~/talk

ML on the Edge for Industry 4.0 with Arm Ethos N-78 Neural Processing Unit

on YouTube 28m
Talkat embedded world 2021 DIGITAL · 2021-03-01..03-05· [machine-learning, embedded-and-hardware]

speaker

name:Raviraj Mahatme
links:none on file

resources

video:https://youtu.be/WuMgQXDPr9g
slides:none on file

summary

Arm Ethos-N78 claims scalable edge inference from 1–10 TOPS with over 100 configurable variants (tunable SRAM, vector engines, clock) and supports int8/int16 workloads; on-chip lossless weight compression yields >50% weight compression for many layers. Driver stack is upgradable post-silicon; software flows include TVM AOT compilation and Arm NN on-device interpretation. Hardware tolerates up to 256-cycle read-latency tolerance, runs YOLOv3 at 416/608 and can handle Full‑HD real‑time inference by offloading work from a Cortex‑A host.

AI-written summary · about

chapters (22)

  1. 0:00Intro
  2. 0:24Applications of Machine Learning in Industry 4.0 • Visual Inspection
  3. 1:57Typical "Edge" Deployment Stack for ML Inference
  4. 4:17ML is a Two-Sided Problem That Benefits from a Scalable Platform
  5. 5:54Arm's Full Stack Solution for Deploying ML on the Edge
  6. 7:07Ethos-N78 NPU: Advances in ML Performance and Efficiel
  7. 8:45Ethos-N78: Unprecedented Flexibility for SoC Architects
  8. 10:16Run a Diversity of Networks
  9. 11:33Ethos-N78: Better Than Linear Scaling
  10. 12:40Ethos-N78 NPU: System Context
  11. 14:10Arm NPU Architecture : Flexible, Scalable, Programmable
  12. 16:05Processing Multiple Operators in Parallel in a Single Pass
  13. 18:59Designed to Tolerate Interconnect Latency Ethos-N7B can tolerate 256/64 cycles of R/W latency
  14. 21:02Ethos-N78 : Software Driver Stack Architecture
  15. 21:16Ethos-N78: Software Driver Stack Architecture
  16. 22:16Ethos-N78 NPU: Upgradable Software Architecture
  17. 22:44Arm NPU ML Software Stack Two flows supported to deploy ML inferencing on Ethos-N NPUS
  18. 23:34Code Generation Software and Tools Platform Offline flow based on TVM compiler for embedded applications under Linux
  19. 24:18Why Model Conditioning is important? Non optimized
  20. 25:19Ethos-N NPU Static Performance Analyzer Tool
  21. 25:51Ethos-N NPU Static Performance Analyzer - Sample Data
  22. 26:10Mapping Use Cases Industry 4.0 to ML Deployment

related talks

~/search
type to search 6,249 conferences, 138,987 speakers and 185,172 talks
~/help
cmdctrl k
search every conference, speaker and series
/
focus the filter search
h
home
c
conferences
s
speakers
e
series
t
light / dark — the status bar button does the same
?
this help
esc
close this, then clear every filter — or just the search, while you are typing in it

Ignored while you are typing. Pages without a filter rail have no search to focus.