don't run unneeded grads

test/test_tensor.py

@@ -16,7 +16,7 @@ class TestTinygrad(unittest.TestCase):
     def test_tinygrad():
       x = Tensor(x_init)
       W = Tensor(W_init)
-      m = Tensor(m_init)
+      m = Tensor(m_init, requires_grad=False)
       out = x.dot(W).relu()
       out = out.logsoftmax()
       out = out.mul(m).add(m).sum()
@@ -64,6 +64,21 @@ class TestTinygrad(unittest.TestCase):
     for x,y in zip(test_tinygrad(), test_pytorch()):
       np.testing.assert_allclose(x, y, atol=1e-5)
 
+  def test_nograd(self):
+    x = Tensor(x_init, requires_grad=False)
+    m = Tensor(m_init, requires_grad=False)
+    W = Tensor(W_init, requires_grad=True)
+    tmp = x.mul(m)
+    mm = tmp.matmul(W)
+    out = mm.relu()
+    out = out.sum()
+    out.backward()
+    assert x.grad is None
+    assert m.grad is None
+    assert tmp.grad is None
+    assert mm.grad is not None
+    assert W.grad is not None
+
   def test_dropout(self):
     Tensor.training = True
     n, rate = 1_000_000, 0.1