Nimitz14

33 ms·

361.
by Nimitz14 6y ago
> It does make it invalid, when the difference between runs is as big as 80,000 estimated deaths which can lead to dramatically different government policies. Wrong. Nobody decides policy based on whether there will be 320k or 400k death
362.
by Nimitz14 6y ago
That's a huge intersection, and it sounds like you were crossing instead of just following Embarcadero, that's obviously a terrible idea.
363.
by Nimitz14 6y ago
> Homeschool kids are exposed to a much wider variety of social situations which develops a broad base of social skills and preparation for adult responsibility much more effectively than the stifling and artificial environment of public
364.
by Nimitz14 7y ago
Hm, I was actually thinking that local authorities should be strengthened. But the perspective I'm coming from is being European and seeing all major EU countries have unpopular governments that are seen as ineffective, and contrasting
365.
by Nimitz14 7y ago
You should take a look at austrian suggestions 2009-2019.
366.
by Nimitz14 7y ago
Yup. Copying my post from reddit: Seems like the title should really be criticisms of STT groups in the industry, work from academic groups is ignored. The original article was kind of silly anyways, it was all engineering work, very little
367.
by Nimitz14 7y ago
I enjoy reading something that is written by someone passionate about the topic, and I enjoyed this perspective, but I disagree based on my admittedly limited experience with DD. I've read white noise and zero-k, and found both underwh
368.
by Nimitz14 7y ago
Firstly, realize you're not a consultant. You're at best a contractor.
369.
by Nimitz14 7y ago
You seem to have falsely generalised based on your experiences with whatever local airlines you use. Norwegian and other european carriers have quite a good reputation, people understand that the low cost has a price.
370.
by Nimitz14 7y ago
Is this satire? I can't tell.
371.
by Nimitz14 7y ago
Check the github and the docs. Funny reading this now, I literally just implemented something in 15min using C++/pybind11 (and operates on numpy arrays) that is an order of magnitude faster than when I used numpy operations.
372.
by Nimitz14 7y ago
It's not the same, when I wrote > (can't breastfeed, milk in a bottle is not the same) that was pumped milk.
373.
by Nimitz14 7y ago
A counterpoint: A friend of mine had a kid a year ago and he found it impossible to take care of the baby when he tried being the caretaker while the mom was away (can't breastfeed, milk in a bottle is not the same). The experience has
374.
by Nimitz14 7y ago
OpenAI doesn't have any products right? So research is a lot more important than production for them. Surprises me it took them this long.
375.
by Nimitz14 7y ago
You are right I've never tried out wav2letter on a large amount of data. Cool project! I'll try out your models.
376.
by Nimitz14 7y ago
The reason those kaldi results are noticeably worse is because they're not using RNN rescoring. The E2E models also take far more compute to train (facebook and google train their models for hundreds of epochs, kaldi for... 4!). All th
377.
by Nimitz14 7y ago
You should look at the latest recipe's for tedlium, librispeech, swbd etc you'll get significantly better results with those models.
378.
by Nimitz14 7y ago
It is very unlikely a E2E toolkit can outperform kaldi if they're trained on the same data. Also I suspect these guys aren't using the latest kaldi architectures.
379.
by Nimitz14 7y ago
I think it's useful for when your feature columns have names. Otherwise yeah not really.
380.
by Nimitz14 7y ago
I think his points are valid. Speaking as someone who has spent the past 18 months doing my own thing. > I also work full-time remotely. I built a community around myself, made new friends and regularly hang out with people I met at coff
381.
by Nimitz14 7y ago
It is misleading to use librispeech WER as a general guide to real world WER. Don't do that.
382.
by Nimitz14 7y ago
Yeah, trading off embedding dim vs hidden layer dim is a common trick that effectively means trading size on disk (and in memory) for inference speed. It's obvious the model will need to be more powerful if the embedding carry less inf
383.
by Nimitz14 7y ago
And I lived in Germany for 12 years, crossing redlights probably every day, and was never reprimanded.
384.
by Nimitz14 7y ago
Economic growth alone is not relevant, what matters is how much there would have been with another system in comparison. A much harder question.
385.
by Nimitz14 7y ago
> If your perfect world is being given a task to do and doing it accurately to a high level of perfection I think it should be obvious if that is what you desire start-ups are not for you. Whoever hired you failed at their job.
386.
by Nimitz14 7y ago
He's obviously taking fitness [boxing] classes and will not actually be sparring anyone.
387.
by Nimitz14 7y ago
> The troops that did go to the Middle East wars have suffered many casualties (6-7K dead, many more wounded, many US contractors dead/wounded, PTSD/suicides/mental health issues afterwards, etc.). That is objectively not
388.
by Nimitz14 7y ago
Interesting, thanks.
389.
by Nimitz14 7y ago
Once pytorch is updated to use XNNPACK (being worked on right now) I think it should be fine to use. That plus QNNPACK makes inference quite low on power usage in my (admittedly limited, just integrated XNNPACK) experience.
390.
by Nimitz14 7y ago
Haha, this really nailed several painpoints of TF. Laughed and nodded at the "there are no experts users, instead there's experts of 2017 TF and experts of 2019 TF".

More ›