Machine Learning with Python

Visualizing the complexity of algorithms

https://t.me/CodeProgrammer ✅

⚡️ Colorizing old black-and-white videos and "bringing faces to life" for FREE

SVFR — a full-fledged framework for restoring faces in videos.

It can:
💬 BFR — improve blurry faces.
💬 Colorization — colorize black-and-white videos.
💬 Inpainting — redraw damaged areas.
💬 and combine all of this in one pass.

Essentially, the model takes old or damaged videos and makes them "as if they were shot yesterday". And it's free and open-source.

⚙️ Installation locally:

1. Create an environment

conda create -n svfr python=3.9 -y
conda activate svfr

2. Install PyTorch (for your CUDA)

pip install torch==2.2.2 torchvision==0.17.2 torchaudio==2.2.2

3. Install dependencies

pip install -r requirements.txt

4. Download models

conda install git-lfs
git lfs install
git clone https://huggingface.co/stabilityai/stable-video-diffusion-img2vid-xt models/stable-video-diffusion-img2vid-xt

5. Start processing videos

python infer.py \
--config config/infer.yaml \
--task_ids 0 \
--input_path input.mp4 \
--output_dir results/ \
--crop_face_region

Where task_ids:

* 0 — face enhancement
* 1 — colorization
* 2 — redrawing damage

An ideal tool if:
🟢you're restoring archival videos;
🟢you're creating historical content;
🟢you're working with neural networks and video effects;
🟢you want a wow result without paid services.

▶️ Demo on Hugging Face

♎️ GitHub/Instructions

#python #soft #github

https://t.me/CodeProgrammer

📱 TorchCode — a PyTorch training tool for preparing for ML interviews

40 tasks for implementing operators and architectures that are actually asked in interviews. Automatic checking, hints, and reference solutions — all in the browser without installation.

If you're preparing for an ML interview, it's useful to go through at least half of them.

Link: https://github.com/duoan/TorchCode

tags: #useful #pytorch

https://t.me/CodeProgrammer ✅

Follow the Machine Learning with Python channel on WhatsApp: https://whatsapp.com/channel/0029VaC7Weq29753hpcggW2A

𝐕𝐢𝐬𝐮𝐚𝐥 𝐛𝐥𝐨𝐠 on Vision Transformers is live.
https://vizuaranewsletter.com/p/vision-transformers?r=5b5pyd&utm_campaign=post&utm_medium=web

Learn how ViT works from the ground up, and fine-tune one on a real classification dataset.

CNNs process images through small sliding filters. Each filter only sees a tiny local region, and the model has to stack many layers before distant parts of an image can even talk to each other.

Vision Transformers threw that whole approach out.

ViT chops an image into patches, treats each patch like a token, and runs self-attention across the full sequence.
Every patch can attend to every other patch from the very first layer. No stacking required.

That global view from layer one is what made ViT surpass CNNs on large-scale benchmarks.

𝐖𝐡𝐚𝐭 𝐭𝐡𝐞 𝐛𝐥𝐨𝐠 𝐜𝐨𝐯𝐞𝐫𝐬:

- Introduction to Vision Transformers and comparison with CNNs
- Adapting transformers to images: patch embeddings and flattening
- Positional encodings in Vision Transformers
- Encoder-only structure for classification
- Benefits and drawbacks of ViT
- Real-world applications of Vision Transformers
- Hands-on: fine-tuning ViT for image classification

The Image below shows

Self-attention connects every pixel to every other pixel at once. Convolution only sees a small local window. That's why ViT captures things CNNs miss, like the optical illusion painting where distant patches form a hidden face.

The architecture is simple. Split image into patches, flatten them into embeddings (like words in a sentence), run them through a Transformer encoder, and the class token collects info from all patches for the final prediction. Patch in, class out.

Inside attention: each patch (query) compares itself to all other patches (keys), softmax gives attention weights, and the weighted sum of values produces a new representation aware of the full image, visualizes what the CLS token actually attends to through attention heatmaps.

The second half of the blog is hands-on code. I fine-tuned ViT-Base from google (86M params) on the Oxford-IIIT Pet dataset, 37 breeds, ~7,400 images.

𝐁𝐥𝐨𝐠 𝐋𝐢𝐧𝐤
https://vizuaranewsletter.com/p/vision-transformers?r=5b5pyd&utm_campaign=post&utm_medium=web

𝐒𝐨𝐦𝐞 𝐑𝐞𝐬𝐨𝐮𝐫𝐜𝐞𝐬
ViT paper dissection
https://youtube.com/watch?v=U_sdodhcBC4

Build ViT from Scratch
https://youtube.com/watch?v=ZRo74xnN2SI

Original Paper
https://arxiv.org/abs/2010.11929

https://t.me/CodeProgrammer

Do you want to understand the methods used to train LLMs?

The training of large language models (LLMs) is based on various approaches that help models understand and generate text.

Each method shapes the learning process in its own way - from predicting the next word to classifying entire sentences or labeling entities.

Here are 4 common methods of training LLMs in simple language 👇

1. Causal Language Modeling
Predicts the next word in a sequence based on the previous ones. Helps the model master the natural flow of speech and the structure of sentences.
Analogy: how to finish a sentence for another person by guessing the next word.

2. Masked Language Modeling
Learns by guessing the missing words in a sentence based on the surrounding context. Improves the overall understanding of language.
Analogy: how to solve tasks with missing words.

3. Text Classification Modeling
Determines the general class of a sentence (for example, tone or topic) by comparing predictions with actual labels.
Analogy: how to sort letters into folders "Work", "Personal", or "Promotions".

4. Token Classification Modeling
Assigns labels to each word or subword - for example, highlights names, places, or dates in the text.
Analogy: how to highlight words with different colors - names in blue, places in green, dates in yellow.

These methods form the basis of modern LLMs, and each of them plays a role in making AI smarter and more useful.

https://t.me/CodeProgrammer

🎁 23 Years of SPOTO – Claim Your Free IT Certs Prep Kit!

🔥Whether you're preparing for #Python, #AI, #Cisco, #PMI, #Fortinet, #AWS, #Azure, #Excel, #comptia, #ITIL, #cloud or any other in-demand certification – SPOTO has got you covered!

✅ Free Resources :
・Free Python, Excel, Cyber Security, Cisco, SQL, ITIL, PMP, AWS courses: https://bit.ly/4lk4m3c
・IT Certs E-book: https://bit.ly/4bdZOqt
・IT Exams Skill Test: https://bit.ly/4sDvi0b
・Free AI material and support tools: https://bit.ly/46TpsQ8
・Free Cloud Study Guide: https://bit.ly/4lk3dIS


👉 Become Part of Our IT Learning Circle! resources and support:
https://chat.whatsapp.com/Cnc5M5353oSBo3savBl397

💬 Want exam help? Chat with an admin now!
wa.link/rozuuw

The Python + Generative AI series by Azure AI Foundry has ended, but all materials are open

Now you can calmly rewatch the recordings, download the slides, and try the code from each session — from LLM and RAG to AI agents and MCP.

All resources are here: aka.ms/pythonai/resources

👉  @codeprogrammer

🤖 Python libraries for AI agents — what to study

If you want to develop AI agents in Python, it's important to understand the order of studying libraries.

Start with LangChain, CrewAI or SmolAgents — they allow you to quickly assemble simple agents, connect tools, and test ideas.

The next level is LangGraph, LlamaIndex and Semantic Kernel. These tools are already used for production systems: RAG, orchestration, and complex workflows.

The most complex level is AutoGen, DSPy and A2A. They are needed for autonomous multi-agent systems and optimizing LLM pipelines.

LangChain — simple agents, tools, and memory 
github.com/langchain-ai/langchain

CrewAI — multi-agent systems with roles 
github.com/joaomdmoura/crewAI

SmolAgents — lightweight agents for quick experiments 
github.com/huggingface/smolagents

LangGraph — orchestration and stateful workflow 
github.com/langchain-ai/langgraph

LlamaIndex — RAG and knowledge-agents 
github.com/run-llama/llama_index

Semantic Kernel — AI workflow and plugins 
github.com/microsoft/semantic-kernel

AutoGen — autonomous multi-agent systems 
github.com/microsoft/autogen

DSPy — optimizing LLM pipelines 
github.com/stanfordnlp/dspy

A2A — protocol for interaction between agents 
github.com/a2aproject/A2A

https://t.me/CodeProgrammer 🌟

PhD Students - Do you need datasets for your research?

Here are 30 datasets for research from NexData.

Use discount code for 20% off: G5W924C3ZI

1. Korean Exam Question Dataset for AI Training

https://lnkd.in/d_paSwt7

2. Multilingual Grammar Correction Dataset

https://lnkd.in/dV43iqTp

3. High quality video caption dataset

https://lnkd.in/dY9kxkhx

4. 3D models and scenes datasets for AI and simulation

https://lnkd.in/dT-zscH4

5. Image editing datasets – object removal, addition & modification

https://lnkd.in/dd8iCGMS

6. QA dataset – visual & text reasoning

https://lnkd.in/dc3TNWFD

7. English instruction tuning dataset

https://lnkd.in/dTeTgd2M

8. Large scale vision language dataset for AI training

https://lnkd.in/dBJuxazN

9. News dataset

https://lnkd.in/dYBJe5gd

10. Global building photos dataset

https://lnkd.in/dVJsDXnC

11. Facial landmarks dataset

https://lnkd.in/dz_KGCS4

12. 3D Human Pose & Landmarks dataset

https://lnkd.in/dXE9ir8Z

13. 3D Hand Pose & Gesture Recognition dataset

https://lnkd.in/d_QdGGb9

14. 14. Driver monitoring dataset – dangerous, fatigue

https://lnkd.in/d6kF-9PW

15. Japanese handwriting OCR dataset

https://lnkd.in/dHnriqrH

16. American English Male voice TTS dataset

https://lnkd.in/dqyvg862

17. Riddles and brain teasers dataset

https://lnkd.in/dKBHY3DE

18. Chinese test questions text

https://lnkd.in/dQpUd8xC

19. Chinese medical question answering data

https://lnkd.in/dsbWUCpz

20. Multi-round interpersonal dialogues text data

https://lnkd.in/dQiUq_Jg

21. Human activity recognition dataset

https://lnkd.in/dHM52MfV

22. Facial expression recognition dataset

https://lnkd.in/dqQAfMau

23. Urban surveillance dataset

https://lnkd.in/dc2RCnTk

24. Human body segmentation dataset

https://lnkd.in/d6sSrDxS

25. Fashion segmentation – clothing & accessories

https://lnkd.in/dptNUTz8

26. Fight video dataset – action recognition

https://lnkd.in/dnY_m5hZ

27. Gesture recognition dataset

https://lnkd.in/dFVPivYg

28. Facial skin defects dataset

https://lnkd.in/dKCbUvU6

29. Smoke detection and behaviour recognition dataset

https://lnkd.in/ddGg56R4

30. Weight loss transformation video dataset

https://lnkd.in/dqqT4ed9

https://t.me/CodeProgrammer 👾

CNN vs Vision Transformer — The Battle for Computer Vision 👁⚡️

Two architectures. One goal: identify the cat. But they see things differently:

🧠 CNN (Convolutional Neural Network)

· Scans the image with filters
· Detects local patterns first (edges → textures → shapes)
· Builds understanding layer by layer

🔄 Vision Transformer (ViT)

· Splits image into patches (like words in a sentence)
· Detects global patterns from the start
· Sees the whole picture using attention mechanisms

Same input. Same output. Different journey.

CNNs think locally and build up.
Transformers think globally from the get-go.

Which one wins? Depends on the task — but both are shaping the future of how machines see.

https://t.me/CodeProgrammer

How a CNN sees images simplified 🧠

1. Input → Image breaks into pixels (RGB numbers)

2. Feature Extraction

· Convolution → Detects edges/patterns
· ReLU → Kills negatives, adds non-linearity
· Pooling → Shrinks data, keeps what matters

3. Fully Connected → Flattens features into meaning

4. Output → Probability scores: Cat? Dog? Car?

Why powerful: Learns hierarchically — edges → shapes → objects

Pixels to predictions. That's it. 👇

#DeepLearning #CNN #ComputerVision #AI

https://t.me/CodeProgrammer

🚀 𝐓𝐎𝐏 𝐑𝐀𝐆 𝐈𝐍𝐓𝐄𝐑𝐕𝐈𝐄𝐖 𝐐𝐔𝐄𝐒𝐓𝐈𝐎𝐍𝐒 𝐀𝐍𝐃 𝐀𝐍𝐒𝐖𝐄𝐑𝐒 ⁣⁣

🔹 Advanced #RAG engineering concepts⁣⁣
• Multi-stage retrieval pipelines⁣⁣
• Agentic RAG vs classical RAG⁣⁣
• Latency optimization⁣⁣
• Security risks in enterprise RAG systems⁣⁣
• Monitoring and debugging production RAG systems⁣⁣
⁣⁣
📄 𝐓𝐡𝐞 𝐏𝐃𝐅 𝐜𝐨𝐧𝐭𝐚𝐢𝐧𝐬 𝟒𝟎 𝐬𝐭𝐫𝐮𝐜𝐭𝐮𝐫𝐞𝐝 𝐪𝐮𝐞𝐬𝐭𝐢𝐨𝐧𝐬 𝐰𝐢𝐭𝐡 𝐜𝐥𝐞𝐚𝐫 𝐞𝐱𝐩𝐥𝐚𝐧𝐚𝐭𝐢𝐨𝐧𝐬 𝐭𝐨 𝐡𝐞𝐥𝐩 𝐲𝐨𝐮 𝐮𝐧𝐝𝐞𝐫𝐬𝐭𝐚𝐧𝐝 𝐛𝐨𝐭𝐡 𝐜𝐨𝐧𝐜𝐞𝐩𝐭𝐬 𝐚𝐧𝐝 𝐬𝐲𝐬𝐭𝐞𝐦 𝐝𝐞𝐬𝐢𝐠𝐧 𝐭𝐡𝐢𝐧𝐤𝐢𝐧𝐠.⁣⁣
⁣⁣
https://t.me/CodeProgrammer

Horizon Lab 🔭 Джеймс Вебб знаходить галактики, яких не мало б існувати за нашими моделями. Hubble бачить зірки, що вибухнули мільярди років тому. Пишемо про це щодня — українською, на основі наукових публікацій.
👉 http://t.me/horizonlab_space

Python Tip: Operator Overloading

This is a very important concept in Python.

Have you ever wondered how #Python understands what the + operator means? For numbers, it's addition; for strings, it's concatenation; for lists, it's union. This is operator overloading in action.

Operator overloading means defining special behavior for operators (+, -, *, ==, etc.) in your user-defined classes. You determine how these operators should work with your objects.


👉 https://t.me/Python53

Machine Learning with Python pinned a photo

Rocket.new lets you build a full website using prompts with their vibe solutioning platform 🧠⚡️
You describe it, it does the work.

🎁 For the first time on this channel: 100% OFF for 2 months
🛒 Coupon code: X7K2M9P4R1NQ
✔️ Valid on all pricing plans

Go to Rocket.new now, enter the code, claim your 2 months free, or miss out and come back later paying the full subscription. 💸

claim your 2 months free

🛫 ML Roadmap 2026 — a comprehensive guide to entering ML, LLM, and MLOps

A rather insightful ML roadmap has gone viral on GitHub: within it, the author has compiled a path from a foundation in mathematics, NumPy, and Pandas to LLM, agentic RAG, fine-tuning, MLOps, and interview preparation. The repository indeed includes sections on Karpathy, MCP, RLHF, LoRA/PEFT, and system design for AI interviews.

Conveniently, this isn't just a list of random links, but rather a structured route through the topics:
▶️ Foundations and tools;
▶️ Classic ML;
▶️ LLM and agents;
▶️ Engineering and MLOps;
▶️ Interview preparation.

➡️ GitHub link:
https://github.com/loganthorneloe/ml-roadmap

tags: #ml #llm

➡ https://t.me/CodeProgrammer