# Hugging Face Transformers v4.17.0 — Release & Upgrade Intelligence
- **Product:** Hugging Face Transformers (https://huggingface.co)
- **Version:** v4.17.0
- **Released:** Mar 03, 2022
- **Impact Score:** 7/10
- **Breaking Changes:** Source flagged
- **Canonical Analysis:** https://changelogs.us/transformers/updates/v4.17.0/

## Executive Summary (TL;DR)
Hugging Face Transformers v4.17.0. Source highlight: The XGLM model was proposed in Few-shot Learning with Multilingual Language Models by Xi Victoria Lin, Todor Mihaylov, Mikel Artetxe, Tianlu Wang, Shuohui Chen, Daniel Simig, Myle Ott, Naman Goyal, Shruti Bhosale, Jingfei Du, Ramakanth Pasunuru, Sam Shleifer, Punit Singh Koura, Vishrav Chaudhary, Br

## Breaking Changes & Deprecations
- Breaking change Unfortunately, some bugs had crept into `CLIPTokenizerFast` : the tokenization produced by `CLIPTokenizer` and `CLIPTokenizerFast` were not equal. `CLIPTokenizerFast` has been corrected to encode the text with the same strategy as `CLIPTokenizer`.

## New Features
- Add XGLM models
- Add ConvNeXT
- Add TFConvNextModel
- Add PoolFormer
- Add PLBart
- Add missing PLBart entry in README

## Improvements
- FX tracing improvement
- [Test refactor 3/5] Notification service improvement
- [Test refactor 4/5] Improve the scheduled tests

## Bug Fixes
- Case 1 : you use `openai/clip-vit-base-patch32`, `openai/clip-vit-base-patch16` or `openai/clip-vit-large-patch14` , before v4.17.0  the good version of the tokenizer was `CLIPTokenizer`. From v4.17.0, you can use both `CLIPTokenizer` and `CLIPTokenizerFast`.
- Fix tests_fetcher
- Fix code format for Accelerate doc
- Fix `bad_words_ids` not working with sentencepiece-based tokenizers
- [docs] fix wrong file name in `pr_check`
- [Fix doc example] FlaxMarianPreTrainedModel

## Upgrade Command
```bash
python -m pip install 'transformers==4.17.0'
```