Architecture summary: GPU neural network training from scratch with OpenCL @sphaerophoria
Architecture summary: GPU neural network training from scratch with OpenCL @sphaerophoria
Uploaded August 2025 | Updated September 2026, 1 week ago
I kinda wrapped this section up off screen, so here's a little overall summary. Link to code for further inspection

github.com/sphaerophoria/sphmelly/commit/f2b1691fb316cdc7db2d4469958b9f3a93249d28

Please consider supporting. This content WILL end some day, but every dollar I make pushes that day further out

Join on youtube for happy hour vods youtube.com/channel/UCXzL31BCxf8En1KT34gSK6g/join
Or on patreon patreon.com/sphaerophoria
Or on twitch twitch.tv/sphaerophoria
Or sponsor me on github (no benefits) github.com/sponsors/sphaerophoria
Architecture summary: GPU neural network training from scratch with OpenCL3D math hurts my head sometimesHTTP client/server interfaces, a year laterModding a compiler so I can find my bugsI put a window manager in my window managerQuick and dirty HTML draggable listDay 6 has hands, day 7 is easy peasyDay 2 was way easierThis shouldnt be happening2 off by 1s make a rightWebsockets are actually not that badI trained a convolutional neural network with no dependencies
sphaerophoria |

Architecture summary: GPU neural network training from scratch with OpenCL

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER