Megatron-LM vs LLMs-from-scratch
A neutral, constraint-first comparison - live GitHub stats and typed relationships, not marketing.
| Megatron-LM | LLMs-from-scratch | |
|---|---|---|
| Tagline | GPU-optimized library for training transformer models at scale | Implement a ChatGPT-like LLM in PyTorch from scratch |
| Stars | 17k | 99k |
| Forks | 4.2k | 15k |
| Open issues | 990 | 4 |
| Language | Python | Jupyter Notebook |
| License | Other | Other |
| Last pushed | Jul 7, 2026 | Jun 2, 2026 |
| Categories | LLM Frameworks | Model Training, LLM Frameworks |
Megatron-LM
Megatron-LM is a repository containing Megatron Core, which provides GPU-optimized building blocks for custom training frameworks, and includes pre-configured training scripts for quick experimentation.
Python
LLMs-from-scratch
Repository containing code and instructions for developing, pretraining, and finetuning a GPT-like large language model (LLM) using PyTorch.
Jupyter Notebook