a

Author: calderjo

Dockerfile.tmpl

@@ -6,15 +6,12 @@ FROM ${BASE_IMAGE}:${BASE_IMAGE_TAG}
 ADD kaggle_requirements.txt /kaggle_requirements.txt
 
 # Freeze existing requirements from base image for critical packages:
-RUN pip freeze | grep -E 'tensorflow|keras|torch|jax|numpy' > /colab_requirements.txt
+RUN pip freeze | grep -E 'tensorflow|keras|torch|jax' > /colab_requirements.txt
 
 # Merge requirements files:
 RUN cat /colab_requirements.txt >> /requirements.txt
 RUN cat /kaggle_requirements.txt >> /requirements.txt
 
-# TODO: GPU requirements.txt
-# TODO: merge them better (override matching ones).
-
 # Install Kaggle packages
 RUN uv pip install --system -r /requirements.txt
 
@@ -36,8 +33,10 @@ RUN uv pip install --system "tbb>=2022" "libpysal==4.9.2"
 # b/415358158: Gensim removed from Colab image to upgrade scipy
 # b/456239669: remove huggingface-hub pin when pytorch-lighting and transformer are compatible
 # b/315753846: Unpin translate package, currently conflicts with adk 1.17.0
-RUN uv pip install --system --force-reinstall --no-deps torchtune gensim "scipy<=1.15.3" "huggingface-hub==0.36.0" "google-cloud-translate==3.12.1" "numpy==2.0.2"
-
+# b/(xxxxx): Unpin Pandas once cuml/cudf are compatible, version 3.0 causes issues
+# b/(xxxxx): Unpin protobuf, v > 5.29.5 causes issues with numerous packages
+# b/(xxxxx): numpy will auto-upgrade to 2.4.x, which causes issues with numerous packages
+RUN uv pip install --system --force-reinstall --no-deps torchtune gensim "scipy<=1.15.3" "huggingface-hub==0.36.0" "google-cloud-translate==3.12.1" "numpy==2.0.2" "pandas==2.2.2"
 RUN uv pip install --system --force-reinstall "protobuf==5.29.5"
 
 # Adding non-package dependencies:

kaggle_requirements.txt

@@ -9,8 +9,7 @@ Pympler
 Rtree
 shapely
 SimpleITK
-# b/302136621: Fix eli5 import for learntools, newer version require scikit-learn > 1.3
-TPOT==0.12.1
+TPOT
 Wand
 annoy
 arrow
@@ -30,9 +29,6 @@ docker
 easyocr
 emoji
 fastcore
-# b/445960030: Requires a newer version of fastai than the currently used base image.
-# Remove when relying on a newer base image.
-fastai>=2.8.4
 fasttext
 featuretools
 fiona
@@ -126,6 +122,8 @@ scikit-surprise
 seaborn
 git+https://github.com/facebookresearch/segment-anything.git
 shap
+# b/(xxxxx): unpin, currently needed as dependency for tsfresh
+statsmodels==0.14.6
 squarify
 tensorflow-cloud
 tensorflow-io
@@ -136,11 +134,5 @@ torchtune
 transformers>=4.51.0
 vtk
 wavio
-# b/350573866: xgboost v2.1.0 breaks learntools
-xgboost==2.0.3
 xvfbwrapper
 ydata-profiling
-ydf
-protobuf==5.29.5
-numpy==2.0.2
-statsmodels==0.14.6

tests/test_xgboost.py

@@ -17,10 +17,9 @@ def test_classifier(self):
         X_test = np.random.random((100, 28))
         y_test = np.random.randint(10, size=(100, 1))
 
-        xgb1 = XGBClassifier(n_estimators=3, use_label_encoder=False)
+        xgb1 = XGBClassifier(n_estimators=3, use_label_encoder=False, eval_metric='mlogloss')
         xgb1.fit(
             X_train, y_train,
             eval_set=[(X_train, y_train), (X_test, y_test)],
-            eval_metric='mlogloss'
         )
         self.assertIn("validation_0", xgb1.evals_result())