nharada

16 ms·

151.
by nharada 3y ago
Has anyone tried this? I’ve seen the same people hyping this up all over /r/ml with multiple posts a week, marketing material, etc. Is this a legit new technique or just cleverly marketed paper?
152.
by nharada 3y ago
The variation even for very nearby McDonalds is fascinating. In the bay area I see differences of over $1 for locations basically next door to each other.
153.
by nharada 3y ago
That’s awesome, I wonder what the cost per MW came out to.
154.
by nharada 3y ago
Juan Browne's YouTube channel posts about basically every aviation accident, and as usual he somehow already has a video about this: https://www.youtube.com/watch?v=I9EvHpf8jZg
155.
by nharada 3y ago
One interesting consequence of the model consolidation happening in the LLM world is that it's actually easier for AMD's hardware to be competitive. It used to be that you had to support a ton of ops since nobody is going to use y
156.
by nharada 3y ago
You mean you want to remove the images because you get false positives? We've thought about dataset curation and how to manage that, ranging from full on "you can build your own dataset from scratch" to "refine results u
157.
by nharada 3y ago
Not yet! This is something we're thinking about, but couldn't afford to offer for free since there's some amount of pre-processing/caching with GPUs required. Maybe we could instead have users somehow pre-process/ho
158.
by nharada 3y ago
Ahh yeah that one is just a warning but I should see if I can improve it. It's basically saying that it's using CPU not GPU. EDIT: Fixed in 0.1.8 Good ideas too, especially base64 should be fairly easy to implement. I've want
159.
by nharada 3y ago
Thank you! The goal is to make it easier for devs without a bunch of ML background to add CV features to their apps. It goes off the rails when the underlying models have weird quirks, so for example if you search "ginger" you get
160.
Show HN: Build an open-source computer vision model in seconds using text(usezeroshot.com)
64 points by nharada 3y ago|14 comments
161.
by nharada 3y ago
At this point I'll settle for the "good sleep" genes
162.
by nharada 3y ago
It's kinda funny that the latest version of a "less space than a Nomad" comment now holds up Apple specifically as the product of comparison.
163.
by nharada 3y ago
> content of interest could be deleted next second HBO hasn't exactly been great about this either
164.
by nharada 3y ago
I think where this breaks down is when you have something rare that requires immediate attention. My impression is that a lot of that medical training is being able to say "oh that's unusual, you need to see a specialist".
165.
by nharada 3y ago
> everything was just peachy ???
166.
by nharada 3y ago
Yeah it takes time to "save preferences" if you don't give it maximum permissions.
167.
by nharada 3y ago
> that kind of money invites the ESG crowd what does that mean?
168.
by nharada 3y ago
Sometimes I need 512 GPUs for 3 days.
169.
by nharada 3y ago
For large scale ML Jax is pretty nice. It makes the multi-host computation feel like a first class citizen and your programs tend to be designed with that in mind.
170.
by nharada 3y ago
Isn't the counter-example in America cars? It's definitely cheaper to have an econo-box but instead we drive 8 seat tanks to get groceries.
171.
by nharada 3y ago
Is this a feature that some trackers offer? When I worked at Fitbit I noticed this correlation (along with resting heart rate) and developed a model that could predict onset of illness a few days in advance with fairly good accuracy. But un
172.
by nharada 3y ago
Showing roughly 1.5-2.0x speed ups on a couple of example workloads under fp8. Not bad for what's basically just a software change. Just one more thing that will make it hard for AMD to catch up.
173.
by nharada 3y ago
I don't think this is unreasonable since the target audience is other researchers. When you try and reproduce this paper you won't think "wow there's a lot of API errors I must be doing something wrong".
174.
by nharada 3y ago
This is cool -- I built something very close to this and I think it's really useful for researchers to have access to pre-trained models. Training something like this isn't that challenging given the relatively limited scale (100M
175.
by nharada 3y ago
Under a week of work to publish potentially the highest impact paper of your PhD (especially if you're first)? I'm guessing most students would take that chance.
176.
by nharada 3y ago
Neat. I'm assuming these would be fairly limited to niche installations, since the cost would almost certainly not be competitive to current panels in an installation where you have the ability to just add more area (not to mention the
177.
by nharada 3y ago
I was curious how much bigger the real stadium parking was. SimCity: https://oyster.ignimgs.com/mediawiki/apis.ign.com/simcity/e/... Real Stadium: https://i.imgur.com/KwT1T53.png
178.
by nharada 3y ago
What kind of conversion efficiency would you expect from a range extender, like 90% or so?
179.
by nharada 3y ago
> the pain of the procedure itself (such as drilling a cavity) was the same I've not had cavities filled without anesthetic, but are you sure someone wasn't just fucking up the procedure? I've had plenty of dental work and
180.
by nharada 3y ago
Not directly about the WGA strike, but it's interesting seeing the reactions here vs if this had been said by say, a tech exec about software engineers.

More ›