Developer Tools / Machine Learning / LLM Architecture
A PyTorch reference implementation of DeepSeek Sparse Attention from the LLMs-from-scratch project, including a lightweight indexer, top-K token selection, and a GPT-style model with KV-cache support.
The page clearly describes an AI/LLM-focused code project, explains the DSA mechanism, and shows concrete files, classes, and usage flags. It gives enough evidence for a useful public listing and appears aimed at developers studying or experimenting with sparse attention in transformer models.
The snapshot shows a focused implementation example, but not benchmarks, tests beyond a single file listing, or proof of end-to-end training quality. It also does not show the full repo context for how this chapter component fits into the larger project beyond the README excerpt.
Last evaluated 54 days ago. Current rank #767. Down 2 spots in the rankings.
CodeGraph is a local code knowledge graph for AI coding agents like Claude Code, Cursor, Codex, OpenCode, and Hermes Agent. It aims to cut token use, tool calls, and runtime by letting agents query pre-indexed code structure instead of scanning files repeatedly.
Single-binary CLI that builds an AI agent knowledge base from git repos, Claude Code/Codex sessions, and saved links. It generates a portable markdown wiki, runs on cron, supports local Ollama mode, and exposes the result for agents via CLAUDE.md/AGENTS.md and MCP.
A lightweight React Markdown renderer with built-in LaTeX, syntax highlighting, streaming-safe rendering, and security-focused defaults.
Claude Code plugin that blocks dependency edits until a fresh, source-cited version check is recorded, helping prevent hallucinated or stale package versions across npm, pip, Poetry/uv, Cargo, and NuGet.
