demo.git/workspaces/main/shaders/cnn_v2, branch main

refactor(cnn): isolate CNN v2 to cnn_v2/ subdirectory

2026-02-15T17:44:17Z

Move all CNN v2 files to dedicated cnn_v2/ directory to prepare for CNN v3 development. Zero functional changes. Structure: - cnn_v2/src/ - C++ effect implementation - cnn_v2/shaders/ - WGSL shaders (6 files) - cnn_v2/weights/ - Binary weights (3 files) - cnn_v2/training/ - Python training scripts (4 files) - cnn_v2/scripts/ - Shell scripts (train_cnn_v2_full.sh) - cnn_v2/tools/ - Validation tools (HTML) - cnn_v2/docs/ - Documentation (4 markdown files) Changes: - Update CMake source list to cnn_v2/src/cnn_v2_effect.cc - Update assets.txt with relative paths to cnn_v2/ - Update includes to ../../cnn_v2/src/cnn_v2_effect.h - Add PROJECT_ROOT resolution to Python/shell scripts - Update doc references in HOWTO.md, TODO.md - Add cnn_v2/README.md Verification: 34/34 tests passing, demo runs correctly. Co-Authored-By: Claude Sonnet 4.5

Fix: CNN v2 compute shader validation error

2026-02-14T11:40:57Z

Replace textureSample() with textureSampleLevel() in compute shader. textureSample() requires derivative calculations only available in fragment shaders. Compute shaders must explicitly specify mip level. Fixes: DemoEffectsTest CNNv2Effect initialization Co-Authored-By: Claude Sonnet 4.5

CNN v2: bilinear mip-level sampling and UI improvements

2026-02-14T06:24:51Z

**CNN v2 Changes:** - Replace point sampling with bilinear interpolation for mip-level features - Add linear sampler (binding 6) to static features shader - Update CNNv2Effect, cnn_test, and HTML tool **HTML Tool UI:** - Move controls to floating bottom bar in central view - Consolidate video controls + Blend/Depth/Save PNG in single container - Increase left panel width: 300px → 315px (+5%) - Remove per-frame debug messages (visualization, rendering logs) **Technical:** - WGSL: textureSample() with linear_sampler vs textureLoad() - C++: Create WGPUSampler with Linear filtering - HTML: Change sampler from 'nearest' to 'linear' handoff(Claude): CNN v2 now uses bilinear mip-level sampling across all tools

Replace hard clamp with sigmoid activation in CNN v2

2026-02-14T01:12:12Z

Fixes training collapse where p1/p2 channels saturate due to gradient blocking at clamp boundaries. Sigmoid provides smooth [0,1] mapping with continuous gradients. Changes: - Layer 0: clamp(x, 0, 1) → sigmoid(x) - Final layer: clamp(x, 0, 1) → sigmoid(x) - Middle layers: ReLU unchanged (already stable) Updated files: - training/train_cnn_v2.py: PyTorch model activations - workspaces/main/shaders/cnn_v2/cnn_v2_compute.wgsl: WGSL shader - tools/cnn_v2_test/index.html: HTML validation tool - doc/CNN_V2.md: Documentation Validation: - Build clean (no shader errors) - 34/36 tests pass (2 unrelated script tests fail) - 10-epoch training: loss 0.153 → 0.088 (good convergence) - cnn_test processes images successfully Breaking change: Old checkpoints trained with clamp() incompatible. Retrain from scratch required. handoff(Claude): CNN v2 sigmoid activation implemented and validated.

CNN v2: Alpha channel depth handling and layer visualization

2026-02-13T22:17:42Z

Training changes: - Changed p3 default depth from 0.0 to 1.0 (far plane semantics) - Extract depth from target alpha channel in both datasets - Consistent alpha-as-depth across training/validation Test tool enhancements (cnn_test): - Added load_depth_from_alpha() for R32Float depth texture - Fixed bind group layout for UnfilterableFloat sampling - Added --save-intermediates with per-channel grayscale composites - Each layer saved as 4x wide PNG (p0-p3 stacked horizontally) - Global layers_composite.png for vertical layer stack overview Investigation notes: - Static features p4-p7 ARE computed and bound correctly - Sin_20_y pattern visibility difference between tools under investigation - Binary weights timestamp (Feb 13 20:36) vs HTML tool (Feb 13 22:12) - Next: Update HTML tool with canonical binary weights handoff(Claude): HTML tool weights update pending - base64 encoded canonical weights ready in /tmp/weights_b64.txt for line 392 replacement. Co-Authored-By: Claude Sonnet 4.5

CNN v2: Fix WebGPU validation error in uniform buffer alignment

2026-02-13T16:16:56Z

Fix two issues causing validation errors in test_demo: 1. Remove redundant pipeline creation without layout (static_pipeline_) 2. Change vec3 to 3× u32 fields in StaticFeatureParams struct WGSL vec3 aligns to 16 bytes (std140), making struct 32 bytes, while C++ struct was 16 bytes. Explicit fields ensure consistent layout. Co-Authored-By: Claude Sonnet 4.5

CNN v2: Change feature #6 from sin(10x) to sin(20y)

2026-02-13T15:57:29Z

Update positional encoding to use vertical coordinate at higher frequency. Changes: - train_cnn_v2.py: sin10_x → sin20_y (computed from uv_y) - cnn_v2_static.wgsl: sin10_x → sin20_y (computed from uv_y) - index.html: sin10_x → sin20_y (STATIC_SHADER) - CNN_V2.md: Update feature descriptions and examples - CNN_V2_BINARY_FORMAT.md: Update static features documentation Feature vector: [p0, p1, p2, p3, uv_x, uv_y, sin20_y, bias] Rationale: Higher frequency (20 vs 10) + vertical axis provides better spatial discrimination for position encoding. Co-Authored-By: Claude Sonnet 4.5

CNN v2: Add TODO for flexible feature layout in binary format v3

2026-02-13T15:54:47Z

Document future enhancement for arbitrary feature vector layouts. Proposed feature descriptor in binary format v3: - Specify feature types, sources, and ordering - Enable runtime experimentation without shader recompilation - Examples: [R,G,B,dx,dy,uv_x,bias] or [mip1.r,mip2.g,laplacian,uv_x,sin20_x,bias] Added TODOs in: - CNN_V2_BINARY_FORMAT.md: Detailed proposal with struct layout - CNN_V2.md: Future extensions section - train_cnn_v2.py: compute_static_features() docstring - cnn_v2_static.wgsl: Shader header comment - cnn_v2_effect.cc: Version check comment Current limitation: Hardcoded [p0,p1,p2,p3,uv_x,uv_y,sin10_x,bias] layout. Co-Authored-By: Claude Sonnet 4.5

CNN v2: Add mip-level support to runtime effect

2026-02-13T15:48:02Z

Binary format v2 includes mip_level in header (20 bytes, was 16). Effect reads mip_level and passes to static features shader via uniform. Shader samples from correct mip texture based on mip_level. Changes: - export_cnn_v2_weights.py: Header v2 with mip_level field - cnn_v2_effect.h: Add StaticFeatureParams, mip_level member, params buffer - cnn_v2_effect.cc: Read mip_level from weights, create/bind params buffer, update per-frame - cnn_v2_static.wgsl: Accept params uniform, sample from selected mip level Binary format v2: - Header: 20 bytes (magic, version=2, num_layers, total_weights, mip_level) - Backward compatible: v1 weights load with mip_level=0 Co-Authored-By: Claude Sonnet 4.5

CNNv2Effect: Document per-layer kernel sizes support

2026-02-13T11:43:03Z

Updated comments to clarify that per-layer kernel sizes are supported. Code already handles this correctly via LayerInfo.kernel_size field. Changes: - cnn_v2_effect.h: Add comment about per-layer kernel sizes - cnn_v2_compute.wgsl: Clarify LayerParams provides per-layer config Co-Authored-By: Claude Sonnet 4.5