Ecosyste.ms: Issues

An open API service for providing issue and pull request metadata for open source projects.

GitHub / ogrisel/pygbm issues and pull requests

#99 - Parallel splitting fails in nopython mode

Issue - State: open - Opened by stephanegaiffas about 4 years ago

#98 - Recent Numba not usable with pygbm

Issue - State: open - Opened by esc over 4 years ago - 1 comment

#97 - Implement histogram recycling to improve memory efficiency

Issue - State: open - Opened by ogrisel over 4 years ago
Labels: enhancement, help wanted

#95 - Implement native support for missing values

Issue - State: open - Opened by ogrisel over 5 years ago
Labels: enhancement, help wanted

#94 - Status of this project?

Issue - State: closed - Opened by jondoering over 5 years ago - 2 comments

#93 - numba-integration-test failure

Issue - State: open - Opened by esc over 5 years ago - 6 comments

#92 - _update_raw_predictions() throws a deprecation warning

Issue - State: open - Opened by NicolasHug over 5 years ago - 1 comment

#91 - Optimize categorical crossentropy gradient update

Issue - State: open - Opened by NicolasHug about 6 years ago - 3 comments
Labels: perf

#90 - sum_gradient and sum_hessians computation in find_node_split_subtraction

Issue - State: open - Opened by NicolasHug about 6 years ago - 4 comments
Labels: perf

#89 - Remove constant_hessian_value?

Issue - State: open - Opened by NicolasHug about 6 years ago - 1 comment

#88 - [MRG] Reuse grower and splitter memory

Pull Request - State: closed - Opened by NicolasHug about 6 years ago - 8 comments

#87 - Failed attempt at ordered gradients/hessians

Pull Request - State: closed - Opened by maartenbreddels about 6 years ago

#86 - Add flake8 linting and code style check to travis

Pull Request - State: closed - Opened by ogrisel about 6 years ago - 5 comments

#85 - Single threaded split

Pull Request - State: open - Opened by maartenbreddels about 6 years ago - 1 comment

#84 - Avoid ordered_gradients?

Issue - State: open - Opened by maartenbreddels about 6 years ago - 7 comments

#83 - Optionally use left/right indices buffer

Issue - State: open - Opened by maartenbreddels about 6 years ago - 7 comments

#82 - Updating to Scipy 1.2.0 breaks loss tests...

Issue - State: open - Opened by NicolasHug about 6 years ago - 2 comments

#81 - Reuse grower (and thus the splitter) instead of creating a new one

Issue - State: open - Opened by NicolasHug about 6 years ago
Labels: perf

#80 - Remove empty slice check (numba fixed the issue)

Issue - State: open - Opened by NicolasHug about 6 years ago

#79 - More memory friendly

Pull Request - State: closed - Opened by maartenbreddels about 6 years ago - 3 comments

#78 - [MRG] Better LightGBM init parameters

Pull Request - State: closed - Opened by NicolasHug about 6 years ago - 1 comment

#77 - Optimize score loss computation

Issue - State: open - Opened by NicolasHug about 6 years ago
Labels: perf

#76 - Allow score monitoring regardless of early stopping

Issue - State: open - Opened by NicolasHug about 6 years ago
Labels: enhancement

#75 - [MRG] Early stopping now done with loss by default

Pull Request - State: closed - Opened by NicolasHug about 6 years ago - 2 comments

#74 - [MRG] Accept pre-binned data in fit

Pull Request - State: closed - Opened by NicolasHug about 6 years ago - 3 comments

#73 - make lightgbm import lazy

Pull Request - State: closed - Opened by amueller about 6 years ago - 2 comments

#72 - All the examples require lightgbm

Issue - State: closed - Opened by amueller about 6 years ago - 1 comment

#71 - Fixing doc

Pull Request - State: closed - Opened by NicolasHug about 6 years ago

#70 - API documentation is broken

Issue - State: closed - Opened by ogrisel about 6 years ago - 1 comment
Labels: bug

#69 - Use the training (or validation) loss value for early stopping by default

Issue - State: closed - Opened by ogrisel about 6 years ago - 2 comments

#68 - Passing pre-binned data to `fit()`

Issue - State: closed - Opened by NicolasHug about 6 years ago - 4 comments
Labels: perf

#67 - [MRG] More docstrings

Pull Request - State: closed - Opened by NicolasHug about 6 years ago - 2 comments

#66 - Use cache=True on everything?

Issue - State: open - Opened by NicolasHug about 6 years ago - 3 comments
Labels: perf

#65 - [MRG] Fix scoring

Pull Request - State: closed - Opened by NicolasHug about 6 years ago - 3 comments

#64 - [MRG] Initialize predictions to average value or class probabilities

Pull Request - State: closed - Opened by NicolasHug about 6 years ago - 2 comments

#63 - Usage of scorer is incorrect

Issue - State: closed - Opened by NicolasHug about 6 years ago - 1 comment

#62 - [MRG] Update bench_predictor.py and move _get_threads_chunks into utils.py

Pull Request - State: closed - Opened by NicolasHug about 6 years ago - 2 comments

#61 - [MRG] Added pygbm to lightgbm model conversion

Pull Request - State: closed - Opened by NicolasHug about 6 years ago - 2 comments

#60 - [MRG] Pass check_estimator form scikit-learn

Pull Request - State: closed - Opened by NicolasHug about 6 years ago - 2 comments

#59 - [MRG] Fixed early stopping mechanism

Pull Request - State: closed - Opened by NicolasHug about 6 years ago - 6 comments

#58 - [MRG] Renamed features_data and binned_features into X_binned

Pull Request - State: closed - Opened by NicolasHug about 6 years ago - 2 comments

#57 - [MRG] Added readthedocs.yml file

Pull Request - State: closed - Opened by NicolasHug about 6 years ago - 1 comment

#56 - Unify names of X_binned, features_data and binned_features

Issue - State: closed - Opened by NicolasHug about 6 years ago

#54 - [MRG] Documentation: added sphinx support and some first docstrings

Pull Request - State: closed - Opened by NicolasHug about 6 years ago - 6 comments

#53 - Rename y_pred to make it clear that those are the trees values

Issue - State: closed - Opened by NicolasHug about 6 years ago - 1 comment

#52 - [MRG] init parameter validation tests

Pull Request - State: closed - Opened by NicolasHug about 6 years ago - 3 comments

#51 - [MRG] Temporary fix for numba update issue

Pull Request - State: closed - Opened by NicolasHug about 6 years ago - 1 comment

#50 - Updating to numba 0.41 breaks everything

Issue - State: closed - Opened by NicolasHug about 6 years ago - 2 comments

#49 - [MRG] Add support for multiclass classification

Pull Request - State: closed - Opened by NicolasHug about 6 years ago - 7 comments

#48 - Parallelize predictions with static scheduling

Issue - State: closed - Opened by NicolasHug about 6 years ago - 2 comments

#47 - Initialize predictions to something better than just 0

Issue - State: closed - Opened by NicolasHug about 6 years ago

#46 - [MRG] min_samples_leaf now only accepts int values (not None)

Pull Request - State: closed - Opened by NicolasHug about 6 years ago - 2 comments

#45 - [MRG] Parametrized test_boston_dataset after bugfix in #43

Pull Request - State: closed - Opened by NicolasHug about 6 years ago - 2 comments

#44 - [MRG] Minimal optimization for root node and min_samples_leaf

Pull Request - State: closed - Opened by NicolasHug about 6 years ago - 1 comment

#43 - [MRG] Fixed computation of hessian/gradient left/right

Pull Request - State: closed - Opened by NicolasHug about 6 years ago - 2 comments

#42 - [MRG] Fix handling of number of bins

Pull Request - State: closed - Opened by NicolasHug about 6 years ago - 3 comments

#41 - small typo: leave -> leaf

Pull Request - State: closed - Opened by maartenbreddels over 6 years ago - 2 comments

#40 - [MGR] Strict comparison to min_gain_to_split

Pull Request - State: closed - Opened by NicolasHug over 6 years ago - 5 comments

#39 - Enhance binning strategy

Issue - State: open - Opened by NicolasHug over 6 years ago - 5 comments

#38 - Thread scalability is suboptimal

Issue - State: open - Opened by ogrisel over 6 years ago - 13 comments
Labels: perf

#37 - [MRG] Added n-trees and n-leaf-nodes as script parameters to bench_higgs_boson

Pull Request - State: closed - Opened by NicolasHug over 6 years ago - 2 comments

#36 - [WIP] Do not store histogram on SplitInfo

Pull Request - State: closed - Opened by ogrisel over 6 years ago - 10 comments

#35 - Fix handling of the min_samples_leaf hyperparameter

Pull Request - State: closed - Opened by ogrisel over 6 years ago - 15 comments

#34 - mean_samples_leaf does not do what it's suppose to do

Issue - State: closed - Opened by ogrisel over 6 years ago - 1 comment
Labels: bug

#33 - [MRG] Ordered_gradients now in SplittingContext

Pull Request - State: closed - Opened by NicolasHug over 6 years ago - 3 comments

#32 - Investigate the discrepancy in default hyperparams compared to LightGBM

Issue - State: open - Opened by ogrisel over 6 years ago - 4 comments

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

Issue - State: closed - Opened by ogrisel over 6 years ago - 20 comments
Labels: bug, perf

#30 - Benchmark results with better parameters

Issue - State: closed - Opened by Laurae2 over 6 years ago - 23 comments
Labels: perf

#29 - Fix dependency versions

Pull Request - State: closed - Opened by Laurae2 over 6 years ago - 2 comments

#28 - scikit-learn 0.20 is required

Issue - State: closed - Opened by Laurae2 over 6 years ago

#27 - Implement support for missing values

Issue - State: open - Opened by ogrisel over 6 years ago

#26 - Implement support for sparse feature data

Issue - State: open - Opened by ogrisel over 6 years ago

#25 - Implement support for categorical variables

Issue - State: open - Opened by ogrisel over 6 years ago

#24 - Implement ONNX import

Issue - State: open - Opened by ogrisel over 6 years ago

#23 - Implement ONNX export

Issue - State: open - Opened by ogrisel over 6 years ago - 1 comment

#22 - Allocate ordered_gradients and ordered_hessians only once

Issue - State: closed - Opened by NicolasHug over 6 years ago - 1 comment

#21 - [MRG] Avoid multiple array allocations by using a partition array

Pull Request - State: closed - Opened by NicolasHug over 6 years ago - 8 comments

#20 - Parallel update_y_pred

Pull Request - State: closed - Opened by NicolasHug over 6 years ago - 3 comments

#19 - [MRG] Parallelized split_indices()

Pull Request - State: closed - Opened by NicolasHug over 6 years ago - 5 comments

#18 - Added test to compare with LGBM predictions

Pull Request - State: closed - Opened by NicolasHug over 6 years ago - 7 comments

#17 - split_info comparison similar to LGBM

Pull Request - State: closed - Opened by NicolasHug over 6 years ago - 1 comment

#16 - Improvements in pygbm.splitting

Pull Request - State: closed - Opened by ogrisel over 6 years ago - 3 comments

#15 - Disable workaround for numba threading backend fix

Pull Request - State: closed - Opened by ogrisel over 6 years ago - 2 comments

#14 - [MRG] Histogram computation optimization

Pull Request - State: closed - Opened by NicolasHug over 6 years ago - 32 comments

#13 - [MRG] Add support for classification tasks

Pull Request - State: closed - Opened by NicolasHug over 6 years ago - 41 comments

#12 - Added plotting tool to compare trees with LightGBM

Pull Request - State: closed - Opened by NicolasHug over 6 years ago - 2 comments

#11 - [WIP] Gradient boosting

Pull Request - State: closed - Opened by ogrisel over 6 years ago - 4 comments

#10 - Bin mapper and predict from numerical features

Pull Request - State: closed - Opened by ogrisel over 6 years ago

#9 - Time compilation and remove function prototype declarations

Pull Request - State: closed - Opened by ogrisel over 6 years ago

#8 - Numba optimized single tree predictions on binned data

Pull Request - State: closed - Opened by ogrisel over 6 years ago

#7 - Test sample order independence in histograms

Pull Request - State: closed - Opened by ogrisel over 6 years ago

#6 - Special case the root split

Pull Request - State: closed - Opened by ogrisel over 6 years ago

#5 - Grower fixes

Pull Request - State: closed - Opened by ogrisel over 6 years ago

#4 - Basic grower prototype with HistogramSplitter

Pull Request - State: closed - Opened by ogrisel over 6 years ago

#3 - naive implementation of exact greedy gbt

Pull Request - State: closed - Opened by lancifollia over 6 years ago - 4 comments

#2 - [WIP] Compute loss gain on split

Pull Request - State: closed - Opened by ogrisel over 6 years ago

#1 - Use an array of structures instead of a structure of arrays for histograms

Pull Request - State: closed - Opened by ogrisel over 6 years ago - 1 comment