Qalabit


Channel's geo and language: not specified, not specified
Category: not specified


https://nama21yo.github.io/natnael-work
4th SWE | Chess | Sketching
Oriental Orthodox
Movie: ofc anime and documentaries
Series - no but was
Kdrama - once I watched the defendant in my lifetime. It was actually great.
Books: Spiritual & non-fiction

Related channels

Channel's geo and language
not specified, not specified
Category
not specified
Statistics
Posts filter


Choosing the right model for a specific task is one of the problems we face when implementing LLM based projects or basically any RAG systems

So some cool websites I found for that is https://artificialanalysis.ai

It is great to check the latency (the time that u spent before receiving the first chunk of output), the price, the output speed and so on.

the other site is lmarena
https://arena.ai

it is kind of crowdsourcing scores of language models by Humans feedback

For normal dev u can do plug and play but fro production u should design ur system so that u can use 2 or more models and ultimately compare the user satisfaction or some other metrics u get since it's catchy to only trust the benchmarks and leader boards only.

Other than that u have studios like google ai studio, groq ai studio and so on that u can try out the models right there with some limit. U can also adjust the temperature.

for choosing between local models for those who have gpus use lmstudio https://lmstudio.ai/
#resources@qalabit


Yesterday I slept 11hr almost half day 😢

I have gone multiple places for meet up, finished clearance whoo asking multiple places for signature, played pool with my new Grad friend for last time I am still newbie in pool only my 2nd time but atleast when I hit the ball it is still on the pool table 😂. The summation of these effects equals 11 😂.

I even missed the Argentina match 😭.

#failure@qalabit


BTW I am also doing this one I have started like a month ago

Going so slowly 😭 but let's update each other.

If there is anyone who also started or want to start Please share urs progress too.

So I am sharing the most GOATEDD book that I used and using to get the fundamentals to implement some of the concepts.

Let me add also a funfact 😂

It is how reboot() system call works ask llm or search the above question and you will get the answer I am sure u will say madness 😂

Hint: classic Linux Easter egg! (B. Date of his daughters and himself 😂)

#funfact@qalabit
#resources@qalabit




Forward from: sudo jajos
The people this country calls "teachers" and "lecturers" are literally the reason we're producing incompetent and incurious graduates. Today we had a computer graphics project presentation, we built procedural infinite terrain generation from scratch, perlin noise, fBm layering, real time lighting, GLSL water shaders, stuff we tried a lot to actually understood deeply. You know what this man's first question was? "where is the code for the weapon" bro hadn't even looked at the screen, didn't know what perlin noise is, didn't know what a shader is, couldn't even tell the difference between a a monitor and a television. Just sat there nodding at whoever sounded confident enough. And the worst part is this isn't even shocking anymore, I've seen this exact same thing with so many "lecturers". A person who doesn't understand the subject examining students on the subject, and somehow that's just normal here. It's not a one time thing it's the whole system and it's genuinely costing this country something real.

#learning_system
@sudojajos


Challenge Accepted and Started!!!

What fascinates me is ሐዋርያት በአሁኑ ጊዜ እኛ ለስድብ የምንጠቀማቸው ህይወት ያላቸው ነበር። Some were uneducated, some were Poor, some were Sinful While being in this kind of Status ጌታ መርጧቸው ነበር ። What kind of reasoning would u give for this it is totally unconditional love. The only thing he needed from them is there willing and so does from us not being rich, nerd or sth.

#biblestudy@qalabit


"There aren't many Ethiopian founders building global products. For many, it feels almost impossible. So seeing that traction gives hope for other people to try to be more ambitious." Bereket. E



Yup that should sink in. Most of the time when discussing with friends I am always the most pessimistic dude saying No we can't do that with our current knowledge, it has this problem that problem minamn I should stop doing that and just start and keep up with the flow.

#takeaways@qalabit




Melkam Senbet

#senbet@qalabit


Forward from: hi-z-k localhost
እንኳን አደረሳችሁ ፤ የሊቀ መላእኩ የቅዱስ ሚካኤል ረድኤቱ በረከቱ አማላጅነቱ አይለየን 🌹

📍 ደ/መ/ቅድስት ኪዳነ ምህረት ቤተ ክርስትያን

#update


Video is unavailable for watching
Show in Telegram
The power u need when u are just seeing the syllabus today for tomorrow exam.

#exams@qalabit


Fundamentals are a must!!

Rust actually have async runtime library called tokio. Honestly dk how it works.

But just implementing my own It's simple one.
It is basically doing MPSC Channeling (Multi-Producer, Single-Consumer)

The dude who haven't implemented Two sum with rust 😭

#rust@qalabit


Forward from: Newbie
Every cybersecurity example in existence is just alice sending message to bob🙄


Forward from: Yoni Moge | Faith & Ambition
ሰርጋችሁ ጋር ስጠሩኝ እንደዚህ ጥሩኝ😂


😂😂 Cloudflare Lava Lamp Wall

So insane thing I knew today

Cloudflare uses lava lamps as a source of true randomness (entropy) for generating cryptographic seeds used in SSL/TLS encryption keys.

How?

the randomness comes from physics 🙄. The physics is related to buoyancy check it out freaking insane.

Computers are usually deterministic and generate only pseudo-random numbers.

So Cloudflare places cameras in front of a wall of lava lamps. The cameras capture images many times per second. Software then converts details of the images (pixel colors, positions, shapes, brightness patterns, etc.) into random bits.

That's another level of security Wow 🤩

For more detail: Link

I am doing this in middle of sibseba 😂 🤫
#blogs@qalabit




Forward from: Tech World
Happy Monday. What a privilege it is to wake up and have another day to better yourself

@lidsverse


This week too, I haven't get the time to read paper 😭
I was working into AWS DevOps
specifically AWS Cost Optimization for servers

You have an EC2 instance (some server) sitting completely idle, doing absolutely nothing. You want to kill it, but you can’t because you don't want to lose your setup configuration like the docker and other packages installed, and you definitely don't want your public IP address to change too.

If you just leave it running, Jeff Bezos wins 😂, and your money burns alive. So, what are the cost optimizations?

1. Obviously Stop & Start

The simplest trick is just stopping the instance when you aren't using it (like weekends or overnight) and starting it back up when you do. Useful for dev or test servers.

2. Prepaid Packages (Reserved Instances)

You commit to a specific instance type (like a t3.micro) for a 1 or 3-year term. AWS rewards your commitment with at least a 30% discount!

The catch: It’s rigid.
3. Savings Plans

If Reserved Instances feel too restricting, Savings Plans are the cooler, more flexible. Instead of committing to a specific machine, you commit to spending a specific amount of money per hour (e.g., $1/hour). The best part? It automatically applies discounts across EC2, AWS Lambda, and ECS Fargate dynamically.

4. Spot Instances

Want massive discounts up to 60% (or more) off the original price? Use Spot Instances. This is AWS selling their unused spare capacity for dirt cheap.

The catch: If someone else offers to pay full price, AWS will kick you off the server with a 2-minute warning. Great for stateless apps like background running task, but not for database hell no

#devops@qalabit


Welcome bro highly recommended


Forward from: sudo jajos
https://arxiv.org/abs/1711.03705

One of the research papers I will after exams and some work related loads ...

The paper is about online deep learning -> u know the opposite of it which is offline deep learning(if ur familiar with online RL and offline RL it is analogous here).But ofc u dont need to know RL to understand there difference. So, basically the difference between these two is the offline DL is the approach where u have collected and static data then u train ur model then voila u have some learned parameters that are fixed till u have full or partial finetuning here after ... but in online DL u can update ur model on the fly and this is critical for streaming data(I think we can say for example stock price prediction perhaps), non-stationary data ...


I was about to read it because I have something in mind to research about and I will tell u after finals and work loads ... 👍

#papers #random
@sudojajos

20 last posts shown.

135

subscribers
Channel statistics