Tianchen Zhao

Tianchen Zhao 赵天辰 (Ziu Tinsan)

Phd. Student at Tsinghua University

Tsinghua University

Biography

Tianchen Zhao is a Phd. student in NICS-EFC Lab at Dept. EE, Tsinghua University, supervised by Prof. Yu Wang and Dr. Xuefei Ning. He got his bachelor and master degree in the Dept. EE Beihang University in 2020 and 2023. His primary research focus is efficient deep learning and software-hardware co-optimization.

News

Interests
  • Efficient Deep Learning
  • Computer Vision
Education

  • PhD in Electrical Engineering, 2023

    Tsinghua University

  • MS in Electrical Engineering, 2020

    Beihang University

  • BSc in Electrical Engineering, 2016

    Beihang University

Publications

[Arxiv] ViDiT-Q: Efficient and Accurate Quantization of Diffusion Transformers for Image and Video Generation
We introduce ViDiT-Q, a quantization method specialized for diffusion transformers. For popular large-scale models (e.g., open-sora, Latte, Pixart-α, Pixart-Σ) for the video and image generation task, ViDiT-Q could achieve W8A8 quantization without metric degradation, and W4A8 without notable visual quality degradation.
Tianchen Zhao, Tongcheng Fang, Enshu Liu, Rui Wan, Widyadewi Soedarmadji, Shiyao Li, Zinan Lin, Guohao Dai, Shengen Yan, Huazhong Yang, Xuefei Ning, Yu Wang
Cite Code Dataset Project Slides Video Source Document
[Arxiv] ViDiT-Q: Efficient and Accurate Quantization of Diffusion Transformers for Image and Video Generation
[ECCV24] MixDQ: Memory-Efficient Few-Step Text-to-Image Diffusion Models with Metric-Decoupled Mixed Precision Quantization
We design MixDQ, a mixed-precision quantization framework that successfully tackles the challenging few-step text-to-image diffusion model quantization. With negligible visual quality degradation and content change, MixDQ could achieve W4A8, with equivalent 3.4x memory compression and 1.5x latency speedup.
Tianchen Zhao, Xuefei Ning, Tongcheng Fang, Enshu Liu, Guyue Huang, Zinan Lin, Shengen Yan, Guohao Dai, Yu Wang
Cite Code Dataset Project Slides Video Source Document
[ECCV24] MixDQ: Memory-Efficient Few-Step Text-to-Image Diffusion Models with Metric-Decoupled Mixed Precision Quantization
[CVPR 24] FlashEval: Towards Fast and Accurate Evaluation of Text-to-image Diffusion Generative Models
We propose an efficient diffusion model evaluation method that condense the textual evaluation dataset, achieving evaluation quality with 5x larger data size.
Lin Zhao, Tianchen Zhao, Zinan Lin, Xuefei Ning, Guohao Dai, Huazhong Yang, Yu Wang
Cite Code Dataset Project Slides Video Source Document
[CVPR 24] FlashEval: Towards Fast and Accurate Evaluation of Text-to-image Diffusion Generative Models
[ICCV 23] Ada3D: Exploiting the Spatial Redundancy with Adaptive Inference for Efficient 3D Object Detection
We propose an adaptive inference method for 3D perception that could reduce 60% 3D voxels, 80% 2D pixels, achieving 5x FLOPs/Memory and 1.4x Latency.
Tianchen Zhao, Xuefei Ning, Ke Hong, Zhongyuan Qiu, Pu Lu, Linfeng Zhang, Yali Zhao, Lipu Zhou, Guohao Dai, Huazhong Yang, Yu Wang
Cite Code Dataset Project Slides Video Source Document
[ICCV 23] Ada3D: Exploiting the Spatial Redundancy with Adaptive Inference for Efficient 3D Object Detection
[CVPR 22] CodedVTR: Codebook-based Sparse Voxel Transformer with Geometric Guidance
We propose CodedVTR (Codebook-based Voxel TRansformer), which improves data efficiency and generalization ability for 3D sparse voxel transformers.
Tianchen Zhao, Niansong Zhang, Xuefei Ning, He Wang, Li Yi, Yu Wang
Cite Code Dataset Project Slides Video Source Document
[CVPR 22] CodedVTR: Codebook-based Sparse Voxel Transformer with Geometric Guidance
See all publications

Recent Events

Visit at KIT
Visit the ITIV Lab at KIT(Karlsruhe Institute of Technology).
Oct 10, 2023 10:00 AM
Project Slides
Visit at KIT
TechBeat Talk
Shared our ICCV23’ work Ada3D at TechBeat.
Sep 21, 2023 8:00 PM
Project Slides
TechBeat Talk

Gallery

2023_10_iccv_1.jpg
2023_10_iccv_2.jpg
2023_10_kit.jpg
2023_11_expo.jpg
2024_07_cvpr_1.jpg
2024_07_cvpr_2.jpg
2024_07_cvpr_3.jpg