0xbear

19 ms·

181.
by 0xbear 9y ago
Just one instance of our product has to process 39 million inferences per month (number is dictated by frame rate), with predictable latency, 24x7, on both Windows and Linux.
182.
by 0xbear 9y ago
Because TF lets you run _just_ the model, with no Python stuff at all. For us there is no “server side”, our stuff is deployed on prem. The fewer variables there are the easier our lives are going to be. Additionally, PyTorch download page
183.
by 0xbear 9y ago
In my experience (computer vision, deep learning) PyTorch is substantially faster as well, especially in data augmentation where it’s not just a thin layer over cudnn. That said, you’re right. There’s no way I’d deploy it to production.
184.
by 0xbear 9y ago
Being clear about that kind of thing would be a serious career limiting move in all of the companies I worked at. No one likes a whiner, even if whining is justified. Simply put, in a 1:1 you're not in a position of power and therefore
185.
by 0xbear 9y ago
In a small company, maybe. In a large one, not at all. Get that counter, negotiate a promo, change teams 6 months later. Everyone is happy.
186.
by 0xbear 9y ago
Where can I buy a board with this?
187.
by 0xbear 9y ago
You mean his comment where he got out of his way to first repeatedly say he does not approve of KKK in any way? What’s with the feigned outrage?
188.
by 0xbear 9y ago
Except in case of profile-guided optimization (which few people use), the compiler has one massive weakness: it doesn't know the "shape" and distribution of your data. If it did, it would be able to do quite a bit more, but i
189.
by 0xbear 9y ago
No. Just show up to work in a pro-Trump baseball hat, or speak openly against illegal immigration, and see what happens. No one cares about the London office.
190.
by 0xbear 9y ago
Cookie-cutter NDAs are nearly always very one-sided. Some of the NDAs our company signs with other companies take weeks of lawyers bickering about things that to the uninitiated seem like useless bullshit. But the fact that those things are
191.
by 0xbear 9y ago
Data centers aren’t built in such places. And for just a few machines a simple building HVAC will do fine.
192.
by 0xbear 9y ago
The whole “electricity costs a lot” argument is getting tiresome. Where I live, electricity is 11 cents per KWh. That means even if you run that machine full tilt 24x7 (which you won’t), and even if it draws 1KW (which it won’t), it’s still
193.
by 0xbear 9y ago
I wish I knew how to “watch Netflix while working”.
194.
by 0xbear 9y ago
+1 having done this and having witnessed the staggering waste and incompetence on the part of people who should know better, I now feel really bad when paying my taxes. Truly, it’s better to not know how sausage is made.
195.
by 0xbear 9y ago
Next thing to criminalize is thinking “extremist” thoughts. All for the benefit of her majesty’s subjects, of course.
196.
by 0xbear 9y ago
Apple’s iOS 11 voice is pretty amazing too. I wonder what they are using.
197.
by 0xbear 9y ago
That’s not because of language. That’s because of commonality in behavior and cultural background. Translating earbuds don’t solve that.
198.
by 0xbear 9y ago
On the positive side, their chief security officer can compose one heck of a musical composition.
199.
by 0xbear 9y ago
You can’t even use their main product without loading in opaque binary blobs and agreeing to a backdoor.
200.
by 0xbear 9y ago
The words Intel and “open” do not really belong in the same sentence.
201.
by 0xbear 9y ago
This is so obviously not going to fly unless they either start bundling with Xeons, or offer something Xilinx can’t do. Cloud providers aren’t stupid, and there are enough eggs in the Intel basket already.
202.
by 0xbear 9y ago
https://www.bloomberg.com/view/articles/2017-09-30/no-trump-...
203.
by 0xbear 9y ago
Seems like her primary goal is to virtue signal to US liberals,not to help her people. Because the Trump administration is already providing help over and above of the what the US would normally provide.
204.
by 0xbear 9y ago
They’re also mostly doing the kind of work that government aerospace agencies have utterly nailed in the 60s. In fact in some aspects (transporting humans to and fro) they aren’t quite there yet. Reusability just wasn’t a concern when budge
205.
by 0xbear 9y ago
Yup. The only people who like communism are the ones who have never lived under a communist regime. Ask any Eastern European over 40 about what that’s like before you make any rash decisions out of ignorance.
206.
by 0xbear 9y ago
For juveniles, IMO, nothing they say before meeting with their lawyer should be admissible in court. It’s as simple as that. These Miranda rights should just point blank tell them to shut up before they talk to a lawyer. Juveniles are just
207.
by 0xbear 9y ago
Come to think of it, google has pissed away so much user goodwill over the years. There was a time when they could do no wrong. I guess same as MS in early 90s. Now they are nearing a point where with every product release they have to prov
208.
by 0xbear 9y ago
Speaking of which, any possibility of open sourcing this? We are looking to get into FPGAs and would rather not reinvent a bicycle.
209.
by 0xbear 9y ago
If a minute to build and link half a million lines of C++ is “astronomical”, then sure. Try to build that on your local box, then get back to me. Go exists solely because Rob Pike had nothing better to do, and a lot of googlers hate Java.
210.
by 0xbear 9y ago
Bazel is on its way to be more like Google’s internal implementation: distributed builds, object file cache, distributed tests. That means your C++ build and test runs will take barely any time at all, same as they do at Google. That, in tu

More ›