@lerabyte: Let’s build a language model completely from scratch. Our language model can finally speak, and it sounds terrible. In Part 3, we build and train a bigram model that predicts the next character using only the current character. The results are messy, but this simple baseline teaches us the complete training and generation process. Build it with me using the GitHub repository: https://github.com/lerabyte/llm-from-scratch #ai #llm #deeplearning #python #coding

Kimberly :

Oh my goodness, it’s like it’s solving sudokus lol

memnoch :

hey new to LLM stuff, I've made a couple harnesses and have started reading the LLM engineering handbook. if one is looking to make an ai voice assistant for a very specific technology, is it better to build and train your own or try to steer an existing open source model?

evgeny.va5 :

I’m so excited for part 4! I’ve been following along in GitHub 😁

Camilio13 :

Do you always fully understand what's going on behind the scenes? Because I recently started learning machine learning and I'm trying to understand things like Back Propagation, Loss Functions, Gradient Descent and all those mathematical terms, but it's so complex. Do you think it's smart to just continue onto coding or should I fully understand what everything does and how the maths work?

Ahmed Maher :

do u have a telegram channel or discord server

Life Of Freedom :

but where you doing i mean in python

Tatiana :

Love this series!

Youssef Elebiary :

Are u following the evolution of NLP models? start with a FFNN then an RNN then LSTM then Attention?

a_trillion :

Your course is definitely worth buying cause you know what you’re talking about

michael :

ok, I am in 🥰 - a question though - I want to make my own custom llm that works with my very specific 18 agents - can thst be done ?

☆♡Ķîsş♡☆ :

so in short the biagram is like a neural network

_Lionhear7_ :

JC International Resources :

🥰