#1111th of 14 in Open models
unilm
Microsoft research code and weights for pre-trained text, vision, speech and document AI
Live demo ↗
(opens in a new tab)Website ↗
(opens in a new tab)Repository on GitHub ↗
(opens in a new tab)
- Stars
- 22.2k
- License
- MIT
- Last commit
- Sep 2026
- Last release
- May 2024
- Language
- Python
Missed the latest GitHub check; these facts may be older.
Overview
A monorepo of Microsoft Research projects on large-scale self-supervised pre-training across languages, vision, speech and multimodal tasks. Each subdirectory holds code and checkpoints for a separate model family, including UniLM, MiniLM, BEiT, LayoutLM, WavLM, TrOCR, VALL-E, Kosmos and E5. It is a research collection, not a single installable tool.
Who it is for: ML researchers needing pretrained document, speech or multimodal models
Strengths
- Covers text, vision, speech and document models in one repo, from UniLM to Kosmos-2.5
- Document AI line (LayoutLM v1-v3, LayoutXLM, DiT, TrOCR) has pretrained models and code
- MIT license
- Some models have hosted demos, such as TextDiffuser-2 and DiT on Hugging Face Spaces
Weaknesses
- Not one library: each subfolder has its own setup, dependencies and README
- Last tagged release is May 2024, and most listed model releases date from 2021-2023
- Top-level README is a news and paper index with little install or usage guidance
- Hardware requirements are not stated at the top level
What it needs
- Models: PyTorch, Hugging Face Transformers
Also in Open models
See all 14| Rank | Project | Score |
|---|---|---|
| 1 | llmfitRust tool that checks which LLMs fit your RAM, CPU and GPU | 84 out of 100 |
| 2 | layaSingle-pass multilingual classifier for typed choice, score and yes/no decisions | 79 out of 100 |
| 3 | timesfmPretrained time-series foundation model for zero-shot forecasting from Google Research | 72 out of 100 |
| #4 | unlimited-ocrOCR model that parses long documents and multi-page PDFs in one pass | 62 out of 100 |
| #5 | insightfaceFace detection, recognition and alignment toolbox with ArcFace and RetinaFace models | 57 out of 100 |