# Hugging Face Transformers v2.9.0 — Release & Upgrade Intelligence
- **Product:** Hugging Face Transformers (https://huggingface.co)
- **Version:** v2.9.0
- **Released:** May 07, 2020
- **Impact Score:** 7/10
- **Breaking Changes:** Source flagged
- **Canonical Analysis:** https://changelogs.us/transformers/updates/v2.9.0/

## Executive Summary (TL;DR)
Hugging Face Transformers v2.9.0. Source highlight: Trainer & TFTrainer (@julien-c)

## Breaking Changes & Deprecations
- Breaking change This introduces a breaking change, in that it increases the default output length of T5Model and T5ForConditionalGeneration from 4 to 5 (including the past_key_value_states).
- This introduces a breaking change, in that it increases the default output length of T5Model and T5ForConditionalGeneration from 4 to 5 (including the past_key_value_states).

## New Features
- Both the TensorFlow and PyTorch trainers have TPU support (@jplu, @LysandreJik, @julien-c). An additional utility is added so that the TPU scripts may be launched in a similar manner to `torch.distributed`.
- Additional tests and support has been added to `huggingface/tokenizers` tokenizers. (@mfuntowicz, @thomwolf)
- Clean Encoder-Decoder models with Bart/T5-like API and add generate possibility (@patrickvonplaten)

## Improvements
- TensorFlow models work out-of-the-box with the new tokenizers (@LysandreJik)
- Apply Encoder Decoder 1.5GB memory savings to TF as well (@patrickvonplaten, translation of same work on PyTorch models by @sshleifer)
- BART Summarization fine-tuning script now works for T5 as well (@sshleifer)

## Bug Fixes
- Clean the generate testing functions (@patrickvonplaten)
- Notebooks updated in the documentation (@LysandreJik)
- Fix RoBERTa/XLNet Pad Token in run_multiple_choice.py (@ethanjperez)
- Fixed RoBERTa conversion script (@myleott)
- Speedup torch summarization tests (@sshleifer)
- Optimize causal mask using torch.where (@Akababa)

## Upgrade Command
```bash
python -m pip install 'transformers==2.9.0'
```