Add GpuDnnBatchNorm tests

theano/sandbox/cuda/tests/test_dnn.py

@@ -715,6 +715,111 @@ class test_DnnSoftMax(test_nnet.test_SoftMax):
         utt.assert_allclose(f(inp), f_ref(inp))
 
 
+def test_batchnorm_train_op():
+    if not cuda.dnn.dnn_available():
+        raise SkipTest(cuda.dnn.dnn_available.msg)
+    if cuda.dnn.version() < (5000, 5000):
+        raise SkipTest("batch normalization requires cudnn v5+")
+    utt.seed_rng()
+
+    for mode in ('per-activation', 'spatial'):
+        x, scale, bias = T.tensor4s('x', 'scale', 'bias')
+        eps = 5e-3  # some non-standard value to test if it's used
+
+        # forward pass
+        batchnorm_op = cuda.dnn.GpuDnnBatchNorm(mode=mode, epsilon=eps)
+        out, x_mean, x_invstd = batchnorm_op(x, scale, bias)
+        # reference forward pass
+        if mode == 'per-activation':
+            axes = (0,)
+        elif mode == 'spatial':
+            axes = (0, 2, 3)
+        x_mean2 = x.mean(axis=axes, keepdims=True)
+        x_invstd2 = T.inv(T.sqrt(x.var(axis=axes, keepdims=True) + eps))
+        scale2 = T.addbroadcast(scale, *axes)
+        bias2 = T.addbroadcast(bias, *axes)
+        out2 = (x - x_mean2) * (scale2 * x_invstd2) + bias2
+        # backward pass
+        dy = T.tensor4('dy')
+        grads = T.grad(None, wrt=[x, scale, bias], known_grads={out: dy})
+        # reference backward pass
+        grads2 = T.grad(None, wrt=[x, scale, bias], known_grads={out2: dy})
+        # compile
+        f = theano.function([x, scale, bias, dy],
+                            [out, x_mean, x_invstd, out2, x_mean2, x_invstd2] +
+                            grads + grads2, mode=mode_with_gpu)
+        # run
+        floatX = theano.config.floatX
+        for data_shape in ((10, 20, 30, 40), (4, 3, 1, 1), (1, 1, 5, 5)):
+            param_shape = tuple(1 if d in axes else s
+                                for d, s in enumerate(data_shape))
+            X = 4 + 3 * numpy.random.randn(*data_shape).astype(floatX)
+            Dy = -1 + 2 * numpy.random.randn(*data_shape).astype(floatX)
+            Scale = numpy.random.randn(*param_shape).astype(floatX)
+            Bias = numpy.random.randn(*param_shape).astype(floatX)
+            outputs = f(X, Scale, Bias, Dy)
+            # compare outputs
+            utt.assert_allclose(outputs[0], outputs[0 + 3])  # out
+            utt.assert_allclose(outputs[1], outputs[1 + 3])  # mean
+            utt.assert_allclose(outputs[2], outputs[2 + 3])  # invstd
+            # compare gradients
+            utt.assert_allclose(outputs[6], outputs[6 + 3])  # dx
+            utt.assert_allclose(outputs[7], outputs[7 + 3], rtol=1e-3)  # dscale
+            utt.assert_allclose(outputs[8], outputs[8 + 3])  # dbias
+
+
+def test_batchnorm_inf_op():
+    if not cuda.dnn.dnn_available():
+        raise SkipTest(cuda.dnn.dnn_available.msg)
+    if cuda.dnn.version() < (5000, 5000):
+        raise SkipTest("batch normalization requires cudnn v5+")
+    utt.seed_rng()
+
+    for mode in ('per-activation', 'spatial'):
+        x, scale, bias, mean, var = T.tensor4s('x', 'scale', 'bias', 'mean', 'var')
+        eps = 5e-3  # some non-standard value to test if it's used
+
+        # forward pass
+        batchnorm_op = cuda.dnn.GpuDnnBatchNormInference(mode=mode, epsilon=eps)
+        out = batchnorm_op(x, scale, bias, mean, var)
+        # reference forward pass
+        if mode == 'per-activation':
+            axes = (0,)
+        elif mode == 'spatial':
+            axes = (0, 2, 3)
+        scale2, bias2, mean2, var2 = (T.addbroadcast(t, *axes)
+                                      for t in (scale, bias, mean, var))
+        out2 = (x - mean2) * (scale2 / T.sqrt(var2 + eps)) + bias2
+        # backward pass
+        dy = T.tensor4('dy')
+        grads = T.grad(None, wrt=[x, scale, bias, mean, var], known_grads={out: dy})
+        # reference backward pass
+        grads2 = T.grad(None, wrt=[x, scale, bias, mean, var], known_grads={out2: dy})
+        # compile
+        f = theano.function([x, scale, bias, mean, var, dy],
+                            [out, out2] + grads + grads2, mode=mode_with_gpu)
+        # run
+        floatX = theano.config.floatX
+        for data_shape in ((10, 20, 30, 40), (4, 3, 1, 1), (1, 1, 5, 5)):
+            param_shape = tuple(1 if d in axes else s
+                                for d, s in enumerate(data_shape))
+            X = 4 + 3 * numpy.random.randn(*data_shape).astype(floatX)
+            Dy = -1 + 2 * numpy.random.randn(*data_shape).astype(floatX)
+            Scale = numpy.random.randn(*param_shape).astype(floatX)
+            Bias = numpy.random.randn(*param_shape).astype(floatX)
+            Mean = numpy.random.randn(*param_shape).astype(floatX)
+            Var = numpy.random.rand(*param_shape).astype(floatX)
+            outputs = f(X, Scale, Bias, Mean, Var, Dy)
+            # compare outputs
+            utt.assert_allclose(outputs[0], outputs[1])  # out
+            # compare gradients
+            utt.assert_allclose(outputs[2], outputs[2 + 5])  # dx
+            utt.assert_allclose(outputs[3], outputs[3 + 5])  # dscale
+            utt.assert_allclose(outputs[4], outputs[4 + 5])  # dbias
+            utt.assert_allclose(outputs[5], outputs[5 + 5])  # dmean
+            utt.assert_allclose(outputs[6], outputs[6 + 5])  # dvar
+
+
 def test_dnn_tag():
     """
     Test that if cudnn isn't avail we crash and that if it is avail, we use it.