def test_squeeze_grad():
data = relay.var("data", shape=(2, 1, 1, 3, 4, 1), dtype="float64")
fwd_func = relay.Function([data], relay.squeeze(data))
fwd_func_subset = relay.Function([data], relay.squeeze(data, axis=[1, -1]))
check_grad(fwd_func)
check_grad(fwd_func_subset)
def test_reshape_like_grad():
data = relay.var("data", shape=(2, 3, 4), dtype="float32")
shape_like = relay.var("shape_like", shape=(6, 2, 2), dtype="float32")
fwd_func = relay.Function([data, shape_like],
relay.reshape_like(data, shape_like))
check_grad(fwd_func)
def verify_max_grad(d_shape, axis=None, keepdims=False, exclude=False):
data = relay.var("data", relay.TensorType(d_shape, "float32"))
fwd_func = relay.Function([data], relay.max(data, axis=axis, keepdims=keepdims, exclude=exclude))
check_grad(fwd_func, scale=1e-3)
def test_stack_grad():
args = [relay.var(c, shape=(2, 3, 4), dtype="float64") for c in "xyz"]
fwd_func = relay.Function(args, relay.stack(args, axis=0))
check_grad(fwd_func)
def test_log_softmax_grad():
data = relay.var("data", relay.TensorType((2, 16), "float64"))
fwd_func = relay.Function([data], relay.nn.log_softmax(data))
check_grad(fwd_func, scale=1)
def verify_bias_add(d_shape, b_shape, axis=1):
data = relay.var("data", relay.TensorType(d_shape, "float32"))
bias = relay.var("bias", relay.TensorType(b_shape, "float32"))
fwd_func = relay.Function([data, bias],
relay.nn.bias_add(data, bias, axis=axis))
check_grad(fwd_func)
def verify_dense_grad(d_shape, w_shape):
data = relay.var("data", relay.TensorType(d_shape, "float32"))
weight = relay.var("weight", relay.TensorType(w_shape, "float32"))
fwd_func = relay.Function([data, weight], relay.nn.dense(data, weight))
check_grad(fwd_func)
def verify_batch_flatten_grad(d_shape):
data = relay.var("data", relay.TensorType(d_shape, "float32"))
fwd_func = relay.Function([data], relay.nn.batch_flatten(data))
check_grad(fwd_func)
def test_copy_grad():
data = relay.var("data", relay.TensorType((10, 4), "float64"))
fwd_func = relay.Function([data], relay.copy(data))
check_grad(fwd_func)
def test_negative_grad():
data = relay.var("data", relay.TensorType((10, 4), "float32"))
fwd_func = relay.Function([data], relay.negative(data))
check_grad(fwd_func)
def verify_transpose_grad(d_shape, axes=None):
data = relay.var("data", relay.TensorType(d_shape, "float32"))
fwd_func = relay.Function([data], relay.transpose(data, axes=axes))
check_grad(fwd_func)
def test_cast_like_grad(executor_kind):
data = relay.var("data", shape=(10, 4), dtype="float32")
like = relay.var("like", shape=(1, ), dtype="float64")
fwd_func = relay.Function([data, like], relay.cast_like(data, like))
check_grad(fwd_func, executor_kind=executor_kind)
def test_cast_grad(executor_kind):
data = relay.var("data", relay.TensorType((10, 4), "float32"))
fwd_func = relay.Function([data], relay.cast(data, "float64"))
check_grad(fwd_func, executor_kind=executor_kind)
