Shipped · AI and LLM tooling

What huggingface/transformers shipped

The public repository of Hugging Face · github.com/huggingface/transformers

Written by FoxPlug from the repository's own releases; not affiliated with Hugging Face. Paste your own repository in the box below and FoxPlug writes the same weekly update for your product. This page is an automatic summary of the releases, pull requests and commits in github.com/huggingface/transformers; Hugging Face did not write it and does not use or endorse FoxPlug. Every line links to the change it describes.

Get a weekly update like this for your product, free

Or Use it as a GitHub Action

Follow transformers's weekly shipped digest

Week of October 5, 2026

What shipped

Why it matters

This week brought new model support with Molmo2 and PP-DocLayoutV4, performance improvements through torchcodec image decoding and optimized kernel selection, and infrastructure enhancements like SubConfigSpec for better configuration management. Numerous bug fixes address generation pipelines, cache handling, and model-specific issues, while CI improvements and torch version bumps maintain code quality and compatibility.

Changelog entry

Example posts FoxPlug drafted from these changes. Not written or posted by the project.

Post for X

New week, new models: Molmo2 and PP-DocLayoutV4 joined the repo. We improved image decoding with torchcodec, optimized inference kernels, and fixed critical generation and cache bugs.

Post for LinkedIn

Transformers shipped big updates this week: Molmo2 and PP-DocLayoutV4 models are now available, image decoding improved through torchcodec integration, and we optimized inference kernels for CUDA. Infrastructure wise, SubConfigSpec centralizes subconfig initialization. Multiple fixes target generation pipelines, quantized caches, and model-specific issues. Torch version bumped for better hardware support.

Week of September 28, 2026

What shipped

Why it matters

Version 5.18.0 brings new model support and infrastructure improvements for distributed training and attention mechanisms. The addition of per-layer cache configuration and continuous batching support improves inference flexibility and performance. Bug fixes in mixture-of-experts models and distributed training address issues for users deploying these configurations.

Changelog entry

Example posts FoxPlug drafted from these changes. Not written or posted by the project.

Post for X

Transformers 5.18.0 is out! New Nemotron 3 Diarization model, GTE support, per-layer cache config, and improved continuous batching. Plus fixes for MoE and distributed training setups.

Week of September 21, 2026

What shipped

Why it matters

This week includes two new model architectures (Nemotron Omni and Nemotron3Diarization), expanded deployment options through MLX recipes and GGUF support in Chat CLI, and fixes for critical issues across attention mechanisms, configuration handling, and compatibility layers. New trainer functionality and improved device handling strengthen the core training pipeline.

Changelog entry

Example posts FoxPlug drafted from these changes. Not written or posted by the project.

Post for X

This week: Nemotron Omni & Nemotron3Diarization support, MLX & GGUF integration, Trainer.end method, PEFT tensor parallelism, plus fixes for attention, configs & compatibility [20,26,10,12,36,33]

Post for LinkedIn

Transformers now supports Nemotron Omni and Nemotron3Diarization architectures, expands deployment with MLX recipes and GGUF loading in Chat CLI, introduces Trainer.end for better training control, and ships PEFT integration with Dtensor-based tensor parallelism. This week also includes critical fixes for attention mechanisms, configuration handling, and cross-version compatibility to strengthen both inference and training workflows.

Weeks with too little public activity are left out rather than filled in. Last updated 2026-10-12.

Is this your repo? Ask us to remove this page.