Tensorflow benchmarks show ResNet-152 resulting in 2.4x lower throughput than ResNet-50. [0]
728 karma · joined January 13, 2011
github.com/dougal @douglasfshearer
Tensorflow benchmarks show ResNet-152 resulting in 2.4x lower throughput than ResNet-50. [0]
See also Mailinator. [0]
[0] http://mailinator.blogspot.ae/2007/01/architecture-of-mailin...
https://en.wikipedia.org/wiki/SEC_Rule_10b5-1#A_possible_loo...
- It's free (though you can become a supporter and get some extra benefits) [1]
- It's fast, since it uses it's own CDN. [2]
- It's secure, all pages support SSL, even with custom domains. [3]
- It has a command line tool [4] that can be wrapped to automate upload of pages, or used in a Git hook.
- It has a plethora of learning resources [5].
[1] https://neocities.org/supporter
[2] https://www.youtube.com/watch?v=-i6wvix6buI
Replace IBM with Intel for the 2010s version.
You mention you use SQS to queue incoming metrics. What server/framework/language are you using to do this? IME using SQS requires tools with good parallelism to deal with it's high-latency/high-throughput performance characteristics.
[0] https://segment.com/blog/spotting-a-million-dollars-in-your-...
[1] http://tomforsyth1000.github.io/blog.wiki.html#%5B%5BWhy%20d...
Can the title be edited?
I live with a view of the bridges, it has been fascinating watching them construct the new bridge.
[0] https://en.wikipedia.org/wiki/Forth_Road_Bridge
[1] http://www.bbc.co.uk/news/uk-scotland-edinburgh-east-fife-32...
Compared to the Titan X Pascal it has more cores, a higher clock, and more memory bandwidth.
[1] - https://en.wikipedia.org/wiki/GeForce_10_series#GeForce_10_....
TLDR: Shard at a time was replicated over the network, then live traffic was cut over to the Google Cloud replica.
[1] https://blog.evernote.com/tech/2017/02/08/part-1-evernote-se...
Custom silicon, rather than the FPGA or ASICS they were using when they started the project in 2012.
They tout not only the freed CPU resources, but lower latencies and increased security of isolating the networking from the CPUs running hosted VMs.
[0] https://github.com/dailymotion/hls.js
[1] http://open.blogs.nytimes.com/2016/02/08/flash-free-video-in...
I do lots of this in Ruby [0], and often write a custom parser when I know that dates or times will be of regular form. As a bonus I get to golf different methods of doing the parsing against each other, using microbenchmarks.
[0] https://gist.github.com/dougal/e972b26cd0293f99b41896e7d89c0...
Perhaps it mitigates against proxies stripping the headers, or it simplifies caching.
They traded outright frontend performance, so as to increase the iterations on what is an internal tool, and thus make it more useful to the company.
If so, is there a timeframe for multi-DC redundancy?
This is the only thing stopping me from moving large amounts of data into B2 right now.
The article describes results for using a Xeon CPU. The linked paper cites other papers where a GPU was used.
The machine never entered production due to high manufacturing and maintenance costs, as well as the speed at which it could cover ground.
The prototype is now housed in The Museum of Lusto [1] in Finland (near bottom of page for pictures).
[1] http://www.unusuallocomotion.com/pages/museums/museum-of-lus...
There was no pre-release of the results to contributors, meaning that frameworks with unusually high error rates were not investigated. I am thinking particularly of Phoenix [2], as this is the first round in which it was included.
Hopefully for round 13 they will have better availability of hardware, and be able to follow to the release process they have previously used.
For anyone able to help, they are looking for a donor for test machines [3].
[1] https://groups.google.com/d/msg/framework-benchmarks/Hq4qxj0...
[2] https://groups.google.com/d/msg/phoenix-talk/6cjJx61FPBg/LGs...
[3] https://groups.google.com/d/msg/framework-benchmarks/IiDBC6l...
I work on a project where ET isn't flexible enough (MPEG-DASH), and wondered whether Lambda would make for a good alternative to EC2 + SQS + Scaling Groups.
No mention of HTTP caching though, which for a whole class of applications is a great way to minimise rendering time.
Rails has great support[1] for etag based content expiry.
[1] http://api.rubyonrails.org/classes/ActionController/Conditio...
Similar style charts to that used by Randall Munroe in XKCD.
See also this D3 style for producing similar charts by Dan Foreman-Mackey. [2]