Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

transformers

AI overview

Sign in and the AI will write an overview from our coverage.

Headlines · 41

  1. Transformers v5.21.0 Adds OCR, MoE, Vision and TTS Models

    Transformers v5.21.0 adds support for OpenBMB’s MiniCPM-OCR document-understanding model, Recursive’s Flint mixture-of-experts language model, UltraVBert for visual document understanding and retrieval, Gradium Phonon TTS and Kyutai’s Pocket TTS. The release notes describe MiniCPM-OCR as specialized for text-dense images such as scanned pages, screenshots, receipts and tables. Quantization changes address DeepSeek-V4.1’s FP4 experts and prevent TimesFM 2.5 from casting float inputs to integer weight dtypes when loaded with 4-bit or 8-bit quantization. The release also includes bug fixes and infrastructure updates.

    Transformers Releases · 🔥 0
  2. MiniMax releases MiniMax-M2.7 open-weight model

    MiniMax released MiniMax-M2.7, a 229B-parameter text-generation model with open weights. The company says the model supports Agent Teams, complex Skills and dynamic tool search, and can be served with SGLang, vLLM, Transformers, ModelScope or NVIDIA NIM. MiniMax also reports that an internal version updated its memory, built skills for reinforcement-learning experiments and optimized a programming scaffold over more than 100 rounds, improving its performance by 30%. The report provides company-reported benchmark results for software engineering and professional work.

    MiniMax · Hugging Face · 🔥 0
  3. MiniMax Releases Open-Weight MiniMax-M2.1 Model

    MiniMax released MiniMax-M2.1 under a Modified-MIT license, enabling local deployment and use through supported inference frameworks. The model is also available through the MiniMax Open Platform, while MiniMax Code, built on M2.1, is publicly available. MiniMax says the model targets coding, tool use, instruction following, and long-horizon planning, with benchmark improvements based on its own evaluations.

    MiniMax · Hugging Face · 🔥 0
  4. MiniMax releases MiniMax-M2 for coding and agentic workflows

    MiniMax released MiniMax-M2, a mixture-of-experts model with 230 billion total parameters and 10 billion active parameters, and made its weights available for local deployment. The model is designed for coding and agentic workflows. MiniMax recommends SGLang, vLLM, MLX and Transformers for serving it, while explicitly citing day-zero support for SGLang and vLLM. The company also reported benchmark and efficiency results, but these claims remain vendor-reported. The model card lists a modified MIT license, while the release is described as open-source.

    MiniMax · Hugging Face · 🔥 0
  5. Tencent releases WeVisOne-VL-4B unified visual perception model

    Tencent WeChat Vision released WeVisOne-VL-4B, a Qwen3-VL-based vision-language model designed to combine object-level perception with pixel-level tasks. Tencent’s model card lists open-vocabulary object detection, referring localization, object counting, object pointing, OCR localization, referring segmentation, and promptable alpha matting. The release includes an inference package and a required 9.6-million-parameter pixel decoder. Tencent says fine alpha boundaries are sensitive to image resolution, severe occlusion, complex transparency, and prompt phrasing, and that the model currently targets single-turn image perception. Unless otherwise noted, Tencent-authored code is released under Apache License 2.0; third-party components retain their original licenses and copyright notices.

    腾讯混元 · Hugging Face · 🔥 0
  6. Hugging Face launches a Trainium model compatibility lookup

    Hugging Face introduced the torch-neuronx model support Space to check whether popular models on the Hub run unmodified on AWS Trainium. It provides compatibility results, tested Neuron and Transformers versions, execution scripts, and the public aws-neuron/torch-neuronx-model-support dataset. Current tests cover text-generation and image-text-to-text Transformers models that fit on four Trainium2 devices, while a PASS does not establish throughput or longer-generation behavior.

    Hugging Face · 🔥 0
  7. IBM updates Granite Speech with three open-weight models

    IBM released Granite Speech 5.0 2B and revised Granite Speech 5.0 Turbo CTC and Granite Speech 4.1 2B Plus, all under Apache 2.0 with native Transformers support. According to IBM, the new 2B model adds multilingual transcription and translation, language-ID prompting and keyword biasing for up to 2,000 terms, while Turbo CTC gains multilingual transcription and native punctuation; 4.1 2B Plus improves speaker attribution, timestamps and long-speech performance. IBM also reports a 5.06% average English WER for Granite Speech 5.0 2B and 19,870 RTFx for Turbo CTC, though these results are vendor-provided and the former ranking is described as unofficial.

    Hugging Face · 🔥 0
  8. Google and Meta Open-Source TorchTPU for PyTorch on TPUs

    Google and Meta’s PyTorch team released TorchTPU as an open-source native backend for running PyTorch workloads on Google Cloud TPUs. It supports standard PyTorch APIs, eager-mode debugging and torch.compile optimization, with integrations spanning Hugging Face Transformers and Diffusers, vLLM, SGLang, llm-d, TorchTitan and LiteRT. Google says its training setup was validated on up to 8,000 Ironwood TPUs and achieved as much as 52% model FLOPS utilization with Qwen3-8B; these figures are vendor claims.

    Hugging Face · 🔥 0
  9. Databricks cuts Qwen 27B fine-tuning time 6x on AI Runtime

    A Databricks blog details systematic performance work that cut supervised fine-tuning of Qwen 3.8 27B on AI Runtime from 7 hours 44 minutes to 1 hour 17 minutes on a fixed dataset, roughly a 6x speedup. The authors note that 98% GPU utilization only shows the GPU was busy, not how efficiently it was used, so they used torch traces to find where time was actually spent. Optimizations included enabling the flash-linear-attention and causal-conv1d fused kernels and reworking batch packing, which alone gave a 1.6x improvement from 2 hours 5 minutes to 1 hour 17 minutes. Example code is published in the AI Runtime cookbook.

    Databricks Blog · 🔥 0
  10. Baidu Releases 3B Unlimited-OCR for Document Parsing

    Baidu released the open-weight, MIT-licensed Unlimited-OCR model, a 3B image-text-to-text system designed for one-shot, long-horizon document parsing across images, multi-page images and PDFs. Its model card lists Transformers inference, vLLM and SGLang deployment, ms-swift training, availability through Baidu Cloud and ModelScope, and a demo on Hugging Face Spaces. Baidu said the model aims to push DeepSeek-OCR further, but the supplied reports do not include independent comparative evaluations.

    百度文心 · Hugging Face · 🔥 0
  11. Hugging Face Transformers tops GitHub Trending

    Hugging Face's Transformers library appeared on GitHub Trending's daily list. The project positions itself as the model-definition framework for state-of-the-art text, vision, audio, video and multimodal models, for both inference and training, and its documentation says there are over 1M Transformers model checkpoints on the Hugging Face Hub. It says the library interoperates with training frameworks such as Axolotl, Unsloth, DeepSpeed, FSDP and PyTorch-Lightning, and with inference engines and modeling libraries including vLLM, SGLang, TGI, llama.cpp and mlx, and requires Python 3.10+ and PyTorch 2.6+.

    GitHub Trending(每日) · 🔥 0

← Previous

Experience and discussion from the community

Share my transformers experienceAsk about transformers

minimind: An open-source project for training a small language model from scratch and understanding its complete training pipeline.

What problem it solves Large language models can be difficult for individuals to train, deploy, or understand because of their scale and the abstractions used by many frameworks. The project addresses this by providing native PyTorch implementations, compact model structures, open datasets, and staged workflows for pretraining, fine-tuning, preference optimization, tool use, and agentic training. …

github.comhttps://github.com/jingyaogong/minimind

Hy-MT2: A family of fast-thinking multilingual translation models for complex real-world translation scenarios.

What problem it solves Real-world translation involves more than converting text between languages: tasks can require terminology, style, personalization, delimiter preservation, and structured-data constraints. Hy-MT2 is designed to support translation among 33 languages and follow translation instructions in multiple languages for complex real-world scenarios. Key features Instruction-driven tas…

github.comhttps://github.com/Tencent-Hunyuan/Hy-MT2

transformers: A model-definition framework for machine learning models across text, computer vision, audio, video, and multimodal tasks, supporting both inference and training.

What problem it solves Machine learning workflows need a shared model definition so that models can work consistently across training frameworks, inference engines, and adjacent modeling libraries. Transformers centralizes these definitions and exposes a unified interface for loading pretrained models, running inference, and adapting model internals. Key features Broad Model Support — The framewor…

github.comhttps://github.com/huggingface/transformers

timesfm: A pretrained time-series foundation model for time-series forecasting.

What problem it solves Time-series forecasting requires models that can forecast univariate and multivariate series across different context lengths. Forecasters also need to incorporate optional covariates and produce quantile forecasts without requiring per-task tuning. Key features Univariate and multivariate forecasting — Forecast individual time series or multiple target variates from NumPy a…

github.comhttps://github.com/google-research/timesfm

Model-Optimizer: A library for composing model optimization techniques and exporting optimized quantized checkpoints.

What problem it solves Deep learning models often need to be adapted for efficient deployment across different inference frameworks. Model Optimizer addresses this by providing a unified set of optimization techniques and export workflows for quantized checkpoints. Key features Post-training quantization — Compress models with quantization techniques available for Hugging Face, Megatron-Bridge, Di…

github.comhttps://github.com/NVIDIA/Model-Optimizer

annotated_deep_learning_paper_implementations: Annotated PyTorch implementations and tutorials for understanding deep learning algorithms.

What problem it solves This project provides simple PyTorch implementations of neural networks and related algorithms. These implementations are documented with explanations and rendered as side-by-side formatted notes to help readers understand the algorithms better. Key features PyTorch Implementations — The project provides implementations of neural networks and related algorithms using PyTorch…

github.comhttps://github.com/labmlai/annotated_deep_learning_paper_implementations