Bugfix: _like factories now default to same device as input DNDarray (

#1443) * bugfix * removed print * set appropriate default device in __factory_like * undo explicit device setting * fix cond dtype setting * fix typo * deleting benchmarks 2020 * pacify pre-commit * reinstate original dtype call * bypass numpy() call in test_random * debug test_random * Update CISupport.yml * Update CIBase.yml --------- Co-authored-by: Hoppe <[email protected]> Co-authored-by: Michael Tarnawa <[email protected]> Co-authored-by: Claudia Comito <[email protected]>
helmholtz-analytics · May 7, 2024 · 421c868 · 421c868 · github-actions · May 7, 2024
1 parent f465ef0
commit 421c868
Show file tree

Hide file tree

Showing 12 changed files with 45 additions and 38 deletions.
diff --git a/.github/workflows/CIBase.yml b/.github/workflows/CIBase.yml
@@ -18,11 +18,15 @@ jobs:
         with:
           egress-policy: audit
 
+      - name: Get branch names
+        id: branch-names
+        uses: tj-actions/branch-names@v8
       - name: 'start test'
         run: |
          curl -s -X POST \
            --fail \
            -F token=${{ secrets.CB_PIPELINE }} \
-           -F "ref=heat/base" \
+           -F "ref=heat/support" \
            -F "variables[SHA]=$GITHUB_SHA" \
+           -F "variables[GHBRANCH]=${{ steps.branch-names.outputs.current_branch }}" \
            https://codebase.helmholtz.cloud/api/v4/projects/7605/trigger/pipeline -o /dev/null
diff --git a/.github/workflows/CISupport.yml b/.github/workflows/CISupport.yml
@@ -9,11 +9,15 @@ jobs:
   starter:
     runs-on: ubuntu-latest
     steps:
+      - name: Get branch names
+        id: branch-names
+        uses: tj-actions/branch-names@v8
       - name: 'start test'
         run: |
          curl -s -X POST \
            --fail \
            -F token=${{ secrets.CB_PIPELINE }} \
            -F "ref=heat/support" \
            -F "variables[SHA]=$GITHUB_SHA" \
+           -F "variables[GHBRANCH]=${{ steps.branch-names.outputs.current_branch }}" \
            https://codebase.helmholtz.cloud/api/v4/projects/7605/trigger/pipeline -o /dev/null
diff --git a/benchmarks/2020/__init__.py b/benchmarks/2020/__init__.py
diff --git a/benchmarks/2020/distance_matrix/__init__.py b/benchmarks/2020/distance_matrix/__init__.py
diff --git a/benchmarks/2020/distance_matrix/numpy-cpu.py b/benchmarks/2020/distance_matrix/numpy-cpu.py
@@ -24,4 +24,4 @@
         start = time.perf_counter()
         dist = cdist(data, data)
         end = time.perf_counter()
-        print(f"\t{end-start}s")
+        print(f"\t{end - start}s")
diff --git a/benchmarks/2020/distance_matrix/torch-cpu.py b/benchmarks/2020/distance_matrix/torch-cpu.py
@@ -22,4 +22,4 @@
         start = time.perf_counter()
         dist = torch.cdist(data, data)
         end = time.perf_counter()
-        print(f"\t{end-start}s")
+        print(f"\t{end - start}s")
diff --git a/benchmarks/2020/kmeans/__init__.py b/benchmarks/2020/kmeans/__init__.py
diff --git a/benchmarks/2020/lasso/__init__.py b/benchmarks/2020/lasso/__init__.py
diff --git a/benchmarks/2020/statistical_moments/__init__.py b/benchmarks/2020/statistical_moments/__init__.py
diff --git a/heat/core/factories.py b/heat/core/factories.py
@@ -598,13 +598,13 @@ def empty_like(
 ) -> DNDarray:
     """
     Returns a new uninitialized :class:`~heat.core.dndarray.DNDarray` with the same type, shape and data distribution
-    of given object. Data type and data distribution strategy can be explicitly overriden.
+    of given object. Data type, data distribution axis, and device can be explicitly overridden.
 
     Parameters
     ----------
     a : DNDarray
-        The shape and data-type of ``a`` define these same attributes of the returned array. Uninitialized array with
-        the same shape, type and split axis as ``a`` unless overriden.
+        The shape, data-type, split axis and device of ``a`` define these same attributes of the returned array. Uninitialized array with
+        the same shape, type, split axis and device as ``a`` unless overriden.
     dtype : datatype, optional
         Overrides the data type of the result.
     split: int or None, optional
@@ -794,8 +794,7 @@ def __factory_like(
     factory : function
         Function that creates a DNDarray.
     device : str
-        Specifies the :class:`~heat.core.devices.Device` the array shall be allocated on, defaults to globally set
-        default device.
+        Specifies the :class:`~heat.core.devices.Device` the array shall be allocated on, defaults to the same device as ``a``.
     comm: Communication
         Handle to the nodes holding distributed parts or copies of this array.
     order: str, optional
@@ -834,6 +833,13 @@ def __factory_like(
             # do not split at all
             pass
 
+    # infer the device, otherwise default to a.device
+    if device is None:
+        try:
+            device = a.device
+        except AttributeError:
+            device = devices.get_device()
+
     # use the default communicator, if not set
     comm = sanitize_comm(comm)
 
@@ -1057,21 +1063,20 @@ def full_like(
     order: str = "C",
 ) -> DNDarray:
     """
-    Return a full :class:`~heat.core.dndarray.DNDarray` with the same shape and type as a given array.
+    Return a full :class:`~heat.core.dndarray.DNDarray` with the same shape and type as a given array. Data type, data distribution axis, and device can be explicitly overridden.
 
     Parameters
     ----------
     a : DNDarray
-        The shape and data-type of ``a`` define these same attributes of the returned array.
+        The shape, data-type, split axis and device of ``a`` define these same attributes of the returned array.
     fill_value : scalar
         Fill value.
     dtype : datatype, optional
-        Overrides the data type of the result.
+        The data type of the result, defaults to `a.dtype`.
     split: int or None, optional
-        The axis along which the array is split and distributed; ``None`` means no distribution.
+        The axis along which the array is split and distributed; defaults to `a.split`.
     device : str or Device, optional
-        Specifies the :class:`~heat.core.devices.Device` the array shall be allocated on, defaults to globally set
-        default device.
+        Specifies the :class:`~heat.core.devices.Device` the array shall be allocated on, defaults to `a.device`.
     comm: Communication, optional
         Handle to the nodes holding distributed parts or copies of this array.
     order: str, optional
@@ -1386,19 +1391,18 @@ def ones_like(
 ) -> DNDarray:
     """
     Returns a new :class:`~heat.core.dndarray.DNDarray` filled with ones with the same type,
-    shape and data distribution of given object. Data type and data distribution strategy can be explicitly overriden.
+    shape, data distribution and device of the input object. Data type, data distribution axis, and device can be explicitly overridden.
 
     Parameters
     ----------
     a : DNDarray
-        The shape and data-type of ``a`` define these same attributes of the returned array.
+        The shape, data-type, split axis and device of ``a`` define these same attributes of the returned array.
     dtype : datatype, optional
         Overrides the data type of the result.
     split: int or None, optional
-        The axis along which the array is split and distributed; ``None`` means no distribution.
+        The axis along which the array is split and distributed; defaults to `a.split`.
     device : str or Device, optional
-        Specifies the :class:`~heat.core.devices.Device` the array shall be allocated on, defaults to globally set
-        default device.
+        Specifies the :class:`~heat.core.devices.Device` the array shall be allocated on, defaults to `a.device`.
     comm: Communication, optional
         Handle to the nodes holding distributed parts or copies of this array.
     order: str, optional
@@ -1482,20 +1486,19 @@ def zeros_like(
     order: str = "C",
 ) -> DNDarray:
     """
-    Returns a new :class:`~heat.core.dndarray.DNDarray` filled with zeros with the same type, shape and data
-    distribution of given object. Data type and data distribution strategy can be explicitly overriden.
+    Returns a new :class:`~heat.core.dndarray.DNDarray` filled with zeros with the same type, shape, data
+    distribution, and device of the input object. Data type, data distribution axis, and device can be explicitly overridden.
 
     Parameters
     ----------
     a : DNDarray
-        The shape and data-type of ``a`` define these same attributes of the returned array.
+        The shape, data-type, split axis, and device  of ``a`` define these same attributes of the returned array.
     dtype : datatype, optional
         Overrides the data type of the result.
     split: int or None, optional
-        The axis along which the array is split and distributed; ``None`` means no distribution.
+        The axis along which the array is split and distributed; defaults to `a.split`.
     device : str or Device, optional
-        Specifies the :class:`~heat.core.devices.Device` the array shall be allocated on, defaults to globally set
-        default device.
+        Specifies the :class:`~heat.core.devices.Device` the array shall be allocated on, defaults to `a.device`.
     comm: Communication, optional
         Handle to the nodes holding distributed parts or copies of this array.
     order: str, optional

diff --git a/heat/core/tests/test_random.py b/heat/core/tests/test_random.py
@@ -350,10 +350,9 @@ def test_randn(self):
         shape = (5, 10, 13, 23, 15, 20)
         a = ht.random.randn(*shape, split=0, dtype=ht.float64)
         self.assertEqual(a.dtype, ht.float64)
-        a = a.numpy()
-        mean = np.mean(a)
-        median = np.median(a)
-        std = np.std(a)
+        mean = ht.mean(a)
+        median = ht.median(a)
+        std = ht.std(a)
         self.assertTrue(-0.01 < mean < 0.01)
         self.assertTrue(-0.01 < median < 0.01)
         self.assertTrue(0.99 < std < 1.01)
@@ -362,18 +361,18 @@ def test_randn(self):
         ht.random.seed(54321)
         elements = np.prod(shape)
         b = ht.random.randn(elements, split=0, dtype=ht.float64)
-        b = b.numpy()
         a = a.flatten()
-        self.assertTrue(np.allclose(a, b))
+        self.assertTrue(ht.allclose(a, b))
 
         # Creating the same array two times without resetting seed results in different elements
         c = ht.random.randn(elements, split=0, dtype=ht.float64)
-        c = c.numpy()
         self.assertEqual(c.shape, b.shape)
-        self.assertFalse(np.allclose(b, c))
+        self.assertFalse(ht.allclose(b, c))
 
         # All the created values should be different
-        d = np.concatenate((b, c))
+        d = ht.concatenate((b, c))
+        d.resplit_(None)
+        d = d.numpy()
         _, counts = np.unique(d, return_counts=True)
         self.assertTrue((counts == 1).all())
 
@@ -383,9 +382,6 @@ def test_randn(self):
         ht.random.seed(12345)
         b = ht.random.randn(*shape, split=5, dtype=ht.float64)
         self.assertTrue(ht.equal(a, b))
-        a = a.numpy()
-        b = b.numpy()
-        self.assertTrue(np.allclose(a, b))
 
         # Tests with float32
         ht.random.seed(54321)

diff --git a/scripts/numpy_coverage_tables.py b/scripts/numpy_coverage_tables.py
@@ -548,7 +548,7 @@
 # create Table of Contents
 f.write("## Table of Contents\n")
 for i, header in enumerate(headers):
-    f.write(f"{i+1}. [{headers[header]}](#{headers[header].lower().replace(' ', '-')})\n")
+    f.write(f"{i + 1}. [{headers[header]}](#{headers[header].lower().replace(' ', '-')})\n")
 f.write("\n")
 
 for i, function_list in enumerate(numpy_functions):