AGENT SKILL

autoresearch

Autonomous ML research agent inspired by Karpathy's autoresearch. Given a training codebase, the agent runs continuous experimentation loops: modify model code, train for 5 minutes, check if val_bpb improved, keep or discard changes. Use when the user mentions 'start autoresearch', 'auto experiment', 'train agent', 'autonomous research', '自动实验', '自动训练', '自动优化模型'.

License 许可
MIT
最近更新
2026-11-01
Stars
★ 3
Forks
1

01 前言 Front Matter

name
autoresearch
description
Autonomous ML research agent inspired by Karpathy's autoresearch. Given a training codebase, the agent runs continuous experimentation loops: modify model code, train for 5 minutes, check if val_bpb improved, keep or discard changes. Use when the user mentions 'start autoresearch', 'auto experiment', 'train agent', 'autonomous research', '自动实验', '自动训练', '自动优化模型'.

02 资源文件 3

03 SKILL.md 正文

Autoresearch — Autonomous ML Research Agent

Overview

Karpathy's autoresearch gives an AI agent a small but real LLM training setup and lets it experiment autonomously. The agent modifies code, trains, checks results, keeps or discards changes, and repeats indefinitely.

Files

  • prepare.py — Fixed constants, data prep, tokenizer, dataloader, evaluation. Do NOT modify.
  • train.py — The single file the agent edits. Model architecture, optimizer, training loop.
  • program.md — Full agent instructions for the autonomous research loop (read this to understand how to run).

How to Use

  1. Read program.md for the complete autonomous research workflow
  2. Set up the environment (NVIDIA GPU, Python 3.10+, uv)
  3. Run uv sync and uv run prepare.py to prepare data
  4. Start the autonomous loop — read program.md and follow the experiment loop instructions
  5. Let it run — the agent should NEVER stop on its own

Key Constraints

  • Only modify train.py — never prepare.py
  • Each experiment runs for exactly 5 minutes (fixed time budget)
  • Log results to results.tsv (tab-separated)
  • Keep only improvements (lower val_bpb), discard regressions
  • Use git branches (autoresearch/<tag>) for experiment isolation

Reference

See program.md for the complete workflow including setup, experimentation loop, output format, and logging.

04 同仓库更多 Skills

shengdabai
shengdabai/Tony-Claude-Code-Skills

seedance-prompt-en

Write effective prompts for Jimeng Seedance 2.0 multimodal AI video generation. Use when users want to create video prompts using text, images, videos, and audio inputs with the @ reference system. Covers camera movements, effects replication, video extension, editing, music beat-matching, e-commerce ads, short dramas, and educational content.

★ 3
shengdabai
shengdabai/Tony-Claude-Code-Skills

onepage-pdf

Convert an HTML page into a single continuous-page PDF (one tall page, no pagination breaks), preserving desktop layout, backgrounds and selectable text. Use when the user asks to turn an HTML file/report/proposal into a "single-page PDF", "long PDF", "不分页 PDF", "单页 PDF", "长图式 PDF", or complains that an HTML-to-PDF export breaks into pages or cuts content. Supports optional string redaction with leak verification.

★ 3
shengdabai
shengdabai/Tony-Claude-Code-Skills

native-feel-cross-platform-desktop

Use when the user is designing, prototyping, or rewriting a desktop app that must run on multiple OSes (macOS + Windows, optionally Linux) AND feel indistinguishable from a native app to its users — fast launch, native windowing, native input handling, native materials. Trigger words include "cross-platform desktop", "Electron alternative", "Tauri vs native", "WebView wrapper", "near-native performance", "Raycast architecture", "WebKit/WebView2 quirks", "WKWebView", "system tray app", "global hotkey app", "launcher app". Do NOT trigger this skill for pure web apps, pure mobile apps, or for greenfield projects that have no native-feel requirement.

★ 3
shengdabai
shengdabai/Tony-Claude-Code-Skills

gstack

Router for the gstack skill suite. (gstack)

★ 3
shengdabai
shengdabai/Tony-Claude-Code-Skills

fireworks-tech-graph

Use when the user wants to create any technical diagram - architecture, data flow, flowchart, sequence, agent/memory, or concept map - and export as SVG+PNG. Trigger on: "画图" "帮我画" "生成图" "做个图" "架构图" "流程图" "可视化一下" "出图" "generate diagram" "draw diagram" "visualize" or any system/flow description the user wants illustrated.

★ 3
shengdabai
shengdabai/Tony-Claude-Code-Skills

dbs-video-extract

接收抖音、小红书、微信视频号等短视频链接或分享文案,查询作品/账号数据并提取语音文字稿,按作者和标题归档为 Markdown。用户希望完整解析短视频、取得数据或文字稿时使用;运行前检查 TikHub 与轻抖使用凭证,缺少时用新手能理解的方式引导购买和安全配置。

★ 3