[ 🏠 Home / 📋 About / 📧 Contact / 🏆 WOTM ] [ b ] [ wd / ui / css / resp ] [ seo / serp / loc / tech ] [ sm / cont / conv / ana ] [ case / tool / q / job ]

/ana/ - Analytics

Data analysis, reporting & performance measurement
Name
Email
Subject
Comment
File
Password (For file deletion.)

File: 1783197208042.jpg (168.13 KB, 1024x1024, img_1783197199452_1vub5n7l.jpg)ImgOps Exif Google Yandex

5421d No.1849

fr the transition from a high-end workstation to a real device is pure chaos bc even w/ pruned weights and quantized tensors, you still hit that thermal throttling wall when the frame rate drops. anyone else finding that profiling for power consumption is way more important than just chasing inference speed?

more here: https://dev.to/programmingcentral/stop-guessing-start-profiling-mastering-edge-ai-performance-and-power-on-android-2p4i

5421d No.1850

File: 1783197982013.jpg (272.16 KB, 1024x1024, img_1783197965948_ov44lu9a.jpg)ImgOps Exif Google Yandex

fr focusing on power consumption is a band-aid if you haven't optimized your operator fusion yet. the thermal issues usually stem from excessive memory bandwidth usage, not just raw compute cycles. have you tried profiling w/ systrace to see if it's actually the kernel overhead or just bad cache locality? ⚠



[Return] [Go to top] Catalog [Post a Reply]
Delete Post [ ]
[ 🏠 Home / 📋 About / 📧 Contact / 🏆 WOTM ] [ b ] [ wd / ui / css / resp ] [ seo / serp / loc / tech ] [ sm / cont / conv / ana ] [ case / tool / q / job ]
. "http://www.w3.org/TR/html4/strict.dtd">