internal static NDArrayOrSymbol[] GetBeginState(RecurrentCell cell, NDArrayOrSymbol[] begin_state,
NDArrayOrSymbol inputs, int batch_size)
{
if (begin_state != null)
{
if (inputs.IsNDArray)
{
var ctx = inputs.NdX.Context;
var args = new FuncArgs();
args.Add("ctx", ctx);
begin_state = cell.BeginState(batch_size, "nd.Zeros", args);
}
else
{
begin_state = cell.BeginState(batch_size, "sym.Zeros");
}
}
return(begin_state);
}
public override KerasSymbol[] Invoke(KerasSymbol[] inputs, FuncArgs kwargs = null)
{
List <KerasSymbol> result = new List <KerasSymbol>();
bool training = kwargs.Get <bool>("training");
foreach (var input in inputs)
{
Func <KerasSymbol> dropped_inputs = () => {
return(K.Dropout(input, this.rate, noise_shape, seed: this.seed));
};
if ((0 < this.rate) && (this.rate < 1.0))
{
var noise_shape = this.GetNoiseShape(input);
result.Add(K.InTrainPhase(dropped_inputs, input, training: training));
}
}
return(result.ToArray());
}
public StateInfo(FuncArgs args)
{
foreach (var arg in args)
{
if (arg.Value == null)
{
continue;
}
switch (arg.Key.ToLower())
{
case "shape":
Shape = (Shape)arg.Value;
break;
case "layout":
Layout = arg.Value.ToString();
break;
case "in_layout":
Layout = arg.Value.ToString();
break;
case "mean":
Mean = Convert.ToSingle(arg.Value);
break;
case "std":
Mean = Convert.ToSingle(arg.Value);
break;
case "dtype":
DataType = (DType)arg.Value;
break;
case "ctx":
Ctx = (Context)arg.Value;
break;
}
}
}
public override KerasSymbol[] Invoke(KerasSymbol[] inputs, FuncArgs kwargs = null)
{
List <KerasSymbol> result = new List <KerasSymbol>();
foreach (var input in inputs)
{
var output = K.Dot(input, this.kernel);
if (this.use_bias)
{
output = K.BiasAdd(output, this.bias, data_format: "channels_last");
}
result.Add(output);
}
if (this.activation != null)
{
return(this.activation.Invoke(result.ToArray()));
}
return(result.ToArray());
}
public FastSEResNet(string architecture, string norm_layer = "BatchNorm", FuncArgs kwargs = null, string prefix = "", ParameterDict @params = null) : base(prefix, @params)
{
throw new NotImplementedException();
}
public DarknetV3(int[] layers, int[] channels, int classes = 1000, string norm_layer = "BatchNorm", FuncArgs norm_kwargs = null, string prefix = null, ParameterDict @params = null) : base(prefix, @params)
{
Debug.Assert(layers.Length == channels.Length - 1, $"len(channels) should equal to len(layers) + 1, given {channels.Length} vs {layers.Length}");
this.features = new HybridSequential();
// first 3x3 conv
this.features.Add(Conv2d(channels[0], 3, 1, 1, norm_layer: norm_layer, norm_kwargs: norm_kwargs));
for (int i = 0; i < layers.Length; i++)
{
int nlayer = layers[i];
int channel = channels[i];
Debug.Assert(channel % 2 == 0, $"channel {channel} cannot be divided by 2");
// add downsample conv with stride=2
this.features.Add(Conv2d(channel, 3, 1, 2, norm_layer: norm_layer, norm_kwargs: norm_kwargs));
// add nlayer basic blocks
foreach (var _ in Enumerable.Range(0, nlayer))
{
this.features.Add(new DarknetBasicBlockV3(channel / 2, norm_layer: norm_layer, norm_kwargs: norm_kwargs));
}
}
// output
this.output = new Dense(classes);
RegisterChild(features);
RegisterChild(output);
}
public IDAUp(int out_channels, int in_channels, float[] up_f, bool use_dcnv2 = false, string norm_layer = "BatchNorm", FuncArgs norm_kwargs = null, string prefix = "", ParameterDict @params = null) : base(prefix, @params)
{
throw new NotImplementedException();
}
private HybridSequential MakeBasicConv(int in_channels, int channels, string norm_layer = "BatchNorm", FuncArgs norm_kwargs = null)
{
throw new NotImplementedException();
}
public override NDArrayOrSymbol[] BeginState(int batch_size = 0, string func = null, FuncArgs args = null)
{
BaseCell._modified = false;
var begin = BaseCell.BeginState(batch_size, func, args);
BaseCell._modified = true;
return(begin);
}
public static CIFARResNetV2 Cifar_ResNet110_V2(bool pretrained = false, Context ctx = null, string root = "~/mxnet", string norm_layer = "BatchNorm", FuncArgs norm_kwargs = null)
{
throw new NotImplementedException();
}
public static HybridSequential MakeLayerSlow(int inplanes, int planes, int num_blocks, int?num_block_temp_kernel_slow = null, string block = "Bottleneck",
int strides = 1, int head_conv = 1, string norm_layer = "BatchNorm", FuncArgs norm_kwargs = null, string layer_name = "")
{
throw new NotImplementedException();
}
public NASNetALarge(int repeat = 6, int penultimate_filters = 4032, int stem_filters = 96, int filters_multiplier = 2, int classes = 1000, bool use_aux = true,
string norm_layer = "BatchNorm", FuncArgs norm_kwargs = null, string prefix = "", ParameterDict @params = null) : base(prefix, @params)
{
throw new NotImplementedException();
}
public ResidualBlock(int channels, int?in_channels = null, int stride = 1, string norm_layer = "BatchNorm", FuncArgs norm_kwargs = null, string prefix = null, ParameterDict @params = null) : base(prefix, @params)
{
throw new NotImplementedException();
}
public BasicBlock(int inplanes, int planes, int spatial_stride = 1, int temporal_stride = 1, int dilation = 1, bool downsample = false, bool if_inflate = true, string inflate_style = "", string norm_layer = "BatchNorm", FuncArgs norm_kwargs = null, string layer_name = "", string prefix = "", ParameterDict @params = null) : base(prefix, @params)
{
throw new NotImplementedException();
}
public Bottleneck(int inplanes, int planes, int spatial_stride = 1, int temporal_stride = 1, int dilation = 1, bool downsample = false, bool if_inflate = true, string inflate_style = "3x1x1", string norm_layer = "BatchNorm", FuncArgs norm_kwargs = null, string layer_name = "", string prefix = "", ParameterDict @params = null) : base(prefix, @params)
{
}
public CIFARResNetV2(HybridBlock block, int[] layers, int[] channels, int classes = 10, string norm_layer = "BatchNorm", FuncArgs norm_kwargs = null, string prefix = "", ParameterDict @params = null) : base(prefix, @params)
{
throw new NotImplementedException();
}
private HybridSequential MakeLayer(HybridBlock block, int[] layers, int[] channels, int stride, int stage_index, int in_channels = 0, string norm_layer = "BatchNorm", FuncArgs norm_kwargs = null)
{
throw new NotImplementedException();
}
public Bottleneck(int inplanes, int planes, int strides = 1, bool downsample = false, int head_conv = 1, string norm_layer = "BatchNorm", FuncArgs norm_kwargs = null,
string layer_name = "", string prefix = null, ParameterDict @params = null) : base(prefix, @params)
{
throw new NotImplementedException();
}
public Tree(int levels, HybridBlock block, int in_channels, int out_channels, int stride = 1, bool level_root = false, int root_dim = 0, int root_kernel_size = 1,
int dilation = 1, bool root_residual = false, string norm_layer = "BatchNorm", FuncArgs norm_kwargs = null, string prefix = null, ParameterDict @params = null) : base(prefix, @params)
{
throw new NotImplementedException();
}
public SE_BasicBlockV2(int channels, int stride, bool downsample = false, int in_channels = 0, string norm_layer = "BatchNorm", FuncArgs norm_kwargs = null, string prefix = "", ParameterDict @params = null) : base(prefix, @params)
{
}
public GoogLeNet(int classes = 1000, string norm_layer = "BatchNorm", float dropout_ratio = 0.4f, bool aux_logits = false, FuncArgs norm_kwargs = null, bool partial_bn = false, bool pretrained_base = true, Context ctx = null, string prefix = null, ParameterDict @params = null) : base(prefix, @params)
{
throw new NotImplementedException();
}
public override KerasSymbol[] Invoke(KerasSymbol[] inputs, FuncArgs kwargs = null)
{
throw new NotImplementedException();
}
private HybridSequential MakeBranch(bool use_pool, string norm_layer = "BatchNorm", FuncArgs norm_kwargs = null, int?channels = null, (int, int)?kernel_size = null, (int, int)?strides = null, (int, int)?padding = null)
public NormalCell(int out_channels_left, int out_channels_right, string norm_layer, FuncArgs norm_kwargs, string prefix = null, ParameterDict @params = null) : base(prefix, @params)
{
throw new NotImplementedException();
}
public MaskRCNN(HybridBlock features, HybridBlock top_features, string[] classes, int mask_channels = 256, int rcnn_max_dets = 1000, int rpn_test_pre_nms = 6000, int rpn_test_post_nms = 1000, int target_roi_scale = 1, int num_fcn_convs = 0, string norm_layer = "", FuncArgs norm_kwargs = null, string prefix = null, ParameterDict @params = null) : base(features: features, top_features: top_features, classes: classes, rpn_test_pre_nms: rpn_test_pre_nms, rpn_test_post_nms: rpn_test_post_nms, additional_output: true, prefix: prefix, @params: @params)
{
throw new NotImplementedException();
}
private HybridSequential ASPPConv(int in_channels, int out_channels, float atrous_rate, string norm_layer, FuncArgs norm_kwargs)
{
throw new NotImplementedException();
}
internal static HybridSequential Conv2d(int channel, int kernel, int padding, int stride, string norm_layer = "BatchNorm", FuncArgs norm_kwargs = null)
{
var cell = new HybridSequential(prefix: "");
cell.Add(new Conv2D(channel, kernel_size: (kernel, kernel), strides: (stride, stride), padding: (padding, stride), use_bias: false));
cell.Add(LayerUtils.NormLayer(norm_layer, norm_kwargs));
cell.Add(new LeakyReLU(0.1f));
return(cell);
}
public _DeepLabHead(int nclass, int c1_channels = 128, string norm_layer = "BatchNorm", FuncArgs norm_kwargs = null, int height = 240, int width = 240, string prefix = "", ParameterDict @params = null) : base(prefix, @params)
{
throw new NotImplementedException();
}
public Node(Layer outbound_layer, Layer[] inbound_layers, int[] node_indices, int[] tensor_indices, KerasSymbol[] input_tensors, KerasSymbol[] output_tensors, KerasSymbol[] input_masks, KerasSymbol[] output_masks, Shape[] input_shapes, Shape[] output_shapes,
FuncArgs arguments = null)
{
throw new NotImplementedException();
}
public _ASPP(int in_channels, int out_channels, string norm_layer, FuncArgs norm_kwargs, int height = 60, int width = 60, string prefix = "", ParameterDict @params = null) : base(prefix, @params)
{
throw new NotImplementedException();
}
