Le Khac Tuan - My Blog

Intro to Large Language Models: A Summary of Andrej Karpathy’s Talk

andrej karpathy

llm

neural networks

Summary of Andrej Karpathy’s “Intro to Large Language Models” talk, deeply diving into the core concepts, current state, future directions, and security challenges surrounding LLMs.

NN-Z2H Lesson 6: Building makemore part 5 - Building a `WaveNet`

til

python

andrej karpathy

nn-z2h

neural networks

CNN/WaveNet and a lot of reading of documentation, keeping track of multidimensional tensor shapes, moving between jupyter notebooks and repository code, …

Advanced SQL

til

sql

SQL things that I wish I could know earlier.

NN-Z2H Lesson 5: Building makemore part 4 - Becoming a Backprop Ninja

til

python

andrej karpathy

nn-z2h

neural networks

backpropagation

Become swole doge: build a 2-layer MLP and deep dive to how gradients flow backwards with cross entropy loss, 2nd linear layer, tanh, batchnorm, 1st linear layer, and the embedding table.

NN-Z2H Lesson 4: Building makemore part 3 - Activations & Gradients, BatchNorm

til

python

andrej karpathy

nn-z2h

neural networks

dive into the internals of MLPs, scrutinize the statistics of the forward pass activations, backward pass gradients, understand the health of your deep network, introduce batch normalization

NN-Z2H Lesson 3: Building makemore part 2 - MLP

til

python

andrej karpathy

nn-z2h

neural networks

implement a multilayer perceptron (MLP) character-level language model, introduce model training, learning rate tuning, hyperparameters, evaluation, train/dev/test splits, under/overfitting, etc.

NN-Z2H Lesson 2: The spelled-out intro to language modeling - building makemore

til

python

andrej karpathy

nn-z2h

bigram

neural networks

implement a bigram character-level language model, focus on (1) introducing torch, and (2) the overall framework of language modeling that includes model training, sampling, and the evaluation of a loss

Data Science Project with Mamba, Python, and VS Code on Window

python

In search for the best environment & dependencies management tool for data project

Keynote on: Causal Effect Estimation in Practice - lessons learned from E-commerce & Banking

causal inference

by Danial Senejohnny, PyData Amsterdam 2024. Thumbnail credit to this post

Một số khuôn mẫu tư duy trong Brazilian Jiu-jitsu

bjj

a quick and short note on “Mechanic Models of BJJ: A Crash Course” by Steve Kwan

Note on Mathematics as the Universal Language of Nature

math

talk by Robert Seiringer (IST Austria), hosted by VIASM, Hue University, August 8 2024

A Prefect Workshop

python

pydata

prefect

by Dr. Adam Hill in PyData London 2024

A lesson of advanced python from Juan Rodríguez

python

pydata

The lecture by Juan Luis Cano Rodríguez in Master in Business Analytics and Big Data, 2021-2022, which focused on modern SWE practices using Python and the insight we can gain on how Data Science projects are put into production.

Now I know `docker init` is a thing

docker

python

How to properly use Docker to containerize your Python project

BJJ Beginners Guide

bjj

Thumbnail image credit to this post

Another James Powell’s impromptu talk, summary by Clauder 3.5 Sonnet

python

pydata

Look how he has confidently presenting and live coding with broken laptop, bad keyboard

Recap: So you want to be a Python expert?

python

pydata

by James Powell, PyData Seattle 2017

Vim 101

python

vim

actually NeoVim which is, actually Vim

Về quy ước đặt tên dữ liệu

data quality

name convention

controlled vocabulary

metadata

Danh pháp quy ước để đặt tên cho dữ liệu dạng bảng, tăng tốc độ hiểu biết và sử dụng dữ liệu

Random youtube video: How are memories stored in neural networks?

neural networks

bro made a channel dedicated to neural networks, posted one video, gained 24k subs, decided this was enough and left - @ginqus

Hi Docker 🐳

docker

python

A single-member full-stack data team me learning Docker 😂

I am starting training Brazilian Jiu-jitsu

life

bjj

When you do Jiu-jitsu, you won’t look at a bigger guy and say: Oh No!, you’ll look and say: How Interesting! - Rener Gracie.

Bitcoin tour in Python

til

crypto

bitcoin

python

andrej karpathy

Nhân việc thánh Andrej Karpathy ra MV tutorial mới về build GPT-2 from scratch, tôi lại ghé thăm trang web cá nhân của anh ấy.

Observable Framework try-out

Dashboard

Observable

til

My first Observable Framework project

NN-Z2H Lesson 1: The spelled-out intro to neural networks and backpropagation - building micrograd

til

python

andrej karpathy

nn-z2h

backpropagation

neural networks

backpropagation, from neuron to neural network, from micro grad to pytorch, and more

TIL: Xử lý dynamic-ragged array trong Excel

excel

lambda

til

Từ khi các hàm mảng như VTACK() xuất hiện, mình thường xuyên sử dụng các patterns như REDUCE/VSTACK để mở rộng khả năng xử lí các phần tử của mảng: thay vì chỉ tính toán với đầu ra là aggregate value, nay đã có thể tạo ra các mảng con. Tuy nhiên không phải lúc nào cũng có thể làm việc với mảng trong mảng một cách đơn giản

Storytelling with Data: Let’s Practice, Improve this table!

python

visualization

Practicing to improve a cross-tab report table using Python

Python is cool ❄

python

lambda

Less-known Python features

Kinh tế lượng với Julia 1: Làm quen với Julia

QEJ

Julia

Quant Econ

Đây là bài khởi đầu, làm quen với ngôn ngữ Julia trong chuỗi series các mô hình Kinh tế lượng - nguồn

Cách đọc Guitar Tab đơn giản

Guitar

Bạn có thể không biết đọc nhạc nhưng ít nhất bạn phải biết cách đọc tab guitar!

A tuturial of ggplot2

R

ggplot2

Data Visualization

This is a again a lesson on ggplot2