create-llm

The fastest way to build and start training your own LLM. CLI tool that scaffolds production-ready PyTorch training projects in seconds. Like create-next-app bu

What is create-llm?

The fastest way to build and start training your own LLM. CLI tool that scaffolds production-ready PyTorch training projects in seconds. Like create-next-app but for language models.

Overview of create-llm

What it is

create-llm is a command-line tool that scaffolds production-ready PyTorch training projects for language models. It generates a complete project structure including model architecture, data preprocessing, tokenizer training, training loops, evaluation, and deployment scripts. The tool is designed to be used like create-next-app but for language model training. It is available as an npm package and can be run via npx without installation. The project is licensed under MIT and is open source.

Key features

  • Right-Sized Templates: Four templates (NANO ~1M params, TINY ~6M params, SMALL ~100M params, BASE ~1B params) optimized for different hardware and use cases.
  • Complete Toolkit: Includes PyTorch training infrastructure, data preprocessing pipeline, tokenizer training (BPE, WordPiece, Unigram), checkpoint management with auto-save, TensorBoard integration, interactive chat interface, model comparison tools, and deployment scripts.
  • Smart Defaults: Auto-detects vocab size from tokenizer, handles sequence length mismatches, warns about model/data size mismatches, detects overfitting, suggests optimal hyperparameters, handles cross-platform paths, and provides detailed diagnostic messages.
  • Plugin System: Optional integrations for WandB (experiment tracking) and HuggingFace (model sharing).
  • Docker Support: Run create-llm without installing Node.js or Python locally; includes Docker Compose support for GPU training and chat interface.

Use cases

  • Learning and quick experiments (NANO template, any CPU, 100+ examples).
  • Prototyping and small projects (TINY template, CPU or basic GPU, 1,000+ examples).
  • Production applications (SMALL template, RTX 3060+ GPU, 10,000+ examples).
  • Research and high-quality models (BASE template, A100 or multi-GPU, 100,000+ examples).

Getting started

bash npx create-llm my-awesome-llm cd my-awesome-llm pip install -r requirements.txt python training/train.py For interactive setup: bash npx create-llm For quick mode with all options specified: bash npx create-llm my-llm --template tiny --tokenizer bpe --skip-install

License and pricing

The project is licensed under MIT and is open source. Pricing is not stated beyond the open source license.

Based on the project's public README and repository metadata.

Add this badge to your README

Featured on YunoDB Badge styles & help

Maintain this project? Show that it is featured on YunoDB.

Markdown[![Featured on YunoDB](https://yunodb.com/badge/theaniketgiricreate-llm.svg)](https://yunodb.com/tools/theaniketgiricreate-llm?utm_source=github&utm_medium=badge)
HTML<a href="https://yunodb.com/tools/theaniketgiricreate-llm?utm_source=github&utm_medium=badge"><img src="https://yunodb.com/badge/theaniketgiricreate-llm.svg" alt="Featured on YunoDB" height="20"></a>