eclectic29

8 ms·

31.
by eclectic29 3y ago
FWIW Ollama has no concurrency support even though llama.cpp's server component (the thing that Ollama actually uses) supports it. Besides, you can't have more than 1 model running. Unloading and loading models is not free. Again,
32.
by eclectic29 3y ago
Few hundred on the planet? Are you kidding me? We're asking enterprises to run LLMs on-premise (I'm intentionally discounting the cloud scenario where the traffic rates are much higher). That's way more than a hundred and sor
33.
by eclectic29 3y ago
I'm not sure why Ollama garners so much attention. It has limited value - used for only experimenting with models + cannot support more than 1 model at a time. It's not meant for production deployments. Granted that it makes the e
34.
by eclectic29 3y ago
I migrated to Joplin from Evernote. I can live with a slightly less polished UI. However the main issue is the time it takes to come up (launch) on my Windows 32GB RAM, Intel i7 laptop. It’s terribly slow to come up.
35.
by eclectic29 3y ago
Will Sundar take true ownership of all the debacles after getting rushed to produce something in AI and resign?
36.
Scaling ChatGPT: Five Real-World Engineering Challenges(newsletter.pragmaticengineer.com)
2 points by eclectic29 3y ago|0 comments
37.
by eclectic29 3y ago
Please excuse me for asking this. I know Andrej is an excellent instructor. I've watched his videos. But what has been his contribution to the industry (besides teaching of course)?
38.
by eclectic29 3y ago
Did you try it? Did it work for you? I would first try it before looking for alternatives. This is what is called analysis-paralysis. You're simply wasting your precious time looking for alternatives without even giving it a shot.
39.
by eclectic29 3y ago
Genuine question: why does kalman filter come up so frequently on HN? Is this something I'm missing? I'm a machine learning engineer, not a data scientist.
40.
by eclectic29 3y ago
Why do we allow someone non technical, manipulative and a glorified salesman to run a deeply technological initiative like AI! The stupid media has made him the God of AI. He has no background in AI , no success stories to beat his chest ot
41.
by eclectic29 3y ago
Any googlers chime in? Is this any good? Or just a dump of one of their internal unused products?
42.
by eclectic29 3y ago
Btw, where did you see that line of text in the docs? I didn't find it.
43.
by eclectic29 3y ago
If you’re serious about production deployments vLLM is the best open source product out there. (I’m not affiliated with it)
44.
by eclectic29 3y ago
Thanks. That’s fair. Just shocked by the extreme hype.
45.
AskHN: What's a real world production use case for Ollama?
11 points by eclectic29 3y ago|6 comments
46.
by eclectic29 3y ago
Once you've tested to your heart's content, you'll deploy your model in production. So, looks like this is really just a dev use case, not a production use case.
47.
by eclectic29 3y ago
What's the use case of Ollama? Why should I not use llama.cpp directly?
48.
by eclectic29 3y ago
Wow! Another configuration language post. 2nd day in a row. And that too on the front page. It's 2024 and the software industry has not solved the problem of configuration yet. Umpteen solutions every year - languages, formats, etc - a
49.
by eclectic29 3y ago
It's 2024 and the software industry has not solved the problem of configuration yet. Umpteen solutions every year - languages, formats, etc - all for just doing configuration. Just wow! It amazes me. It would be so much better if all t
50.
by eclectic29 3y ago
I mean you still need to learn the material but you cannot pass these interviews without memorizing.
51.
by eclectic29 3y ago
I've 20 yrs of work experience in small to large tech companies and thinking of switching to a new company. Even I'm expected to be swift with leetcode style questions. Now, it's not like I can't do it; I've been an
52.
by eclectic29 3y ago
I moved to Joplin a few months ago. The editor is not as good as Evernote but does the job. I can use it on as many machines as I want but I always remember to manually sync to avoid loss. Overall, Joplin is decent and functional without th
53.
by eclectic29 3y ago
"Open"AI and for-profit. This company is a troubled mess and if it were not for the "potential" money, employees wouldn't be putting up with this nonsense. Sad to see Ilya getting sidelined as shown by this "..
54.
by eclectic29 3y ago
I'm sorry, impressed by what?
55.
by eclectic29 3y ago
The media and the VCs are treating Sam like some hero and savior of AI. I’m not getting it. What has he done in life and/or AI to deserve so much respect and admiration? Why don’t top researchers and scientists get equivalent (if not m
56.
OpenAI's unique corporate structure left Sam Altman vulnerable. Now he's out(cnbc.com)
2 points by eclectic29 3y ago|0 comments
57.
by eclectic29 3y ago
Viaweb was acquired by Yahoo and rebranded as Yahoo Stores. Happened to read his code when I was at Yahoo Stores. Lisp at its best! He wrote the templating engine running behind the UIs. It was used until 2008 when I left. Not sure if Yahoo
58.
by eclectic29 3y ago
Is BM25 still used by "modern" search engines? I wasn't aware.
59.
by eclectic29 3y ago
Nit: use "enum class". See https://isocpp.org/wiki/faq/cpp11-language-types#enum-class .
60.
Lessons Learnt from Consolidating ML Models in Large Scale Recommendation System(netflixtechblog.medium.com)
2 points by eclectic29 3y ago|0 comments

More ›