Anagha

100 posts

Anagha banner
Anagha

Anagha

@Lilac_code_

Turning bugs into lessons & code into fun! 💜💻 | Tips, projects & coding chaos 😎 | #100DaysOfCode

Indianapolis,IN Katılım Mart 2025
116 Takip Edilen19 Takipçiler
Anagha retweetledi
Python Hub
Python Hub@PythonHub·
City2Graph Transform geospatial relations into graphs for Graph Neural Networks and network analysis. github.com/c2g-dev/city2g…
English
1
92
513
21.7K
Anagha retweetledi
Noth1ng Real
Noth1ng Real@noth1ng_real·
The word "car" is meaningless to an LLM. It is just a string of characters. The first thing a transformer does is convert every token into a vector. Somehow this high-dimensional vector encodes the meaning of the word. GPT-3 uses 12,288 dimensions. But here’s the cool part: the geometry of the space actually means something. E(king) − E(queen) ≈ E(man) − E(woman) In this example, one specific direction in 12,288-dimensional space encodes gender. Another encodes plurality, and so on. Nobody explicitly programmed this. The model learned it simply by reading enough text. There is a special matrix called the embedding matrix, which stores the “definition” of all words as vectors. In GPT-3 its size is 12,288 × 50,257: one column per word in the vocabulary.
Noth1ng Real tweet mediaNoth1ng Real tweet media
English
4
4
60
8.3K
Aditi
Aditi@aditilovess·
🚀 Hey @X algorithm, I'm looking to #CONNECT with people interested in: 🎨 Frontend 💼 Backend 👩‍💻 Gen AI ✨ Full Stack 🧑‍💻 DevOps ✅ DSA 🧠 AI/ML 🧱 Web3 📊 Data Science 💸 Freelancing 🐍 Python 🔨 Startups Let’s grow, learn & build together! 🔥
English
185
3
265
9.1K
Tekraj Awasthi🧑‍💻🇳🇵
A popular curated list of resources to get into LLMs (>65k GitHub ⭐) The LLM course is divided into 3 parts: 🧩 LLM Fundamentals: Covers basics about Maths for ML, Python, and Neural Networks. 🧑‍🔬 The LLM Scientist: Focuses on LLM architectures and the techniques to build LLM(Pre-training, Post-training, SFT, Evaluation, Quantization, etc.) 👷 The LLM Engineer: focuses on creating LLM-based applications(Building Vector DB, RAG, Advanced RAG, Agents, Inference Optimization) and deploying them. (link in the comment)
Tekraj Awasthi🧑‍💻🇳🇵 tweet media
English
13
5
66
14.8K
Zephyr
Zephyr@Zephyr_hg·
I spent weeks researching 2025's hottest AI opportunities. Found 50 business ideas that actually make money using current tech. Each idea includes revenue paths, MVP scope, and exact tech stack needed. Perfect for solopreneurs ready to capitalize on the AI boom before everyone else catches on. Skip months of market research and failed experiments. Comment "IDEAS" and I'll DM it to you (must be following)
Zephyr tweet media
English
1.4K
235
2K
280.4K
Anagha retweetledi
Mohit Goyal (Harness arc)
Mohit Goyal (Harness arc)@ByteMohit·
Bad ML dev → grinds through theory & code → starts training small models → applies them to real data → builds & deploys useful AI systems → great ML dev
English
0
2
11
516
atul
atul@atullchaurasia·
Today is my birthday
atul tweet media
English
250
2
520
21K
Anagha
Anagha@Lilac_code_·
Late-night grind session: code, calm, and coffee. Because dreams don't debug themselves. 💻🌙
Anagha tweet media
English
0
0
0
15
Seyram
Seyram@__theSeyram·
You must delete 1 app, which one?
Seyram tweet media
English
14.2K
2.1K
34.4K
4.9M
Anagha
Anagha@Lilac_code_·
One of the coolest tricks in computer vision for spotting fake images 👀 Check the light sources and reflections Mirrors and shadows don’t lie, if perception says “something’s off,” the pixels probably are 🪞✨
English
0
0
0
22
Anagha
Anagha@Lilac_code_·
Scikit-learn is your go-to for fast ML workflows and experimentation, while XGBoost is the champ for squeezing every drop of performance out of structured data.
English
1
0
1
22
Anagha
Anagha@Lilac_code_·
⚔️Scikit-learn vs. XGBoost: The Ultimate ML Smackdown Building a killer predictive model in 2025? Which tool’s your MVP?
English
1
0
1
31
Anagha
Anagha@Lilac_code_·
The hardest part of coding is not the syntax It’s explaining to your future self what you were thinking when you wrote that line 🫠
English
0
0
0
20