v0.2.3

0.2.3 (2026-08-17)

Bug Fixes

• align PyPI trusted publishing identity (1f5483b)
• align PyPI trusted publishing identity (29a7116)
• align PyPI trusted publishing identity (bf4127f)
• verify draft release assets by id (f231604)
• verify draft release assets by id (48c5d36)
• verify draft release assets by id (201b97e)

Documentation

• fix three-surface portability and accuracy (370b50c)
• fix three-surface portability and accuracy (04e6f15)
• promote documentation audit remediation (d26f4c1)

v0.2.2

0.2.2 (2026-08-09)

Features

• dataset: add regression targets to NNTabularDataset (4380b03)
• dataset: add regression-target support to NNTabularDataset via target_dtype (2ee37a6)
• dpo: emit implicit reward metrics (4e9c6d8)
• dpo: surface reward_accuracy in dpo_train_step_factory evaluation extras (d7b1bb7)
• generation: add on_token streaming callback to GenerativeNNModel.generate() (660b06b)
• generation: add per-token streaming callback (48bea6d)
• promote DPO implicit reward metrics (a8ca47b)
• promote generation streaming callback (1450ce5)
• promote optional run identity salt (6aea935)
• promote tabular regression support (a339107)
• run: add optional run identity salt (52989c9)
• run: add optional salt to NNRun identity for distinct identical-config runs (bc19cbf)

Bug Fixes

• dataset: build targets from validated numeric values (634da5b)
• enable publishing from reusable release calls (356d8ab)
• enable publishing from reusable release calls (9d6c256)
• harden training persistence and contracts (d308bfb)
• promote reusable release publication repair (8b082e8)
• run: validate salt runtime type (59d561c)
• synchronize release version in uv lock (a882137)
• synchronize release version in uv lock (96dfd9e)

Documentation

• center the shared product opener (e88322d)
• correct poster verification plan (2c3f595)
• design audit remediation (fe2a548)
• design NNx poster wordmark (8cee8cb)
• design opener remediation (51ad504)
• generation: describe callback contract precisely (7ebce00)
• integrate NNx wordmark into poster (5d79519)
• integrate NNx wordmark into poster (5c55fe0)
• plan NNx poster wordmark (c6c197b)
• plan opener remediation (8422466)
• promote the NNx opener remediation (6f2e288)
• promote the NNx poster wordmark (9e51199)
• rebuild the NNx project opener (096cd35)
• replace poster with brand banner (4e20ea4)
• resolve three-surface audit findings (dc203d5)
• resolve three-surface audit findings (a1ffa3b)
• resolve three-surface audit findings (8032cab)
• separate navigation and brand headings (9301944)
• synchronize project publication surfaces (28101b3)

v0.2.1

0.2.1 (2026-07-22)

Features

• Add ConvNN, NNConvParams, and Nets.CONV (#89).
• Add FeedFwdMoENN, NNMoEParams, and Nets.FEED_FWD_MOE (#88).
• Add full-batch graph node-classification loading without pyg-lib.
• Add per-layer activation and dropout configuration (#85).
• Add pluggable eval_step_fn validation hooks (#86).

Bug Fixes

• Persist callback-driven network mutations in the final checkpoint (#87).
• Reconstruct converted QAT checkpoints from saved transform metadata.
• Harden training lifecycle, parameter validation, and pandas 2 compatibility.
• Refresh the dependency lock to remove the vulnerable GitPython resolution.
• Complete specialized network exports and reject invalid ConvNN/MoE configurations.

Documentation

• Add and update runnable examples for the new model and training surfaces.
• Link the rendered documentation site and provide a themed architecture view.
• Synchronize architecture and publishing documentation.