Update constants.ts

constants.ts

@@ -9,6 +9,7 @@ export const LAYER_DEFINITIONS: Record<LayerType, LayerDefinition> = {
     description: 'Entry point for data tensors',
     category: 'Core',
     parameters: [
+      { name: 'name', type: 'string', label: 'Name', default: 'input_1', description: 'Variable name in forward()' },
       { name: 'modality', type: 'select', label: 'Modality', default: 'Tensor', options: ['Tensor', 'Image', 'Text', 'Audio', 'Video', 'Latent', 'State', '3D Volume', 'Point Cloud'] },
       { name: 'shape', type: 'string', label: 'Shape (e.g. 3,224,224)', default: '3, 224, 224' },
       { name: 'batch_size', type: 'number', label: 'Batch Size', default: 32 },
@@ -23,7 +24,8 @@ export const LAYER_DEFINITIONS: Record<LayerType, LayerDefinition> = {
     parameters: [
       { name: 'in_features', type: 'number', label: 'In Features (Opt)', default: 0, description: "0 = Auto-infer" },
       { name: 'out_features', type: 'number', label: 'Output Features', default: 128 },
-      { name: 'bias', type: 'boolean', label: 'Use Bias', default: true }
+      { name: 'bias', type: 'boolean', label: 'Use Bias', default: true },
+      { name: 'activation', type: 'select', label: 'Fused Activation', default: 'None', options: ['None', 'ReLU', 'GELU', 'Sigmoid'] }
     ]
   },
   [LayerType.OUTPUT]: {
@@ -83,6 +85,16 @@ export const LAYER_DEFINITIONS: Record<LayerType, LayerDefinition> = {
       { name: 'k', type: 'number', label: 'Kernel Size', default: 5 }
     ]
   },
+  [LayerType.DARKNET_BLOCK]: {
+    type: LayerType.DARKNET_BLOCK,
+    label: 'Darknet Block',
+    description: 'Residual block used in Darknet',
+    category: 'Detection',
+    parameters: [
+      { name: 'channels', type: 'number', label: 'Channels', default: 64 },
+      { name: 'num_repeats', type: 'number', label: 'Repeats', default: 1 }
+    ]
+  },
   [LayerType.DETECT_HEAD]: {
     type: LayerType.DETECT_HEAD,
     label: 'Detection Head',
@@ -103,8 +115,29 @@ export const LAYER_DEFINITIONS: Record<LayerType, LayerDefinition> = {
       { name: 'scales', type: 'text', label: 'Scales', default: '[32, 64, 128]' }
     ]
   },
+  [LayerType.NMS]: {
+    type: LayerType.NMS,
+    label: 'NMS',
+    description: 'Non-Maximum Suppression',
+    category: 'Detection',
+    parameters: [
+      { name: 'iou_threshold', type: 'number', label: 'IoU Thresh', default: 0.5 },
+      { name: 'score_threshold', type: 'number', label: 'Score Thresh', default: 0.25 }
+    ]
+  },
 
   // --- AUDIO / SPEECH ---
+  [LayerType.STFT]: {
+    type: LayerType.STFT,
+    label: 'STFT',
+    description: 'Short-Time Fourier Transform',
+    category: 'Audio',
+    parameters: [
+      { name: 'n_fft', type: 'number', label: 'N_FFT', default: 1024 },
+      { name: 'hop_length', type: 'number', label: 'Hop Length', default: 256 },
+      { name: 'win_length', type: 'number', label: 'Window Length', default: 1024 }
+    ]
+  },
   [LayerType.MEL_SPECTROGRAM]: {
     type: LayerType.MEL_SPECTROGRAM,
     label: 'MelSpectrogram',
@@ -116,6 +149,16 @@ export const LAYER_DEFINITIONS: Record<LayerType, LayerDefinition> = {
       { name: 'n_mels', type: 'number', label: 'Num Mels', default: 80 }
     ]
   },
+  [LayerType.SPEC_AUGMENT]: {
+    type: LayerType.SPEC_AUGMENT,
+    label: 'SpecAugment',
+    description: 'Time/Freq masking for Audio',
+    category: 'Audio',
+    parameters: [
+      { name: 'freq_mask_param', type: 'number', label: 'Freq Mask', default: 27 },
+      { name: 'time_mask_param', type: 'number', label: 'Time Mask', default: 100 }
+    ]
+  },
   [LayerType.CONFORMER_BLOCK]: {
     type: LayerType.CONFORMER_BLOCK,
     label: 'Conformer Block',
@@ -138,6 +181,26 @@ export const LAYER_DEFINITIONS: Record<LayerType, LayerDefinition> = {
       { name: 'kernel_size', type: 'number', label: 'Kernel', default: 3 }
     ]
   },
+  [LayerType.WAV2VEC2_ENC]: {
+    type: LayerType.WAV2VEC2_ENC,
+    label: 'Wav2Vec2 Encoder',
+    description: 'Self-supervised Speech Encoder',
+    category: 'Audio',
+    parameters: [
+      { name: 'output_dim', type: 'number', label: 'Output Dim', default: 768 },
+      { name: 'extractor_mode', type: 'select', label: 'Mode', default: 'default', options: ['default', 'layer_norm'] }
+    ]
+  },
+  [LayerType.RVC_ENCODER]: {
+    type: LayerType.RVC_ENCODER,
+    label: 'RVC Hubert',
+    description: 'Content Encoder for Voice Cloning',
+    category: 'Audio',
+    parameters: [
+      { name: 'model_type', type: 'select', label: 'Model', default: 'hubert-soft', options: ['hubert-soft', 'vec256', 'vec768'] },
+      { name: 'freeze', type: 'boolean', label: 'Freeze', default: true }
+    ]
+  },
   [LayerType.VOCODER]: {
     type: LayerType.VOCODER,
     label: 'Vocoder',
@@ -181,6 +244,26 @@ export const LAYER_DEFINITIONS: Record<LayerType, LayerDefinition> = {
       { name: 'out_channels', type: 'number', label: 'Out Channels', default: 64 }
     ]
   },
+  [LayerType.POINT_TRANSFORMER]: {
+    type: LayerType.POINT_TRANSFORMER,
+    label: 'PointTransformer',
+    description: 'Self-Attention for Point Clouds',
+    category: '3D',
+    parameters: [
+      { name: 'dim', type: 'number', label: 'Dim', default: 32 },
+      { name: 'num_neighbors', type: 'number', label: 'Neighbors (k)', default: 16 }
+    ]
+  },
+  [LayerType.TRIPLANE_ENC]: {
+    type: LayerType.TRIPLANE_ENC,
+    label: 'Triplane Enc',
+    description: 'Project 3D to 3x2D Planes',
+    category: '3D',
+    parameters: [
+      { name: 'plane_res', type: 'number', label: 'Resolution', default: 256 },
+      { name: 'channels', type: 'number', label: 'Channels', default: 32 }
+    ]
+  },
   [LayerType.GAUSSIAN_SPLAT]: {
     type: LayerType.GAUSSIAN_SPLAT,
     label: 'Gaussian Splat',
@@ -330,9 +413,11 @@ export const LAYER_DEFINITIONS: Record<LayerType, LayerDefinition> = {
       { name: 'kernel_size', type: 'number', label: 'Kernel Size', default: 3 },
       { name: 'stride', type: 'number', label: 'Stride', default: 1 },
       { name: 'padding', type: 'number', label: 'Padding', default: 1 },
+      { name: 'padding_mode', type: 'select', label: 'Pad Mode', default: 'zeros', options: ['zeros', 'reflect', 'replicate', 'circular'] },
       { name: 'dilation', type: 'number', label: 'Dilation', default: 1 },
       { name: 'groups', type: 'number', label: 'Groups', default: 1, description: "For depthwise separable" },
-      { name: 'bias', type: 'boolean', label: 'Bias', default: true }
+      { name: 'bias', type: 'boolean', label: 'Bias', default: true },
+      { name: 'activation', type: 'select', label: 'Fused Activation', default: 'None', options: ['None', 'ReLU', 'LeakyReLU', 'SiLU'] }
     ]
   },
   [LayerType.CONV3D]: {
@@ -359,6 +444,17 @@ export const LAYER_DEFINITIONS: Record<LayerType, LayerDefinition> = {
       { name: 'padding', type: 'number', label: 'Padding', default: 0 }
     ]
   },
+  [LayerType.DEFORMABLE_CONV]: {
+    type: LayerType.DEFORMABLE_CONV,
+    label: 'Deformable Conv',
+    description: 'Deformable Convolution v2',
+    category: 'Convolution',
+    parameters: [
+      { name: 'out_channels', type: 'number', label: 'Filters', default: 64 },
+      { name: 'kernel_size', type: 'number', label: 'Kernel Size', default: 3 },
+      { name: 'offset_groups', type: 'number', label: 'Offset Groups', default: 1 }
+    ]
+  },
   [LayerType.MAXPOOL]: {
     type: LayerType.MAXPOOL,
     label: 'MaxPool2D',
@@ -613,7 +709,8 @@ export const LAYER_DEFINITIONS: Record<LayerType, LayerDefinition> = {
       { name: 'embed_dim', type: 'number', label: 'Embed Dim', default: 512 },
       { name: 'num_heads', type: 'number', label: 'Num Heads', default: 8 },
       { name: 'dropout', type: 'number', label: 'Dropout', default: 0.1 },
-      { name: 'batch_first', type: 'boolean', label: 'Batch First', default: true }
+      { name: 'batch_first', type: 'boolean', label: 'Batch First', default: true },
+      { name: 'causal', type: 'boolean', label: 'Causal Mask', default: false }
     ]
   },
   [LayerType.CROSS_ATTENTION]: {
@@ -703,6 +800,27 @@ export const INITIAL_EDGES = [
 ];
 
 export const TEMPLATES: Record<string, GraphTemplate> = {
+  'rvc_voice': {
+    id: 'rvc_voice',
+    name: 'Voice Cloning (RVC)',
+    description: 'Retrieval-based Voice Conversion backbone.',
+    nodes: [
+        { id: 'audio', type: 'custom', position: {x: 200, y: 0}, data: {label: 'Source Audio', type: LayerType.INPUT, params: {modality: 'Audio'}} },
+        { id: 'hubert', type: 'custom', position: {x: 200, y: 100}, data: {label: 'HuBERT Soft', type: LayerType.RVC_ENCODER, params: {}} },
+        { id: 'f0', type: 'custom', position: {x: 450, y: 0}, data: {label: 'Pitch (F0)', type: LayerType.INPUT, params: {modality: 'Tensor'}} },
+        { id: 'emb', type: 'custom', position: {x: 450, y: 100}, data: {label: 'F0 Embed', type: LayerType.EMBEDDING, params: {num_embeddings: 256}} },
+        { id: 'cat', type: 'custom', position: {x: 325, y: 200}, data: {label: 'Merge Features', type: LayerType.CONCAT, params: {}} },
+        { id: 'wn', type: 'custom', position: {x: 325, y: 300}, data: {label: 'WaveNet Stack', type: LayerType.WAVENET_BLOCK, params: {channels: 256, dilation: 2}} },
+        { id: 'voc', type: 'custom', position: {x: 325, y: 400}, data: {label: 'HiFiGAN', type: LayerType.VOCODER, params: {}} },
+        { id: 'out', type: 'custom', position: {x: 325, y: 500}, data: {label: 'Cloned Audio', type: LayerType.OUTPUT, params: {}} },
+    ],
+    edges: [
+        { id: '1', source: 'audio', target: 'hubert' }, { id: '2', source: 'f0', target: 'emb' },
+        { id: '3', source: 'hubert', target: 'cat' }, { id: '4', source: 'emb', target: 'cat' },
+        { id: '5', source: 'cat', target: 'wn' }, { id: '6', source: 'wn', target: 'voc' },
+        { id: '7', source: 'voc', target: 'out' }
+    ]
+  },
   'yolo_v8': {
     id: 'yolo_v8',
     name: 'YOLO (Detection)',