Source: https://seerepo.com/r/hpcaitech/ColossalAI/built
Markdown source: https://seerepo.com/r/hpcaitech/ColossalAI/built.md
Description: ColossalAI is a Python repo built with FastAPI, uvicorn, anyio, and Ray. It reaches it with SQLAlchemy. It has 317,403 written lines, with 505 test files.

# What is ColossalAI built with?

ColossalAI is a Python repo built with FastAPI, uvicorn, anyio, and Ray. It reaches it with SQLAlchemy, uses AI through ModelScope, openai, scikit-learn, sentence-transformers, tiktoken, and 20 more, makes its content with jinja2, markdown-it-py, markupsafe, mdurl, pypdf, and 3 more, validates its data with annotated-types, pydantic-core, and Pydantic, is tested with Pytest, deploys with psutil, and uses chromadb and Hugging Face. It has 317,403 written lines, with 505 test files.

Read on November 6, 2026, at commit da8f5c2.

Repository: https://github.com/hpcaitech/ColossalAI

## Frontend

### Interface

- **[Gradio](https://seerepo.com/uses/gradio)**
- **seaborn**
- **[matplotlib](https://seerepo.com/uses/matplotlib)**: 2 files
- **[Rich](https://seerepo.com/uses/rich)**: 1 file

## Backend

### The base

- **[FastAPI](https://seerepo.com/uses/fastapi)**: 3 files
- **[uvicorn](https://seerepo.com/uses/uvicorn)**: 1 file

### Data

- **[chromadb](https://seerepo.com/uses/chromadb)**

### Data access

- **[SQLAlchemy](https://seerepo.com/uses/sqlalchemy)**: 2 files

### AI

- **[ModelScope](https://seerepo.com/uses/modelscope)**
- **[openai](https://seerepo.com/uses/openai)**
- **[scikit-learn](https://seerepo.com/uses/scikit-learn)**
- **[sentence-transformers](https://seerepo.com/uses/sentence-transformers)**
- **[tiktoken](https://seerepo.com/uses/tiktoken)**
- **[torch](https://seerepo.com/uses/torch)**: 649 files
- **[transformers](https://seerepo.com/uses/transformers)**: 135 files
- **[LangChain](https://seerepo.com/uses/langchain)**: 15 files
- **triton**: 11 files
- **[peft](https://seerepo.com/uses/peft)**: 10 files
- **[diffusers](https://seerepo.com/uses/diffusers)**: 9 files
- **flash-attn**: 9 files
- **[datasets](https://seerepo.com/uses/datasets)**: 7 files
- **timm**: 6 files
- **[safetensors](https://seerepo.com/uses/safetensors)**: 5 files
- **[bitsandbytes](https://seerepo.com/uses/bitsandbytes)**: 3 files
- **vLLM**: 3 files
- **[einops](https://seerepo.com/uses/einops)**: 2 files
- **[accelerate](https://seerepo.com/uses/accelerate)**: 1 file
- **galore-torch**: 1 file
- **math-verify**: 1 file
- **[sentencepiece](https://seerepo.com/uses/sentencepiece)**: 1 file
- **[tokenizers](https://seerepo.com/uses/tokenizers)**: 1 file
- **[torchaudio](https://seerepo.com/uses/torchaudio)**: 1 file
- **[torchvision](https://seerepo.com/uses/torchvision)**: 1 file

### Services

- **[Hugging Face](https://seerepo.com/uses/hugging-face)**: 1 file · [colossalai/legacy/moe/openmoe/train.py, line 9](https://github.com/hpcaitech/ColossalAI/blob/da8f5c2dd56f97d3d23d3d222b114e1a3d4710f5/colossalai/legacy/moe/openmoe/train.py#L9)

## Support

### The base

- **[anyio](https://seerepo.com/uses/anyio)**
- **Ray**: 20 files

### Content

- **[jinja2](https://seerepo.com/uses/jinja2)**
- **[markdown-it-py](https://seerepo.com/uses/markdown-it-py)**
- **[markupsafe](https://seerepo.com/uses/markupsafe)**
- **mdurl**
- **[pypdf](https://seerepo.com/uses/pypdf)**
- **unstructured**
- **[pillow](https://seerepo.com/uses/pillow)**: 3 files
- **jsonlines**: 1 file

### Validation

- **[annotated-types](https://seerepo.com/uses/annotated-types)**
- **[pydantic-core](https://seerepo.com/uses/pydantic-core)**
- **[Pydantic](https://seerepo.com/uses/pydantic)**: 5 files

### Testing

- **Pytest**: 265 files

### Deploy

- **[psutil](https://seerepo.com/uses/psutil)**: 3 files

### Tools

- **annotated-doc**
- **autoflake**
- **Black**
- **coverage**
- **iniconfig**
- **ninja**
- **pip**
- **pluggy**
- **pygments**
- **shellingham**
- **tensorboard**
- **Typer**
- **typing-extensions**
- **wheel**
- **packaging**: 28 files
- **click**: 7 files
- **Setuptools**: 1 file

### Other

- **certifi**
- **fsspec**
- **fuzzywuzzy**
- **[google](https://seerepo.com/uses/google)**
- **h11**
- **hf-xet**
- **httpcore**
- **httpx**
- **idna**
- **jieba**
- **mpmath**
- **networkx**
- **protobuf**
- **regex**
- **six**
- **sse-starlette**
- **SymPy**
- **typing-inspection**
- **NumPy**: 35 files
- **tqdm**: 20 files
- **wandb**: 11 files
- **requests**: 6 files
- **tabulate**: 3 files
- **filelock**: 1 file
- **latex2sympy2-extended**: 1 file
- **pandas**: 1 file
- **pyyaml**: 1 file

## Other repos with a similar stack

- [ray-project/ray](https://seerepo.com/r/ray-project/ray/built): FastAPI, uvicorn, Hugging Face, anyio, Ray, Gradio, matplotlib, Rich, SQLAlchemy, scikit-learn, tiktoken, torch, transformers, LangChain, triton, peft, diffusers, datasets, timm, safetensors, bitsandbytes, einops, accelerate, sentencepiece, tokenizers, torchaudio, torchvision, jinja2, markdown-it-py, markupsafe, mdurl, pillow, jsonlines, annotated-types, pydantic-core, Pydantic, psutil · [Compare](https://seerepo.com/compare/hpcaitech/ColossalAI/vs/ray-project/ray)
- [run-llama/llama_index](https://seerepo.com/r/run-llama/llama_index/built): chromadb, anyio, Ray, matplotlib, Rich, SQLAlchemy, ModelScope, OpenAI, scikit-learn, sentence-transformers, tiktoken, torch, transformers, LangChain, peft, datasets, vLLM, einops, sentencepiece, tokenizers, jinja2, markupsafe, pypdf, unstructured, pillow, annotated-types, pydantic-core, Pydantic, psutil · [Compare](https://seerepo.com/compare/hpcaitech/ColossalAI/vs/run-llama/llama_index)
- [unslothai/unsloth](https://seerepo.com/r/unslothai/unsloth/built): FastAPI, uvicorn, Hugging Face, anyio, matplotlib, Rich, OpenAI, scikit-learn, sentence-transformers, tiktoken, torch, transformers, peft, diffusers, datasets, timm, safetensors, einops, accelerate, sentencepiece, tokenizers, jinja2, markdown-it-py, mdurl, pillow, Pydantic, psutil · [Compare](https://seerepo.com/compare/hpcaitech/ColossalAI/vs/unslothai/unsloth)
- [mudler/LocalAI](https://seerepo.com/r/mudler/LocalAI/built): FastAPI, uvicorn, Hugging Face, anyio, Gradio, matplotlib, ModelScope, OpenAI, sentence-transformers, tiktoken, torch, transformers, peft, diffusers, flash-attn, datasets, safetensors, bitsandbytes, vLLM, einops, accelerate, sentencepiece, tokenizers, torchaudio, torchvision, jinja2, pillow · [Compare](https://seerepo.com/compare/hpcaitech/ColossalAI/vs/mudler/LocalAI)
- [datawhalechina/hello-agents](https://seerepo.com/r/datawhalechina/hello-agents/built): FastAPI, uvicorn, chromadb, Hugging Face, anyio, Gradio, matplotlib, Rich, SQLAlchemy, OpenAI, scikit-learn, sentence-transformers, tiktoken, torch, transformers, peft, datasets, safetensors, accelerate, tokenizers, jinja2, pypdf, pillow, annotated-types, pydantic-core, Pydantic · [Compare](https://seerepo.com/compare/datawhalechina/hello-agents/vs/hpcaitech/ColossalAI)
- [bojieli/ai-agent-book](https://seerepo.com/r/bojieli/ai-agent-book/built): FastAPI, uvicorn, chromadb, Hugging Face, anyio, seaborn, matplotlib, Rich, OpenAI, scikit-learn, sentence-transformers, tiktoken, torch, transformers, triton, peft, datasets, bitsandbytes, vLLM, accelerate, sentencepiece, torchaudio, pypdf, pillow, Pydantic, psutil · [Compare](https://seerepo.com/compare/bojieli/ai-agent-book/vs/hpcaitech/ColossalAI)

## Frequently asked questions

### What language is ColossalAI written in?

ColossalAI is written in Python, with 317,403 written lines.

### What framework does ColossalAI use?

ColossalAI is built with FastAPI, uvicorn, anyio, and Ray.

### Does ColossalAI use AI?

Yes. ColossalAI uses AI through ModelScope, openai, scikit-learn, sentence-transformers, tiktoken, torch, transformers, LangChain, triton, peft, diffusers, flash-attn, datasets, timm, safetensors, bitsandbytes, vLLM, einops, accelerate, galore-torch, math-verify, sentencepiece, tokenizers, torchaudio, and torchvision.

### What services does ColossalAI use?

ColossalAI uses chromadb and Hugging Face.

### Does ColossalAI have tests?

Yes. ColossalAI has 505 test files and is tested with Pytest.

### Which repos have a stack like ColossalAI's?

The ones that share most of its stack: ray-project/ray, run-llama/llama_index, and unslothai/unsloth.

## What is your repo built with?

Any public repo, with no account: https://seerepo.com/r/owner/repo/built, with the owner and name it has on GitHub.