Chen Zhao
  • Bio
  • News
  • Publications
  • Talks
  • Teaching
  • Awards
ESC

Searching...

No results found

↑↓ Navigate ↵ Select
Powered by Hugo Blox
  • Talks
    • Invertible Diffusion Models for Inverse Problems
    • Video Understanding for Embodied AI
    • Long-form Video Understanding in the 2020s
    • 神经网络可逆化：大模型微调的高效显存优化策略
    • Long-form Video Understanding in the 2020s
    • Reversifying Neural Networks: Efficient Memory Optimization Strategies for Finetuning Large Models
    • Towards More Realistic Continual Learning at Scale
    • Optimizing Memory Efficiency in Pretrained Model Finetuning
    • Optimizing Memory Efficiency in Pretrained Model Finetuning
    • Toward Long-form Video Understanding
    • Challenges and Advances in Long-form Video Understanding
    • Towards Long-form Video Understanding
    • Challenges and Innovation for Long-form Video Understanding: Compute, Algorithm, and Data
    • Research Highlights at IVUL with a Focus on Video Understanding
    • Towards Long-form Video Understanding
    • A Simple and Effective Approach for Long-form Video Understanding
    • Detecting Actions in Videos via Graph Convolutional Networks
    • Let the Computer See the World as Humans
    • Image/Video Cloud Coding
    • Making a Lighter Encoder: Image/Video Compressive Sensing
    • Compressive Sensing-based Image/Video Coding
  • Projects
    • Pandas
    • PyTorch
    • scikit-learn
  • Awards
    • First place, Ego4D Visual Queries 3D
    • First place, Epic-Kitchens action detection
    • First place, Epic-Kitchens action recognition
    • First place, Epic-Kitchens audio-based interaction detection
    • Recipient of Grant
    • Best Paper Award
    • First place, Visual Queries 3D Localization Challenge in Ego4D Workshop
    • First place, Visual Queries 3D Localization Challenge in Ego4D Workshop
    • Outstanding Reviewer
    • Finalist
    • Second place, HACS Temporal Action Localization Challenge
    • Finalist
    • Outstanding Graduate
    • Scholarship of Outstanding Talent
    • Best Paper Award
    • First Prize, Qualcomm Innovation Fellowship Contest (QInF)
    • Outstanding Individual in the Summer Social Practice
    • Outstanding Graduate Leader
    • Goldman Sachs Global Leaders Award
    • First-Class Scholarship
    • National Scholarship
  • Slides
    • Example Talk: Recent Work
  • Experience
  • Courses
    • Hugo Blox
      • Getting Started
      • Guide
        • Project Structure
        • Configuration
        • Formatting
          • Embed Media
          • Buttons
          • Callouts
          • Cards
          • Spoilers
          • Steps
      • Reference
        • Customization
        • Internationalization (i18n)
  • News
  • Publications
    • BOLT: Boost Large Vision-Language Model Without Training for Long-form Video Understanding
    • SMILE: Infusing Spatial and Motion Semantics in Masked Video Learning
    • OSMamba: Omnidirectional Spectral Mamba with Dual-Domain Prior Generator for Exposure Correction
    • Invertible Diffusion Models for Compressed Sensing
    • SEVERE++: Evaluating Benchmark Sensitivity in Generalization of Video Representation Learning
    • Effectiveness of Max-Pooling for Fine-Tuning CLIP on Videos
    • OpenTAD: A Unified Framework and Comprehensive Study of Temporal Action Detection
    • Ego4D: Around the World in 3,000 Hours of Egocentric Video
    • Towards Automated Movie Trailer Generation
    • Dr<sup>2</sup>Net: Dynamic Reversible Dual-Residual Networks for Memory-Efficient Finetuning
    • Ego-Exo4D: Understanding Skilled Human Activity from First-and Third-Person Perspectives
    • End-to-End Temporal Action Detection with 1B Parameters Across 1000 Frames
    • Re<sup>2</sup>TAL: Rewiring Pretrained Video Backbones for Reversible Temporal Action Localization
    • EgoLoc: Revisiting 3D Object Localization from Egocentric Videos with Visual Queries
    • FreeDoM: Training-Free Energy-Guided Conditional Diffusion Model
    • A Unified Continual Learning Framework with General Parameter-Efficient Tuning
    • Large-capacity and Flexible Video Steganography via Invertible Neural Network
    • ETAD: Training Action Detection End to End on a Laptop
    • Just a Glimpse: Rethinking Temporal Information for Video Continual Learning
    • Owl (observe, watch, listen): Localizing actions in egocentric video via audiovisual temporal context
    • R-DFCIL: Relation-Guided Representation Learning for Data-Free Class Incremental Learning
    • End-to-End Active Speaker Detection
    • Evaluation of Diverse Convolutional Neural Networks and Training Strategies for Wheat Leaf Disease Identification with Field-Acquired Photographs
    • When NAS Meets Trees: An Efficient Algorithm for Neural Architecture Search
    • MAD: A Scalable Dataset for Language Grounding in Videos from Movie Audio Descriptions
    • Ego4D: Around the World in 3,000 Hours of Egocentric Video
    • SegTAD: Precise Temporal Action Detection via Semantic Segmentation
    • Video Self‑Stitching Graph Network for Temporal Action Localization
    • ThumbNet: One Thumbnail Image Contains All You Need for Recognition
    • Improve Baseline for Temporal Action Detection: HACS Challenge 2020 Solution of IVUL‑KAUST team
    • G‑TAD: Sub‑Graph Localization for Temporal Action Detection
    • Optimization‑Inspired Compact Deep Compressive Sensing
    • Logistic Regression is Still Alive and Effective: The 3rd YouTube 8M Challenge Solution of the IVUL‑KAUST team
    • CREAM: CNN-REgularized ADMM framework for compressive-sensed image reconstruction
    • BoostNet: A Structured Deep Recursive Network to Boost Image Deblocking
    • Better and Faster, when ADMM Meets CNN: Compressive-sensed Image Reconstruction
    • Reducing Image Compression Artifacts by Structural Sparse Representation and Quantization Constraint Prior
    • Video Compressive Sensing Reconstruction via Reweighted Residual Sparsity
    • CONCOLOR: COnstrained Non-Convex Low-Rank Model for Image Deblocking
    • Nonconvex Lp Nuclear Norm based ADMM Framework for Compressive Sensing
    • Compressive-Sensed Image Coding via Stripe-based DPCM
    • A Dual Structured-Sparsity Model for Compressive-Sensed Video Reconstruction
    • 基于云数据的高效图像编码方法
    • Thousand to one: An image compression system via cloud search
    • Adaptive intra-refresh for low-delay error-resilient video coding
    • Video Compressive Sensing via Structured Laplacian Modelling
    • Image Compressive-Sensing Recovery Using Structured Laplacian Sparsity in DCT Domain and Multi-Hypothesis Prediction
    • Image Compressive Sensing Recovery Using Adaptively Learned Sparsifying Basis via L0 Minimization
    • Weakly Supervised Photo Cropping
    • Wavelet Inpainting Driven Image Compression via Collaborative Sparsity at Low Bit Rates
    • A Highly Effective Error Concealment Method for Whole Frame Loss
  • Recent & Upcoming Talks
    • Create Beautiful Presentations with Markdown
  • Teaching
    • Machine Learning (CS229), KAUST
    • Deep Learning for Visual Computing (CS323), KAUST
    • Video understanding, KAUST
    • Machine learning, KAUST
    • Deep learning, KAUST
    • Deep learning, KAUST
    • Deep learning, KAUST
  • Thanks for reaching out

Optimizing Memory Efficiency in Pretrained Model Finetuning

June 11, 2024 · 1 min read
talks

Talk, Berkeley Artificial Intelligence Research Lab (BAIR), UC Berkeley, US

Last updated on January 9, 2026
Chen Zhao
Authors
Chen Zhao
Research Scientist
← Optimizing Memory Efficiency in Pretrained Model Finetuning June 14, 2024
Toward Long-form Video Understanding May 5, 2024 →

© All rights reserved by Chen Zhao, 2026