返回全部动态

Transformers v5.21.0 发布:新增多款模型支持

原标题:Release: v5.21.0

Transformers Releases一手来源开源质量 76

AI 摘要

Hugging Face Transformers 发布 v5.21.0,新增 MiniCPM-OCR、Flint、UltraVBert、Gradium Phonon TTS、Pocket TTS 及其音频分词器等多个模型支持。其中 MiniCPM-OCR 面向文档理解,Flint 为 Recursive 的混合注意力 MoE 语言模型,Pocket TTS 为 Kyutai 的 100M 参数端侧语音合成模型。该版本还修复了 Gemma4 处理器导入问题,并为 QuantizedCache 新增 fp8 后端以存储 KV 状态。

以上摘要由 AI 生成,可能存在误差。事实请以原文为准。

正文节选

# Release v5.21.0 ## New Model additions ### MiniCPM-OCR MiniCPM-OCR is a document-understanding model from OpenBMB's MiniCPM-V family. It keeps the MiniCPM-V 4.6 architecture, which pairs a SigLIP vision encoder and a window-attention merger with a Qwen3.5 language model. It is specialized for reading text-dense images such as scanned pages, screenshots, receipts and tables. **Links:** [Documentation](https://huggingface.co/docs/transformers/main/en/model_doc/minicpm_ocr) | [Paper


发布时间:2026-10-27 02:17
抓取时间:2026-10-27 02:29
来源机构:Hugging Face
阅读原文github.com