[0 I know there are some people who managed to ascend (win) without reading up, but they are probably in the < 0.01% so it's definitely not the norm.
533 karma · joined November 4, 2015
[0 I know there are some people who managed to ascend (win) without reading up, but they are probably in the < 0.01% so it's definitely not the norm.
But you can just ask them.
I have to say that not everything was 100% smooth - sometimes the proprietary NVidia driver crashed (you have to use the right CUDA and driver combination) my Linux instance and hanged the system, so I had to hard-reboot it (which is supported via their admin console) which takes some minutes. However that's not their fault as I heard the driver is a big pile of crap shit anyway because NVidia is too embarrassed to post it to LKML.
GTX1080 for 100$ a month. Grantend, it is older, but it still works for DL. Let's say you do 10 experiments a month for ~20 hours. Thats 0.5$/hour and I don't think it is 3 times faster.
If you then want to do even more learning the price goes even down.
//DISCLAIMER: I do not work for them, but used it for DL in the past and it was for sure cheaper than GCP or AWS. If you have to do lots of experiments (>year) go with your own hardware, but do not underestimate the convenience of >100MByte/s if you download many big training sets.
EDIT: 's/strong/powerful'
"These kind of errors." Master Typikos said - and she told the student the story about the time when a bug in a runtime caused Math.Round to do IO although it's type was `Integer -> Float32 -> Float32`.
The student did not believe this story. Later he was enlightened.
EDIT: refined type
If you know your project is for sure _personal_ (i.e. not shared) and a small _proof-of-concept_ (i.e. if you go further in that direction you will do a rewrite anyway), I think it is overkill. Your code and data will be in so much flux, that good old copying is faster than dealing with the VC.
Thinking about it, I think it makes sense to add VC to a project, when it delivers its first satisfying result, I would call that the proof-of-concept program. At that point you throw it away/archive it, or you have bigger plans, then you probably add version control knowing you have this working prototype and rewriting stuff.
You do not have the history to that point, but it would not be helpful anyway.
If this is a good or bad development will (or already is, as we have seen in discussions concerning Go) be the topic of many flame wars.
Bam! You immediately know it's outdated (emphasized by the headline) and switch to https://docs.djangoproject.com/en/1.9/topics/auth/default/. This might seem minor, but I highly appreciate such attention to detail. And I really like the official page's layout, look and feel which concentrates on information, rather than stupid, flashy animations. Ha, looks like I'm getting old.
If they would make their stack FOSS they would be absolute heroes and shape the future of applied math and computation software with a historic impact. Otherwise they will fade away, when other interfaces (Jupyter & Co.) catch up and/or surpass it in its capabilities. More and more people will just not accept the closed nature as it inherently contradicts with the idea of exploration.
There are basically 3 scenarios:
- they go FOSS now: best outcome for people and Wolfram
- they go FOSS later: good outcome for people, more difficult for Wolfram, as they lose developers/community
- the don't go FOSS (or not in near future): Wolfram SW won't be used anymore as open alternatives surpass it. People won't care.
Even Microsoft got it.
And haven't this people been right? :)
It was (and still is, but people are fortunately more skeptic) completely oversold like REST now.
Of course, there is pythonanywhere or you might want to use PAAS like Heroku, but I find it interesting non the less. Is it so much more work for shared hosting providers to provide Python?