Pre-trained models for natural language understanding (NLU) and generation (NLG) tasks
The family of UniLM:
v1@NeurIPS'19 | v2@ICML'20): unified pre-training for language understanding and generation
new): multilingual/cross-lingual pre-trained models for language understanding and generation
v1@NeurIPS'20): small pre-trained models for language understanding and generation
v1@KDD'20 | v2): multimodal (text + layout/format + image) pre-training for document understanding (e.g. scanned documents, PDF, etc.)
NEW): Domain adaptation of pre-trained models
NEW): Language model pre-training for multilingual NMT
s2s-ft: sequence-to-sequence fine-tuning toolkit
- October 8th, 2020: T-ULRv2 (aka InfoXLM) as the SOTA on the XTREME leaderboard. // Blog
- September, 2020: MiniLM was accepted by NeurIPS 2020.
- July 16, 2020 (
NEW): InfoXLM (Multilingual UniLM) arXiv
- June, 2020: UniLMv2 was accepted by ICML 2020; LayoutLM was accepted by KDD 2020.
- April 5, 2020: Multilingual MiniLM released!
- September, 2019: UniLMv1 was accepted by NeurIPS 2019.
- LayoutLM 1.0 (February 18, 2020): pre-trained models for document (image) understanding (e.g. receipts, forms, etc.) . It achieves new SOTA results in several downstream tasks, including form understanding (the FUNSD dataset from 70.72 to 79.27), receipt understanding (the ICDAR 2019 SROIE leaderboard from 94.02 to 95.24) and document image classification (the RVL-CDIP dataset from 93.07 to 94.42). "LayoutLM: Pre-training of Text and Layout for Document Image Understanding
- s2s-ft 1.0 (February 26, 2020): A PyTorch package used to fine-tune pre-trained Transformers for sequence-to-sequence language generation. "s2s-ft: Fine-Tuning Pre-Trained Transformers for Sequence-to-Sequence Learning"
- MiniLM 1.0 (February 26, 2020): deep self-attention distillation is all you need (for task-agnostic knowledge distillation of pre-trained Transformers). MiniLM (12-layer, 384-hidden) achieves 2.7x speedup and comparable results over BERT-base (12-layer, 768-hidden) on NLU tasks as well as strong results on NLG tasks. The even smaller MiniLM (6-layer, 384-hidden) obtains 5.3x speedup and produces very competitive results. "MiniLM: Deep Self-Attention Distillation for Task-Agnostic Compression of Pre-Trained Transformers
- UniLM 2.0 (February 28, 2020): unified pre-training of bi-directional LM (via autoencoding) and sequence-to-sequence LM (via partially autoregressive) w/ Pseudo-Masked Language Model for language understanding and generation. UniLM v2 achieves new SOTA in a wide range of natural language understanding and generation tasks. "UniLMv2: Pseudo-Masked Language Models for Unified Language Model Pre-Training
***** October 1st, 2019: UniLM v1 release *****
- UniLM v1 (September 30, 2019): the code and pre-trained models for the
NeurIPS 2019paper entitled "Unified Language Model Pre-training for Natural Language Understanding and Generation". UniLM (v1) achieves the new SOTA results in NLG (especially sequence-to-sequence generation) tasks, including abstractive summarization (the Gigaword and CNN/DM datasets), question generation (the SQuAD QG dataset), etc.
This project is licensed under the license found in the LICENSE file in the root directory of this source tree. Portions of the source code are based on the transformers project.
For help or issues using UniLM, please submit a GitHub issue.
For other communications related to UniLM, please contact Li Dong (
firstname.lastname@example.org), Furu Wei (