Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
krona
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
101 ms
·
1201.
▲
by
krona
11y ago
Given the current memory utilization of TF during training, I doubt it's ready for embedded systems.
1202.
▲
by
krona
11y ago
why CUDA, and not OpenCL? The library appears to be using cuBLAS; what's the OpenCL alternative with comparable run-time performance and functionality?