# Hugging Face Transformers v4.13.0 — Release & Upgrade Intelligence
- **Product:** Hugging Face Transformers (https://huggingface.co)
- **Version:** v4.13.0
- **Released:** Dec 09, 2021
- **Impact Score:** 5/10
- **Breaking Changes:** Not verified
- **Canonical Analysis:** https://changelogs.us/transformers/updates/v4.13.0/

## Executive Summary (TL;DR)
Hugging Face Transformers v4.13.0. Source highlight: Eight new models are released as part of the Perceiver implementation: `PerceiverModel`, `PerceiverForMaskedLM`, `PerceiverForSequenceClassification`, `PerceiverForImageClassificationLearned`, `PerceiverForImageClassificationFourier`, `PerceiverForImageClassificationConvProcessing`, `PerceiverForOpt

## New Features
- Add Perceiver IO
- Add mLUKE
- Add ImageGPT
- Add QDQBert model and quantization examples of SQUAD task
- Add SegFormer
- Add BeitForSemanticSegmentation

## Improvements
- LayoutLMv2FeatureExtractor now supports non-English languages when applying Tesseract OCR.
- `--bf16` - do training or eval in mixed precision of bfloat16
- `--bf16_full_eval` - do eval in full bfloat16
- `--tf32` control having TF32 mode on/off

## Bug Fixes
- Replace assertions with RuntimeError exceptions
- Adding `batch_size` support for (almost) all pipelines
- Remove n_ctx from configs
- Add `BlenderbotTokenizerFast`
- Adding `handle_long_generation` paramters for `text-generation` pipeline.
- Fix pipeline tests env and fetch

## Upgrade Command
```bash
python -m pip install 'transformers==4.13.0'
```