greefea

@Madisonkanna An instant classic

@mdancho84 Nice info! With tuning in two separate steps, can you be sure the optimal learning rate you found in step 1 is still the optimal one after you tuned other parameters in step 2?

For years, I was hyperparameter tuning XGBoost models wrong. In 3 minutes, I'll share one secret that took me 3 years to figure out. When I did, it cut my training time 10X (and gave an instant 25% boost in model accuracy). Let's dive in. 🧵

@pokateo_maps Excel messing up their columns (character -> integer) in table join

Experienced GIS users, what's a common mistake that you see new users make? #GISChat

@bearliu 😂

完美诠释了将AI融入日常工作流的现状

@435hz 章工这鸡汤霸道，点个赞👍

大多数人对自律的认识，来自学生时代的学业成功。 进社会后，这种自律与事业成功基本没什么关系，不过是在无关紧要的事上设置规则、自我激励（自我欺骗）。 自律就是每天干最重要的事，至于几点到几点、什么时间管理，这其实一点不重要。如果需要，连续几天持续突击也在所不惜。一切只为最重要的目标。

@BasedMikeLee 18 U.S. Code § 611: law.cornell.edu/uscode/text/18…

Americans don’t vote in foreign elections Noncitizens shouldn’t vote in ours We need the SAVE ACT — attached to the September spending bill — to stop that from happening Share if you agree

@Uniswap @ourZORA minted

Now available on @ourZORA for 7 days ⤵️ zora.co/collect/zora:0…

Today marks 5 years since Unisocks became the world's first dynamically priced sock, powered by the AMM So we're celebrating with a limited-time mint "5OCKS" 🧦

@wondercraft_ai, enjoying the wonderful hacker news recap podcast. But can you pls fix one thing: occasionally an episode would include a tiny segment with abnormal volume and strange (distressed?) AI voice. Example 4/19 episode beginning 3:06-18.

@YosefBlockchain 查理芒格有一条praise by name, criticize by category，表扬的时候具体到人，批评的时候批评一类行为/人

在职场沟通中，高情商讲话非常重要，今天聊一下我的几点心得体会：

@JefferyTatsuya 碰巧看到推上的另一个开源TTS的总结，大部分重合：twitter.com/reach_vb/statu…

现在有哪些顶级的开源TTS模型？ 整理一下我收集到的。有没有补充？ 🔊 XTTS - huggingface.co/coqui/XTTS-v2 🗣️ YourTTS - github.com/Edresson/YourT… 🚀 FastSpeech2 - github.com/DigitalPhoneti… 🎵 VITS - huggingface.co/docs/transform… 🐢 TorToiSe - github.com/neonbjb/tortoi… (1/3)

@realrenmin 还有著名的Yann Lecun prompt: twitter.com/stanislavfort/…

13/ 下一个thread，将记录用CoT完成api和工具使用的paper读书笔记，如果你喜欢我的读书笔记，请关注我 @realrenmin ，每周会写一到两个长thread跟大家分享NLP的知识。

1/ 从Chain-of-Thought (CoT) 到 Let's think step by step 最近每天都可以看到大语言模型进展的文章，LLM的能力从简单文字的处理，逐步演化到复杂的reasoning推断，甚至可以做decision making来完成api调用及tool的使用。 此条thread回顾了两篇重要的CoT文章作为读书笔记。 🧵👇

@AiBreakfast It is impressive that Chincilla doing so well with only 70B parameters

Meta's new AI model LLaMA came out today. The paper was a gentle reminder of how capable Google's 540-billion parameter model PaLM will be.

@kevinyang Would love access to try it out

This is also just the beginning... I have even crazier shit cooking 👀 twitter.com/RealKevinYang/…

OMG it finally works!! 🚀🚀🚀 I got @OpenAI to draft emails 📧 for me in the background 🤯 This christmas I built an email assistant (EmailTriager.com) that automatically drafts email replies behind the scenes no chrome extension necessary, here's the story 👇

@yishan @awilkinson What makes you think that AI cannot/won't improve on "traditional robot things"? When Internet first came out, it could not do many things we use it for today.

@awilkinson Many people don’t understand this because they don’t think rigorously themselves, so they’re just unable to even discern the difference and think it’s all one magic robot. It’s not. Turns out the robot is bad at “traditional” robot things.

The year is 1996 and you own a newspaper. The internet just came out, but you don’t really get it. Which businesses are similar in 2023 with GPT3 + AI looming?

@amasad Dan Carlin's Hardcore History feeds.feedburner.com/dancarlin/hist…

Shill me your fav niche podcast. Ideally something with fresh weird ideas you don’t hear elsewhere.

@StokesNeuro Bless u 🙏

Hi folks, I’m afraid it’s time for me to say goodbye. Not just leaving Twitter, but the whole show. I’ve been battling cancer last 2 years, but now only have a few days left now. Thank you wonderful people, I leave this crazy world with much love in my heart ❤️

@GnomesMindverse @nbaschez Can I have an invite to try it out? Thanks!

@cyber_porter Interested

Announcement: I will teach you Zero Knowledge Proofs. Eight Zoom classes over the next 4 weeks. 100% free (I'm doing this for fun in my free time). Reply or dm if interested.

@NikkiSiapno Another mistake I am aware of: PR with conflicts - do a proper rebase before PR, related to mistake 9, but different

Don’t waste hours programming / coding just to get your pull request rejected. 9 common mistakes I see in pull requests ( fixes provided ) :

@shuiduoduo_eth @zksync @PhezzanProtocol 0xE315B3321eB012170C9f7d564e7a9E75034ccb74 谢谢～