The article details “autoresearch,” a project by Karpathy where an AI agent autonomously experiments with training a small language model (nanochat) to improve its performance. The agent modifies the `train.py` file, trains for a fixed 5-minute period, and evaluates the results, repeating this process to iteratively refine the model. The project aims to demonstrate autonomous AI research, focusing on a simplified, single-GPU setup with a clear metric (validation bits per byte).
* **Autonomous Research:** The core concept of AI-driven experimentation.
* **nanochat:** The small language model used for training.
* **Fixed Time Budget:** Each experiment runs for exactly 5 minutes.
* **program.md:** The file containing instructions for the AI agent.
* **Single-File Modification:** The agent only edits `train.py`.
* **Autonomous Research:** The core concept of AI-driven experimentation.
* **nanochat:** The small language model used for training.
* **Fixed Time Budget:** Each experiment runs for exactly 5 minutes.
* **program.md:** The file containing instructions for the AI agent.
* **Single-File Modification:** The agent only edits `train.py`.
This repository provides the official implementation of the STATIC (Sparse Transition-Accelerated Trie Index for Constrained decoding) framework, as described in Su et al., 2026. STATIC is a high-performance method for enforcing outputs to stay within a prespecified set during autoregressive decoding from large language models, designed for maximum efficiency on modern hardware accelerators like GPUs and TPUs.
Learn how to design, develop, deploy and iterate on production-grade ML applications.
oLLM is a Python library for running large-context Transformers on NVIDIA GPUs by offloading weights and KV-cache to SSDs. It supports models like Llama-3, GPT-OSS-20B, and Qwen3-Next-80B, enabling up to 100K tokens of context on 8-10 GB GPUs without quantization.
This tutorial introduces the essential topics of the PyTorch deep learning library in about one hour. It covers tensors, training neural networks, and training models on multiple GPUs.
Unify your existing devices into one powerful GPU: iPhone, iPad, Android, Mac, NVIDIA, Raspberry Pi, pretty much any device!
The article discusses the competition Nvidia faces from Intel and AMD in the GPU market. While these competitors have introduced new accelerators that match or surpass Nvidia's offerings in terms of memory capacity, performance, and price, Nvidia maintains a strong advantage through its CUDA software ecosystem. CUDA has been a significant barrier for developers switching to alternative hardware due to the effort required to port and optimize existing code. However, both Intel and AMD have developed tools to ease this transition, like AMD's HIPIFY and Intel's SYCL. Despite these efforts, the article notes that the majority of developers now write higher-level code using frameworks like PyTorch, which can run on different hardware with varying levels of support and performance. This shift towards higher-level programming languages has reduced the impact of Nvidia's CUDA moat, though challenges still exist in ensuring compatibility and performance across different hardware platforms.
efficient method for fine-tuning LLM using LoRA and QLoRA, making it possible to train them even on consumer hardware
