ramoz

64 ms·

931.
by ramoz 5y ago
x2 Gunicorn workers, MKL mapped to half physical cores for each... for some reason the model (tensorflow) performs better on half vs 1 worker
932.
by ramoz 5y ago
CPU, GPU doesn't work out well in our case. There is the data transfer cost as well as memory constraint & the model blows up in memory for every inference call.
933.
by ramoz 5y ago
I spent the past year and a half deploying a distributed backend for bert-like models & we ultimately chose a K8s architecture & "precise" affinity mapped out, which is still hard due to cpu pinning issues. On the frontend
934.
by ramoz 5y ago
There is a "hidden", for lack of better word, world of commercial and political pysops most people are never exposed to beyond their social media feeds. Often masked in marketing, but even more so in blanketed initiatives run by i
935.
by ramoz 5y ago
Need one for TikTok. please help.
936.
by ramoz 5y ago
I do think this turns into a modern stack overflow however. If observing local runtime errors, debugging process, and tidying/fixing code it produces itself. Plus train on all the SO questions out there.
937.
by ramoz 5y ago
How is GitHub/OpenAI ensuring this tech doesn’t throw the industry towards a flawed human/tech capability or dystopian of sorts? Like… a bubble of developing code from yesterday’s code (plenty flawed itself) and with exponential g
938.
by ramoz 5y ago
Nothing crazy, but something I always laugh at. I was so excited to meet a legit/professional dev team the first day of my career. I was paired with a Sr dev and sat in his cubicle so he could show me some backend update to a prod app
939.
by ramoz 5y ago
This, itself, is dangerous thinking.
940.
by ramoz 5y ago
Im not a researcher, but why would any study of some social environment not be worth studying? Especially as it relates to present day & macro-scale social dilemmas, and more importantly a robust history to learn from vs repeat –– Thoug
941.
by ramoz 5y ago
Wonder if that includes books about burning books, struggle sessions, or psychoanalysis within political sociology, and ... how radicalized progressive movements develop — likely not a part of any curriculum anyway vs redefining them as new
942.
by ramoz 5y ago
Is it though? I checked my initial bias of assuming the rest of the blog would be well written after the author declared himself a skilled writer along with the accompanying validation of another person deeming so as well. I was still a bit
943.
by ramoz 5y ago
Kubernetes https://www.cncf.io/the-childrens-illustrated-guide-to-kuber...
944.
by ramoz 5y ago
I'd pay for the magical place sequel.
945.
by ramoz 5y ago
Why walnuts & almonds?
946.
by ramoz 5y ago
Random perspective: I think the author is unintentionally speaking of a future not quite here yet & specifically with AI integrated into applications & their infrastructure. Essentially the need for low data gravity with how AI work
947.
Ask HN: Who Is Looking for Investors?
3 points by ramoz 5y ago|3 comments
948.
by ramoz 6y ago
I feel ya, and I work at ACN Fed :) The fluff is a thing... Luckily I get to lead a lot of what I do and in the applied intelligence organization where we're doing some interesting things as well as some fairly advanced engineering.
949.
by ramoz 6y ago
You're explaining your understanding of commercial consultancy, but are obviously (and you note at the end) not detailing your knowledge of systems integrators that don't always happen off-shore. For instance, many large companies
950.
by ramoz 6y ago
I think that's an intriguing point - and I also feel we generalize "artificial general intelligence" to the point & in such a way where humans have yet to even achieve that level of intelligence. How can we build smart sy
951.
by ramoz 6y ago
What an impressive site Tech details: https://www.datamuse.com/api/
952.
Rhymezone(rhymezone.com)
2 points by ramoz 6y ago|1 comments
953.
by ramoz 6y ago
chill https://nvidianews.nvidia.com/news/nvidia-provides-u-s-posta... recent NVIDIA conference talks will detail some use cases. are you interested in federal contract jobs tho? hmu
954.
Using NFTs, IPwe and IBM seek to unlock $1T or more in patents(cnbc.com)
1 points by ramoz 6y ago|0 comments
955.
by ramoz 6y ago
I think we assume too much about where infrastructure is headed. Serverless, containers... they're fueling this notion of increased productivity but I don't think they themselves are keeping up with where Software is headed. E.g.
956.
by ramoz 6y ago
I align with most of these purists. But I don't think the verbosity matters. Its not a bottleneck when you're using Java because you shouldn't be in a position to care about such things. Hell, it's probably beneficial in
957.
by ramoz 6y ago
I love your comment,"YOLO-ing" gave me a laugh, and a laugh at myself in some previously stressful/complex deployment periods. Terraform was not a friend there. That said, I've worked on/with a team enabling some f
958.
by ramoz 6y ago
The only reason I prefer Mac is for the trackpad. Windows pad has always seemed like a terrible flat version of some mouse/scroll interaction. Is this haptic or not? Personally, no other detail matters.
959.
by ramoz 6y ago
I was in some of those CH convos with you. I was actually suspended for a little while and tried clearing it up with them. Sent all the details I had and the original google doc I published w/ a lot of styprs work. They never responde
960.
by ramoz 6y ago
True. I know it exists for inference though. Wondering where/when solutions like MKL might work for training.

More ›