Add test for create random permute function

torchhd/functional.py

@@ -716,7 +716,10 @@ def permute(input: VSATensor, *, shifts=1) -> VSATensor:
     return input.permute(shifts)
 
 
-def create_random_permute(dim: int) -> Callable[[VSATensor, int], VSATensor]:
+class create_random_permute(torch.nn.Module):
+    forward_indices: LongTensor
+    backward_indices: LongTensor
+
     r"""Creates random permutation functions.
 
     Args:
@@ -741,21 +744,28 @@ def create_random_permute(dim: int) -> Callable[[VSATensor, int], VSATensor]:
 
     """
 
-    forward = torch.randperm(dim)
-    backward = torch.empty_like(forward)
-    backward[forward] = torch.arange(dim)
+    def __init__(self, dim: int) -> None:
+        super().__init__()
+
+        forward = torch.randperm(dim)
+        backward = torch.empty_like(forward)
+        backward[forward] = torch.arange(dim)
+
+        self.register_buffer("forward_indices", forward)
+        self.register_buffer("backward_indices", backward)
 
-    def permute(input: VSATensor, shifts: int = 1) -> VSATensor:
+    def __call__(self, input: VSATensor, shifts: int = 1) -> VSATensor:
         y = input
+
         if shifts > 0:
-            for _ in range(shifts):
-                y = y[..., forward]
+            for _ in range(abs(shifts)):
+                y = y[..., self.forward_indices]
+
         elif shifts < 0:
-            for _ in range(shifts):
-                y = y[..., backward]
-        return y
+            for _ in range(abs(shifts)):
+                y = y[..., self.backward_indices]
 
-    return permute
+        return y.clone()
 
 
 def inverse(input: VSATensor) -> VSATensor:

torchhd/structures.py

@@ -944,7 +944,7 @@ def node_neighbors(self, input: VSATensor, outgoing=True) -> VSATensor:
             return functional.bind(self.value, input.inverse())
 
     def contains(self, input: VSATensor) -> Tensor:
-        """Returns the cosine similarity of the input vector against the graph.
+        """Returns the normalized dot similarity of the input vector against the graph.
 
         Args:
             input (Tensor): Hypervector to compare against the multiset.
@@ -955,7 +955,7 @@ def contains(self, input: VSATensor) -> Tensor:
             >>> G.contains(e)
             tensor(1.)
         """
-        return functional.cosine_similarity(input, self.value)
+        return functional.dot_similarity(input, self.value) / self.value.size(-1)
 
     def clear(self) -> None:
         """Empties the graph.

torchhd/tests/basis_hv/test_circular_hv.py

@@ -135,7 +135,7 @@ def test_device(self, dtype):
 
         device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
         hv = functional.circular(3, 52, device=device, dtype=dtype)
-        assert hv.device == device
+        assert hv.device.type == device.type
 
     @pytest.mark.parametrize("dtype", torch_dtypes)
     @pytest.mark.parametrize("vsa", vsa_tensors)
@@ -148,7 +148,7 @@ def test_device(self, dtype, vsa):
 
         device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
         hv = functional.circular(3, 52, vsa, device=device, dtype=dtype)
-        assert hv.device == device
+        assert hv.device.type == device.type
 
     def test_uses_default_dtype(self):
         hv = functional.circular(3, 52, "BSC")

torchhd/tests/basis_hv/test_empty_hv.py

@@ -77,7 +77,7 @@ def test_device(self, dtype, vsa):
 
         device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
         hv = functional.empty(3, 52, vsa, device=device, dtype=dtype)
-        assert hv.device == device
+        assert hv.device.type == device.type
 
     def test_uses_default_dtype(self):
         hv = functional.empty(3, 52, "BSC")

torchhd/tests/basis_hv/test_identity_hv.py

@@ -82,7 +82,7 @@ def test_device(self, dtype, vsa):
 
         device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
         hv = functional.identity(3, 52, vsa, device=device, dtype=dtype)
-        assert hv.device == device
+        assert hv.device.type == device.type
 
     def test_uses_default_dtype(self):
         hv = functional.identity(3, 52, "BSC")

torchhd/tests/basis_hv/test_level_hv.py

@@ -29,7 +29,7 @@
 
 from ..utils import torch_dtypes, supported_dtype, vsa_tensors
 
-seed = 2147483644
+seed = 2147483643
 
 
 class Testlevel:
@@ -129,7 +129,7 @@ def test_device(self, dtype, vsa):
 
         device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
         hv = functional.level(3, 52, vsa, device=device, dtype=dtype)
-        assert hv.device == device
+        assert hv.device.type == device.type
 
     def test_uses_default_dtype(self):
         hv = functional.level(3, 52, "BSC")

torchhd/tests/basis_hv/test_random_hv.py

@@ -137,7 +137,7 @@ def test_device(self, dtype, vsa):
 
         device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
         hv = functional.random(3, 52, vsa, device=device, dtype=dtype)
-        assert hv.device == device
+        assert hv.device.type == device.type
 
     def test_uses_default_dtype(self):
         hv = functional.random(3, 52, "BSC")

torchhd/tests/structures/test_distinct_sequence.py

@@ -34,15 +34,15 @@
 class TestBindSequence:
     def test_creation_dim(self):
         S = structures.BindSequence(10000)
-        assert torch.equal(S.value, torch.ones(10000))
+        assert torch.allclose(S.value, torch.ones(10000))
 
     def test_creation_tensor(self):
         generator = torch.Generator()
         generator.manual_seed(seed)
         hv = functional.random(len(letters), 10000, generator=generator)
 
         S = structures.BindSequence(hv[0])
-        assert torch.equal(S.value, hv[0])
+        assert torch.allclose(S.value, hv[0])
 
     def test_generator(self):
         generator = torch.Generator()

torchhd/tests/structures/test_fsa.py

@@ -26,6 +26,7 @@
 import string
 
 from torchhd import structures, functional
+from torchhd import MAPTensor
 
 seed = 2147483644
 seed1 = 2147483643
@@ -35,7 +36,7 @@
 class TestFSA:
     def test_creation_dim(self):
         F = structures.FiniteStateAutomata(10000)
-        assert torch.equal(F.value, torch.zeros(10000))
+        assert torch.allclose(F.value, torch.zeros(10000))
 
     def test_generator(self):
         generator = torch.Generator()
@@ -49,37 +50,81 @@ def test_generator(self):
         assert (hv1 == hv2).min().item()
 
     def test_add_transition(self):
-        generator = torch.Generator()
-        generator1 = torch.Generator()
-        generator.manual_seed(seed)
-        generator1.manual_seed(seed1)
-        tokens = functional.random(10, 10, generator=generator)
-        actions = functional.random(10, 10, generator=generator1)
+        tokens = MAPTensor(
+            [
+                [1.0, 1.0, -1.0, 1.0, 1.0, 1.0, -1.0, 1.0, -1.0, 1.0],
+                [1.0, -1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, -1.0, 1.0],
+                [-1.0, 1.0, -1.0, -1.0, 1.0, -1.0, 1.0, -1.0, 1.0, -1.0],
+                [-1.0, -1.0, 1.0, 1.0, -1.0, -1.0, 1.0, -1.0, 1.0, -1.0],
+                [1.0, -1.0, 1.0, -1.0, 1.0, -1.0, -1.0, 1.0, -1.0, -1.0],
+                [1.0, 1.0, 1.0, 1.0, -1.0, -1.0, -1.0, -1.0, 1.0, -1.0],
+                [1.0, 1.0, 1.0, 1.0, -1.0, 1.0, 1.0, 1.0, 1.0, -1.0],
+                [-1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, -1.0, -1.0, 1.0],
+                [-1.0, -1.0, 1.0, -1.0, -1.0, -1.0, 1.0, -1.0, 1.0, 1.0],
+                [-1.0, -1.0, 1.0, -1.0, 1.0, 1.0, 1.0, -1.0, -1.0, 1.0],
+            ]
+        )
+        states = MAPTensor(
+            [
+                [1.0, 1.0, -1.0, -1.0, 1.0, 1.0, -1.0, -1.0, 1.0, 1.0],
+                [-1.0, -1.0, -1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, -1.0],
+                [1.0, 1.0, -1.0, -1.0, 1.0, -1.0, 1.0, 1.0, 1.0, 1.0],
+                [1.0, -1.0, 1.0, -1.0, 1.0, -1.0, -1.0, -1.0, 1.0, 1.0],
+                [1.0, 1.0, 1.0, 1.0, -1.0, -1.0, -1.0, -1.0, -1.0, 1.0],
+                [-1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, -1.0, 1.0, -1.0],
+                [1.0, -1.0, -1.0, -1.0, 1.0, 1.0, 1.0, -1.0, -1.0, -1.0],
+                [1.0, -1.0, 1.0, 1.0, 1.0, -1.0, -1.0, -1.0, -1.0, -1.0],
+                [1.0, 1.0, 1.0, -1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0],
+                [1.0, 1.0, 1.0, -1.0, 1.0, -1.0, 1.0, -1.0, 1.0, 1.0],
+            ]
+        )
 
         F = structures.FiniteStateAutomata(10)
 
-        F.add_transition(tokens[0], actions[1], actions[2])
-        assert torch.equal(
+        F.add_transition(tokens[0], states[1], states[2])
+        assert torch.allclose(
             F.value,
-            torch.tensor([1.0, 1.0, -1.0, 1.0, 1.0, 1.0, -1.0, -1.0, -1.0, 1.0]),
+            MAPTensor([-1.0, -1.0, 1.0, -1.0, -1.0, 1.0, 1.0, 1.0, -1.0, -1.0]),
         )
-        F.add_transition(tokens[1], actions[1], actions[3])
-        assert torch.equal(
-            F.value, torch.tensor([0.0, 0.0, -2.0, 2.0, 0.0, 2.0, 0.0, -2.0, -2.0, 0.0])
+        F.add_transition(tokens[1], states[1], states[3])
+        assert torch.allclose(
+            F.value, MAPTensor([-2.0, 0.0, 2.0, 0.0, -2.0, 2.0, 0.0, 0.0, 0.0, -2.0])
         )
-        F.add_transition(tokens[2], actions[1], actions[3])
-        assert torch.equal(
+        F.add_transition(tokens[2], states[1], states[3])
+        assert torch.allclose(
             F.value,
-            torch.tensor([1.0, 1.0, -3.0, 1.0, 1.0, 3.0, -1.0, -1.0, -1.0, 1.0]),
+            MAPTensor([-1.0, -1.0, 1.0, -1.0, -3.0, 1.0, -1.0, 1.0, -1.0, -1.0]),
         )
 
     def test_transition(self):
-        generator = torch.Generator()
-        generator1 = torch.Generator()
-        generator.manual_seed(seed)
-        generator1.manual_seed(seed1)
-        tokens = functional.random(10, 10, generator=generator)
-        states = functional.random(10, 10, generator=generator1)
+        tokens = MAPTensor(
+            [
+                [1.0, 1.0, -1.0, 1.0, 1.0, 1.0, -1.0, 1.0, -1.0, 1.0],
+                [1.0, -1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, -1.0, 1.0],
+                [-1.0, 1.0, -1.0, -1.0, 1.0, -1.0, 1.0, -1.0, 1.0, -1.0],
+                [-1.0, -1.0, 1.0, 1.0, -1.0, -1.0, 1.0, -1.0, 1.0, -1.0],
+                [1.0, -1.0, 1.0, -1.0, 1.0, -1.0, -1.0, 1.0, -1.0, -1.0],
+                [1.0, 1.0, 1.0, 1.0, -1.0, -1.0, -1.0, -1.0, 1.0, -1.0],
+                [1.0, 1.0, 1.0, 1.0, -1.0, 1.0, 1.0, 1.0, 1.0, -1.0],
+                [-1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, -1.0, -1.0, 1.0],
+                [-1.0, -1.0, 1.0, -1.0, -1.0, -1.0, 1.0, -1.0, 1.0, 1.0],
+                [-1.0, -1.0, 1.0, -1.0, 1.0, 1.0, 1.0, -1.0, -1.0, 1.0],
+            ]
+        )
+        states = MAPTensor(
+            [
+                [1.0, 1.0, -1.0, -1.0, 1.0, 1.0, -1.0, -1.0, 1.0, 1.0],
+                [-1.0, -1.0, -1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, -1.0],
+                [1.0, 1.0, -1.0, -1.0, 1.0, -1.0, 1.0, 1.0, 1.0, 1.0],
+                [1.0, -1.0, 1.0, -1.0, 1.0, -1.0, -1.0, -1.0, 1.0, 1.0],
+                [1.0, 1.0, 1.0, 1.0, -1.0, -1.0, -1.0, -1.0, -1.0, 1.0],
+                [-1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, -1.0, 1.0, -1.0],
+                [1.0, -1.0, -1.0, -1.0, 1.0, 1.0, 1.0, -1.0, -1.0, -1.0],
+                [1.0, -1.0, 1.0, 1.0, 1.0, -1.0, -1.0, -1.0, -1.0, -1.0],
+                [1.0, 1.0, 1.0, -1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0],
+                [1.0, 1.0, 1.0, -1.0, 1.0, -1.0, 1.0, -1.0, 1.0, 1.0],
+            ]
+        )
 
         F = structures.FiniteStateAutomata(10)
 
@@ -121,6 +166,6 @@ def test_clear(self):
         F.add_transition(tokens[2], states[1], states[5])
 
         F.clear()
-        assert torch.equal(
+        assert torch.allclose(
             F.value, torch.tensor([0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0])
         )

torchhd/tests/structures/test_graph.py

@@ -28,22 +28,22 @@
 from torchhd import structures, functional
 from torchhd.tensors.map import MAPTensor
 
-seed = 2147483644
+seed = 2147483645
 letters = list(string.ascii_lowercase)
 
 
 class TestGraph:
     def test_creation_dim(self):
         G = structures.Graph(10000, directed=True)
-        assert torch.equal(G.value, torch.zeros(10000))
+        assert torch.allclose(G.value, torch.zeros(10000))
 
     def test_creation_tensor(self):
         generator = torch.Generator()
         generator.manual_seed(seed)
         hv = functional.random(len(letters), 10000, generator=generator)
         g = functional.bind(hv[0], hv[1])
         G = structures.Graph(g)
-        assert torch.equal(G.value, g)
+        assert torch.allclose(G.value, g)
 
     def test_generator(self):
         generator = torch.Generator()
@@ -68,22 +68,22 @@ def test_add_edge(self):
         ).as_subclass(MAPTensor)
 
         G.add_edge(hv[0], hv[1])
-        assert torch.equal(
+        assert torch.allclose(
             G.value, torch.tensor([-1.0, -1.0, -1.0, 1.0, -1.0, -1.0, 1.0, -1.0])
         )
         G.add_edge(hv[2], hv[3])
-        assert torch.equal(
+        assert torch.allclose(
             G.value, torch.tensor([-2.0, -2.0, 0.0, 2.0, -2.0, 0.0, 2.0, -2.0])
         )
 
         GD = structures.Graph(8, directed=True)
 
         GD.add_edge(hv[0], hv[1])
-        assert torch.equal(
+        assert torch.allclose(
             GD.value, torch.tensor([-1.0, 1.0, -1.0, -1.0, -1.0, -1.0, 1.0, -1.0])
         )
         GD.add_edge(hv[2], hv[3])
-        assert torch.equal(
+        assert torch.allclose(
             GD.value, torch.tensor([0.0, 0.0, 0.0, -2.0, 0.0, -2.0, 2.0, -2.0])
         )
 
@@ -99,23 +99,23 @@ def test_encode_edge(self):
         ).as_subclass(MAPTensor)
 
         e1 = G.encode_edge(hv[0], hv[1])
-        assert torch.equal(
+        assert torch.allclose(
             e1, torch.tensor([-1.0, -1.0, -1.0, 1.0, -1.0, -1.0, 1.0, -1.0])
         )
         e2 = G.encode_edge(hv[2], hv[3])
-        assert torch.equal(
+        assert torch.allclose(
             e2, torch.tensor([-1.0, -1.0, 1.0, 1.0, -1.0, 1.0, 1.0, -1.0])
         )
 
         GD = structures.Graph(8, directed=True)
 
         e1 = GD.encode_edge(hv[0], hv[1])
-        assert torch.equal(
+        assert torch.allclose(
             e1, torch.tensor([-1.0, 1.0, -1.0, -1.0, -1.0, -1.0, 1.0, -1.0])
         )
         e2 = GD.encode_edge(hv[2], hv[3])
         print(e2)
-        assert torch.equal(
+        assert torch.allclose(
             e2, torch.tensor([1.0, -1.0, 1.0, -1.0, 1.0, -1.0, 1.0, -1.0])
         )
 
@@ -157,8 +157,8 @@ def test_node_neighbors(self):
     def test_contains(self):
         generator = torch.Generator()
         generator.manual_seed(seed)
-        hv = functional.random(4, 8, generator=generator)
-        G = structures.Graph(8)
+        hv = functional.random(4, 1000, generator=generator)
+        G = structures.Graph(1000)
 
         e1 = G.encode_edge(hv[0], hv[1])
         e2 = G.encode_edge(hv[0], hv[2])
@@ -172,7 +172,7 @@ def test_contains(self):
         assert G.contains(e2) > torch.tensor([0.6])
         assert G.contains(e3) < torch.tensor(0.6)
 
-        GD = structures.Graph(8, directed=True)
+        GD = structures.Graph(1000, directed=True)
 
         ee1 = GD.encode_edge(hv[0], hv[1])
         ee2 = GD.encode_edge(hv[0], hv[2])
@@ -200,16 +200,16 @@ def test_clear(self):
 
         G.clear()
 
-        assert torch.equal(
+        assert torch.allclose(
             G.value, torch.tensor([0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0])
         )
 
     def test_from_edges(self):
         generator = torch.Generator()
         generator.manual_seed(seed)
 
-        hv = functional.random(4, 8, generator=generator)
-        edges = torch.empty(2, 3, 8).as_subclass(MAPTensor)
+        hv = functional.random(4, 1000, generator=generator)
+        edges = torch.empty(2, 3, 1000).as_subclass(MAPTensor)
         edges[0, 0] = hv[0]
         edges[1, 0] = hv[1]
         edges[0, 1] = hv[0]