dslighting 1.7.1__py3-none-any.whl → 1.7.8__py3-none-any.whl

mlebench/competitions/plant-pathology-2021-fgvc8/prepare_val.py

@@ -0,0 +1,130 @@
+import shutil
+from pathlib import Path
+from typing import Dict
+
+import pandas as pd
+from sklearn.model_selection import train_test_split
+from tqdm.auto import tqdm
+
+from mlebench.utils import extract, read_csv
+
+
+def _process_and_save_split(
+    source_df: pd.DataFrame,
+    public_path: Path,
+    private_path: Path,
+    raw_images_path: Path,
+    test_size: float,
+    random_state: int,
+) -> pd.DataFrame:
+    """
+    Helper function to perform a data split, save CSVs, and copy image files.
+
+    Args:
+        source_df: The DataFrame to split.
+        public_path: The public output directory.
+        private_path: The private output directory.
+        raw_images_path: The path to the source images.
+        test_size: The proportion of the dataset to allocate to the test split.
+        random_state: The random state for reproducibility.
+
+    Returns:
+        The training portion of the split DataFrame.
+    """
+    # Create output directories
+    public_path.mkdir(exist_ok=True, parents=True)
+    private_path.mkdir(exist_ok=True, parents=True)
+
+    # Create train, test from the source dataframe
+    new_train, answers = train_test_split(
+        source_df, test_size=test_size, random_state=random_state
+    )
+
+    # Create a sample submission file
+    submission_df = answers.copy()
+    submission_df["labels"] = "healthy"
+
+    # Checks
+    assert len(answers) == len(submission_df), "Answers and submission should have the same length"
+    assert not set(new_train["image"]).intersection(
+        set(answers["image"])
+    ), "new_train and answers should not share any image"
+    assert (
+        "image" in new_train.columns and "labels" in new_train.columns
+    ), "Train DataFrame must have 'image' and 'labels' columns"
+    assert (
+        "image" in submission_df.columns and "labels" in submission_df.columns
+    ), "Sample submission DataFrame must have 'image' and 'labels' columns"
+    assert len(new_train) + len(answers) == len(
+        source_df
+    ), "The combined length of new_train and answers should equal the length of the source dataframe"
+
+    # Write CSVs using the required standard filenames
+    answers.to_csv(private_path / "answers.csv", index=False)
+    new_train.to_csv(public_path / "train.csv", index=False)
+    submission_df.to_csv(public_path / "sample_submission.csv", index=False)
+
+    # Copy files
+    (public_path / "test_images").mkdir(exist_ok=True)
+    (public_path / "train_images").mkdir(exist_ok=True)
+
+    for file_id in tqdm(new_train["image"], desc=f"Copying Train Images to {public_path.name}"):
+        shutil.copyfile(
+            src=raw_images_path / f"{file_id}",
+            dst=public_path / "train_images" / f"{file_id}",
+        )
+
+    for file_id in tqdm(answers["image"], desc=f"Copying Test Images to {public_path.name}"):
+        shutil.copyfile(
+            src=raw_images_path / f"{file_id}",
+            dst=public_path / "test_images" / f"{file_id}",
+        )
+
+    # Checks
+    assert len(list(public_path.glob("train_images/*.jpg"))) == len(
+        new_train
+    ), f"Public train images in {public_path.name} should have the same number of images as the train DataFrame"
+    assert len(list(public_path.glob("test_images/*.jpg"))) == len(
+        answers
+    ), f"Public test images in {public_path.name} should have the same number of images as the answers DataFrame"
+
+    return new_train
+
+
+def prepare(raw: Path, public: Path, private: Path):
+    """
+    Splits the data in raw into public/private datasets for the main competition,
+    and creates a parallel validation split in public_val/private_val directories.
+    """
+    # Define paths for the new validation split
+    public_val = public.parent / "public_val"
+    private_val = private.parent / "private_val"
+
+    old_train = read_csv(raw / "train.csv")
+    raw_images_path = raw / "train_images"
+
+    # 1. Create the original train/test split for the competition.
+    # This creates the `public` and `private` directories. The outputs here
+    # will be identical to the original script's outputs.
+    # The returned `competition_train_df` is the 80% training set from this first split.
+    competition_train_df = _process_and_save_split(
+        source_df=old_train,
+        public_path=public,
+        private_path=private,
+        raw_images_path=raw_images_path,
+        test_size=0.2,
+        random_state=0,
+    )
+
+    # 2. Create the new validation split from the competition's training data.
+    # We split the `competition_train_df` (80% of original data) again.
+    # To get a validation set of the same size as the original test set (20% of total),
+    # we take 25% from this new pool of data (0.25 * 0.8 = 0.2).
+    _process_and_save_split(
+        source_df=competition_train_df,
+        public_path=public_val,
+        private_path=private_val,
+        raw_images_path=raw_images_path,
+        test_size=0.25,
+        random_state=0,
+    )

mlebench/competitions/plant-seedlings-classification/grade.py

@@ -0,0 +1,39 @@
+import pandas as pd
+from sklearn.metrics import f1_score
+
+from mlebench.grade_helpers import InvalidSubmissionError
+
+
+def prepare_for_metric(submission: pd.DataFrame, answers: pd.DataFrame):
+    """
+    `submission` and `answers` are pd.DataFrame with "file" and "species" columns
+    """
+
+    # Answer checks
+    assert "file" in answers.columns, "Answers must have 'file' column"
+    assert "species" in answers.columns, "Answers must have 'species' column"
+
+    # Submission checks
+    if "file" not in submission.columns:
+        raise InvalidSubmissionError("Submission must have 'file' column")
+    if "species" not in submission.columns:
+        raise InvalidSubmissionError("Submission must have 'species' column")
+    if len(submission) != len(answers):
+        raise InvalidSubmissionError(
+            f"Submission length {len(submission)} != answers length {len(answers)}"
+        )
+
+    # Sort submission and answers by "file"
+    submission = submission.sort_values("file")
+    answers = answers.sort_values("file")
+    if (submission["file"].values != answers["file"].values).any():
+        raise InvalidSubmissionError("Submission and answers have mismatched 'file' columns")
+
+    y_true = answers["species"].tolist()
+    y_pred = submission["species"].tolist()
+    return y_true, y_pred
+
+
+def grade(submission: pd.DataFrame, answers: pd.DataFrame) -> float:
+    y_true, y_pred = prepare_for_metric(submission, answers)
+    return f1_score(y_true=y_true, y_pred=y_pred, average="micro")

mlebench/competitions/plant-seedlings-classification/prepare.py

@@ -0,0 +1,91 @@
+import os
+import shutil
+from pathlib import Path
+
+import pandas as pd
+from sklearn.model_selection import train_test_split
+from tqdm import tqdm
+
+from mlebench.utils import extract, read_csv
+
+
+def create_dataframe_from_directory(directory: str) -> pd.DataFrame:
+    """
+    Creates a DataFrame from a directory of images.
+
+    Args:
+        directory (str): The path to the directory containing subdirectories of images.
+
+    Returns:
+        pd.DataFrame: A DataFrame with two columns: 'image' and 'label'. The 'image' column contains the file paths to the images, and the 'label' column contains the corresponding labels (subdirectory names).
+    """
+    data = []
+    for label in sorted(os.listdir(directory)):  # Sort labels for determinism
+        label_path = os.path.join(directory, label)
+        if os.path.isdir(label_path):
+            for file_name in sorted(os.listdir(label_path)):  # Sort files for determinism
+                if file_name.endswith(".png"):
+                    file_path = os.path.join(label_path, file_name)
+                    data.append({"file": os.path.basename(file_path), "species": label})
+    return pd.DataFrame(data)
+
+
+def prepare(raw: Path, public: Path, private: Path):
+    """
+    Splits the data in raw into public and private datasets with appropriate test/train splits.
+    """
+
+    # Directory containing the training images
+    train_dir = raw / "train"
+    old_train = create_dataframe_from_directory(train_dir)
+    test_ratio = 0.14  # 794/(4750+794) = 0.14
+    train_df, test_df = train_test_split(old_train, test_size=test_ratio, random_state=0)
+
+    # Create a sample submission file
+    submission_df = test_df.copy()
+    submission_df["species"] = "Sugar beet"
+
+    # Checks
+    assert len(test_df) == len(submission_df), "Answers and submission should have the same length"
+    assert not set(train_df["file"]).intersection(
+        set(test_df["file"])
+    ), "new_train and answers should not share any image"
+    assert (
+        "file" in train_df.columns and "species" in train_df.columns
+    ), "Train DataFrame must have 'file' and 'species' columns"
+    assert (
+        "file" in submission_df.columns and "species" in submission_df.columns
+    ), "Sample submission DataFrame must have 'file' and 'species' columns"
+    assert len(train_df) + len(test_df) == len(
+        old_train
+    ), "The combined length of new_train and answers should equal the length of old_train"
+
+    # Write CSVs
+    test_df.to_csv(private / "answers.csv", index=False)
+    submission_df.to_csv(public / "sample_submission.csv", index=False)
+
+    # Copy files
+    (public / "test").mkdir(exist_ok=True)
+    (public / "train").mkdir(exist_ok=True)
+
+    # Create nested folder structure for train
+    for species in train_df["species"].unique():
+        (public / "train" / species).mkdir(parents=True, exist_ok=True)
+
+    for _, row in tqdm(train_df.iterrows(), desc="Copying Train Images", total=len(train_df)):
+        src_path = train_dir / row["species"] / row["file"]
+        dst_path = public / "train" / row["species"] / row["file"]
+        shutil.copyfile(src=src_path, dst=dst_path)
+
+    for _, row in tqdm(test_df.iterrows(), desc="Copying Test Images", total=len(test_df)):
+        src_path = train_dir / row["species"] / row["file"]
+        dst_path = public / "test" / row["file"]
+        shutil.copyfile(src=src_path, dst=dst_path)
+
+    # Checks
+    assert len(list(public.glob("train/**/*.png"))) == len(
+        train_df
+    ), f"Public train images should have the same number of images as the train DataFrame: number of files {len(list(public.glob('train/**/*.png')))} != len(train_df)={len(train_df)}"
+    assert len(list(public.glob("test/*.png"))) == len(
+        test_df
+    ), f"Public test images should have the same number of images as the answers DataFrame: number of files {len(list(public.glob('test/*.png')))} != len(test_df)={len(test_df)}"

mlebench/competitions/plant-seedlings-classification/prepare_val.py

@@ -0,0 +1,158 @@
+import os
+import shutil
+from pathlib import Path
+
+import pandas as pd
+from sklearn.model_selection import train_test_split
+from tqdm import tqdm
+
+from mlebench.utils import extract, read_csv
+
+
+def create_dataframe_from_directory(directory: str) -> pd.DataFrame:
+    """
+    Creates a DataFrame from a directory of images.
+
+    Args:
+        directory (str): The path to the directory containing subdirectories of images.
+
+    Returns:
+        pd.DataFrame: A DataFrame with two columns: 'image' and 'label'. The 'image' column contains the file paths to the images, and the 'label' column contains the corresponding labels (subdirectory names).
+    """
+    data = []
+    for label in sorted(os.listdir(directory)):  # Sort labels for determinism
+        label_path = os.path.join(directory, label)
+        if os.path.isdir(label_path):
+            for file_name in sorted(os.listdir(label_path)):  # Sort files for determinism
+                if file_name.endswith(".png"):
+                    file_path = os.path.join(label_path, file_name)
+                    data.append({"file": os.path.basename(file_path), "species": label})
+    return pd.DataFrame(data)
+
+
+def _process_split(
+    train_df: pd.DataFrame,
+    test_df: pd.DataFrame,
+    public_path: Path,
+    private_path: Path,
+    source_images_dir: Path,
+):
+    """
+    Helper function to process a train/test split and write all necessary files and folders.
+    This function creates the public and private directories, generates CSVs, and copies image files.
+    """
+    # Ensure destination directories exist
+    public_path.mkdir(exist_ok=True)
+    private_path.mkdir(exist_ok=True)
+
+    # Create a sample submission file
+    submission_df = test_df.copy()
+    submission_df["species"] = "Sugar beet"
+
+    # Checks
+    assert len(test_df) == len(submission_df), "Answers and submission should have the same length"
+    assert not set(train_df["file"]).intersection(
+        set(test_df["file"])
+    ), "new_train and answers should not share any image"
+    assert (
+        "file" in train_df.columns and "species" in train_df.columns
+    ), "Train DataFrame must have 'file' and 'species' columns"
+    assert (
+        "file" in submission_df.columns and "species" in submission_df.columns
+    ), "Sample submission DataFrame must have 'file' and 'species' columns"
+
+    # Write CSVs
+    test_df.to_csv(private_path / "answers.csv", index=False)
+    submission_df.to_csv(public_path / "sample_submission.csv", index=False)
+
+    # Prepare image directories
+    public_test_images_path = public_path / "test"
+    public_train_images_path = public_path / "train"
+    public_test_images_path.mkdir(exist_ok=True)
+    public_train_images_path.mkdir(exist_ok=True)
+
+    # Create nested folder structure for train
+    for species in train_df["species"].unique():
+        (public_train_images_path / species).mkdir(parents=True, exist_ok=True)
+
+    # Use public path name for progress bar description
+    desc_prefix = public_path.name.capitalize()
+
+    for _, row in tqdm(
+        train_df.iterrows(), desc=f"Copying {desc_prefix} Train Images", total=len(train_df)
+    ):
+        src_path = source_images_dir / row["species"] / row["file"]
+        dst_path = public_train_images_path / row["species"] / row["file"]
+        shutil.copyfile(src=src_path, dst=dst_path)
+
+    for _, row in tqdm(
+        test_df.iterrows(), desc=f"Copying {desc_prefix} Test Images", total=len(test_df)
+    ):
+        src_path = source_images_dir / row["species"] / row["file"]
+        dst_path = public_test_images_path / row["file"]
+        shutil.copyfile(src=src_path, dst=dst_path)
+
+    # Final checks on copied files
+    assert len(list(public_train_images_path.glob("**/*.png"))) == len(
+        train_df
+    ), f"Public train images should have the same number of images as the train DataFrame: number of files {len(list(public_train_images_path.glob('**/*.png')))} != len(train_df)={len(train_df)}"
+    assert len(list(public_test_images_path.glob("*.png"))) == len(
+        test_df
+    ), f"Public test images should have the same number of images as the answers DataFrame: number of files {len(list(public_test_images_path.glob('*.png')))} != len(test_df)={len(test_df)}"
+
+
+def prepare(raw: Path, public: Path, private: Path):
+    """
+    Splits the data in raw into public and private datasets with appropriate test/train splits.
+    It then creates a secondary validation split (public_val, private_val) from the main training data.
+    """
+    # Directory containing the original images
+    train_dir = raw / "train"
+    old_train = create_dataframe_from_directory(train_dir)
+    test_ratio = 0.14  # 794/(4750+794) = 0.14
+
+    # --- 1. Original Split: Create main train and test sets ---
+    train_df, test_df = train_test_split(old_train, test_size=test_ratio, random_state=0)
+
+    # Check for original split integrity
+    assert len(train_df) + len(test_df) == len(
+        old_train
+    ), "The combined length of new_train and answers should equal the length of old_train"
+
+    # Process and save the original split to public/ and private/
+    # This generates the original, unmodified competition output.
+    _process_split(
+        train_df=train_df,
+        test_df=test_df,
+        public_path=public,
+        private_path=private,
+        source_images_dir=train_dir,
+    )
+
+    # --- 2. New Validation Split: Split the main train_df again ---
+    # Define paths for the new validation dataset, parallel to the original ones.
+    public_val = public.parent / "public_val"
+    private_val = private.parent / "private_val"
+
+    # Calculate the new test ratio to make the validation set (test_val_df)
+    # approximately the same size as the original test set (test_df).
+    new_test_ratio = len(test_df) / len(train_df)
+
+    # Split the training data to create a new, smaller training set and a validation set.
+    # We use the same random_state for reproducibility and consistency with the first split.
+    train_val_df, test_val_df = train_test_split(train_df, test_size=new_test_ratio, random_state=0)
+
+    # Check for new split integrity
+    assert (
+        len(train_val_df) + len(test_val_df) == len(train_df)
+    ), "The combined length of train_val_df and test_val_df should equal the length of train_df"
+
+    # Process and save the new validation split to public_val/ and private_val/
+    # Using the same helper ensures an identical file and directory structure.
+    _process_split(
+        train_df=train_val_df,
+        test_df=test_val_df,
+        public_path=public_val,
+        private_path=private_val,
+        source_images_dir=train_dir,
+    )

mlebench/competitions/playground-series-s3e1/grade.py

@@ -0,0 +1,52 @@
+import numpy as np
+import pandas as pd
+from sklearn.metrics import mean_squared_error
+
+from mlebench.grade_helpers import InvalidSubmissionError
+
+
+def prepare_for_metric(submission: pd.DataFrame, answers: pd.DataFrame):
+    """Prepare submission and answers for RMSE calculation."""
+    if len(submission) != len(answers):
+        raise InvalidSubmissionError(
+            f"Submission length ({len(submission)}) != answers length ({len(answers)})"
+        )
+
+    if "id" not in submission.columns:
+        raise InvalidSubmissionError("Submission must have an 'id' column")
+
+    if "MedHouseVal" not in submission.columns:
+        raise InvalidSubmissionError("Submission must have a 'MedHouseVal' column")
+
+    # Sort by id
+    submission = submission.sort_values("id").reset_index(drop=True)
+    answers = answers.sort_values("id").reset_index(drop=True)
+
+    # Check id alignment
+    if not (submission["id"] == answers["id"]).all():
+        raise InvalidSubmissionError("Submission and answers id mismatch")
+
+    # Extract predictions and true values
+    y_pred = submission["MedHouseVal"].values
+    y_true = answers["MedHouseVal"].values
+
+    # Validate predictions
+    try:
+        y_pred = y_pred.astype(float)
+    except (ValueError, TypeError):
+        raise InvalidSubmissionError("Predictions must be numeric")
+
+    if np.any(np.isnan(y_pred)):
+        raise InvalidSubmissionError("Predictions cannot contain NaN values")
+
+    return y_true, y_pred
+
+
+def grade(submission: pd.DataFrame, answers: pd.DataFrame) -> float:
+    """Calculate RMSE score."""
+    y_true, y_pred = prepare_for_metric(submission, answers)
+
+    # RMSE
+    rmse = np.sqrt(mean_squared_error(y_true, y_pred))
+
+    return rmse

mlebench/competitions/playground-series-s3e1/prepare.py

@@ -0,0 +1,25 @@
+from pathlib import Path
+import pandas as pd
+
+
+def prepare(raw: Path, public: Path, private: Path):
+    """
+    Prepare playground-series-s3e1 dataset.
+    """
+    # Read data
+    train = pd.read_csv(raw / "train.csv")
+    test = pd.read_csv(raw / "test.csv")
+    sample_submission = pd.read_csv(raw / "sample_submission.csv")
+    test_answer = pd.read_csv(raw / "test_answer.csv")
+
+    # Public files (visible to agents)
+    train.to_csv(public / "train.csv", index=False)
+    test.to_csv(public / "test.csv", index=False)
+    sample_submission.to_csv(public / "sample_submission.csv", index=False)
+
+    # Private files (for grading)
+    test_answer.to_csv(private / "test.csv", index=False)
+
+    # Validation checks
+    assert len(test_answer) == len(sample_submission), \
+        f"Test answer ({len(test_answer)}) and sample submission ({len(sample_submission)}) must have same length"

mlebench/competitions/playground-series-s3e11/grade.py

@@ -0,0 +1,55 @@
+import numpy as np
+import pandas as pd
+from sklearn.metrics import mean_squared_log_error
+
+from mlebench.grade_helpers import InvalidSubmissionError
+
+
+def prepare_for_metric(submission: pd.DataFrame, answers: pd.DataFrame):
+    """Prepare submission and answers for RMSLE calculation."""
+    if len(submission) != len(answers):
+        raise InvalidSubmissionError(
+            f"Submission length ({len(submission)}) != answers length ({len(answers)})"
+        )
+
+    if "id" not in submission.columns:
+        raise InvalidSubmissionError("Submission must have an 'id' column")
+
+    if "cost" not in submission.columns:
+        raise InvalidSubmissionError("Submission must have a 'cost' column")
+
+    # Sort by id
+    submission = submission.sort_values("id").reset_index(drop=True)
+    answers = answers.sort_values("id").reset_index(drop=True)
+
+    # Check id alignment
+    if not (submission["id"] == answers["id"]).all():
+        raise InvalidSubmissionError("Submission and answers id mismatch")
+
+    # Extract predictions and true values
+    y_pred = submission["cost"].values
+    y_true = answers["cost"].values
+
+    # Validate predictions
+    try:
+        y_pred = y_pred.astype(float)
+    except (ValueError, TypeError):
+        raise InvalidSubmissionError("Predictions must be numeric")
+
+    if np.any(np.isnan(y_pred)):
+        raise InvalidSubmissionError("Predictions cannot contain NaN values")
+
+    if np.any(y_pred < 0):
+        raise InvalidSubmissionError("Predictions cannot be negative")
+
+    return y_true, y_pred
+
+
+def grade(submission: pd.DataFrame, answers: pd.DataFrame) -> float:
+    """Calculate RMSLE score."""
+    y_true, y_pred = prepare_for_metric(submission, answers)
+
+    # RMSLE
+    rmsle = np.sqrt(mean_squared_log_error(y_true, y_pred))
+
+    return rmsle

mlebench/competitions/playground-series-s3e11/prepare.py

@@ -0,0 +1,25 @@
+from pathlib import Path
+import pandas as pd
+
+
+def prepare(raw: Path, public: Path, private: Path):
+    """
+    Prepare playground-series-s3e11 dataset.
+    """
+    # Read data
+    train = pd.read_csv(raw / "train.csv")
+    test = pd.read_csv(raw / "test.csv")
+    sample_submission = pd.read_csv(raw / "sample_submission.csv")
+    test_answer = pd.read_csv(raw / "test_answer.csv")
+
+    # Public files (visible to agents)
+    train.to_csv(public / "train.csv", index=False)
+    test.to_csv(public / "test.csv", index=False)
+    sample_submission.to_csv(public / "sample_submission.csv", index=False)
+
+    # Private files (for grading)
+    test_answer.to_csv(private / "test.csv", index=False)
+
+    # Validation checks
+    assert len(test_answer) == len(sample_submission), \
+        f"Test answer ({len(test_answer)}) and sample submission ({len(sample_submission)}) must have same length"

mlebench/competitions/playground-series-s3e18/grade.py

@@ -0,0 +1,39 @@
+import pandas as pd
+from sklearn.metrics import roc_auc_score
+
+from mlebench.grade_helpers import InvalidSubmissionError
+
+
+def prepare_for_metric(submission: pd.DataFrame, answers: pd.DataFrame) -> dict:
+
+    if "id" not in submission.columns or "id" not in answers.columns:
+        raise InvalidSubmissionError(
+            "Both submission and answers DataFrames must contain an 'id' column."
+        )
+
+    if "EC1" not in submission.columns or "EC2" not in submission.columns:
+        raise InvalidSubmissionError("Submission DataFrame must contain 'EC1' and 'EC2' columns.")
+
+    assert (
+        "EC1" in answers.columns and "EC2" in answers.columns
+    ), "Answers DataFrame must contain 'EC1' and 'EC2' columns."
+
+    if len(submission) != len(answers):
+        raise InvalidSubmissionError("Submission must be the same length as the answers.")
+
+    answers_sorted = answers.sort_values("id")
+    submission_sorted = submission.sort_values("id")
+
+    if (submission_sorted["id"].values != answers_sorted["id"].values).any():
+        raise InvalidSubmissionError("Submission and answers have mismatched 'id' columns")
+
+    y_true = answers_sorted[["EC1", "EC2"]].to_numpy()
+    y_score = submission_sorted[["EC1", "EC2"]].to_numpy()
+
+    return {"y_true": y_true, "y_score": y_score}
+
+
+def grade(submission: pd.DataFrame, answers: pd.DataFrame) -> float:
+    roc_auc_inputs = prepare_for_metric(submission, answers)
+    # macro average: calc metric for each class, and then average, as outlined in comp description
+    return roc_auc_score(**roc_auc_inputs, average="macro")

mlebench/competitions/playground-series-s3e18/prepare.py

@@ -0,0 +1,36 @@
+from pathlib import Path
+
+from sklearn.model_selection import train_test_split
+
+from mlebench.utils import read_csv
+
+
+def prepare(raw: Path, public: Path, private: Path):
+
+    # Create train and test splits from train set
+    old_train = read_csv(raw / "train.csv")
+    new_train, new_test = train_test_split(old_train, test_size=0.1, random_state=0)
+
+    # Create sample submission
+    sample_submission = new_test.copy()
+    sample_submission["EC1"] = 0.5
+    sample_submission["EC2"] = 0.5
+    sample_submission.drop(
+        sample_submission.columns.difference(["id", "EC1", "EC2"]), axis=1, inplace=True
+    )
+    sample_submission.to_csv(public / "sample_submission.csv", index=False)
+
+    # Create private files
+    new_test.to_csv(private / "test.csv", index=False)
+
+    # Create public files visible to agents
+    new_train.to_csv(public / "train.csv", index=False)
+    new_test.drop(["EC1", "EC2", "EC3", "EC4", "EC5", "EC6"], axis=1, inplace=True)
+    new_test.to_csv(public / "test.csv", index=False)
+
+    # Checks
+    assert new_test.shape[1] == 32, "Public test set should have 32 columns"
+    assert new_train.shape[1] == 38, "Public train set should have 38 columns"
+    assert len(new_train) + len(new_test) == len(
+        old_train
+    ), "Length of new_train and new_test should equal length of old_train"