Skip to content

History / Networks

Revisions

  • Pages for the libtorch engine and for importing a model, and every page brought to 0.5.0 TorchSharp backend and Importing a model are new, linked from the sidebar and Home. Home, Getting Started, Pipeline, Networks, Notebook, Architecture, Quality, Security, Roadmap and PDD now describe 0.5.0: the Parquet, Excel and JSON readers, what a learner needs and a run for it, checkpoints and the engine they name, served rows nothing in training resembled, the notebook's install from the Extensions panel, its guard on saving and opening and the hosts that ask it, an encoded category unticked in the grid, the pickle reader's bounds, and rows 7 and 8 done. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

    @xkqg xkqg committed Oct 1, 2026
  • A Networks page covers the layers, the loop, the two ways to describe a network, its one file and the charts, and the roadmap marks rows 2 to 6 done - Networks.md is new: one network from the Titanic rows to its file, Keras's words and a network written as code, the layers, the losses, optimizers and schedules, the loop with early stopping and checkpoints, a network behind a pipeline and what it predicts, the report and the charts, the file and its faults, and what is not here. - Roadmap: rows 2 to 6 are done. Home, Getting-Started, PDD, Pipeline, Notebook, Quality, Security and the sidebar say what 0.4.0 holds. Architecture carries the operations the layers need and what one machine measured for a training step. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

    @xkqg xkqg committed Sep 29, 2026