feat(tests): add tests for fixed point linear layer

elasticai/creator/nn/fixed_point/linear/layer/linear_test.py

@@ -0,0 +1,56 @@
+import torch
+
+from .linear import Linear
+
+
+def test_inference_of_multidimensional_data() -> None:
+    linear = Linear(
+        total_bits=16, frac_bits=8, in_features=3, out_features=2, bias=False
+    )
+    linear.weight.data = torch.ones_like(linear.weight.data)
+
+    inputs = torch.tensor([1.0, 2.0, 3.0])
+    expected = [6.0, 6.0]
+    actual = linear(inputs).tolist()
+
+    assert expected == actual
+
+
+def test_overflow_behaviour() -> None:
+    linear = Linear(
+        total_bits=4, frac_bits=1, in_features=2, out_features=1, bias=False
+    )
+    linear.weight.data = torch.ones_like(linear.weight.data) * 2
+
+    inputs = torch.tensor([2.5, -1.0])
+    expected = [3.0]  # quantize(2.5 * 2 - 1.0 * 2)
+    actual = linear(inputs).tolist()
+
+    assert expected == actual
+
+
+def test_underflow_behaviour() -> None:
+    linear = Linear(
+        total_bits=4, frac_bits=1, in_features=1, out_features=1, bias=False
+    )
+    linear.weight.data = torch.ones_like(linear.weight.data) * 0.5
+
+    inputs = torch.tensor([0.5])
+    expected = [0.0]
+    actual = linear(inputs).tolist()
+
+    assert expected == actual
+
+
+def test_bias_addition() -> None:
+    linear = Linear(
+        total_bits=16, frac_bits=8, in_features=1, out_features=1, bias=True
+    )
+    linear.weight.data = torch.ones_like(linear.weight.data)
+    linear.bias.data = torch.ones_like(linear.bias.data) * 2
+
+    inputs = torch.tensor([3.0])
+    expected = [5.0]
+    actual = linear(inputs).tolist()
+
+    assert expected == actual