Stories by xcodevn

Implementing DeepSeek R1's GRPO algorithm from scratch

The LLM pre-training data wall

PodcastLM: An open-source AI podcast creator

Scaling up self-attention inference

Scaling up self-attention inference

Letter from Professors Bengio, Hinton, Lessig, & Russell

Logit Prisms: Decomposing Transformer Outputs for Mechanistic Interpretability

Exploring MLP neurons inside Llama3 model

Self-reference is so cool

Pax: a PyTorch-like pytree module on top of Jax

Google Colab Pro+ ($50/month)

Prompting GTP-3 is like saying in your head

GPT 9000 for Speech Signal

Generative MNIST-RNN

Parallel Worlds Probably Exist. Here’s Why [video]

GPT-2: 6-Month Follow-Up

Finding the Brain’s Compass

Reinforcement learning: Fast and slow – Matthew Botvinick [video]

All About That Bayes: Probability, Statistics [video]