Spaces:
Running
Running
Commit History
19 days ago
-
Faster WebGPU sampling: fp16 compute, GPU-resident steps, non-blocking thinker preview
Stefatorus +1: Claude Opus 5.5- 002/003 and 001 load the v2 build from webgpu/v2/ in each model repo: one graph run per Euler step (CFG mix + update inside the graph), z stays on the GPU between steps, fp16 compute (fp32 for graph I/O, CFG/Euler, timestep/resolution/count embeddings and Range), 1x1 conv + depth-to-space instead of ConvTranspose. - 003 @ 512: 275 -> 161 ms/step, 14 -> 8.1 s/image; 256 px and 001/002: ~6 -> ~3.4 s (RTX 4060, Edge). - Show thinker: async readback, colouring in a Web Worker, frames dropped while busy; no cost to sampling. - "Fast" toggle: 25 steps (~4.1 s at 512); default stays 50. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018EtDckYfNgMbkwqVmTsdha146ba97 -
Preview 002 and 003 (512 px), version selector, Show thinker, AI-output marking
Stefatorus +1: Claude Opus 5.5- Model selector (003 default for a first visit, last choice remembered per browser); 003 at 512 or 256 px with its prompt pipeline (normaliser, spelled-out quoted text, count code, "no X" -> negative) and SD3 shift. - The ONNX files now load from each release's model repo (Logolabs/agate-preview-00X, webgpu/); the graphs also output the thinker plan. models/ (the original 001 build) stays for ?models=./models/. - Show thinker (default on, ~2-4% per step): the 16 x 16 plan (PCA colours, as the ComfyUI live preview) and the expected image x1 at every step. - Every image is marked: invisible-watermark dwtDctSvd (payload AGATE + release, bit-identical JS port), PNG provenance text chunks on save (no prompt), and a visible "AI-generated" note. - Parity fixtures for 002 and 003, and a 38-prompt golden test of the prompt pipeline (?golden=1). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018EtDckYfNgMbkwqVmTsdha2dc8d66