summaryrefslogtreecommitdiff
path: root/workspaces/main
AgeCommit message (Expand)Author
9 hourssome tweakingskal
12 hoursrefactor(wgsl): consolidate SDF shapes into single common fileskal
12 hoursrefactor(wgsl): replace inline fullscreen_vs with common includeskal
12 hoursrefactor(wgsl): modularize common shader functionsskal
14 hoursFix: CNN v2 compute shader validation errorskal
20 hoursCNN v2: bilinear mip-level sampling and UI improvementsskal
25 hoursReplace hard clamp with sigmoid activation in CNN v2skal
26 hoursFix --mix option: blend prev layer with static p4-p7, not p0-p3skal
28 hoursCNN v2: Alpha channel depth handling and layer visualizationskal
34 hoursCNN v2: Fix WebGPU validation error in uniform buffer alignmentskal
34 hoursCNN v2: Change feature #6 from sin(10*x) to sin(20*y)skal
34 hoursCNN v2: Add TODO for flexible feature layout in binary format v3skal
34 hoursCNN v2: Add mip-level support to runtime effectskal
38 hoursAdd pop-punk drum track from documented sequenceskal
38 hoursCNNv2Effect: Document per-layer kernel sizes supportskal
39 hoursCNN v2: Refactor to uniform 12D→4D architectureskal
43 hoursRemediation: Implement shared common/shaders/ directoryskal
43 hoursUpdate workspace.cfg for new directory structureskal
43 hoursAdd weights/ subdirectory to workspaces for CNN training outputsskal
43 hoursRefactor: Reorganize workspaces and remove assets/ directoryskal
43 hoursRefactor: Move application entry points to src/app/skal
2 daystest_demo: Add beat-synchronized CNN post-processing with version selectionskal
3 daysTimeline editor: CPU load visualization + UX improvementsskal
3 daysRefine training script output and validationskal
3 daysTODO: 8-bit weight quantization for 2× size reductionskal
3 daysCNN v2: storage buffer architecture foundationskal
3 daysCNN v2: parametric static features - Phases 1-4skal
3 dayscleanup: remove test-only files from main workspaceskal
3 daysfix: update shader files to use beat_phase instead of beatskal
3 daysfeat: implement beat-based timing systemskal
3 daysadd trained layersskal
3 daysdocs: Update CNN comments and add bias fix summaryskal
3 daysfix: CNN bias accumulation and output format improvementsskal
3 daysupdate cnn codeskal
3 daysfix: Register cnn_conv1x1 snippet and add verificationskal
4 daysfix: Move sigmoid activation to call site in CNN layer shaderskal
4 daysfix: Replace clamp with sigmoid in CNN final layerskal
4 daysformat .wgsl layer code (cosmetics)skal
4 daysopt: Move invariant in1 calculation outside CNN convolution loopsskal
4 daysopt: Vec4-optimize CNN convolution shaders for SIMDskal
4 dayschore: Update CNN architecture to 3×3×3 with new trained weightsskal
4 daysfix: Correct UV coordinate computation to match PyTorch linspaceskal
4 daysfix: Add clamp to CNN final layer to match PyTorch trainingskal
4 daysrefactor: Optimize CNN grayscale computationskal
4 daysupdate train_cnn.py and shaderskal
4 daysfix: CNN training normalization pipeline consistencyskal
4 daysudpate CNN shader code.skal
4 daysrefactor: Optimize CNN normalization to eliminate redundant conversionsskal
4 daysupdate timeline.seqskal
4 daysfix: Flip Y-axis to match ShaderToy coordinate conventionskal