v0.0.14

cebeuq/Synthid-Bypassv0.0.14May 11, 2026by github-actions[bot]

AI Summary

Focuses on the smart-turn pipeline, adding publishing capabilities for fine-tunes and fixing quantization issues.

Key Highlights

  • Added `wk-st publish` command to stage and push fine-tunes to HuggingFace.
  • Fixed int8 quantization to ensure the ONNX model actually shrinks in size.

New Features

  • wk-st publish — stage + push fine-tunes to HuggingFace

Full Release Notes

## [0.0.14](https://github.com/wavekat/wavekat-lab/compare/v0.0.13...v0.0.14) (2026-05-11)


### Features

* **smart-turn:** wk-st publish — stage + push fine-tunes to HuggingFace ([a9a3a44](https://github.com/wavekat/wavekat-lab/commit/a9a3a44f4d039371715adf266bb5f3c42972bb3a))
* **smart-turn:** wk-st publish — stage + push fine-tunes to HuggingFace ([#42](https://github.com/wavekat/wavekat-lab/issues/42)) ([2d4ce7e](https://github.com/wavekat/wavekat-lab/commit/2d4ce7e78605e533f13f5c9f60a84de199810436))


### Bug Fixes

* **smart-turn:** enable constant folding so int8 quant actually shrinks the ONNX ([#43](https://github.com/wavekat/wavekat-lab/issues/43)) ([dc0ba56](https://github.com/wavekat/wavekat-lab/commit/dc0ba563821a09d2633b2384cc569a2aa2adb752))