learndeeply

8 ms·

61.
by learndeeply 4y ago
ResNets have been around for 7 years...
62.
by learndeeply 4y ago
It's a strange ML term to describe the data's metadata.
63.
by learndeeply 4y ago
I was expecting to be pessimistic, but Google actually releases the datasets under a permissive license (CC-By 4.0). Awesome! https://research.google/tools/datasets/open-images-extended-... https://gith
64.
by learndeeply 4y ago
Since the DuckDB people are here, just want to say that what you're doing is going to be a complete game-changer in the next few years, much like SQLite changed the game. Thanks for making it open source!
65.
by learndeeply 4y ago
Why the interest in PyScript? Pyodide offered something similar and has been around for a lot longer, right?
66.
by learndeeply 4y ago
Since both Fly.io and Litestream founders are here - why not disclose the price?
67.
by learndeeply 4y ago
Skimming through this, a lot of it has to deal with bad GPU hosts. > CSP fat fingered and deleted our entire cluster when trying to replenish our buffer nodes. Ouch.
68.
by learndeeply 4y ago
OpenAI released their large GPT-2 models weights a couple months after making that post: https://openai.com/blog/gpt-2-1-5b-release/
69.
by learndeeply 4y ago
Depends which model, but assuming the largest: 175B * 16 bits = 350GB. Half of that if it's quantized to 8 bits. Good luck finding a GPU that can fit that in memory.
70.
by learndeeply 4y ago
Can someone explain why I was downvoted?
71.
by learndeeply 4y ago
Why is this open source? Sourcegraph could make this a successful paid, enterprise product.
72.
by learndeeply 4y ago
Doesn't seem like you can. From Virgin's website: > We do not allow animals in the main cabin unless they are assistance animals. https://help.virginatlantic.com/gb/en/how-to-book-your-pet-o...
73.
by learndeeply 4y ago
Why would facebook block an open source initiative, considering that they've contributed a ton of open source software (react, graphql, pytorch, etc)? The more simple explanation is that an automated detector misclassified the link bec
74.
by learndeeply 4y ago
This seems obviously fake, since there's no proof. Can someone steelman a good argument for why this could be feasible? Also, if you had achieved true fusion (more energy emitted than put in), wouldn't the more profitable thing to
75.
by learndeeply 4y ago
> Please don't post insinuations about astroturfing, shilling, bots, brigading, foreign agents and the like. It degrades discussion and is usually mistaken. If you're worried about abuse, email hn@ycombinator.com and we'll
76.
by learndeeply 4y ago
No it doesn't. DALL-E 2 is purposely vague on model architecture to make it impossible to reproduce from the paper alone.
77.
by learndeeply 5y ago
Diffusion models existed long before this announcement. I have no idea who this person is, but they did not invent this idea. Edit: Diffusion models guided by CLIP*
78.
by learndeeply 5y ago
Anyone have any ideas on how they're doing text insertion using an auto-regressive model?
79.
by learndeeply 5y ago
Incidentally, the author has admitted that GPT-3 has passed the Turing test, if he thinks the answers were given by "armies of humans".
80.
by learndeeply 5y ago
There's no restrictions on investing in YC startups if you participate on HN.
81.
by learndeeply 5y ago
It's not ransomware, but just extortion. Pay up or your files will be released. https://www.wired.com/story/lapsus-hacking-group-extortion-n...
82.
by learndeeply 5y ago
There are thousands of organizations spending a lot of money on okta that are demanding answers. If they didn't respond, their business wouldn't survive.
83.
by learndeeply 5y ago
The open standard version of NVLink is CXL. They're available in latest gen CPUs.
84.
by learndeeply 5y ago
GPT-3 can't fit in 80GB of RAM.