Expired
Milestone Mar 14, 2023–Apr 11, 2023

Adopt new tooling (phase 3)

  1. Analyse “basetorch” and common patterns inside “deepfind”, “deepdraw” and “ptbench”, see what can be moved into alternative, STABLE frameworks such as pytorch-lightning, or “weights-and-biases” wandb, or even scikit-learn. (1 week)
  2. Adapt code on the 3 packages to use new patterns from a framework builder such as pytorch-lightning first (1 week)
  3. Add support for pytorch-segmentation-models to “deepdraw”. (1 week)
  4. Adopt code on the 3 packages to use new patterns from wandb later, upon analysis and reflection (1 week)
  5. Points one may consider:
    1. Loops for training, prediction and evaluation (PyTorch-lightning)
    2. Logging (measures while training) (pytorch-lightning)
    3. Measures of performance evaluation (scikit-learn? Other frameworks?)
    4. Data loading and database description (?)
    5. Configuration support? (?)