itissid

22 ms·

271.
by itissid 3y ago
Read the article by atul gawande linke in thee. As a programmer i do wonder what has caused my performance in terms of learning longer, massive subjects to peak. I think as a kid I used to just storm and power through learning things, even
272.
by itissid 3y ago
Lists to live by...
273.
by itissid 3y ago
DataColada people already had a lot of work prior to this.
274.
by itissid 3y ago
Another noob Question: So a 50% size reduction in BERT? let's see if I am getting these numbers right. At inference time you need a fraction of the neurons in the FF layer to do the inference based on the input data and the previous do
275.
by itissid 3y ago
For those not familiar with Bert transformer arch. You can read a bunch of their torch benchmark code to measure speed up in just the FFF: https://github.com/pbelcak/UltraFastBERT/blob/main/benchmark...
276.
by itissid 3y ago
Noob Question: So is the idea to load only specific branches (and by extension log(n) order neurons), right based on the input data. Would this be something that a compiler would do using a JIT trick(because the input needs to be known to g
277.
by itissid 3y ago
> We fine-tuned a PaLM 2-XS-Otter model based on our available data for 20k steps and choose the checkpoint with the best validation results. What is the "available data for 20k steps"?
278.
by itissid 3y ago
Re Educasso. what kind of schools are you thinking of public private charter? Also do teachers create their own content? I thought the content always came from a tie up of the school to a publisher that supplies all of it. It's hard to
279.
by itissid 3y ago
I feel, if you use a pre-trained model to do these things without knowing the set intersection of the test and that dataset, makes it very tough to know weather inference is in the transitive closure of generated text the models were traine
280.
by itissid 3y ago
Oh man what a time to be alive. The creativity is amazing. Bravo.
281.
by itissid 3y ago
Also referenced https://www.nytimes.com/2023/11/15/world/europe/orcas-attack... Good to see people are trying to finding solutions other than to kill them.
282.
A group to help sailors with the Orca problem(orcas.pt)
1 points by itissid 3y ago|1 comments
283.
by itissid 3y ago
And remember there is no removing/uninstalling this kind of a service short of wiping the entire OS and installing another Playstore/Open Source alternative.
284.
by itissid 3y ago
I will NEVER buy a samsung phone. They have all these backroom deals to put in Services like additional LocationProvider(the underlying service that can be subscribed to for loc updates) for companies that will pay them to get access to loc
285.
by itissid 3y ago
TBH I found this article a bit poorly written, especially around the core of the symptoms*. It does not even list out what exactly are the kinds of things to look for when you are in say your 30's or something and suspect you have ADHD
286.
by itissid 3y ago
> Some people mainly have symptoms of inattention. Its actually the opposite. I found that its to do with lots of attention just not at the task at hand.
287.
by itissid 3y ago
Hi is the benchmarking code available anywhere? I could not see it.
288.
by itissid 3y ago
Clickable link: http://drophere.co
289.
http://drophere.co A meditative take on urban exploration
1 points by itissid 3y ago|1 comments
290.
by itissid 3y ago
This is great. Ive been using GPT based tools extensively for my side project which I have a demo for( http://drophere.co see code at https://github.com/itissid/drop_webdemo ) one thing I think that is missin
291.
by itissid 3y ago
FWIW 1. Freeze all your credit with experian, equifax and transunion 2. Opt out of them selling your info: https://consumerprivacy.experian.com/ https://myprivacy.equifax.com/opt-in-opt-out/personal-in
292.
by itissid 3y ago
https://news.ycombinator.com/item?id=29834753 I was shocked to learn that last year about the level of detail they had. 1. All your mortgage, credit inquiries and bank account names 2. All your previous addesses and perevi
293.
by itissid 3y ago
Sometimes calmly thinking about what you want to type next is 1/2 the joy...
294.
by itissid 3y ago
Interesting. They are taking the pain out of two things 1. RAG based infra building for people(regular joes). 2. Compartmentalize or Work around context lengths by introducing threads.
295.
by itissid 3y ago
I actually find grouping tabs into one tab in the UI and then color coding helps ADD people who see all the work piled up. But in reality most tabs are either distractions or we dont return to them for a long time until other tasks are done
296.
by itissid 3y ago
Is the idea here to also compile pytorch to this(other than CUDA) so they can leverage those APIs(Vulkan, OpenCL etc) for other platforms?
297.
by itissid 3y ago
Noob Question: Isn't this going to be an great source for training Language models? Is it safe to assume that OpenAI/Google/Meta etc already have these? In any case great work!
298.
by itissid 3y ago
Noob Q: Would this library be a (good?) candidate to be integrated with a wrappers/extension libraries to have all the graph based clustering algorithms in one place(assuming they are not already)? Or do(better?) frameworks for the sam
299.
by itissid 3y ago
I mean meta's first gen(they had glasses last year but they were not that good and were more of a test run).
300.
by itissid 3y ago
They can but remember, having location always on, collected when the app is backgrounded, is a deeply buried permission, once turned off it's hard to turn it back on(or in apples case reminding user with a map to tell you "xyz app

More ›