daniel
0 followers · 0 following
  • Joined on 2026-08-07
Repositories
2
 Projects Packages Public activity Starred repositories
daniel pushed to vae-decode-optimization at daniel/h3-blackwell-runtime 2026-08-14 19:26:51 +00:00
b3a2544936 Match NVFP4 block scale swizzle
daniel pushed to vae-decode-optimization at daniel/h3-blackwell-runtime 2026-08-14 19:24:10 +00:00
57b84b8d6b Match NVFP4 block scale padding
daniel pushed to vae-decode-optimization at daniel/h3-blackwell-runtime 2026-08-14 19:22:39 +00:00
0e855b8e55 Prototype Vortex native NVFP4 quantizer
daniel pushed to vae-decode-optimization at daniel/h3-blackwell-runtime 2026-08-14 19:15:22 +00:00
1fec77d2e9 Tune Vortex NVFP4 scale geometry
daniel pushed to vae-decode-optimization at daniel/h3-blackwell-runtime 2026-08-14 19:11:48 +00:00
1cc0ffdd00 Add Vortex NVFP4 scale v1
daniel pushed to vae-decode-optimization at daniel/h3-blackwell-runtime 2026-08-14 18:57:27 +00:00
684b645546 Match Torch NVFP4 scale precision
daniel pushed to vae-decode-optimization at daniel/h3-blackwell-runtime 2026-08-14 18:54:04 +00:00
0fda84502a Add Vortex NVFP4 scale kernel
daniel pushed to vae-decode-optimization at daniel/h3-blackwell-runtime 2026-08-14 18:47:04 +00:00
03dbe1456d Prototype Vortex NVFP4 quantizer seam
daniel pushed to vae-decode-optimization at daniel/h3-blackwell-runtime 2026-08-14 18:37:45 +00:00
d6eabb150e Add NVFP4 linear profiler
daniel pushed to vae-decode-optimization at daniel/h3-blackwell-runtime 2026-08-14 18:32:20 +00:00
7cc03f3e57 Add H3 block profiler plan
daniel pushed to vae-decode-optimization at daniel/h3-blackwell-runtime 2026-08-14 14:07:46 +00:00
3d0c093168 Add Sol-Attn backend
75ee9ba4ca Document backend implementation status
c3cc04e98d Add approximate H3 cache modes
fb257ef982 Add exact memory backend options
eaf9324145 Add KJ Sage attention backends
Compare 9 commits »
daniel pushed to vae-decode-optimization at daniel/h3-blackwell-runtime 2026-08-14 07:58:12 +00:00
ea7712b4b2 Use paired kitchen RoPE in VAE fast path
daniel pushed to vae-decode-optimization at daniel/h3-blackwell-runtime 2026-08-14 07:54:40 +00:00
8595dd875e Add optional kitchen VAE fast ops
daniel created branch vae-decode-optimization in daniel/h3-blackwell-runtime 2026-08-14 07:36:45 +00:00
daniel pushed to vae-decode-optimization at daniel/h3-blackwell-runtime 2026-08-14 07:36:45 +00:00
f6166ede68 Add VAE decode benchmark tool
daniel pushed to master at daniel/h3-blackwell-runtime 2026-08-14 07:20:47 +00:00
ea4f4a2dad Add Comfy cat benchmark submitter
daniel pushed to master at daniel/h3-blackwell-runtime 2026-08-14 07:11:38 +00:00
53cd8bd4f4 Make runtime diagnostics opt in
daniel pushed to master at daniel/h3-blackwell-runtime 2026-08-14 06:57:31 +00:00
fea69673cd Return native H3 audio latents
daniel pushed to master at daniel/h3-blackwell-runtime 2026-08-14 06:39:47 +00:00
59b8302d4e Add direct H3 audio VAE decode
daniel pushed to master at daniel/h3-blackwell-runtime 2026-08-14 06:33:10 +00:00
c5cf42850b Document H3 audio VAE path