set nil as the initial value of the struct reference type fields (#51)

autograd/gates_basic.v

@@ -75,8 +75,8 @@ pub fn (g &SubstractGate[T]) cache[T](mut result Variable[T], args ...CacheParam
 
 pub struct MultiplyGate[T] {
 pub:
-	a &Variable[T]
-	b &Variable[T]
+	a &Variable[T] = unsafe { nil }
+	b &Variable[T] = unsafe { nil }
 }
 
 pub fn multiply_gate[T](a &Variable[T], b &Variable[T]) &MultiplyGate[T] {
@@ -119,8 +119,8 @@ pub fn (g &MultiplyGate[T]) cache[T](mut result Variable[T], args ...CacheParam)
 
 pub struct DivideGate[T] {
 pub:
-	a &Variable[T]
-	b &Variable[T]
+	a &Variable[T] = unsafe { nil }
+	b &Variable[T] = unsafe { nil }
 }
 
 pub fn divide_gate[T](a &Variable[T], b &Variable[T]) &DivideGate[T] {

autograd/gates_blas.v

@@ -5,8 +5,8 @@ import vtl.la
 
 pub struct MatMulGate[T] {
 pub:
-	a &Variable[T]
-	b &Variable[T]
+	a &Variable[T] = unsafe { nil }
+	b &Variable[T] = unsafe { nil }
 }
 
 pub fn matmul_gate[T](a &Variable[T], b &Variable[T]) &MatMulGate[T] {

autograd/gates_exp.v

@@ -4,7 +4,7 @@ import vtl
 
 pub struct ExpGate[T] {
 pub:
-	a &Variable[T]
+	a &Variable[T] = unsafe { nil }
 }
 
 pub fn exp_gate[T](a &Variable[T]) &ExpGate[T] {

autograd/gates_pow.v

@@ -4,8 +4,8 @@ import math
 import vtl
 
 pub struct PowGate[T] {
-	a &Variable[T]
-	b &Variable[T]
+	a &Variable[T] = unsafe { nil }
+	b &Variable[T] = unsafe { nil }
 }
 
 pub fn pow_gate[T](a &Variable[T], b &Variable[T]) &PowGate[T] {

autograd/gates_trig.v

@@ -4,7 +4,7 @@ import vtl
 
 pub struct SinGate[T] {
 pub:
-	a &Variable[T]
+	a &Variable[T] = unsafe { nil }
 }
 
 pub fn sin_gate[T](a &Variable[T]) &SinGate[T] {
@@ -37,7 +37,7 @@ pub fn (g &SinGate[T]) cache[T](mut result Variable[T], args ...CacheParam) ! {
 
 pub struct CosGate[T] {
 pub:
-	a &Variable[T]
+	a &Variable[T] = unsafe { nil }
 }
 
 pub fn cos_gate[T](a &Variable[T]) &CosGate[T] {
@@ -70,7 +70,7 @@ pub fn (g &CosGate[T]) cache[T](mut result Variable[T], args ...CacheParam) ! {
 
 pub struct TanGate[T] {
 pub:
-	a &Variable[T]
+	a &Variable[T] = unsafe { nil }
 }
 
 pub fn tan_gate[T](a &Variable[T]) &TanGate[T] {

autograd/node.v

@@ -13,7 +13,7 @@ pub mut:
 	// The variables that created this node
 	parents []&Variable[T]
 	// Wrapper around a Tensor, contains operation data
-	payload &Payload[T]
+	payload &Payload[T] = unsafe { nil }
 	// Debug use only, contains a name for a node
 	name string
 }

autograd/payload.v

@@ -7,7 +7,7 @@ module autograd
 pub struct Payload[T] {
 pub:
 	// Contents of the paylod
-	variable &Variable[T]
+	variable &Variable[T] = unsafe { nil }
 }
 
 pub fn payload[T](variable &Variable[T]) &Payload[T] {

autograd/variable.v

@@ -15,15 +15,15 @@ pub mut:
 	// The value of the Variable.  This should not be edited outside
 	// of Variable operations, as other edits will not be tracked
 	// and will lead to incorrect results
-	value &vtl.Tensor[T]
+	value &vtl.Tensor[T] = unsafe { nil }
 	// The graph the variable is associated with.  This is a reference,
 	// as a variable does not own its context
-	context &Context[T]
+	context &Context[T] = unsafe { nil }
 	// The gradient of the Variable.  This is set as a reference to
 	// the value of a Variable unless `backprop` has been called, in
 	// which case all related Variables will have their gradient
 	// updated correctly
-	grad &vtl.Tensor[T]
+	grad &vtl.Tensor[T] = unsafe { nil }
 	// If set to true, this variable will track its operations,
 	// otherwise it will act similar to a vtl.Tensor, only calculating
 	// forward operations

datasets/imdb.v

@@ -12,10 +12,10 @@ pub const (
 // ImdbDataset is a dataset for sentiment analysis.
 pub struct ImdbDataset {
 pub:
-	train_features &vtl.Tensor[string]
-	train_labels   &vtl.Tensor[int]
-	test_features  &vtl.Tensor[string]
-	test_labels    &vtl.Tensor[int]
+	train_features &vtl.Tensor[string] = unsafe { nil }
+	train_labels   &vtl.Tensor[int]    = unsafe { nil }
+	test_features  &vtl.Tensor[string] = unsafe { nil }
+	test_labels    &vtl.Tensor[int]    = unsafe { nil }
 }
 
 // load_imdb_helper loads the IMDB dataset for a given split.

datasets/mnist.v

@@ -14,10 +14,10 @@ pub const (
 // MnistDataset is a dataset of MNIST handwritten digits.
 pub struct MnistDataset {
 pub:
-	train_features &vtl.Tensor[u8]
-	train_labels   &vtl.Tensor[u8]
-	test_features  &vtl.Tensor[u8]
-	test_labels    &vtl.Tensor[u8]
+	train_features &vtl.Tensor[u8] = unsafe { nil }
+	train_labels   &vtl.Tensor[u8] = unsafe { nil }
+	test_features  &vtl.Tensor[u8] = unsafe { nil }
+	test_labels    &vtl.Tensor[u8] = unsafe { nil }
 }
 
 // load_mnist_helper loads the MNIST dataset from the given filename.

nn/gates/activation/elu.v

@@ -6,7 +6,7 @@ import vtl.nn.internal
 
 pub struct EluGate[T] {
 pub:
-	cache &vtl.Tensor[T]
+	cache &vtl.Tensor[T] = unsafe { nil }
 }
 
 pub fn elu_gate[T](cache &vtl.Tensor[T]) &EluGate[T] {

nn/gates/activation/leaky_relu.v

@@ -6,7 +6,7 @@ import vtl.nn.internal
 
 pub struct LeakyReluGate[T] {
 pub:
-	cache &vtl.Tensor[T]
+	cache &vtl.Tensor[T] = unsafe { nil }
 }
 
 pub fn leaky_relu_gate[T](cache &vtl.Tensor[T]) &LeakyReluGate[T] {

nn/gates/activation/relu.v

@@ -6,7 +6,7 @@ import vtl.nn.internal
 
 pub struct ReLUGate[T] {
 pub:
-	cache &vtl.Tensor[T]
+	cache &vtl.Tensor[T] = unsafe { nil }
 }
 
 pub fn relu_gate[T](cache &vtl.Tensor[T]) &ReLUGate[T] {

nn/gates/activation/sigmoid.v

@@ -6,7 +6,7 @@ import vtl.nn.internal
 
 pub struct SigmoidGate[T] {
 pub:
-	cache &vtl.Tensor[T]
+	cache &vtl.Tensor[T] = unsafe { nil }
 }
 
 pub fn sigmoid_gate[T](cache &vtl.Tensor[T]) &SigmoidGate[T] {

nn/gates/layers/dropout.v

@@ -6,7 +6,7 @@ import vtl.autograd
 pub struct DropoutGate[T] {
 pub:
 	prob f64
-	mask &vtl.Tensor[T]
+	mask &vtl.Tensor[T] = unsafe { nil }
 }
 
 pub fn dropout_gate[T](mask &vtl.Tensor[T], prob f64) &DropoutGate[T] {

nn/gates/layers/flatten.v

@@ -5,7 +5,7 @@ import vtl.autograd
 
 pub struct FlattenGate[T] {
 pub:
-	input        &autograd.Variable[T]
+	input        &autograd.Variable[T] = unsafe { nil }
 	cached_shape []int
 }
 

nn/gates/layers/linear.v

@@ -7,9 +7,9 @@ import vtl.stats
 
 pub struct LinearGate[T] {
 pub:
-	input  &autograd.Variable[T]
-	weight &autograd.Variable[T]
-	bias   &autograd.Variable[T]
+	input  &autograd.Variable[T] = unsafe { nil }
+	weight &autograd.Variable[T] = unsafe { nil }
+	bias   &autograd.Variable[T] = unsafe { nil }
 }
 
 pub fn linear_gate[T](input &autograd.Variable[T], weight &autograd.Variable[T], bias &autograd.Variable[T]) &LinearGate[T] {

nn/gates/layers/maxpool.v

@@ -6,7 +6,7 @@ import vtl.nn.internal
 
 pub struct MaxPool2DGate[T] {
 pub:
-	max_indices &vtl.Tensor[int]
+	max_indices &vtl.Tensor[int] = unsafe { nil }
 	kernel      []int
 	shape       []int
 	stride      []int

nn/gates/loss/mse.v

@@ -6,8 +6,8 @@ import vtl.nn.internal
 
 pub struct MseGate[T] {
 pub:
-	cache  &autograd.Variable[T]
-	target &vtl.Tensor[T]
+	cache  &autograd.Variable[T] = unsafe { nil }
+	target &vtl.Tensor[T]        = unsafe { nil }
 }
 
 pub fn mse_gate[T](cache &autograd.Variable[T], target &vtl.Tensor[T]) &MseGate[T] {

nn/gates/loss/sigmoid_cross_entropy.v

@@ -6,8 +6,8 @@ import vtl.nn.internal
 
 pub struct SigmoidCrossEntropyGate[T] {
 pub:
-	cache  &autograd.Variable[T]
-	target &vtl.Tensor[T]
+	cache  &autograd.Variable[T] = unsafe { nil }
+	target &vtl.Tensor[T]        = unsafe { nil }
 }
 
 pub fn sigmoid_cross_entropy_gate[T](cache &autograd.Variable[T], target &vtl.Tensor[T]) &SigmoidCrossEntropyGate[T] {

nn/gates/loss/softmax_cross_entropy.v

@@ -6,8 +6,8 @@ import vtl.nn.internal
 
 pub struct SoftmaxCrossEntropyGate[T] {
 pub:
-	cache  &autograd.Variable[T]
-	target &vtl.Tensor[T]
+	cache  &autograd.Variable[T] = unsafe { nil }
+	target &vtl.Tensor[T]        = unsafe { nil }
 }
 
 pub fn softmax_cross_entropy_gate[T](cache &autograd.Variable[T], target &vtl.Tensor[T]) &SoftmaxCrossEntropyGate[T] {

nn/layers/linear.v

@@ -9,8 +9,8 @@ import vtl.nn.types
 
 // LinearLayer is a layer that applies a linear transformation to its input.
 pub struct LinearLayer[T] {
-	weights &autograd.Variable[T]
-	bias    &autograd.Variable[T]
+	weights &autograd.Variable[T] = unsafe { nil }
+	bias    &autograd.Variable[T] = unsafe { nil }
 }
 
 pub fn linear_layer[T](ctx &autograd.Context[T], input_dim int, output_dim int) types.Layer[T] {

nn/models/sequential.v

@@ -11,7 +11,7 @@ fn init() {
 
 pub struct Sequential[T] {
 pub mut:
-	info &SequentialInfo[T]
+	info &SequentialInfo[T] = unsafe { nil }
 }
 
 // sequential creates a new sequential network with a new context.

nn/models/sequential_info.v

@@ -6,7 +6,7 @@ import vtl.nn.loss
 import vtl.nn.types
 
 pub struct SequentialInfo[T] {
-	ctx &autograd.Context[T]
+	ctx &autograd.Context[T] = unsafe { nil }
 pub mut:
 	layers []types.Layer[T]
 	loss   types.Loss

src/iter.v

@@ -12,7 +12,7 @@ pub enum IteratorStrategy {
 @[heap]
 pub struct TensorIterator[T] {
 pub:
-	tensor       &Tensor[T]
+	tensor       &Tensor[T] = unsafe { nil }
 	next_handler IteratorStrategy
 pub mut:
 	iteration int

src/iter_axis.v

@@ -5,7 +5,7 @@ module vtl
 @[heap]
 pub struct TensorAxisIterator[T] {
 pub:
-	tensor &Tensor[T]
+	tensor &Tensor[T] = unsafe { nil }
 pub mut:
 	shape     []int
 	strides   []int

src/tensor.v

@@ -12,7 +12,7 @@ pub enum MemoryFormat {
 @[heap]
 pub struct Tensor[T] {
 pub mut:
-	data    &storage.CpuStorage[T]
+	data    &storage.CpuStorage[T] = unsafe { nil }
 	memory  MemoryFormat
 	size    int
 	shape   []int

src/tensor_vcl_d_vcl.v

@@ -6,7 +6,7 @@ import vtl.storage
 @[heap]
 pub struct VclTensor[T] {
 pub mut:
-	data    &storage.VclStorage[T]
+	data    &storage.VclStorage[T] = unsafe { nil }
 	memory  MemoryFormat
 	size    int
 	shape   []int