Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
zfountas
searching Neon…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
4 ms
·
1.
▲
by
zfountas
1y ago
Hi MacsHeadroom, first author here. Thanks for the great questions about compute/memory trade-offs. The quick take: To give you an example of processing speed, with a 7B model on an NVIDIA V100, EM-LLM processes (or generates) about 32
2.
▲
by
zfountas
7y ago
I found this quite handy actually, good stuff!