talolard

9 ms·

31.
by talolard 5y ago
A quick summary/translation for those of us who don't speak ML. We keep hearing about these giant models like GPT3 with 1.5 billion paramaters. Parameters are the things that change when we train a model, you can think about them
32.
by talolard 5y ago
I think your last paragraph is the sales pitch for AWS. Hiring that level of expertise doesn’t scale. Easier and cheaper to hire 10x as many “developers” and pay the AWS bill than headhunt performance gurus that understand hardware and reta
33.
by talolard 5y ago
I think AWS Aurora is the most prominent thing that comes to mind. The key value proposition is the separation of storage from compute. That unlocks many promising features, “serverless” , a better parallelization story of OLAP queries etc
34.
by talolard 5y ago
A non ml way to approach this is to use phonetic distance, e.g. qanoon and kanun sound the same so they are close. There is an algorithm called Soundex with python implementations you can try.
35.
No-code tools at Lambda School(twitter.com)
6 points by talolard 5y ago|0 comments
36.
by talolard 5y ago
Simple can mean a few different things: fewest moving parts, easiest to code, minimal amount of stuff to learn… I’m good with Django. You can implement the app in Django and use Celery to do the scheduled work. I think that’s easiest to cod
37.
by talolard 5y ago
Someone on the internet is wrong!!! Since onsite is usually implied and rarely mentioned explicitly in the text, taking a corpus that mentions either term will bias heavily towards remote
38.
by talolard 5y ago
I subscribe to an alternative premise - "The browser is effectively an OS and thus a standardized distribution target". There are cases where the people I want to distribute to don't suffer from the constraints you mentioned
39.
by talolard 5y ago
Worked at a large bank and saw /worked on a few interesting projects. Obviously a lot of work on trading. One group of market makers had an RL agent built that would trade small sizes. The value here wasn't in profitability, it wa
40.
by talolard 5y ago
Hi HN, I wrote this and wanted to share a few more words. This post discusses a trick for training neural networks. Usually when we train a NN, we just show it what the correct answer is, but we don't say anything explicit about the
41.
Complement Objective Training with Pytorch Lightning(lighttag.io)
4 points by talolard 5y ago|2 comments
42.
by talolard 5y ago
Isn’t the middle ground writing custom allocators ? Allocators allocate blocks continuously and developers keep developing with standard OOP?
43.
by talolard 5y ago
I made a dev tool into a small business. Came hear to rant: invest in ui upfront. At the end, customers visit your site and you have very little time to make them go wow. Your spending a lot of attention capital when you don’t have much
44.
by talolard 5y ago
I'll offer an illustration. We sell our SaaS at X. We have a customer that needed stronger SLAs and servers located in Europe. They pay 8X and (I think) both sides feel like they got a great deal. So to paraphrase Arubis, this is a bus
45.
by talolard 5y ago
Congrats! I frequently deploy sklearn models to lambda with API gateway so this is spot on for me. I hate integrating lambda with gateway, building custom layers etc so this scratches a real itch. I’m not a fan of deploying my models on you
46.
by talolard 5y ago
While a great effort, the author conveniently forgot to mention their commercial competitors and even some open source ones. I’m from LightTag, we should be there as should explosion’s Prodi.gy and Labelbox. And where are docanno, Webanno a
47.
by talolard 5y ago
This is a profound insight.
48.
by talolard 5y ago
Dataish person here. If you search for Ml/ds you’ll get technical content that won’t help you much or fluff. You’d be better served by looking at companies you know do DS and searching for business problems they solved with it. Exampl
49.
by talolard 5y ago
As a data scientist who does more frontend, I think this is a really valuable concept. Hello by users/stakeholders engage with our work is the way to push it forward in the org and a dash of frontend can do wonders for getting that mes
50.
by talolard 6y ago
As a small vendor, I hate procurement platforms. Vendor UX always seems to be an after thought, since you’re the pipe between our customer and getting paid, why care about us. I hope you guys do it differently, maybe be so good to vendors t
51.
by talolard 6y ago
I have a question on the compute aspect regarding your business model, hope I’m not being to nosy.. I tried HL, the experience was stellar (well done!) and it made me think... To get AL working with a great user experience you need quite a
52.
by talolard 6y ago
I wrote a blog post a few years ago about possible downsides to active learning. https://www.lighttag.io/blog/active-learning-optimization-is... To be fair to the Humanloop folks, the criticisms probably don’t apply to
53.
by talolard 6y ago
I have a few suggestions regarding search performance. As others have mentioned, there isn't one "right answer" but there are a few nifty tricks you can use to boost performance. You could try a trigram index[0] as the primar
54.
Database Multi-Tenancy for SaaS(lighttag.io)
1 points by talolard 6y ago|0 comments
55.
by talolard 6y ago
Question for OP Can you share how you manage the borders between the open source product and the infrastructure for offering a SaaS As a trivial example, I’m assuming youre using plausible in plausible, do you have protections ensuring secr
56.
by talolard 6y ago
Yeah, Gatsby-image was what drew us into the platform to begin with. That's what made it fast to get started. I found the plugin system confusing, and poorly documented so it was hard to know how to get the most out of a plugin and adj
57.
by talolard 6y ago
Poor Gatsby. We run a Gatsby site, it was great to setup fast fast fast and has been a pain in the butt ever since. Next, from our limited experimentation, has much better ergonomics and it looks like the company has a stronger grasp on go
58.
by talolard 6y ago
If you have schools (e.g. institutions, not individuals) already paying, maybe experiment if you can make a pricing tier for institutions. E.g. Special School package 3k/yr includes unlimited tracks, group by classroom, some other sch
59.
by talolard 6y ago
Hi, Congrats on hitting that milestone . As others have said, the site looks great and it's clear you've put a lot of time into it. I can offer a few feedback points: The pricing tiers could have better names, from a quick glance
60.
by talolard 6y ago
If I understand the paper (which is questionable), that's what the author is aiming for. E.g. he's saying 1) We can make these amazing black boxes 2) We don't really understand them 3) But when we make them with gradient des

More ›