Merge pull request #3430 from riya-singh28/unsorted_segment_sum

Unsorted_segment_sum and unittest

deepchem/utils/pytorch_utils.py

@@ -1,7 +1,7 @@
 """Utility functions for working with PyTorch."""
 
 import torch
-from typing import Callable, Union
+from typing import Callable, Union, List
 
 
 def get_activation(fn: Union[Callable, str]):
@@ -14,3 +14,52 @@ def get_activation(fn: Union[Callable, str]):
     if isinstance(fn, str):
         return getattr(torch.nn.functional, fn)
     return fn
+
+
+def unsorted_segment_sum(data: torch.Tensor, segment_ids: torch.Tensor,
+                         num_segments: int) -> torch.Tensor:
+    """Computes the sum along segments of a tensor. Analogous to tf.unsorted_segment_sum.
+
+    Parameters
+    ----------
+    data: torch.Tensor
+        A tensor whose segments are to be summed.
+    segment_ids: torch.Tensor
+        The segment indices tensor.
+    num_segments: int
+        The number of segments.
+
+    Returns
+    -------
+    tensor: torch.Tensor
+
+    Examples
+    --------
+    >>> segment_ids = torch.Tensor([0, 1, 0]).to(torch.int64)
+    >>> data = torch.Tensor([[1, 2, 3, 4], [5, 6, 7, 8], [4, 3, 2, 1]])
+    >>> num_segments = 2
+    >>> result = unsorted_segment_sum(data=data,
+                                  segment_ids=segment_ids,
+                                  num_segments=num_segments)
+    >>> result
+    tensor([[5., 5., 5., 5.],
+        [5., 6., 7., 8.]])
+
+    """
+    # length of segment_ids.shape should be 1
+    assert len(segment_ids.shape) == 1
+
+    # Shape of segment_ids should be equal to first dimension of data
+    assert segment_ids.shape[-1] == data.shape[0]
+
+    s = torch.prod(torch.tensor(data.shape[1:])).long()
+    segment_ids = segment_ids.repeat_interleave(s).view(segment_ids.shape[0],
+                                                        *data.shape[1:])
+
+    # data.shape and segment_ids.shape should be equal
+    assert data.shape == segment_ids.shape
+    shape: List[int] = [num_segments] + list(data.shape[1:])
+    tensor: torch.Tensor = torch.zeros(*shape).scatter_add(
+        0, segment_ids, data.float())
+    tensor = tensor.type(data.dtype)
+    return tensor

deepchem/utils/test/test_pytorch_utils.py

@@ -0,0 +1,29 @@
+import deepchem as dc
+import numpy as np
+import pytest
+try:
+    import torch
+    has_torch = True
+except ModuleNotFoundError:
+    has_torch = False
+
+
+@pytest.mark.torch
+def test_unsorted_segment_sum():
+
+    segment_ids = torch.Tensor([0, 1, 0]).to(torch.int64)
+    data = torch.Tensor([[1, 2, 3, 4], [5, 6, 7, 8], [4, 3, 2, 1]])
+    num_segments = 2
+
+    # length of segment_ids.shape should be 1
+    assert len(segment_ids.shape) == 1
+
+    # Shape of segment_ids should be equal to first dimension of data
+    assert segment_ids.shape[-1] == data.shape[0]
+    result = dc.utils.pytorch_utils.unsorted_segment_sum(
+        data=data, segment_ids=segment_ids, num_segments=num_segments)
+
+    assert np.allclose(
+        np.array(result),
+        np.load("deepchem/utils/test/assets/result_segment_sum.npy"),
+        atol=1e-04)

docs/source/api_reference/utils.rst

@@ -286,3 +286,8 @@ The utilites here are used to create an object that contains information about a
   :members:
 
 .. autofunction:: deepchem.utils.dftutils.hashstr
+
+Pytorch Utilities
+-----------------
+
+.. autofunction:: deepchem.utils.pytorch_utils.unsorted_segment_sum