theblackcat1002

“About US$2 million was earmarked for the protest movement in Hong Kong, but has now been frozen as part of a general overhaul and restructuring by a new agency boss.”

so the funding doesn't really give out?

As someone who has work with all three NLP toolkit: huggingface, openmt-py and fairseq. I always have trouble juggling through the heavy abstraction of openmt-py.

For example in openmt-py you need to write fields, reader and raw datasets before you even load into their complex dataset class. Each item is heavily abstracted through several layers of classes. I understand this improve code reuse, but introduce a huge steep curve for newcomer.

Huggingface approach on the other hand is slightly more "messy" [2] but easier to understand and add your own tweak.

[1] https://github.com/OpenNMT/OpenNMT-py/blob/master/onmt/bin/p...

[2] https://github.com/huggingface/transformers/tree/master/src/...

Frechet Inception Distance using language model is a solution (commonly used in text GAN). I think the core idea is that you should never rely on one metrics for evaluation but rather a mix of them ( even statistic ones: unique word count, word distribution similarity etc )

"There was a problem talking to Instagram. Please try again in a moment." blocked by Instagram?

This will be quite useful if ported to database

Arxiv pdf link : https://arxiv.org/pdf/2006.03575.pdf

It seems to be a rewrite of ngxtop

https://github.com/lebinh/ngxtop

On the other side, AMD APU code name was found in latest macOS release which should hint the other side of the story.

https://www.tomshardware.com/news/apple-may-start-selling-ma...

This is actually what I did, but with web application and services. I tried app development in the one of my first few projects but finally gave up for web apps because it allows faster trial and error ( mainly I can push changes in UI and UX faster to a wider audience ). My take away is that stay away from "hot" topics ( news, game app ) unless you have a solid background because you need a really polish product to compete.

The first thing I saw the name immediately relates to the Malay word Lagu. But no you don't see a word like this in real world, but you can certainly add them together.

You usually ended up in more recaptcha challenge (from form submission and cloudflare )than usual.

Hope they update the non touchbar Macbook Pro as well. Which has been sitting dust from all the upgrades(new keyboard, 8th gen CPU).

I think this works for university student when 99.99% uptime isn't your highest priority. My university has static IP and near 700 up/down speed to each dormitory. So by investing in an old workstation I can easily work on various ideas and built MVP for it. I finally ended up running a crawler service for months before I migrated to a vps.

You have to look at this issue from a Hong Kong resident point of view (I am not a Hong Kong resident). Since the handover to the China government, the average living cost is higher than ever before [1], real estate price skyrocket due to huge amount of investment from China[2]. Some Hong Kongers may blame these issues to "mainlanders" for "disrupting" their live.

Also in 2013 "mainlanders" rush to Hong Kong to buy milk powder due to the poisonous milk power scandal in China, causing a shortage of milk power in Hong Kong [3]. I certainly wouldn't be happy if suddenly I can't buy milk power for my child because my neighbour had bought all of them.

[1] https://www.scmp.com/news/hong-kong/article/1143157/hong-kon...

[2] https://www.scmp.com/business/article/3002396/hong-kong-home...

[3] https://www.worldjournal.com/6148953/article-%E9%A6%99%E6%B8...

I tried feeding a random CNN news I found [1], with the first few paragraph as follows:

"London (CNN)The interminable US process for picking a President has its faults but as a democratic exercise it outshines the closed-door system being used in Britain to choose a new prime minister.

Theresa May resigned as Conservative leader last week after an ill-starred premiership and triggered a party leadership race that will select the next resident of 10 Downing Street by late July."

as I added more paragraphs it started to shift from "quite sure this was written by a machine." to "quite sure this was written by a human.".

Does this mean the model has learned to take advantage of short text being generated by machine while longer are human written text. A similar effect can be replicate from this article as well[2]

[1] https://edition.cnn.com/2019/06/09/politics/uk-prime-ministe...

[2] https://nypost.com/2019/06/08/mexican-troop-deployment-may-h...

Probably due to using newspaper instead of dragnet as content scrapper. Newspaper tend to get pretty deformed content from my experience.

You can check out Sonic[1] and Tantivy[2] both are lightweight search engine

[1] https://github.com/valeriansaliou/sonic

[2] https://github.com/tantivy-search/tantivy

This looks a bit like Neon, which to me the UX is a bit of gimmick. Since Opera is sold to a Chinese company, I may consider this if Opera's going to open source it's code. Btw, I believe nothing is free, you are just paying it in a non monetary form.

> Overworked, underpaid, extremely competent was my experience of Taiwanese engineering

As an undergrad student studying both Electrical Engineering and CS in Taiwan, this is what I heard from my seniors. For most of the tech job, you are held "responsible" for the progress, so overtime is very common.

Note the emergency breaking system isn't part of the Uber self driving system, rather it came from the car manufacturer. Uber simply disabled all the settings, including all the safety features.

[1] https://www.npr.org/sections/thetwo-way/2018/05/24/614200117...

I think a better and elegant way to implement is the pytorch way. By defining a forward and backward function for easy chaining, you can easily chain multiple layers into a single model. Placing the training loop inside the layer will be hard to write a multi layer version.

Just wrote a similar from scratch code few months back: https://github.com/theblackcat102/DL_2018/blob/master/HW2/la...

I wrote a news website[https://todayheadlines.live/] which show news with similar topic in a cluster form. My pipeline do collect content data as well. However, I am not very sure how to properly summarized all the perspectives.

I heard this comes with forced inserted ads when browsing as well as limited internet access ( only facebook, twitter is available ), is it still valid today?

Well if the first day you shipped your product and it's already worth 1k, yeah it makes sense to do everything the "correct" way. But for most of the side projects and startup products this makes no sense at all.

The reasons I choose not to use any pre-existing workflow are complexity and extra compute overhead. I think there's a need for micro workflow where it's simple enough you can setup a simple production workflow in a single python script without requiring a ton of heavy package. d6tflow seems light enough for me, using luigi and two dataframe libraries.

FYI, the top 6 results are mainly based on BERT which is proposed by Google[1]. [1] https://arxiv.org/abs/1810.04805

this is a joke right?

USB with and without thunderbolt was already confusing enough for your normal joe. Now you have 3 version of USB 3.2 with require you to do a detail specsheet lookup before plugging in?

I use Linode to host all my side project. It's similar to Digital Ocean offerings and so far work pretty fine.

I wonder how do you choose which news or story to deliver to users? From your recent link[1] it seems like the news is bit on the short side.

1. http://getthecrypt.com/recent/