0 — amortized in-context predictor (S19) + training substrate (P5) by @thankrandomness in #103
• S16-G — declared-edge graph-neighbourhood context assembly by @thankrandomness in #104

Full Changelog: v0.21.0...v0.24.0