@cline Tried GPT-5.5 with cline, keeps on crashing with 400 error. PLEASE ALSO ADD GROK 4.3!!!!
English
Muhammad Khalid
104 posts
Muhammad Khalid
@MKhalidTaher
Entrepreneur | AI Scientist
Mannheim, Germany Katılım Nisan 2021
238 Takip Edilen25 Takipçiler
@sh_reya Observability for you as an Engineer makes sense but for people who don’t know coding it doesn’t. They just want to make something that’s fun and that works. This is how you get people to be interested in AI, maths and other stuff!
English
Once we collectively realize that users need observability, because all of this is just data and ML after all, and observability is more important than whatever “moat” or competitive advantage through obscurity, then we have to figure out how to present all this unstructured content and make it queryable so that users can fix their issues. Sounds fun I’m excited
English
My comment is not about replit but this debugging mindset simply does not work for most AI tools/agent bc they have no observability. No way to inspect *all* tokens, see what context is even available in the app let alone the subset of context the AI consumes. Todays experience is like a game of whack a mole where an error comes up and you just copy paste that back into the chat and pray
Andrew Ng@AndrewYNg
Good tip from Replit’s @mattyp at AI Dev 25 on debugging while vibe coding: Large part of it is looking at outputs to figure out what context you have that LLM does not, so that you can give it that context and help it get unstuck. Sometimes pasting in the error messages is enough, but also sometimes not.
English
Muhammad Khalid retweetledi
The @NVIDIA Project DIGITS personal AI supercomputer will enable developers to prototype, fine-tune and inference some of our most advanced Llama models locally.
NVIDIA AI Developer@NVIDIAAIDev
"By making @AIatMeta Llama models open source, we're committed to democratizing access to cutting-edge AI technology. With Project DIGITS, developers can harness the power of Llama locally, unlocking new possibilities for innovation and collaboration.” -- @Ahmad_Al_Dahle, VP, GenAI at Meta
English
@elonmusk your robo-taxis need lidar and other sensors to perform well in the real world. You argument “Humans can do it with just eyes so why not robo-taxis” is quite stupid. For your information humans have others senses(sensors) as well to understand the world around them!!
English
@ferjadnaeem @CVPR Man this glass ceiling actually exists everywhere for us as a citizen of a developing country... 💔
Kaiserslautern, Germany 🇩🇪 English
Muhammad Khalid retweetledi
Yesterday my advisor gave a tutorial on how *not* to describe your model, using bigram language models as an example.
I have taken his lesson to heart, and have written the worst explanation of a bigram language model that I could. 😊
English
Muhammad Khalid retweetledi
Notebook ML workflows have a lot of exploration & data wrangling, mixed w/modeling. Relevant artifacts (dataframe, model) are only a sliver of the workflow. So @stephen_macke & I developed nbslicer, a dynamic slicer for Python notebooks, built on ipyflow (github.com/ipyflow/ipyflow)
English
Muhammad Khalid retweetledi
I need to preach the wonders of using VSCode as a Latex editor with the LaTeX Workshop extension. It's a surprisingly lightweight program (opens faster than TexStudio), is highly customizable, has a built-in pdf viewer, BibTex and Git integration, and it's really pretty! 1/3
English
@mkhalidtaher Hello, according to my checks, your feedback is in progress. Our unit will contact you as soon as possible. Thank you. ^Berk
English
@TurkishAirlines I booked my ticket in June and on 2nd November you guys canceled my ticket and refunded me 0 Euros even though I paid more than 500 Euros. My feedback number TK-7458540 has been reported on 2nd November and still, my issue has not been resolved.
English
@ylecun Please do! I would like to give it a try.
Kaiserslautern, Germany 🇩🇪 English
Muhammad Khalid retweetledi
Muhammad Khalid retweetledi
Matrix multiplication is not easy to understand.
Even looking at the definition used to make me sweat, let alone trying to comprehend the pattern. Yet, there is a stunningly simple explanation behind it.
Let's pull back the curtain!
English
Muhammad Khalid retweetledi
This map shows how far you can travel from each station in Europe in less than 5 hours!
👉🏼 chronotrains-eu.vercel.app
English
@alfcnz Just missed you by a few days... It would've been nice to meet you professor 🙂
Kaiserslautern, Germany 🇩🇪 English
@muniza_ateeq I hope you know horoscopes are never true 😛
Kaiserslautern, Germany 🇩🇪 English
Abandoning yourself to sensual pleasure might be a smart choic... More for Capricorn bit.ly/A5KmeJ
English
Muhammad Khalid retweetledi
Wow! Exciting thread of yet another free course on Math for ML! 🥳🥳🥳
✖️➕➖➗
«Math4ML: in the class, we cover core ideas from linear algebra, calculus, and probability that are useful in ML.»
Charles 🎉 Frye@charles_irl
New video series out this week (and into next!) on the @wandb YouTube channel. They're Socratic livecoding sessions where @_ScottCondron and I work through the exercise notebooks for the Math4ML class. Details in 🧵⤵️
English
Muhammad Khalid retweetledi
Amplify is hiring our first Associate to focus on AI/ML. Upside is you get to work with me AND @sarahcat21! Check it out👇 amplifypartners.com/associate/
English
Muhammad Khalid retweetledi
Why to pretrain, by @kchonyc! 🤩🤩🤩
• SGD can't move very far from where it starts.
• The travelling dist depends on the nb of labels.
• Due to its stochasticity, it escapes narrow minima and land at a wide minimum.
• Pretraining: choose a better starting point.
English