Skip to content

chore(master): release 0.230.0 #1814

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
ooples wants to merge 1 commit into
base: master
Choose a base branch
from
Open

chore(master): release 0.230.0 #1814

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion .release-please-manifest.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,3 +1,3 @@
{
".": "0.229.2"
".": "0.230.0"
}
43 changes: 42 additions & 1 deletion CHANGELOG.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,6 +19,48 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0

---

## [0.230.0](https://github.com/ooples/AiDotNet/compare/v0.229.2...v0.230.0) (2026-07-11)


### Features

* **checkpoint:** typed model-state restore via ICheckpointableModel sidecar ([#1811](https://github.com/ooples/AiDotNet/issues/1811)) ([e5395d8](https://github.com/ooples/AiDotNet/commit/e5395d8d4298108de3fff9054ddb21159c2addbc))
* **facade:** pluggable credit-assignment rules (Feedback Alignment / DFA / Sign-Symmetric) ([#1805](https://github.com/ooples/AiDotNet/issues/1805)) ([e5f25be](https://github.com/ooples/AiDotNet/commit/e5f25be7304c896b90583bc3cfd5ac14171783ed))
* **licensing:** asymmetric public-key signatures (aidn2) — replace extractable symmetric HMAC ([#1808](https://github.com/ooples/AiDotNet/issues/1808)) ([794c717](https://github.com/ooples/AiDotNet/commit/794c717ad52c53f390f6ad5a79f35a99179d53aa))
* **metrics:** language-model perplexity + top-k accuracy metrics ([#1791](https://github.com/ooples/AiDotNet/issues/1791)) ([b09e399](https://github.com/ooples/AiDotNet/commit/b09e399eec3119fddefb18b27d0617d5294f2070))
* **training:** GPU-resident fused step for non-TS single-net models ([#1843](https://github.com/ooples/AiDotNet/issues/1843)) ([a5e69ca](https://github.com/ooples/AiDotNet/commit/a5e69cafb03318459389df42eb22267aa8a9a62c))
* **transformer:** opt-in numerically-stable log-softmax-cross-entropy head (default OFF) ([#1828](https://github.com/ooples/AiDotNet/issues/1828)) ([ae6f645](https://github.com/ooples/AiDotNet/commit/ae6f645632acbae5d8e4be572563b17b16f36c78))


### Bug Fixes

* **ci:** green Diffusion ModelFamily shards — fix DeepFloydIF shape + defer verified foundation-scale OOM models ([#1706](https://github.com/ooples/AiDotNet/issues/1706)) ([#1758](https://github.com/ooples/AiDotNet/issues/1758)) ([3522f76](https://github.com/ooples/AiDotNet/commit/3522f76ff4c3fbe62ba54030da8d6efb2159a3dc))
* consolidated AiDotNet fixes + excellence goals + audit pass ([#1832](https://github.com/ooples/AiDotNet/issues/1832), [#1833](https://github.com/ooples/AiDotNet/issues/1833), [#1834](https://github.com/ooples/AiDotNet/issues/1834), [#1835](https://github.com/ooples/AiDotNet/issues/1835), [#1836](https://github.com/ooples/AiDotNet/issues/1836), [#1837](https://github.com/ooples/AiDotNet/issues/1837)) ([#1838](https://github.com/ooples/AiDotNet/issues/1838)) ([1ca524d](https://github.com/ooples/AiDotNet/commit/1ca524d4e224efba9c7b7585818c20d851d7f039))
* **determinism:** seed minibatch shuffle under SetDeterministicMode (real cause of run-to-run training nondeterminism) ([#1819](https://github.com/ooples/AiDotNet/issues/1819)) ([a43234e](https://github.com/ooples/AiDotNet/commit/a43234eb4ccdbadcd611da72d1b3f1ed4ef2e8c3))
* **diffusion:** predictNoiseBatched must not drop the batch dim ([#1843](https://github.com/ooples/AiDotNet/issues/1843) regression) ([#1850](https://github.com/ooples/AiDotNet/issues/1850)) ([d83f043](https://github.com/ooples/AiDotNet/commit/d83f0433340b52f41ff69d2012ac57b3bd623a5a))
* **diffusion:** preserve fp16-resident weights across clone/param round-trip ([#1764](https://github.com/ooples/AiDotNet/issues/1764)) ([#1788](https://github.com/ooples/AiDotNet/issues/1788)) ([c6f0aee](https://github.com/ooples/AiDotNet/commit/c6f0aee5af354589bbb2af08e629b13d0b5f623f))
* **facade:** revert unneeded transformer routing; assert REAL learning ([#1803](https://github.com/ooples/AiDotNet/issues/1803)) ([35d55f1](https://github.com/ooples/AiDotNet/commit/35d55f114a29791891aa1b2281356ef36ab9c721))
* **facade:** unblock BuildAsync for radiance-field models ([#1826](https://github.com/ooples/AiDotNet/issues/1826)) ([#1829](https://github.com/ooples/AiDotNet/issues/1829)) ([8909159](https://github.com/ooples/AiDotNet/commit/8909159c60b24239104c64ecdd300652d28aa071))
* **finance:** tFT/Informer train through the genuine tape forward (ForwardNativeForTraining) ([#1849](https://github.com/ooples/AiDotNet/issues/1849)) ([47d493e](https://github.com/ooples/AiDotNet/commit/47d493e57d1c7d40ef95ac6f2691670e2aaa89cc))
* **generators:** scope AIDN001 model-metadata validation to the AiDotNet library ([#1825](https://github.com/ooples/AiDotNet/issues/1825)) ([8c6256b](https://github.com/ooples/AiDotNet/commit/8c6256b16a4cd7495fc3fce6264a445350b047d8))
* **gpu:** invalidate resident weight buffers after in-place optimizer update (GPU transformer training was stale) ([#1817](https://github.com/ooples/AiDotNet/issues/1817)) ([5c19829](https://github.com/ooples/AiDotNet/commit/5c1982908ac5f0507fe9b40195c305d4a75d7531))
* **inference:** memory-bounded transformer forward — arena recycles per-layer scratch ([#1824](https://github.com/ooples/AiDotNet/issues/1824)) ([#1824](https://github.com/ooples/AiDotNet/issues/1824)) ([413628d](https://github.com/ooples/AiDotNet/commit/413628d19b366d856892072f1353ce194252769f))
* **license:** only classify aidn. keys as offline-HMAC when the signature is 32 bytes ([#1807](https://github.com/ooples/AiDotNet/issues/1807)) ([9e71dea](https://github.com/ooples/AiDotNet/commit/9e71dea6a23a504d2f5a9b14c0078188bda76a56))
* **licensing:** require a prior successful online validation before honouring ValidationPending ([#1802](https://github.com/ooples/AiDotNet/issues/1802)) ([d1afcd0](https://github.com/ooples/AiDotNet/commit/d1afcd0769622e5356f9ee7e8de8797f6f4d2b0c))
* **optimizer:** bound eval cache + add O(tokens) mini-batch-loss fitness mode ([#1820](https://github.com/ooples/AiDotNet/issues/1820)) ([e1e2394](https://github.com/ooples/AiDotNet/commit/e1e239411e29a6855d2a0fa17df316d670b64a7a))
* **optimizer:** bump Tensor.Version after in-place tape Step so GPU re-uploads updated weights ([#1810](https://github.com/ooples/AiDotNet/issues/1810)) ([ca3a1e8](https://github.com/ooples/AiDotNet/commit/ca3a1e8e6057624e804592af7afdcd1cdc67e1d6))
* **optimizers:** bound DefaultGradientCache to stop unbounded training-loop memory leak ([#1831](https://github.com/ooples/AiDotNet/issues/1831)) ([58b6834](https://github.com/ooples/AiDotNet/commit/58b68349ca20a724a2f18ea01b276e4ca8d4bdc0))
* **training:** Transformer.Train() silent no-op — fused compiled step didn't persist to live params ([#1822](https://github.com/ooples/AiDotNet/issues/1822)) ([#1823](https://github.com/ooples/AiDotNet/issues/1823)) ([803330c](https://github.com/ooples/AiDotNet/commit/803330ca95ae25a3a844cf12f1e3ff81a97bb8d7))
* **transformer:** label smoothing (paper eps=0.1) un-freezes batched training ([#1559](https://github.com/ooples/AiDotNet/issues/1559)) ([#1818](https://github.com/ooples/AiDotNet/issues/1818)) ([f135e7b](https://github.com/ooples/AiDotNet/commit/f135e7b3ae384d7ebf3b9a8497a141a571b22dd8))
Comment thread
ooples marked this conversation as resolved.


### Performance

* activate TensorArena for training across all model paths (zero-alloc) ([#1809](https://github.com/ooples/AiDotNet/issues/1809)) ([a39fb1b](https://github.com/ooples/AiDotNet/commit/a39fb1bae59b737904696e94b03ff2452c06c4e6))
* cache Adam per-param backing arrays across steps (+ NBEATS bias-col reuse) ([#1816](https://github.com/ooples/AiDotNet/issues/1816)) ([29d2ad4](https://github.com/ooples/AiDotNet/commit/29d2ad468a6c30bb49764a48f767f00b6d4148b7))
* **optimizer:** parallelize eager Adam Step across parameters ([#1806](https://github.com/ooples/AiDotNet/issues/1806)) ([0b3adac](https://github.com/ooples/AiDotNet/commit/0b3adac800f0804534519eec16be7f08d0deb213))
* **optimizer:** wire eager fp32 Adam step onto the shared SIMD kernel ([#1815](https://github.com/ooples/AiDotNet/issues/1815)) ([1a5125c](https://github.com/ooples/AiDotNet/commit/1a5125ca9e73f9a57e96ac9777efa5e51ff67235))

## [v0.207.0] - 2026-05-21

_Release v0.207.0_
Expand Down Expand Up @@ -1272,4 +1314,3 @@ See https://github.com/ooples/AiDotNet/releases/tag/v0.0.3-preview
## [v0.0.1-preview] - 2023-09-23 (pre-release)

See https://github.com/ooples/AiDotNet/releases/tag/v0.0.1-preview