A lightweight vLLM implementation built from scratch.
- 🚀 Fast offline inference - Comparable inference speeds to vLLM
- 📖 Readable codebase - Clean implementation in ~ 1,200 lines of Python code
- ⚡ Optimization Suite - Prefix caching, Torch compilation, CUDA graph, etc.
pip install git+https://github.com/GeeeekExplorer/nano-vllm.gitSee example.py for usage. The API mirrors vLLM's interface with minor differences in the LLM.generate method.
See bench.py for benchmark.
Test Configuration:
- Hardware: RTX 4070
- Model: Qwen3-0.6B
- Total Requests: 256 sequences
- Input Length: Randomly sampled between 100–1024 tokens
- Output Length: Randomly sampled between 100–1024 tokens
Performance Results:
| Inference Engine | Output Tokens | Time (s) | Throughput (tokens/s) |
|---|---|---|---|
| vLLM | 133,966 | 98.95 | 1353.86 |
| Nano-vLLM | 133,966 | 101.90 | 1314.65 |