ogrisel/pygbm

Issues

• Parallel splitting fails in nopython mode

  #99 opened 4 years ago by stephanegaiffas
  0

• Recent Numba not usable with pygbm

  #98 opened 4 years ago by esc
  1

• did you stopped development since since can not do better than lightGBM pr Xgboost pr catboost?

  #96 opened 4 years ago by Sandy4321
  4

• Implement histogram recycling to improve memory efficiency

  #97 opened 4 years ago by ogrisel
  0

• numba-integration-test failure

  #93 opened 5 years ago by esc
  6

• Implement native support for missing values

  #95 opened 5 years ago by ogrisel
  0

• Status of this project?

  #94 opened 5 years ago by jondoering
  2

• _update_raw_predictions() throws a deprecation warning

  #92 opened 5 years ago by NicolasHug
  1

• Optimize categorical crossentropy gradient update

  #91 opened 6 years ago by NicolasHug
  3

• sum_gradient and sum_hessians computation in find_node_split_subtraction

  #90 opened 6 years ago by NicolasHug
  4

• Remove constant_hessian_value?

  #89 opened 6 years ago by NicolasHug
  1

• Thread scalability is suboptimal

  #38 opened 6 years ago by ogrisel
  13

• Updating to Scipy 1.2.0 breaks loss tests...

  #82 opened 6 years ago by NicolasHug
  2

• Avoid ordered_gradients?

  #84 opened 6 years ago by maartenbreddels
  7

• Passing pre-binned data to `fit()`

  #68 opened 6 years ago by NicolasHug
  4

• Optionally use left/right indices buffer

  #83 opened 6 years ago by maartenbreddels
  7

• Reuse grower (and thus the splitter) instead of creating a new one

  #81 opened 6 years ago by NicolasHug
  0

• Remove empty slice check (numba fixed the issue)

  #80 opened 6 years ago by NicolasHug
  0

• Optimize score loss computation

  #77 opened 6 years ago by NicolasHug
  0

• Use the training (or validation) loss value for early stopping by default

  #69 opened 6 years ago by ogrisel
  2

• Allow score monitoring regardless of early stopping

  #76 opened 6 years ago by NicolasHug
  0

• Use cache=True on everything?

  #66 opened 6 years ago by NicolasHug
  3

• All the examples require lightgbm

  #72 opened 6 years ago by amueller
  1

• API documentation is broken

  #70 opened 6 years ago by ogrisel
  1

• Usage of scorer is incorrect

  #63 opened 6 years ago by NicolasHug
  1

• Initialize predictions to something better than just 0

  #47 opened 6 years ago by NicolasHug
  0

• Parallelize predictions with static scheduling

  #48 opened 6 years ago by NicolasHug
  2

• Write tests for the held out validation data and early stopping

  #55 opened 6 years ago by NicolasHug
  0

• Rename y_pred to make it clear that those are the trees values

  #53 opened 6 years ago by NicolasHug
  1

• Unify names of X_binned, features_data and binned_features

  #56 opened 6 years ago by NicolasHug
  0

• Updating to numba 0.41 breaks everything

  #50 opened 6 years ago by NicolasHug
  2

• Enhance binning strategy

  #39 opened 6 years ago by NicolasHug
  5

• Benchmark results with better parameters

  #30 opened 6 years ago by Laurae2
  23

• Excessive memory usage when the number of trees is large (e.g. 500 trees)

  #31 opened 6 years ago by ogrisel
  20

• mean_samples_leaf does not do what it's suppose to do

  #34 opened 6 years ago by ogrisel
  1

• Allocate ordered_gradients and ordered_hessians only once

  #22 opened 6 years ago by NicolasHug
  1

• Investigate the discrepancy in default hyperparams compared to LightGBM

  #32 opened 6 years ago by ogrisel
  4

• scikit-learn 0.20 is required

  #28 opened 6 years ago by Laurae2
  0

• Implement ONNX export

  #23 opened 6 years ago by ogrisel
  1

• Implement support for missing values

  #27 opened 6 years ago by ogrisel
  0

• Implement support for sparse feature data

  #26 opened 6 years ago by ogrisel
  0

• Implement support for categorical variables

  #25 opened 6 years ago by ogrisel
  0

• Implement ONNX import

  #24 opened 6 years ago by ogrisel
  0