腾讯混元开源的大规模视频生成框架，基于扩散 Transformer 把文本提示直接生成连贯视频，适合做文生视频研究和应用落地。
用 LoRA、Adapter 等少量参数微调大模型和扩散模型，大幅降低显存与训练成本，是 Hugging Face 生态里做参数高效微调的标准库。
把训练好的深度学习模型优化并部署到 Intel CPU、GPU 等硬件上做推理，支持视觉、LLM 和生成式模型，省去自己折腾底层加速的麻烦。
高性能大模型与多模态模型推理服务框架，通过优化注意力机制和调度策略提升吞吐与延迟表现，支持 Llama、DeepSeek、MoE 等主流模型部署。
LTX-Video 是 Lightricks 开源的视频生成模型，支持文生视频和图生视频，基于 DiT 扩散架构，可快速生成高质量短视频。
… diffusion LoRA checkpoint fusion into standalone transformer/pipeline artifacts for adapter-free downstream inference. ( 581, 676, 681) - Added merged LoRA weight sync su
…-Flash ( 4077 by @ved1beta). Dependency Updates - transformers 5.17.0 → 5.20.1 , peft 0.21.0 → 0.21.2 , tokenizers 0.23.1 → 0.23.3 , and kernels 0.16 → =0.17.2 ( 4097, 415
…xt trajectory, Qwen3-Omni audio, and WAN 2.1 full-transformer SFT ( 234, 352, 345, 320). - Agentic multi-turn rollout and training on Sample lineages ( 214, 315, 363). - D
…y . Packaging - Relaxed the pymeshlab pin; capped transformers<5 and huggingface hub<1.0 (transformers 5.x removed CLIPFeatureExtractor ); moved preprocessing-only deps (
…al AI — a single omni-model built on a Mixture-of-Transformers (MoT) architecture that combines world generation, physical reasoning, and action generation, replacing the
… ( 8473) Updated supported version of Huggingface Transformers ( 8574) New Contributors @kitsiosk made their first contribution in https://github.com/Project-MONAI/MONAI/p