【Hackathon 7th No.31】NO.31 为 paddle.sparse.sparse_csr_tensor进行功能增强 (#68281)

monster1015 · web-flow · commit d2744a06385f · 2024-09-27T16:35:00.000+08:00
* 【Hackathon 7th No.31】NO.31 为 paddle.sparse.sparse_csr_tensor进行功能增强

* update

* update
diff --git a/python/paddle/sparse/creation.py b/python/paddle/sparse/creation.py
@@ -199,13 +199,29 @@ def sparse_coo_tensor(
         return out
 
 
+def _infer_dense_csr_shape(crows, cols):
+    crows_numpy = crows.numpy()
+    cols_numpy = cols.numpy()
+    batchs = np.sum(crows_numpy[:-1] > crows_numpy[1:]) + 1
+    if (int(len(crows_numpy) / batchs) * batchs) != len(crows_numpy):
+        raise ValueError(
+            f"The calculated original matrix batch size is {batchs}, but it cannot correctly split the row data. Please carefully check the data or the input shape."
+        )
+    col = cols_numpy.max() + 1
+    row = int(len(crows_numpy) / batchs) - 1
+    if batchs == 1:
+        return [row, col]
+    else:
+        return [batchs, row, col]
+
+
 # TODO: need to support shape is None
 @dygraph_only
 def sparse_csr_tensor(
     crows: list[int] | tuple[int, ...] | npt.NDArray[np.int_] | Tensor,
     cols: list[int] | tuple[int, ...] | npt.NDArray[np.int_] | Tensor,
     values: NumbericSequence | npt.NDArray[Any] | Tensor,
-    shape: ShapeLike,
+    shape: ShapeLike | None = None,
     dtype: DTypeLike | None = None,
     place: CPUPlace | CUDAPinnedPlace | CUDAPlace | str | None = None,
     stop_gradient: bool = True,
@@ -268,10 +284,14 @@ def sparse_csr_tensor(
     _check_indices_dtype(crows.dtype)
     _check_indices_dtype(cols.dtype)
 
-    if len(shape) != 2 and len(shape) != 3:
-        raise ValueError(
-            f"SparseCsrTensor only support 2-D or 3-D matrix. but get shape {shape}"
-        )
+    if shape is not None:
+        if len(shape) != 2 and len(shape) != 3:
+            raise ValueError(
+                f"SparseCsrTensor only support 2-D or 3-D matrix. but get shape {shape}"
+            )
+    else:
+        shape = _infer_dense_csr_shape(crows, cols)
+
     rows = shape[len(shape) - 2]
 
     if not crows.place._equals(place):
diff --git a/test/legacy_test/test_sparse_utils_op.py b/test/legacy_test/test_sparse_utils_op.py
@@ -125,6 +125,37 @@ def test_create_coo_no_shape(self):
         coo = paddle.sparse.sparse_coo_tensor(indices, values)
         assert [2, 2] == coo.shape
 
+    def test_create_csr_no_shape(self):
+        # 2D sparse tensor
+        crows = [0, 2, 3, 5]
+        clos = [1, 3, 2, 0, 1]
+        values = [1.0, 2.0, 3.0, 4.0, 5.0]
+        crows = paddle.to_tensor(crows, dtype='int32')
+        clos = paddle.to_tensor(clos, dtype='int32')
+        values = paddle.to_tensor(values, dtype='float32')
+        csr = paddle.sparse.sparse_csr_tensor(crows, clos, values)
+        assert [3, 4] == csr.shape
+
+        # 3D sparse tensor
+        crows = [0, 2, 2, 0, 1, 1, 0, 0, 0]
+        clos = [0, 1, 1]
+        values = [1, 2, 5]
+        crows = paddle.to_tensor(crows, dtype='int32')
+        clos = paddle.to_tensor(clos, dtype='int32')
+        values = paddle.to_tensor(values, dtype='float32')
+        csr = paddle.sparse.sparse_csr_tensor(crows, clos, values)
+        assert [3, 2, 2] == csr.shape
+
+        # 3D sparse tensor
+        crows = [0, 1, 2, 0, 1, 1, 0, 1, 2]
+        clos = [0, 2, 1, 0, 1]
+        values = [1, 2, 3, 4, 5]
+        crows = paddle.to_tensor(crows, dtype='int32')
+        clos = paddle.to_tensor(clos, dtype='int32')
+        values = paddle.to_tensor(values, dtype='float32')
+        csr = paddle.sparse.sparse_csr_tensor(crows, clos, values)
+        assert [3, 2, 3] == csr.shape
+
 
 class TestSparseConvert(unittest.TestCase):
     def test_to_sparse_coo(self):
@@ -489,6 +520,16 @@ def test_dtype(self):
                 crows, cols, values, shape
             )
 
+    def test_error_crows(self):
+        with self.assertRaises(ValueError):
+            crows = [0, 2, 2, 0, 1, 1, 0, 0, 0, 0]
+            clos = [0, 1, 1]
+            values = [1, 2, 5]
+            crows = paddle.to_tensor(crows, dtype='int32')
+            clos = paddle.to_tensor(clos, dtype='int32')
+            values = paddle.to_tensor(values, dtype='float32')
+            coo = paddle.sparse.sparse_csr_tensor(crows, clos, values)
+
 
 devices = []
 if paddle.device.get_device() != "cpu":
