# Hugging Face Transformers v3.1.0 — Release & Upgrade Intelligence
- **Product:** Hugging Face Transformers (https://huggingface.co)
- **Version:** v3.1.0
- **Released:** Sep 01, 2020
- **Impact Score:** 7/10
- **Breaking Changes:** Source flagged
- **Canonical Analysis:** https://changelogs.us/transformers/updates/v3.1.0/

## Executive Summary (TL;DR)
Hugging Face Transformers v3.1.0. Source highlight: Pegasus, mBART, DPR, self-documented outputs and new pipelines

## Breaking Changes & Deprecations
- One **breaking behavior** of these new outputs (which is the reason you have to opt-in to use these new outputs:

## New Features
- add pegasus finetuning script #6811 script. (warning very slow)
- Add DPR model #5279 (@lhoestq)
- Add DeeBERT (entropy-based early exiting for *BERT) #5477 (@Ji-Xin)
- [EncoderDecoder] Add encoder-decoder for roberta/ vanilla longformer #6411 (@patrickvonplaten)
- [EncoderDecoder] Add Cross Attention for GPT2 #6415 (@patrickvonplaten)
- [EncoderDecoder] Add functionality to tie encoder decoder weights #6538 (@patrickvonplaten)

## Improvements
- [s2s] tiny QOL improvement: run_eval prints scores #6341 (@sshleifer)
- Improvements to PretrainedConfig documentation #5642 (@sgugger)

## Bug Fixes
- Fix tests imports dpr #5576 (@lhoestq)
- [Almost all TF models] TF clean up: add missing CLM / MLM loss; fix T5 naming and keras compile #5395 (@patrickvonplaten)
- [fix] mbart_en_ro_generate test now identical to fairseq #5731 (@sshleifer)
- Fix #6096: MBartTokenizer's mask token #6098 (@sshleifer)
- [examples (seq2seq)] fix preparing decoder_input_ids for T5 #5994 (@patil-suraj)
- s2s: fix LR logging, remove some dead code. #6205 (@sshleifer)

## Upgrade Command
```bash
python -m pip install 'transformers==3.1.0'
```