minerva-ml/steppy

Issues

• Step docstring conflicts with implementation

  #115 opened 6 years ago by thomasjpfan
  2

• consider merging steppy, steppy-toolkit. snd steppy-examples into one project. documentation will be rundant otherwise. will

  #109 opened 6 years ago by bcottman
  0

• Does stepper have a CONTRIBUTING?

  #93 opened 7 years ago by bcottman
  13

• FAQ DOC

  #95 opened 7 years ago by bcottman
  1

• Structure of steps - ideas for making it cleaner

  #87 opened 7 years ago by stared
  2

• difficult debugging

  #84 opened 7 years ago by taraspiotr
  1

• How to evaluate each step only once?

  #72 opened 7 years ago by denyslazarenko
  4

• tests and informative assertions

  #22 opened 7 years ago by kamil-kaczmarek
  1

• Let's pass `experiment_directory` only once per pipeline

  #85 opened 7 years ago by apyskir
  1

• have to clean cache manually

  #39 opened 7 years ago by mromaniukcdl
  2

• Maybe load_saved_input?

  #77 opened 7 years ago by apyskir
  0

• setup Travis CI

  #80 opened 7 years ago by kamil-kaczmarek
  0

• add is_trainable flag

  #79 opened 7 years ago by jakubczakon
  1

• steppy logger prints wrong data

  #78 opened 7 years ago by kamil-kaczmarek
  1

• Do we really need a transformer to be cached to perform transform()?

  #73 opened 7 years ago by apyskir
  1

• cache outputs in memory

  #46 opened 7 years ago by taraspiotr
  1

• Concat features

  #71 opened 7 years ago by denyslazarenko
  8

• Difference between cache and persist

  #88 opened 7 years ago by denyslazarenko
  2

• Should Kaggle solutions become step-examples?

  #96 opened 7 years ago by bcottman
  1

• API design

  #21 opened 7 years ago by kamil-kaczmarek
  0

• documentation

  #23 opened 7 years ago by kamil-kaczmarek
  0

• Not fitting in 0.1.6

  #89 opened 7 years ago by kstrzala
  1

• Customizable logging level

  #94 opened 7 years ago by mtad
  2

• logger's path and filename

  #90 opened 7 years ago by fengyonglv
  2

• Logging

  #92 opened 7 years ago by bcottman
  2

• Let's make it possible to transform based on checkpoints

  #91 opened 7 years ago by apyskir
  0

• Do all Steps execute parallel?

  #82 opened 7 years ago by denyslazarenko
  2

• improve docstrings in the adapter.py

  #70 opened 7 years ago by kamil-kaczmarek
  1

• tutorial: notebook 5 (Keras)

  #29 opened 7 years ago by kamil-kaczmarek
  1

• consider adding abstract image and abstract batch

  #34 opened 7 years ago by kamil-kaczmarek
  0

• type annotations in Step class

  #35 opened 7 years ago by kamil-kaczmarek
  1

• API for ensembling

  #37 opened 7 years ago by kamil-kaczmarek
  1

• For deep learning pipelines -> creating batches not in loader

  #38 opened 7 years ago by kamil-kaczmarek
  1

• cpu and gpu utilization in the logs / callback / neptune graph

  #43 opened 7 years ago by kamil-kaczmarek
  0

• remove graphviz from the project

  #45 opened 7 years ago by kamil-kaczmarek
  1

• check shapes of inputs/outputs

  #44 opened 7 years ago by kamil-kaczmarek
  0

• benchmark tools for building pipelines and reproducible reseach

  #51 opened 7 years ago by kamil-kaczmarek
  1

• refactor adapter.py

  #63 opened 7 years ago by kamil-kaczmarek
  2

• setup readthedocs

  #67 opened 7 years ago by kamil-kaczmarek
  1

• refactor all examples to work with v0.1.2

  #69 opened 7 years ago by kamil-kaczmarek
  1

• One `fit_transform` call on the pipeline may result in multiple `fit_transform` calls on some steps

  #24 opened 7 years ago by grzes314
  0

• UNet error

  #27 opened 7 years ago by apyskir
  1

• Add checking that the adapter definition matches the list of input steps and input data

  #31 opened 7 years ago by mromaniukcdl
  1

• prepare v0.1.3 with readme.md and index.rst

  #74 opened 7 years ago by kamil-kaczmarek
  0

• correct names connected with cache/save

  #36 opened 7 years ago by kamil-kaczmarek
  1

• make steppy pip installable

  #53 opened 7 years ago by kamil-kaczmarek
  1

• Unintuitive adapter syntax

  #28 opened 7 years ago by grzes314
  2

• PyTorch model is never saved as checkpoint after first epoch

  #48 opened 7 years ago by apyskir
  2

• Validation loss is calculated by a few callbacks and sometimes it takes a lot of time

  #49 opened 7 years ago by apyskir
  1

• Change time format in CLI output

  #30 opened 7 years ago by mromaniukcdl
  1