Add line profiler

Author: IlyaMuravjov

cfpq_decomposer/abstract_decomposer.py

@@ -2,6 +2,7 @@
 from typing import Tuple
 
 import graphblas
+import line_profiler
 from graphblas.core.dtypes import BOOL
 from graphblas.core.matrix import Matrix
 
@@ -32,6 +33,7 @@ def column_based_decompose(self, matrix: Matrix):
         left_transposed, right_transposed = self.row_based_decompose(matrix.T.new())
         return right_transposed.T.new(), left_transposed.T.new()
 
+    @line_profiler.profile
     def decompose(self, matrix: Matrix) -> Tuple[Matrix, Matrix]:
         residual = matrix
         if residual.nvals == 0:

cfpq_decomposer/high_performance_decomposer.py

@@ -1,6 +1,7 @@
 from dataclasses import dataclass
 from typing import Tuple, Any
 
+import line_profiler
 import numpy as np
 import numpy.typing as npt
 from graphblas import semiring
@@ -34,12 +35,14 @@ def rows_in_buckets(self):
 
 
 class HighPerformanceDecomposer(AbstractDecomposer):
+    @line_profiler.profile
     def row_based_decompose(self, matrix: Matrix) -> Tuple[Matrix, Matrix]:
         build_left_factor_result = self._build_left_factor(matrix)
         right_factor = self._build_right_factor(matrix, build_left_factor_result)
         return build_left_factor_result.left_factor, right_factor
 
     @staticmethod
+    @line_profiler.profile
     def _build_left_factor(matrix: Matrix) -> BuildLeftFactorResult:
         num_rows, num_cols = matrix.shape
 
@@ -60,11 +63,13 @@ def _build_left_factor(matrix: Matrix) -> BuildLeftFactorResult:
         return BuildLeftFactorResult(left_factor, bucket_sizes)
 
     @staticmethod
+    @line_profiler.profile
     def _group_rows_to_buckets(matrix: Matrix) -> RowGroupingResult:
         row_hashes = HighPerformanceDecomposer._compute_row_hashes(matrix)
         return HighPerformanceDecomposer._group_row_hashes_to_buckets(row_hashes)
 
     @staticmethod
+    @line_profiler.profile
     def _compute_row_hashes(matrix: Matrix) -> npt.NDArray[Any]:
         row_signatures_matrix = HighPerformanceDecomposer._compute_row_signatures_matrix(matrix)
         hash_weights = np.random.default_rng().integers(
@@ -77,6 +82,7 @@ def _compute_row_hashes(matrix: Matrix) -> npt.NDArray[Any]:
         return row_hashes
 
     @staticmethod
+    @line_profiler.profile
     def _compute_row_signatures_matrix(matrix: Matrix) -> Matrix:
         num_rows, num_cols = matrix.shape
         hash_coefficients = np.random.randint(1, HASH_PRIME_MODULUS, size=HASH_FUNCTIONS_COUNT, dtype=np.int64)
@@ -90,6 +96,7 @@ def _compute_row_signatures_matrix(matrix: Matrix) -> Matrix:
         return row_signatures_matrix
 
     @staticmethod
+    @line_profiler.profile
     def _group_row_hashes_to_buckets(row_hashes: npt.NDArray[Any]) -> RowGroupingResult:
 
         _, row_to_bucket, new_bucket_sizes = np.unique(
@@ -111,6 +118,7 @@ def _group_row_hashes_to_buckets(row_hashes: npt.NDArray[Any]) -> RowGroupingRes
         return RowGroupingResult(row_to_bucket=new_row_to_bucket, bucket_sizes=new_bucket_sizes)
 
     @staticmethod
+    @line_profiler.profile
     def _build_right_factor(matrix: Matrix, build_left_factor_result: BuildLeftFactorResult) -> Matrix:
         """
         This function essentially computes the value of

cfpq_decomposer/prototype_decomposer.py

@@ -4,6 +4,7 @@
 
 import random
 
+import line_profiler
 from graphblas.binary import plus
 from graphblas.core.dtypes import BOOL, INT32
 from graphblas.core.matrix import Matrix
@@ -26,6 +27,7 @@ class BucketFactor:
 
 
 class PrototypeDecomposer(AbstractDecomposer):
+    @line_profiler.profile
     def row_based_decompose(self, matrix: Matrix) -> Tuple[Matrix, Matrix]:
         row_to_column_sets = self._extract_row_to_column_sets(matrix)
         row_minhash_signatures = self._compute_row_minhash_signatures(row_to_column_sets)
@@ -34,6 +36,7 @@ def row_based_decompose(self, matrix: Matrix) -> Tuple[Matrix, Matrix]:
         return self._build_factor_matrices(bucket_factors, matrix)
 
     @staticmethod
+    @line_profiler.profile
     def _extract_row_to_column_sets(matrix: Matrix) -> Dict[int, Set[int]]:
         row_to_column_sets: Dict[int, Set[int]] = defaultdict(set)
         rows, cols, _ = matrix.to_coo()
@@ -42,6 +45,7 @@ def _extract_row_to_column_sets(matrix: Matrix) -> Dict[int, Set[int]]:
         return row_to_column_sets
 
     @staticmethod
+    @line_profiler.profile
     def _generate_hash_coefficients_and_offsets() -> List[Tuple[int, int]]:
         coefficients_and_offsets: List[Tuple[int, int]] = []
         for _ in range(HASH_FUNCTIONS_COUNT):
@@ -68,6 +72,7 @@ def _compute_row_minhash_signatures(
         return row_minhash_signatures
 
     @staticmethod
+    @line_profiler.profile
     def _group_rows_by_master_hash(
         row_minhash_signatures: Dict[int, Tuple[int, ...]],
     ) -> Dict[int, List[int]]:
@@ -81,6 +86,7 @@ def _group_rows_by_master_hash(
         }
 
     @staticmethod
+    @line_profiler.profile
     def _build_bucket_factors(
         master_hash_to_rows: Dict[int, List[int]],
         row_to_column_sets: Dict[int, Set[int]],
@@ -94,6 +100,7 @@ def _build_bucket_factors(
         return bucket_factors
 
     @staticmethod
+    @line_profiler.profile
     def _filter_rows_by_frequency(
         candidate_rows: List[int],
         row_to_column_sets: Dict[int, Set[int]],
@@ -114,6 +121,7 @@ def _filter_rows_by_frequency(
         return frequency_signature, surviving_rows
 
     @staticmethod
+    @line_profiler.profile
     def _build_bucket_factor(
         bucket_rows: List[int],
         row_to_column_sets: Dict[int, Set[int]],
@@ -142,6 +150,7 @@ def _build_bucket_factor(
         )
 
     @staticmethod
+    @line_profiler.profile
     def _build_factor_matrices(
         bucket_factors: List[BucketFactor],
         matrix: Matrix,

requirements.txt

@@ -9,3 +9,4 @@ pandas==2.0.3
 numpy==1.23.5
 psutil==5.9.8
 tabulate==0.9.0
+line_profiler==4.1.0