Deebash D @mr_deebash

@DhivCM 😂🤣😂🤣😂 You have 5 years to cry. Take your time.

For those who think that's bad, you should write a compiler ;p

Valentin Ignatev @valigo

2 months ago

You look at source code of best games ever made, and they are mostly for loops and if conditions over big fat structs. Really makes you think.

101 116 4K 307K 638

@TheGoldenDays Windows 98 old

- you are - a random CS grad with 0 clue how LLMs work - get tired of people gatekeeping with big words and tiny GPUs - decide to go full monk mode - 2 years later i can explain attention mechanisms at parties and ruin them - here’s the forbidden knowledge map - top to bottom, how LLMs *actually* work - start at the beginning - text → tokens - tokens → embeddings - you are now a floating point number in 4D space - vibe accordingly - positional embeddings: - absolute: “i am position 5” - rotary (RoPE): “i am a sine wave” - alibi: “i scale attention by distance like a hater” - attention is all you need - self-attention: “who am i allowed to pay attention to?” - multihead: “what if i do that 8 times in parallel?” - QKV: query, key, value - sounds like a crypto scam - actually the core of intelligence - transformers: - take your inputs - smash them through attention layers - normalize, activate, repeat - dump the logits - congratulations, you just inferred a token - sampling tricks for the final output: - temperature: how chaotic you want to be - top-k: only sample from the top K options - top-p: sample from the smallest group of tokens whose probabilities sum to p - beam search? never ask about beam search - kv cache = cheat code - saves past keys & values - lets you skip reprocessing old tokens - turns a 90B model from “help me I’m melting” to “real-time genius” - long context hacks: - sliding window: move the attention like a scanner - infini attention: attend sparsely, like a laser sniper - memory layers: store thoughts like a diary with read access - mixture of experts (MoE): - not all weights matter - route tokens to different sub-networks - only activate ~3B params out of 80B - “only the experts reply” energy - grouped query attention (GQA): - fewer keys/values than queries - improves inference speed - “i want to be fast without being dumb” - normalization & activations: - layernorm, RMSnorm - gelu, silu, relu - they all sound like failed Pokémon - but they make the network stable and smooth - training goals: - causal LM: guess the next word - masked LM: guess the missing word - span prediction, fill-in-the-middle, etc - LLMs trained on the art of guessing and got good at it - tuning flavors: - finetuning: new weights - instruction tuning: “please act helpful” - rlhf: reinforcement from vibes and clickbait prompts - dpo: direct preference optimization — basically “do what humans upvote” - scaling laws: - more data, more parameters, more compute - loss goes down predictably - intelligence is now a budget line item - bonus round: - quantization: - post-training quantization (PTQ) - quant-aware training (QAT) - models shrink, inference gets cheaper - gguf, awq, gptq — all just zip files with extra spice - training vs inference stacks: - deepspeed, megatron, fschat — for pain - vllm, tgi, tensorRT-LLM — for speed - everyone has a repo - nobody reads the docs - synthetic data: - generate your own training set - model teaches itself - feedback loop of knowledge and hallucination - welcome to the ouroboros era - final boss secret: - you can learn *all of this* in ~2 years - no PhD - no 10x compute - just relentless curiosity, good bookmarks, and late nights - the elite don’t want you to know this - but now that you do - choose to act - start now - build the models

@PraveenByBuy Just curious, how you convince someone who is already using and spent huge on other well known billing systems. Are they ready to replace that? Btw, what you're doing is great, all the best 👍

Steve Jobs : how to design perfect products.

This paper is highly recommended for programmers. Even the god of programming can make mistakes. It also helps deepen your understanding of Linux scheduling and how to solve complex performance issues. Thanks to @Not_Yours_Mom for the recommendation. people.ece.ubc.ca/sasha/papers/e…

@akhan999 @pubity Pakistanis need more Photoshop tutorials. 😂 Can you share original CNN news link here?

@Pk78676803105 @pubity newschecker.in/ai-deepfake/ai… That's fake AI generated image 😂🤣😂🤣

@Varshha4ever வாழ்த்துக்கள்! இனி அனைத்தும் நல்லதே!

@fbhutto Send your kids to school, not to ter*orist camps.

My favorite technical blogs

keyboard_arrow_left Previous keyboard_arrow_right Next

@SavukkuOfficial

Hi 👋 Launching 🎉 tncrimes.com Website for real-time crime stats of Tamilnadu State. AI bot reads multiple news publishers every hour, if it detects certain keywords related to crimes, it classifies the news under different categories, and presents it as numbers.

@jrsikken Tester

@hey_yogini deebash.org

The water quality in Warsaw, the capital city of Poland, is monitored by clams. If the water gets too toxic, they close, and this triggers analyses that can immediately shut off the city's water supply.

@thekaipullai With the capital they have they can even build a OS, browser without chromium, etc etc but they don't. Challenge: Skill. We don't have any original software from India. Everything we make is a rip off. Just like Chinese for manufacturing, India is for softwares.

@Ravisutanjani Govt. website always built on static pages, not frameworks. It'll dramatically decrease bandwidth, faster in lower network, faster in times of tatkal, page will load faster for low spec mobiles. Don't assume everyone will be using iphones with 5G network. Think about rural.

@_shrayanshsingh @IndiGo6E Why there are two charges for travelling? 1. Air fare charges. 2. Seat fee Do I need to travel by standing? Even auto is providing a seat for me to travel. In future we might see 1. Seat fee 2. Seat cover fee 3. Arm rest fee Etc.