dslighting 1.3.9__py3-none-any.whl → 1.7.1__py3-none-any.whl

dslighting/registry/dabench-480-feature-engineering-techniques/prepare.py

@@ -0,0 +1,49 @@
+from pathlib import Path
+import pandas as pd
+
+
+def prepare(raw: Path, public: Path, private: Path):
+    """
+    Prepare the DABench task 480 dataset.
+
+    Args:
+        raw: Path to raw data directory (should contain oecd_education_spending.csv)
+        public: Path to public directory (for participants)
+        private: Path to private directory (for grading)
+    """
+    # Load the data
+    data_file = raw / "oecd_education_spending.csv"
+    if not data_file.exists():
+        raise FileNotFoundError(f"Data file not found: {data_file}")
+
+    df = pd.read_csv(data_file)
+
+    # Save the full dataset to public directory
+    train_file = public / "train.csv"
+    df.to_csv(train_file, index=False)
+    print(f"Saved training data to {train_file} ({len(df)} rows)")
+
+    # Create sample submission file
+    sample_submission = pd.DataFrame({
+        'id': [480],
+        'answer': ['@placeholder[0.00]']  # Placeholder answer
+    })
+    sample_submission_file = public / "sample_submission.csv"
+    sample_submission.to_csv(sample_submission_file, index=False)
+    print(f"Created sample submission: {sample_submission_file}")
+
+    # Create answer file (ground truth)
+    answer = pd.DataFrame({
+        'id': [480],
+        'answer': ['@standard_deviation[1.22]']
+    })
+    answer_file = private / "answer.csv"
+    answer.to_csv(answer_file, index=False)
+    print(f"Created answer file: {answer_file}")
+
+    # Verify
+    assert train_file.exists(), "Training file not created"
+    assert sample_submission_file.exists(), "Sample submission not created"
+    assert answer_file.exists(), "Answer file not created"
+
+    print(f"✓ DABench task 480 prepared successfully")

dslighting/registry/dabench-490-what-mean-percentage/checksums.yaml

@@ -0,0 +1,2 @@
+# Checksums for dabench-490-what-mean-percentage dataset
+zip: ""

dslighting/registry/dabench-490-what-mean-percentage/config.yaml

@@ -0,0 +1,16 @@
+id: dabench-490-what-mean-percentage
+name: DABench Task - What is the mean percentage of graduates in the fi
+competition_type: code
+awards_medals: false
+prizes: null
+description: mlebench/competitions/dabench-490-what-mean-percentage/description.md
+
+dataset:
+  answers: dabench-490-what-mean-percentage/prepared/private/answer.csv
+  sample_submission: dabench-490-what-mean-percentage/prepared/public/sample_submission.csv
+
+grader:
+  name: exact_match
+  grade_fn: mlebench.competitions.dabench-490-what-mean-percentage.grade:grade
+
+preparer: mlebench.competitions.dabench-490-what-mean-percentage.prepare:prepare

dslighting/registry/dabench-490-what-mean-percentage/description.md

@@ -0,0 +1,43 @@
+# DABench Task 490 - What is the mean percentage of graduates in the field of Engineering?
+
+## Task Description
+
+What is the mean percentage of graduates in the field of Engineering?
+
+## Concepts
+
+Summary Statistics
+
+## Data Description
+
+Dataset file: `percent-bachelors-degrees-women-usa.csv`
+
+The data is available in the `train.csv` file in the public directory.
+
+## Constraints
+
+Calculate the mean value using all available data points in the field of Engineering. Round the mean value to two decimal places.
+
+## Files
+
+- `train.csv` - The training data
+- `sample_submission.csv` - A sample submission file in the correct format
+
+## Submission Format
+
+Submit a CSV with columns `id,answer`, containing one row. In the `answer` cell include:
+@mean_value[number]
+
+"number" is between 0 and 100, rounded to two decimal places.
+
+## Evaluation
+
+Submissions are evaluated on **exact match** with the ground truth answer (with tolerance of 0.01 for floating-point values).
+
+## Goal
+
+Provide the correct answer in the specified format.
+
+## Difficulty Level
+
+Easy

dslighting/registry/dabench-490-what-mean-percentage/grade.py

@@ -0,0 +1,66 @@
+import pandas as pd
+import re
+
+
+def grade(submission: pd.DataFrame, answers: pd.DataFrame) -> float:
+    """
+    Grade the DABench submission.
+
+    Args:
+        submission: DataFrame with columns ['id', 'answer']
+        answers: DataFrame with columns ['id', 'answer']
+
+    Returns:
+        float: 1.0 if exact match, 0.0 otherwise
+    """
+    try:
+        # Both should have exactly one row
+        if len(submission) != 1 or len(answers) != 1:
+            return 0.0
+
+        # Get the submission and answer strings
+        submission_str = str(submission.iloc[0]['answer']).strip()
+        answer_str = str(answers.iloc[0]['answer']).strip()
+
+        # Parse all key-value pairs
+        # Expected keys: ['mean_value']
+        pattern = r'@(\w+)\[([^\]]+)\]'
+
+        submission_dict = dict(re.findall(pattern, submission_str))
+        answer_dict = dict(re.findall(pattern, answer_str))
+
+        if not submission_dict or not answer_dict:
+            print(f"Failed to parse: submission='{submission_str}', answer='{answer_str}'")
+            return 0.0
+
+        # Check if all keys match
+        if set(submission_dict.keys()) != set(answer_dict.keys()):
+            print(f"Key mismatch: submission has {set(submission_dict.keys())}, answer has {set(answer_dict.keys())}")
+            return 0.0
+
+        # Compare values
+        all_match = True
+        for key in answer_dict:
+            submission_value = submission_dict[key]
+            answer_value = answer_dict[key]
+
+            # Try to compare as numbers
+            try:
+                sub_float = float(submission_value)
+                ans_float = float(answer_value)
+                if abs(sub_float - ans_float) >= 0.01:
+                    print(f"Value mismatch for {key}: submission={sub_float}, answer={ans_float}")
+                    all_match = False
+                    break
+            except ValueError:
+                # Compare as strings (case-insensitive)
+                if submission_value.lower() != answer_value.lower():
+                    print(f"Value mismatch for {key}: submission='{submission_value}', answer='{answer_value}'")
+                    all_match = False
+                    break
+
+        return 1.0 if all_match else 0.0
+
+    except Exception as e:
+        print(f"Error in grading: {e}")
+        return 0.0

dslighting/registry/dabench-490-what-mean-percentage/prepare.py

@@ -0,0 +1,49 @@
+from pathlib import Path
+import pandas as pd
+
+
+def prepare(raw: Path, public: Path, private: Path):
+    """
+    Prepare the DABench task 490 dataset.
+
+    Args:
+        raw: Path to raw data directory (should contain percent-bachelors-degrees-women-usa.csv)
+        public: Path to public directory (for participants)
+        private: Path to private directory (for grading)
+    """
+    # Load the data
+    data_file = raw / "percent-bachelors-degrees-women-usa.csv"
+    if not data_file.exists():
+        raise FileNotFoundError(f"Data file not found: {data_file}")
+
+    df = pd.read_csv(data_file)
+
+    # Save the full dataset to public directory
+    train_file = public / "train.csv"
+    df.to_csv(train_file, index=False)
+    print(f"Saved training data to {train_file} ({len(df)} rows)")
+
+    # Create sample submission file
+    sample_submission = pd.DataFrame({
+        'id': [490],
+        'answer': ['@placeholder[0.00]']  # Placeholder answer
+    })
+    sample_submission_file = public / "sample_submission.csv"
+    sample_submission.to_csv(sample_submission_file, index=False)
+    print(f"Created sample submission: {sample_submission_file}")
+
+    # Create answer file (ground truth)
+    answer = pd.DataFrame({
+        'id': [490],
+        'answer': ['@mean_value[12.89]']
+    })
+    answer_file = private / "answer.csv"
+    answer.to_csv(answer_file, index=False)
+    print(f"Created answer file: {answer_file}")
+
+    # Verify
+    assert train_file.exists(), "Training file not created"
+    assert sample_submission_file.exists(), "Sample submission not created"
+    assert answer_file.exists(), "Answer file not created"
+
+    print(f"✓ DABench task 490 prepared successfully")

dslighting/registry/dabench-492-which-field-highest/checksums.yaml

@@ -0,0 +1,2 @@
+# Checksums for dabench-492-which-field-highest dataset
+zip: ""

dslighting/registry/dabench-492-which-field-highest/config.yaml

@@ -0,0 +1,16 @@
+id: dabench-492-which-field-highest
+name: DABench Task - Which field has the highest percentage of graduate
+competition_type: code
+awards_medals: false
+prizes: null
+description: mlebench/competitions/dabench-492-which-field-highest/description.md
+
+dataset:
+  answers: dabench-492-which-field-highest/prepared/private/answer.csv
+  sample_submission: dabench-492-which-field-highest/prepared/public/sample_submission.csv
+
+grader:
+  name: exact_match
+  grade_fn: mlebench.competitions.dabench-492-which-field-highest.grade:grade
+
+preparer: mlebench.competitions.dabench-492-which-field-highest.prepare:prepare

dslighting/registry/dabench-492-which-field-highest/description.md

@@ -0,0 +1,43 @@
+# DABench Task 492 - Which field has the highest percentage of graduates in the year 2010?
+
+## Task Description
+
+Which field has the highest percentage of graduates in the year 2010?
+
+## Concepts
+
+Summary Statistics
+
+## Data Description
+
+Dataset file: `percent-bachelors-degrees-women-usa.csv`
+
+The data is available in the `train.csv` file in the public directory.
+
+## Constraints
+
+Compare the last available data points (year 2010) for all fields within the dataset. If fields share the maximum percentage value, return all those fields separated by commas.
+
+## Files
+
+- `train.csv` - The training data
+- `sample_submission.csv` - A sample submission file in the correct format
+
+## Submission Format
+
+Submit a CSV with columns `id,answer`, containing one row. In the `answer` cell include:
+@fields[field_names]
+
+"field_names" is string(s) separated by commas, mentioning field(s) with the highest percentage of graduates in the year 2010.
+
+## Evaluation
+
+Submissions are evaluated on **exact match** with the ground truth answer (with tolerance of 0.01 for floating-point values).
+
+## Goal
+
+Provide the correct answer in the specified format.
+
+## Difficulty Level
+
+Easy

dslighting/registry/dabench-492-which-field-highest/grade.py

@@ -0,0 +1,66 @@
+import pandas as pd
+import re
+
+
+def grade(submission: pd.DataFrame, answers: pd.DataFrame) -> float:
+    """
+    Grade the DABench submission.
+
+    Args:
+        submission: DataFrame with columns ['id', 'answer']
+        answers: DataFrame with columns ['id', 'answer']
+
+    Returns:
+        float: 1.0 if exact match, 0.0 otherwise
+    """
+    try:
+        # Both should have exactly one row
+        if len(submission) != 1 or len(answers) != 1:
+            return 0.0
+
+        # Get the submission and answer strings
+        submission_str = str(submission.iloc[0]['answer']).strip()
+        answer_str = str(answers.iloc[0]['answer']).strip()
+
+        # Parse all key-value pairs
+        # Expected keys: ['fields']
+        pattern = r'@(\w+)\[([^\]]+)\]'
+
+        submission_dict = dict(re.findall(pattern, submission_str))
+        answer_dict = dict(re.findall(pattern, answer_str))
+
+        if not submission_dict or not answer_dict:
+            print(f"Failed to parse: submission='{submission_str}', answer='{answer_str}'")
+            return 0.0
+
+        # Check if all keys match
+        if set(submission_dict.keys()) != set(answer_dict.keys()):
+            print(f"Key mismatch: submission has {set(submission_dict.keys())}, answer has {set(answer_dict.keys())}")
+            return 0.0
+
+        # Compare values
+        all_match = True
+        for key in answer_dict:
+            submission_value = submission_dict[key]
+            answer_value = answer_dict[key]
+
+            # Try to compare as numbers
+            try:
+                sub_float = float(submission_value)
+                ans_float = float(answer_value)
+                if abs(sub_float - ans_float) >= 0.01:
+                    print(f"Value mismatch for {key}: submission={sub_float}, answer={ans_float}")
+                    all_match = False
+                    break
+            except ValueError:
+                # Compare as strings (case-insensitive)
+                if submission_value.lower() != answer_value.lower():
+                    print(f"Value mismatch for {key}: submission='{submission_value}', answer='{answer_value}'")
+                    all_match = False
+                    break
+
+        return 1.0 if all_match else 0.0
+
+    except Exception as e:
+        print(f"Error in grading: {e}")
+        return 0.0

dslighting/registry/dabench-492-which-field-highest/prepare.py

@@ -0,0 +1,49 @@
+from pathlib import Path
+import pandas as pd
+
+
+def prepare(raw: Path, public: Path, private: Path):
+    """
+    Prepare the DABench task 492 dataset.
+
+    Args:
+        raw: Path to raw data directory (should contain percent-bachelors-degrees-women-usa.csv)
+        public: Path to public directory (for participants)
+        private: Path to private directory (for grading)
+    """
+    # Load the data
+    data_file = raw / "percent-bachelors-degrees-women-usa.csv"
+    if not data_file.exists():
+        raise FileNotFoundError(f"Data file not found: {data_file}")
+
+    df = pd.read_csv(data_file)
+
+    # Save the full dataset to public directory
+    train_file = public / "train.csv"
+    df.to_csv(train_file, index=False)
+    print(f"Saved training data to {train_file} ({len(df)} rows)")
+
+    # Create sample submission file
+    sample_submission = pd.DataFrame({
+        'id': [492],
+        'answer': ['@placeholder[0.00]']  # Placeholder answer
+    })
+    sample_submission_file = public / "sample_submission.csv"
+    sample_submission.to_csv(sample_submission_file, index=False)
+    print(f"Created sample submission: {sample_submission_file}")
+
+    # Create answer file (ground truth)
+    answer = pd.DataFrame({
+        'id': [492],
+        'answer': ['@fields[Health Professions]']
+    })
+    answer_file = private / "answer.csv"
+    answer.to_csv(answer_file, index=False)
+    print(f"Created answer file: {answer_file}")
+
+    # Verify
+    assert train_file.exists(), "Training file not created"
+    assert sample_submission_file.exists(), "Sample submission not created"
+    assert answer_file.exists(), "Answer file not created"
+
+    print(f"✓ DABench task 492 prepared successfully")

dslighting/registry/dabench-495-outlier-detection-percentage/checksums.yaml

@@ -0,0 +1,2 @@
+# Checksums for dabench-495-outlier-detection-percentage dataset
+zip: ""

dslighting/registry/dabench-495-outlier-detection-percentage/config.yaml

@@ -0,0 +1,16 @@
+id: dabench-495-outlier-detection-percentage
+name: DABench Task - Perform outlier detection on the percentage of gra
+competition_type: code
+awards_medals: false
+prizes: null
+description: mlebench/competitions/dabench-495-outlier-detection-percentage/description.md
+
+dataset:
+  answers: dabench-495-outlier-detection-percentage/prepared/private/answer.csv
+  sample_submission: dabench-495-outlier-detection-percentage/prepared/public/sample_submission.csv
+
+grader:
+  name: exact_match
+  grade_fn: mlebench.competitions.dabench-495-outlier-detection-percentage.grade:grade
+
+preparer: mlebench.competitions.dabench-495-outlier-detection-percentage.prepare:prepare

dslighting/registry/dabench-495-outlier-detection-percentage/description.md

@@ -0,0 +1,45 @@
+# DABench Task 495 - Perform outlier detection on the percentage of graduates in the field of Architecture over the years using the Z-score method with a threshold of 3. Identify all years with outliers, then calculate the mean and standard deviation for the years without these outliers.
+
+## Task Description
+
+Perform outlier detection on the percentage of graduates in the field of Architecture over the years using the Z-score method with a threshold of 3. Identify all years with outliers, then calculate the mean and standard deviation for the years without these outliers.
+
+## Concepts
+
+Outlier Detection, Summary Statistics
+
+## Data Description
+
+Dataset file: `percent-bachelors-degrees-women-usa.csv`
+
+The data is available in the `train.csv` file in the public directory.
+
+## Constraints
+
+Use the Z-score method with a threshold of 3 for outlier detection.
+Include all years in the dataset for the calculation.
+After identifying the outliers, remove them and then calculate the mean percentage and the standard deviation of the remaining data. Round to two decimal places.
+
+## Files
+
+- `train.csv` - The training data
+- `sample_submission.csv` - A sample submission file in the correct format
+
+## Submission Format
+
+Submit a CSV with columns `id,answer`, containing one row. In the `answer` cell include both markers (space or newline separated) in this order (quote the cell as in the sample to preserve spaces/newlines):
+@std_without_outliers[std_value] @mean_without_outliers[mean_value]
+
+"mean_value" and "std_value" are floating point numbers rounded to two decimals representing the mean and standard deviation after removing outliers.
+
+## Evaluation
+
+Submissions are evaluated on **exact match** with the ground truth answer (with tolerance of 0.01 for floating-point values).
+
+## Goal
+
+Provide the correct answer in the specified format.
+
+## Difficulty Level
+
+Hard

dslighting/registry/dabench-495-outlier-detection-percentage/grade.py

@@ -0,0 +1,66 @@
+import pandas as pd
+import re
+
+
+def grade(submission: pd.DataFrame, answers: pd.DataFrame) -> float:
+    """
+    Grade the DABench submission.
+
+    Args:
+        submission: DataFrame with columns ['id', 'answer']
+        answers: DataFrame with columns ['id', 'answer']
+
+    Returns:
+        float: 1.0 if exact match, 0.0 otherwise
+    """
+    try:
+        # Both should have exactly one row
+        if len(submission) != 1 or len(answers) != 1:
+            return 0.0
+
+        # Get the submission and answer strings
+        submission_str = str(submission.iloc[0]['answer']).strip()
+        answer_str = str(answers.iloc[0]['answer']).strip()
+
+        # Parse all key-value pairs
+        # Expected keys: ['std_without_outliers', 'mean_without_outliers']
+        pattern = r'@(\w+)\[([^\]]+)\]'
+
+        submission_dict = dict(re.findall(pattern, submission_str))
+        answer_dict = dict(re.findall(pattern, answer_str))
+
+        if not submission_dict or not answer_dict:
+            print(f"Failed to parse: submission='{submission_str}', answer='{answer_str}'")
+            return 0.0
+
+        # Check if all keys match
+        if set(submission_dict.keys()) != set(answer_dict.keys()):
+            print(f"Key mismatch: submission has {set(submission_dict.keys())}, answer has {set(answer_dict.keys())}")
+            return 0.0
+
+        # Compare values
+        all_match = True
+        for key in answer_dict:
+            submission_value = submission_dict[key]
+            answer_value = answer_dict[key]
+
+            # Try to compare as numbers
+            try:
+                sub_float = float(submission_value)
+                ans_float = float(answer_value)
+                if abs(sub_float - ans_float) >= 0.01:
+                    print(f"Value mismatch for {key}: submission={sub_float}, answer={ans_float}")
+                    all_match = False
+                    break
+            except ValueError:
+                # Compare as strings (case-insensitive)
+                if submission_value.lower() != answer_value.lower():
+                    print(f"Value mismatch for {key}: submission='{submission_value}', answer='{answer_value}'")
+                    all_match = False
+                    break
+
+        return 1.0 if all_match else 0.0
+
+    except Exception as e:
+        print(f"Error in grading: {e}")
+        return 0.0

dslighting/registry/dabench-495-outlier-detection-percentage/prepare.py

@@ -0,0 +1,49 @@
+from pathlib import Path
+import pandas as pd
+
+
+def prepare(raw: Path, public: Path, private: Path):
+    """
+    Prepare the DABench task 495 dataset.
+
+    Args:
+        raw: Path to raw data directory (should contain percent-bachelors-degrees-women-usa.csv)
+        public: Path to public directory (for participants)
+        private: Path to private directory (for grading)
+    """
+    # Load the data
+    data_file = raw / "percent-bachelors-degrees-women-usa.csv"
+    if not data_file.exists():
+        raise FileNotFoundError(f"Data file not found: {data_file}")
+
+    df = pd.read_csv(data_file)
+
+    # Save the full dataset to public directory
+    train_file = public / "train.csv"
+    df.to_csv(train_file, index=False)
+    print(f"Saved training data to {train_file} ({len(df)} rows)")
+
+    # Create sample submission file
+    sample_submission = pd.DataFrame({
+        'id': [495],
+        'answer': ['@placeholder[0.00]']  # Placeholder answer
+    })
+    sample_submission_file = public / "sample_submission.csv"
+    sample_submission.to_csv(sample_submission_file, index=False)
+    print(f"Created sample submission: {sample_submission_file}")
+
+    # Create answer file (ground truth)
+    answer = pd.DataFrame({
+        'id': [495],
+        'answer': ['@std_without_outliers[9.57] @mean_without_outliers[33.69]']
+    })
+    answer_file = private / "answer.csv"
+    answer.to_csv(answer_file, index=False)
+    print(f"Created answer file: {answer_file}")
+
+    # Verify
+    assert train_file.exists(), "Training file not created"
+    assert sample_submission_file.exists(), "Sample submission not created"
+    assert answer_file.exists(), "Answer file not created"
+
+    print(f"✓ DABench task 495 prepared successfully")

dslighting/registry/dabench-496-feature-engineering-creating/checksums.yaml

@@ -0,0 +1,2 @@
+# Checksums for dabench-496-feature-engineering-creating dataset
+zip: ""

dslighting/registry/dabench-496-feature-engineering-creating/config.yaml

@@ -0,0 +1,16 @@
+id: dabench-496-feature-engineering-creating
+name: DABench Task - Perform feature engineering by creating a new feat
+competition_type: code
+awards_medals: false
+prizes: null
+description: mlebench/competitions/dabench-496-feature-engineering-creating/description.md
+
+dataset:
+  answers: dabench-496-feature-engineering-creating/prepared/private/answer.csv
+  sample_submission: dabench-496-feature-engineering-creating/prepared/public/sample_submission.csv
+
+grader:
+  name: exact_match
+  grade_fn: mlebench.competitions.dabench-496-feature-engineering-creating.grade:grade
+
+preparer: mlebench.competitions.dabench-496-feature-engineering-creating.prepare:prepare

dslighting/registry/dabench-496-feature-engineering-creating/description.md

@@ -0,0 +1,44 @@
+# DABench Task 496 - Perform feature engineering by creating a new feature called "STEM" (Science, Technology, Engineering, and Math). It should be the sum of the percentages of graduates in the fields of Computer Science, Engineering, Math and Statistics, and Physical Sciences. Calculate the mean and range (maximum - minimum) of the "STEM" feature for the years beyond 2000.
+
+## Task Description
+
+Perform feature engineering by creating a new feature called "STEM" (Science, Technology, Engineering, and Math). It should be the sum of the percentages of graduates in the fields of Computer Science, Engineering, Math and Statistics, and Physical Sciences. Calculate the mean and range (maximum - minimum) of the "STEM" feature for the years beyond 2000.
+
+## Concepts
+
+Feature Engineering, Summary Statistics
+
+## Data Description
+
+Dataset file: `percent-bachelors-degrees-women-usa.csv`
+
+The data is available in the `train.csv` file in the public directory.
+
+## Constraints
+
+Calculate the new feature "STEM" as the sum of the percentages of graduates in the fields of Computer Science, Engineering, Math and Statistics, and Physical Sciences.
+Compute the mean and the range (maximum - minimum) of the "STEM" feature for the years 2000 and beyond. Round to two decimal places.
+
+## Files
+
+- `train.csv` - The training data
+- `sample_submission.csv` - A sample submission file in the correct format
+
+## Submission Format
+
+@mean_STEM[mean_value]
+@range_STEM[range_value]
+where "mean_value" is a floating point number rounded to two decimal places representing the mean of the "STEM" feature.
+where "range_value" is a floating point number rounded to two decimal places representing the range of the "STEM" feature.
+
+## Evaluation
+
+Submissions are evaluated on **exact match** with the ground truth answer (with tolerance of 0.01 for floating-point values).
+
+## Goal
+
+Provide the correct answer in the specified format.
+
+## Difficulty Level
+
+Hard