models t2s_lightning_module_onnx.py t2s_model_onnx.py modules activation_onnx.py embedding_onnx.py patched_mha_with_cache_onnx.py transformer_onnx.py 106 changes: 106 additions & 0 deletions 106 GPT_SoVITS/
- 方法:T2V模型,时间信息提取,缩放模块,时间集成机制- 效果:效率高,可控性强 更新 Uni\textbf{F}^2ace: Fine-grained Face Understanding and Generation with Unified Multimodal Models 统一多模态模型进行细粒度人脸理解和生成:UniFace Junzhe Li, Xuerui Qiu, Linrui Xu, Liya Guo, Delin Qu, Tingting ...
Generation: Taming Optimization Dilemma in Latent Diffusion Models 重建与生成:解决潜在扩散模型中的优化困境 Jingfeng Yao, Bin Yang, Xinggang Wang arxiv.org/pdf/2501.0142 [代码]- 问题:优化困境,重建与生成,信息损失,计算成本- 方法:VA-VAE,预训练模型,LightningDiT- 效果:SOTA性能,FID 1.35,收敛速度提升 ...
models t2s_lightning_module_onnx.py t2s_model_onnx.py modules activation_onnx.py embedding_onnx.py patched_mha_with_cache_onnx.py transformer_onnx.py 106 changes: 106 additions & 0 deletions 106 GPT_SoVITS/AR/models/t2s_lightning_module_onnx.py Original file line numberDiff ...
2024-12-07 T2I-FactualBench: Benchmarking the Factuality of Text-to-Image Models with Knowledge-Intensive Concepts T2I-FactualBench:基于知识密集型概念的文本到图像模型事实性基准测试 Ziwei Huang, Wanggui He, Quanyu Long, Yandi Wang, Haoyuan Li, Zhelun Yu, Fangxun Shu, Long Chan arxiv.org/pd...