Merge pull request #922 from Kaggle/upgrade-lightgbm

Upgrade LightGBM to v3.1.1

Author: rosbo

Dockerfile

@@ -79,7 +79,7 @@ RUN apt-get install -y libfreetype6-dev && \
     pip install wordcloud && \
     pip install xgboost && \
     # Pinned to match GPU version. Update version together.
-    pip install lightgbm==2.3.1 && \
+    pip install lightgbm==3.1.1 && \
     pip install keras && \
     pip install keras-tuner && \
     pip install flake8 && \

gpu.Dockerfile

@@ -69,7 +69,7 @@ RUN pip uninstall -y lightgbm && \
     cd /usr/local/src && \
     git clone --recursive https://github.com/microsoft/LightGBM && \
     cd LightGBM && \
-    git checkout tags/v2.3.1 && \
+    git checkout tags/v3.1.1 && \
     mkdir build && cd build && \
     cmake -DUSE_GPU=1 -DOpenCL_LIBRARY=/usr/local/cuda/lib64/libOpenCL.so -DOpenCL_INCLUDE_DIR=/usr/local/cuda/include/ .. && \
     make -j$(nproc) && \

tests/data/lgb_test.bin

@@ -1,15 +1,15 @@
 import unittest
 
 import lightgbm as lgb
+import pandas as pd
 
 from common import gpu_test
 
 class TestLightgbm(unittest.TestCase):
     # Based on the "simple_example" from their documentation:
     # https://github.com/Microsoft/LightGBM/blob/master/examples/python-guide/simple_example.py
     def test_cpu(self):
-        lgb_train = lgb.Dataset('/input/tests/data/lgb_train.bin')
-        lgb_eval = lgb.Dataset('/input/tests/data/lgb_test.bin', reference=lgb_train)
+        lgb_train, lgb_eval = self.load_datasets()
 
         params = {
             'task': 'train',
@@ -35,9 +35,8 @@ def test_cpu(self):
 
     @gpu_test
     def test_gpu(self):
-        lgb_train = lgb.Dataset('/input/tests/data/lgb_train.bin')
-        lgb_eval = lgb.Dataset('/input/tests/data/lgb_test.bin', reference=lgb_train)
-
+        lgb_train, lgb_eval = self.load_datasets()
+        
         params = {
             'boosting_type': 'gbdt',
             'objective': 'regression',
@@ -59,3 +58,17 @@ def test_gpu(self):
                         early_stopping_rounds=1)
 
         self.assertEqual(1, gbm.best_iteration)
+    
+    def load_datasets(self):
+        df_train = pd.read_csv('/input/tests/data/lgb_train.csv', header=None, sep='\t')
+        df_test = pd.read_csv('/input/tests/data/lgb_test.csv', header=None, sep='\t')
+        
+        y_train = df_train[0]
+        y_test = df_test[0]
+        X_train = df_train.drop(0, axis=1)
+        X_test = df_test.drop(0, axis=1)
+
+        lgb_train = lgb.Dataset(X_train, y_train)
+        lgb_eval = lgb.Dataset(X_test, y_test, reference=lgb_train)
+
+        return (lgb_train, lgb_eval)