| Age | Commit message (Collapse) | Author |
|---|
|
1. Loss printed at every epoch with \r (no scrolling)
2. Validation only on final epoch (not all checkpoints)
3. Process all input images (not just img_000.png)
Training output now shows live progress with single line update.
|
|
- Add QAT (quantization-aware training) notes
- Requires training with fake quantization
- Target: ~1.6 KB weights (vs 3.2 KB f16)
- Shader unpacking needs adaptation (4× u8 per u32)
|
|
- Export weights from epoch 70 checkpoint (3.2 KB binary)
- Disable shader template generation (use manual cnn_v2_compute.wgsl)
- Build successful with real weights
- Ready for integration testing
Storage buffer architecture complete:
- Dynamic layer count support
- ~0.3ms overhead vs constants (negligible)
- Single shader, flexible configuration
- Binary format: header + layer info + f16 weights
|
|
- Add binary weight format (header + layer info + packed f16)
- New export_cnn_v2_weights.py for binary weight export
- Single cnn_v2_compute.wgsl shader with storage buffer
- Load weights in CNNv2Effect::load_weights()
- Create layer compute pipeline with 5 bindings
- Fast training config: 100 epochs, 3×3 kernels, 8→4→4 channels
Next: Complete bind group creation and multi-layer compute execution
|
|
Added note for future enhancement: mix salient + random samples.
Rationale:
- Salient point detection focuses on edges/corners
- Random samples improve generalization across entire image
- Prevents overfitting to only high-gradient regions
Proposed implementation:
- Default: 90% salient points, 10% random samples
- Configurable: --random-sample-percent parameter
- Example: 64 patches = 58 salient + 6 random
Location: train_cnn_v2.py
- TODO in _detect_salient_points() method
- TODO in argument parser
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
Salient point detection on original images with patch extraction.
Changes:
- Added PatchDataset class (harris/fast/shi-tomasi/gradient detectors)
- Detects salient points on ORIGINAL images (no resize)
- Extracts 32×32 patches around salient points
- Default: 64 patches/image, harris detector
- Batch size: 16 (512 patches per batch)
Training modes:
1. Patch-based (default): --patch-size 32 --patches-per-image 64 --detector harris
2. Full-image (option): --full-image --image-size 256
Benefits:
- Focuses training on interesting regions
- Handles variable image sizes naturally
- Matches CNN v1 workflow
- Better convergence with limited data (8 images → 512 patches)
Script updated:
- train_cnn_v2_full.sh: Patch-based by default
- Configuration exposed for easy switching
Example:
./scripts/train_cnn_v2_full.sh # Patch-based
# Edit script: uncomment FULL_IMAGE for resize mode
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
Training script now resizes all images to fixed size before batching.
Issue: RuntimeError when batching variable-sized images
- Images had different dimensions (376x626 vs 344x361)
- PyTorch DataLoader requires uniform tensor sizes for batching
Solution:
- Add --image-size parameter (default: 256)
- Resize all images to target_size using LANCZOS interpolation
- Preserves aspect ratio independent training
Changes:
- train_cnn_v2.py: ImagePairDataset now resizes to fixed dimensions
- train_cnn_v2_full.sh: Added IMAGE_SIZE=256 configuration
Tested: 8 image pairs, variable sizes → uniform 256×256 batches
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
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 <noreply@anthropic.com>
|
|
|
|
BREAKING CHANGE: Timeline format now uses beats as default unit
## Core Changes
**Uniform Structure (32 bytes maintained):**
- Added `beat_time` (absolute beats for musical animation)
- Added `beat_phase` (fractional 0-1 for smooth oscillation)
- Renamed `beat` → `beat_phase`
- Kept `time` (physical seconds, tempo-independent)
**Seq Compiler:**
- Default: all numbers are beats (e.g., `5`, `16.5`)
- Explicit seconds: `2.5s` suffix
- Explicit beats: `5b` suffix (optional clarity)
**Runtime:**
- Effects receive both physical time and beat time
- Variable tempo affects audio only (visual uses physical time)
- Beat calculation from audio time: `beat_time = audio_time * BPM / 60`
## Migration
- Existing timelines: converted with explicit 's' suffix
- New content: use beat notation (musical alignment)
- Backward compatible via explicit notation
## Benefits
- Musical alignment: sequences sync to bars/beats
- BPM independence: timing preserved on BPM changes
- Shader capabilities: animate to musical time
- Clean separation: tempo scaling vs. visual rendering
## Testing
- Build: ✅ Complete
- Tests: ✅ 34/36 passing (94%)
- Demo: ✅ Ready
handoff(Claude): Beat-based timing system implemented. Variable tempo
only affects audio sample triggering. Visual effects use physical_time
(constant) and beat_time (musical). Shaders can now animate to beats.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
+misc
|
|
- Fix stale comments: RGBD→RGB (not grayscale)
- Clarify shape transformations in inference
- Add CNN_BIAS_FIX_2026-02.md consolidating recent fixes
- Include regenerated weights with 5x5 kernel for layer 0
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
- Fix bias division bug: divide by num_positions to compensate for
shader loop accumulation (affects all layers)
- train_cnn.py: Save RGBA output preserving alpha channel from input
- Add --debug-hex flag to both tools for pixel-level debugging
- Remove sRGB/linear_png debug code from cnn_test
- Regenerate weights with corrected bias export
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
|
|
Simplify coordinate initialization by generating [-1,1] range directly
instead of [0,1] then normalizing. Mathematically equivalent, clearer.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
Match training forward pass: compute grayscale from original [0,1] RGB
before normalization, then normalize gray to [-1,1].
Previously computed gray from normalized [-1,1] RGB in generated shader,
creating mismatch with train.py which does:
gray = 0.2126*R + 0.7152*G + 0.0722*B # [0,1]
gray = (gray - 0.5) * 2.0 # [-1,1]
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
Root cause: After swapping init/resize order, effects with Renderer3D crashed
because resize() called before init() tried to use uninitialized GPU resources.
Changes:
- Add guards in FlashCubeEffect::resize() and Hybrid3DEffect::resize() to
check ctx_.device before calling renderer_.resize()
- Remove lazy initialization remnants from CircleMaskEffect and CNNEffect
- Register auxiliary textures directly in init() (width_/height_ already set)
- Remove ensure_texture() methods and texture_initialized_ flags
All 36 tests passing. Demo runs without crashes.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
|
|
Conv functions now return raw sum, sigmoid applied at call site.
Matches tanh pattern used for inner layers.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
Final layer used hard clamp causing saturation to white when output > 1.0.
Replaced with sigmoid activation for smooth [0,1] mapping with gradients.
Changes:
- train_cnn.py: torch.sigmoid() in forward pass and WGSL codegen
- WGSL shaders: 1.0/(1.0+exp(-sum)) in cnn_conv3x3/5x5 _7to1 functions
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
Add --early-stop-patience and --early-stop-eps parameters to stop training when loss plateaus. Automatically exports weights when triggered.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
Training now computes loss only on center pixels (excludes conv padding
borders). Inference changed from tiling to full-image sliding window.
Both match cnn_layer.wgsl: each pixel processed from NxN neighborhood.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
|
|
Inference now tiles images into patches matching training patch size,
preventing distribution mismatch between patch training and full-image inference.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
The in1 vector (uv_norm, gray, 1.0) is loop-invariant and doesn't depend on
dx/dy offset. Moving it outside the convolution loop eliminates redundant
computation and enables better SIMD optimization.
Updated both shader files and train.py code generation.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
Restructured CNN weight storage and computation for GPU SIMD efficiency:
**Weight format:**
- Before: array<array<f32, 8>, N> (scalar array)
- After: array<vec4<f32>, N*2> (vec4 pairs)
**Computation:**
- Before: 8 scalar MADs + separate bias add
- After: 2 dot4 instructions (4 parallel MADs each)
- Input: [rgba][uv,gray,1] where 1.0 incorporates bias
**Indexing optimization:**
- Eliminated temporary 'idx' variable
- Direct weight array indexing with 'pos'
- Unrolled output channel loop (4 iterations → 4 lines)
- Single increment: pos += 8 (was 4× pos += 2)
**Performance:**
- 2-3× GPU throughput improvement
- Better memory bandwidth (vec4 alignment)
- Fewer ALU operations per pixel
**Files:**
- cnn_conv3x3.wgsl, cnn_conv5x5.wgsl: All 3 functions per file
- train_cnn.py: Export format + code generation
- cnn_weights_generated.wgsl, cnn_layer.wgsl: Regenerated
- CNN_EFFECT.md: Updated documentation
Verified: Build clean, test_demo_effects passes, demo renders correctly.
handoff(Claude): CNN vec4 SIMD optimization complete
|
|
Changed from 3×5×3 to 3×3×3 architecture for testing.
Changes:
- cnn_layer.wgsl: Use 3×3 conv for all layers
- cnn_weights_generated.wgsl: Regenerated weights
- image_style_processor.py: Made executable
handoff(Claude): CNN mismatch analysis complete, patch extraction added, docs updated
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
Streamlined and updated all training docs with new patch-based approach.
Changes:
- HOWTO.md: Updated training section with patch/full-image examples
- CNN_EFFECT.md: Streamlined training workflow, added detector info
- training/README.md: Complete rewrite with detector comparison table
New sections:
- Detector comparison (harris, fast, shi-tomasi, gradient)
- Practical examples for different use cases
- Tips for patch size and batch size selection
- Benefits of patch-based training
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
Preserve natural pixel scale by extracting patches at salient points
instead of resizing entire images.
Features:
- Multiple detectors: Harris (default), FAST, Shi-Tomasi, gradient
- Configurable patch size (e.g., 32×32) and patches per image
- Automatic fallback to random patches if insufficient features
Usage:
# Patch-based training (preserves scale)
python3 train_cnn.py --input dir/ --target dir/ --patch-size 32 --patches-per-image 64 --detector harris
# Original resize mode (if --patch-size omitted)
python3 train_cnn.py --input dir/ --target dir/
Arguments:
--patch-size: Patch dimension (e.g., 32 for 32×32 patches)
--patches-per-image: Number of patches to extract per image (default: 64)
--detector: harris|fast|shi-tomasi|gradient (default: harris)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
Critical mismatch: shader used pixel-center coordinates while PyTorch
uses pixel-corner coordinates, causing 0.5-pixel offset.
PyTorch: linspace(0, 1, H) → [0, 1/(H-1), ..., 1]
Shader: (p.xy - 0.5) / (resolution - 1.0) to match
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
CNN output mismatch resolved: final layer (7→1) now clamps to [0,1].
Changes:
- Add clamp(sum, 0.0, 1.0) to cnn_conv3x3_7to1 and cnn_conv5x5_7to1
- Add generate_conv_final_function() to train_cnn.py for auto-generation
- Update comments to clarify clamping behavior
- Future exports will auto-generate final layers with correct clamp
PyTorch uses torch.clamp(out, 0.0, 1.0) on final output; shaders
were missing this critical operation, causing range mismatches.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
Compute gray once per fragment using dot() instead of per-layer.
Pass gray as f32 parameter to conv functions instead of vec4 original.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
|
|
- Added --infer flag for single-image inference
- Loads checkpoint, runs forward pass, saves PNG output
- Useful for verifying shader matches trained model
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
**Training changes:**
- Final layer now outputs [0,1] directly with torch.clamp()
- Removed denormalization step (was converting [-1,1] to [0,1])
- Network learns [0,1] output natively
**Shader generation fixes:**
- Layer 0 uses _src variant (5 params, normalizes [0,1] input internally)
- Removed pre-normalization of input texture (handled by _src)
- Final layer blending: gray_out already [0,1], no denormalization needed
- Added generate_conv_src_function() for all kernel sizes
- Auto-generates _src variants when exporting (skips if exists)
**Cleanup:**
- Removed obsolete 4-channel functions from cnn_conv5x5.wgsl
- Keep only 7-channel variants (_7to4, _7to1, _7to4_src)
**Normalization flow:**
[0,1] texture → _src normalizes to [-1,1] → tanh [-1,1] → ... → final conv [0,1] clipped
handoff(Claude): CNN normalization pipeline fixed and consistent with training
|
|
Normalize textures once in fs_main instead of in every conv function.
Keep all intermediate layers in [-1,1] range, denormalize only for final display.
Changes:
- train_cnn.py: Generator normalizes input once, keeps [-1,1] between layers
- cnn_conv*.wgsl: Remove texture normalization (already [-1,1])
- cnn_layer.wgsl: Regenerated with new normalization flow
- CNN_EFFECT.md: Updated documentation
Eliminates redundant [0,1]↔[-1,1] conversions, reducing shader complexity.
handoff(Claude): CNN normalization optimized, all tests passing (35/36).
|
|
Training script was hardcoded to generate cnn_conv3x3_* calls regardless
of actual kernel size, causing shader validation errors when layer 1 used
5×5 kernel (100 weights) but called 3×3 function (expected 36).
Changes:
- train_cnn.py: Generate correct conv function based on kernel_sizes[i]
- cnn_conv5x5.wgsl: Add cnn_conv5x5_7to4 and cnn_conv5x5_7to1 variants
- Regenerate cnn_layer.wgsl with correct function calls for [3,5,3]
- Document kernel size→function mapping in HOWTO.md
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
Upgrade CNN architecture to process RGBD input, output grayscale, with
7-channel layer inputs (RGBD + UV coords + grayscale).
Architecture changes:
- Inner layers: Conv2d(7→4) output RGBD
- Final layer: Conv2d(7→1) output grayscale
- All inputs normalized to [-1,1] for tanh activation
- Removed CoordConv2d in favor of unified 7-channel input
Training (train_cnn.py):
- SimpleCNN: 7→4 (inner), 7→1 (final) architecture
- Forward: Normalize RGBD/coords/gray to [-1,1]
- Weight export: array<array<f32, 8>, 36> (inner), array<f32, 8>, 9> (final)
- Dataset: Load RGBA (RGBD) input
Shaders (cnn_conv3x3.wgsl):
- Added cnn_conv3x3_7to4: 7-channel input → RGBD output
- Added cnn_conv3x3_7to1: 7-channel input → grayscale output
- Both normalize inputs and use flattened weight arrays
Documentation:
- CNN_EFFECT.md: Updated architecture, training, weight format
- CNN_RGBD_GRAYSCALE_SUMMARY.md: Implementation summary
- HOWTO.md: Added training command example
Next: Train with RGBD input data
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
|
|
Implements automatic layer chaining and generic framebuffer capture API for
multi-layer neural network effects with proper original input preservation.
Key changes:
- Effect::needs_framebuffer_capture() - generic API for pre-render capture
- MainSequence: auto-capture to "captured_frame" auxiliary texture
- CNNEffect: multi-layer support via layer_index/total_layers params
- seq_compiler: expands "layers=N" to N chained effect instances
- Shader: @binding(4) original_input available to all layers
- Training: generates layer switches and original input binding
- Blend: mix(original, result, blend_amount) uses layer 0 input
Timeline syntax: CNNEffect layers=3 blend=0.7
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
- Document coordinate-aware layer 0 architecture
- Add checkpointing examples and options table
- Consolidate training workflow with practical examples
- Clarify CoordConv2d usage and size impact
- Streamline training/README.md structure
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
- Save checkpoints every N epochs (--checkpoint-every)
- Resume from checkpoint (--resume)
- Store model, optimizer, epoch, loss, and architecture info
- Auto-create checkpoint directory
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
|
|
|
- Implement CoordConv2d custom layer accepting (x,y) patch center
- Split layer 0 weights: rgba_weights (9x mat4x4) + coord_weights (mat2x4)
- Add *_with_coord() functions to 3x3/5x5/7x7 convolution shaders
- Update training script to generate coordinate grid and export split weights
- Regenerate placeholder weights with new format
Size impact: +32B coord weights + ~100B shader code = +132B total
All 36 tests passing (100%)
handoff(Claude): CNN coordinate awareness implemented, ready for training
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
