From 91d42f2d057e077c267d6775cc109a801aa315c0 Mon Sep 17 00:00:00 2001 From: skal Date: Thu, 12 Feb 2026 11:34:50 +0100 Subject: CNN v2: parametric static features - Phases 1-4 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Infrastructure for enhanced CNN post-processing with 7D feature input. Phase 1: Shaders - Static features compute (RGBD + UV + sin10_x + bias → 8×f16) - Layer template (convolution skeleton, packing/unpacking) - 3 mip level support for multi-scale features Phase 2: C++ Effect - CNNv2Effect class (multi-pass architecture) - Texture management (static features, layer buffers) - Build integration (CMakeLists, assets, tests) Phase 3: Training Pipeline - train_cnn_v2.py: PyTorch model with static feature concatenation - export_cnn_v2_shader.py: f32→f16 quantization, WGSL generation - Configurable architecture (kernels, channels) Phase 4: Validation - validate_cnn_v2.sh: End-to-end pipeline - Checkpoint → shaders → build → test images Tests: 36/36 passing Next: Complete render pipeline implementation (bind groups, multi-pass) Co-Authored-By: Claude Sonnet 4.5 --- workspaces/main/shaders/cnn_v2_layer_template.wgsl | 68 ++++++++++++++++++++++ 1 file changed, 68 insertions(+) create mode 100644 workspaces/main/shaders/cnn_v2_layer_template.wgsl (limited to 'workspaces/main/shaders/cnn_v2_layer_template.wgsl') diff --git a/workspaces/main/shaders/cnn_v2_layer_template.wgsl b/workspaces/main/shaders/cnn_v2_layer_template.wgsl new file mode 100644 index 0000000..1bf6819 --- /dev/null +++ b/workspaces/main/shaders/cnn_v2_layer_template.wgsl @@ -0,0 +1,68 @@ +// CNN v2 Layer Template (placeholder for generated shaders) +// This file documents the structure - actual layers generated by export script + +// Example: Layer 0 (1×1 kernel, 8→16 channels) +// const KERNEL_SIZE: u32 = 1u; +// const IN_CHANNELS: u32 = 8u; // 7 features + bias +// const OUT_CHANNELS: u32 = 16u; +// const weights: array = array(...); + +@group(0) @binding(0) var static_features: texture_2d; +@group(0) @binding(1) var layer_input: texture_2d; // Previous layer output +@group(0) @binding(2) var output_tex: texture_storage_2d; + +fn unpack_static_features(coord: vec2) -> array { + let packed = textureLoad(static_features, coord, 0); + let v0 = unpack2x16float(packed.x); + let v1 = unpack2x16float(packed.y); + let v2 = unpack2x16float(packed.z); + let v3 = unpack2x16float(packed.w); + return array(v0.x, v0.y, v1.x, v1.y, v2.x, v2.y, v3.x, v3.y); +} + +fn unpack_layer_channels(coord: vec2) -> array { + let packed = textureLoad(layer_input, coord, 0); + let v0 = unpack2x16float(packed.x); + let v1 = unpack2x16float(packed.y); + let v2 = unpack2x16float(packed.z); + let v3 = unpack2x16float(packed.w); + return array(v0.x, v0.y, v1.x, v1.y, v2.x, v2.y, v3.x, v3.y); +} + +fn pack_channels(values: array) -> vec4 { + return vec4( + pack2x16float(vec2(values[0], values[1])), + pack2x16float(vec2(values[2], values[3])), + pack2x16float(vec2(values[4], values[5])), + pack2x16float(vec2(values[6], values[7])) + ); +} + +@compute @workgroup_size(8, 8) +fn main(@builtin(global_invocation_id) id: vec3) { + let coord = vec2(id.xy); + let dims = textureDimensions(static_features); + + if (coord.x >= i32(dims.x) || coord.y >= i32(dims.y)) { + return; + } + + // Load static features (always available) + let static_feat = unpack_static_features(coord); + + // Convolution loop (example for generated code) + // var output: array; + // for (var c: u32 = 0u; c < OUT_CHANNELS; c++) { + // var sum: f32 = 0.0; + // for (var ky: i32 = -radius; ky <= radius; ky++) { + // for (var kx: i32 = -radius; kx <= radius; kx++) { + // let sample_coord = coord + vec2(kx, ky); + // // Load static + prev layer, multiply weights, accumulate + // } + // } + // output[c] = max(0.0, sum); // ReLU + // } + + // Placeholder output + textureStore(output_tex, coord, vec4(0u)); +} -- cgit v1.2.3