Merge branch 'master' of github.com:jonasrauber/eagerpy

README.rst

@@ -1,6 +1,9 @@
 .. image:: https://badge.fury.io/py/eagerpy.svg
     :target: https://badge.fury.io/py/eagerpy
 
+.. image:: https://codecov.io/gh/jonasrauber/eagerpy/branch/master/graph/badge.svg
+  :target: https://codecov.io/gh/jonasrauber/eagerpy
+
 .. image:: https://img.shields.io/badge/code%20style-black-000000.svg
     :target: https://github.com/ambv/black
 

eagerpy/framework.py

@@ -167,8 +167,6 @@ def tile(t, multiples):
 
 
 def matmul(x, y):
-    if not istensor(x):
-        return y.matmul(x)
     return x.matmul(y)
 
 

eagerpy/tensor/base.py

@@ -115,6 +115,16 @@ def __truediv__(self, other):
     def __rtruediv__(self, other):
         return self.tensor.__rtruediv__(other)
 
+    @unwrapin
+    @wrapout
+    def __floordiv__(self, other):
+        return self.tensor.__floordiv__(other)
+
+    @unwrapin
+    @wrapout
+    def __rfloordiv__(self, other):
+        return self.tensor.__rfloordiv__(other)
+
     @unwrapin
     @wrapout
     def __lt__(self, other):
@@ -185,8 +195,6 @@ def ndim(self):
 
     @property
     def T(self):
-        if self.ndim < 2:
-            return self
         return self.transpose()
 
     def value_and_grad(self, f, *args, **kwargs):

eagerpy/tensor/jax.py

@@ -154,7 +154,7 @@ def onehot_like(self, indices, *, value=1):
         assert indices.ndim == 1
         x = self.backend.arange(self.tensor.shape[1]).reshape(1, -1)
         indices = indices.reshape(-1, 1)
-        return x == indices
+        return (x == indices) * value
 
     @wrapout
     def from_numpy(self, a):

tests/test_main.py

@@ -70,6 +70,30 @@ def test_astensor_tensor(t):
     assert (ep.astensor(t) == t).all()
 
 
+def test_module():
+    assert ep.istensor(ep.numpy.tanh([3, 5]))
+    assert not ep.istensor(ep.numpy.tanh(3))
+
+
+def test_module_dir():
+    assert "zeros" in dir(ep.numpy)
+
+
+def test_repr(t):
+    assert not repr(t).startswith("<")
+
+
+def test_format(dummy):
+    t = ep.arange(dummy, 5).sum()
+    return f"{t:.1f}" == "10.0"
+
+
+@compare_equal
+def test_item(t):
+    t = t.sum()
+    return t.item()
+
+
 @compare_equal
 def test_len(t):
     return len(t)
@@ -146,25 +170,45 @@ def test_rmul_scalar(t):
 
 
 @compare_allclose
-def test_div(t1, t2):
+def test_truediv(t1, t2):
     return t1 / t2
 
 
 @compare_allclose(rtol=1e-6)
-def test_div_scalar(t):
+def test_truediv_scalar(t):
     return t / 3
 
 
 @compare_allclose
-def test_rdiv_scalar(t):
+def test_rtruediv_scalar(t):
     return 3 / (abs(t) + 1e-8)
 
 
+@compare_allclose
+def test_floordiv(t1, t2):
+    return t1 // t2
+
+
+@compare_allclose(rtol=1e-6)
+def test_floordiv_scalar(t):
+    return t // 3
+
+
+@compare_allclose
+def test_rfloordiv_scalar(t):
+    return 3 // (abs(t) + 1e-8)
+
+
 @compare_all
 def test_getitem(t):
     return t[2]
 
 
+def test_getitem_tuple(dummy):
+    t = ep.arange(dummy, 8).float32().reshape((2, 4))
+    return t[1, 3]
+
+
 @compare_all
 def test_getitem_slice(t):
     return t[1:3]
@@ -386,11 +430,29 @@ def test_logical_and(t):
     return ep.logical_and(t < 3, t > 1)
 
 
+@compare_all
+def test_logical_and_scalar(t):
+    return ep.logical_and(True, t < 3)
+
+
+def test_logical_and_manual(t):
+    assert (ep.logical_and(t < 3, ep.ones_like(t).bool()) == (t < 3)).all()
+
+
 @compare_all
 def test_logical_or(t):
     return ep.logical_or(t > 3, t < 1)
 
 
+@compare_all
+def test_logical_or_scalar(t):
+    return ep.logical_or(True, t < 1)
+
+
+def test_logical_or_manual(t):
+    assert (ep.logical_or(t < 3, ep.zeros_like(t).bool()) == (t < 3)).all()
+
+
 @compare_all
 def test_logical_not(t):
     return ep.logical_not(t > 3)
@@ -440,6 +502,14 @@ def test_full_like(t):
     return ep.full_like(t, 5)
 
 
+@pytest.mark.parametrize("value", [1, -1, 2])
+@compare_all
+def test_onehot_like(dummy, value):
+    t = ep.arange(dummy, 18).float32().reshape((6, 3))
+    indices = ep.arange(t, 6) // 2
+    return ep.onehot_like(t, indices, value=value)
+
+
 @compare_all
 def test_zeros_scalar(t):
     return ep.zeros(t, 5)
@@ -496,10 +566,16 @@ def test_argsort(t):
 
 
 @compare_all
-def test_transpose(t):
+def test_transpose(dummy):
+    t = ep.arange(dummy, 8).float32().reshape((2, 4))
     return ep.transpose(t)
 
 
+def test_transpose_1d(dummy):
+    t = ep.arange(dummy, 8).float32()
+    assert (ep.transpose(t) == t).all()
+
+
 @compare_all
 def test_transpose_axes(dummy):
     t = ep.arange(dummy, 60).float32().reshape((3, 4, 5))
@@ -578,7 +654,7 @@ def test_expand_dims(t, axis):
     return ep.expand_dims(t, axis)
 
 
-@pytest.mark.parametrize("axis", [0, 1, (0, 1)])
+@pytest.mark.parametrize("axis", [None, 0, 1, (0, 1)])
 @compare_all
 def test_squeeze(t, axis):
     t = t.expand_dims(axis=0).expand_dims(axis=1)

tests/test_norms.py

@@ -24,15 +24,16 @@ def test_1d(x1d, p):
     assert_allclose(norms[p](x1d).numpy(), norm(x1d.numpy(), ord=p))
 
 
-@pytest.mark.parametrize("p", [0, 1, 2, ep.inf])
+@pytest.mark.parametrize("p", [0, 1, 2, 3, 4, ep.inf])
 @pytest.mark.parametrize("axis", [0, 1, -1])
 @pytest.mark.parametrize("keepdims", [False, True])
 def test_2d(x2d, p, axis, keepdims):
     assert_allclose(
         lp(x2d, p, axis=axis, keepdims=keepdims).numpy(),
         norm(x2d.numpy(), ord=p, axis=axis, keepdims=keepdims),
     )
-    assert_allclose(
-        norms[p](x2d, axis=axis, keepdims=keepdims).numpy(),
-        norm(x2d.numpy(), ord=p, axis=axis, keepdims=keepdims),
-    )
+    if p in norms:
+        assert_allclose(
+            norms[p](x2d, axis=axis, keepdims=keepdims).numpy(),
+            norm(x2d.numpy(), ord=p, axis=axis, keepdims=keepdims),
+        )