MIT HAN Lab: Hardware, AI and Neural-nets

Accelerate Deep Learning Computing

Group website: hanlab.mit.edu
TinyML Course: efficientml.ai

MLSys'25 - QServe: W4A8KV4 Quantization and System Co-design for Efficient LLM Serving
13:45
MLSys'25 - QServe: W4A8KV4 Quantization and System Co-design for Efficient LLM Serving
MLSys'25 - LServe: Efficient Long-sequence LLM Serving with Unified Sparse Attention
11:36
MLSys'25 - LServe: Efficient Long-sequence LLM Serving with Unified Sparse Attention
ICLR'25 Oral SANA: Efficient High-Resolution Image Synthesis with Linear Diffusion Transformers
13:08
ICLR'25 Oral SANA: Efficient High-Resolution Image Synthesis with Linear Diffusion Transformers
Nunchaku Windows安装使用教程
10:02
Nunchaku Windows安装使用教程
Nunchaku Installation and Usage Tutorial on Windows
10:41
Nunchaku Installation and Usage Tutorial on Windows
HPCA 2025 EMC2 | Accelerating Large Language Models and Generative AI
42:24
HPCA 2025 EMC2 | Accelerating Large Language Models and Generative AI
HPCA 2025 | LEGO: Spatial Accelerator Generation and Optimization for Tensor Applications
15:51
HPCA 2025 | LEGO: Spatial Accelerator Generation and Optimization for Tensor Applications
SVDQuant NVFP4 Demo
1:11
SVDQuant NVFP4 Demo
SVDQuant Demo
0:38
SVDQuant Demo
EfficientML.ai Lecture 19 - Distributed Training Part 1 (MIT 6.5940, Fall 2024)
1:09:54
EfficientML.ai Lecture 19 - Distributed Training Part 1 (MIT 6.5940, Fall 2024)
TinyML Demo on Microcontrollers (short)
0:13
TinyML Demo on Microcontrollers (short)
EfficientML.ai Lecture 17 - GAN, Video, Point Cloud (Zoom Recording) (MIT 6.5940, Fall 2024)
1:10:19
EfficientML.ai Lecture 17 - GAN, Video, Point Cloud (Zoom Recording) (MIT 6.5940, Fall 2024)
EfficientML.ai Lecture 15 - Long-Context LLM (MIT 6.5940, Fall 2024)
1:13:56
EfficientML.ai Lecture 15 - Long-Context LLM (MIT 6.5940, Fall 2024)
EfficientML.ai Lecture 15 - Long-Context LLM (MIT 6.5940, Fall 2024, Zoom Recording)
1:13:53
EfficientML.ai Lecture 15 - Long-Context LLM (MIT 6.5940, Fall 2024, Zoom Recording)
EfficientML.ai Lecture 14 - LLM Post-Training (MIT 6.5940, Fall 2024)
1:17:58
EfficientML.ai Lecture 14 - LLM Post-Training (MIT 6.5940, Fall 2024)
EfficientML.ai Lecture 14 - LLM Post-Training (MIT 6.5940, Fall 2024, Zoom Recording)
1:18:40
EfficientML.ai Lecture 14 - LLM Post-Training (MIT 6.5940, Fall 2024, Zoom Recording)
EfficientML.ai Lecture 13 - LLM Deployment Techniques (MIT 6.5940, Fall 2024, Zoom Recording)
1:16:43
EfficientML.ai Lecture 13 - LLM Deployment Techniques (MIT 6.5940, Fall 2024, Zoom Recording)
EfficientML.ai Lecture 13 - LLM Deployment Techniques (MIT 6.5940, Fall 2024)
1:16:23
EfficientML.ai Lecture 13 - LLM Deployment Techniques (MIT 6.5940, Fall 2024)
DuoAttention Demo: Running LLMs with 3.3 Million Contextual Tokens on a Single A100 GPU
1:01
DuoAttention Demo: Running LLMs with 3.3 Million Contextual Tokens on a Single A100 GPU
On-device demo for HART: Efficient Visual Generation with Hybrid Autoregressive Transformer
0:55
On-device demo for HART: Efficient Visual Generation with Hybrid Autoregressive Transformer
Image Reconstruction Demo of Deep Compression Autoencoder
0:07
Image Reconstruction Demo of Deep Compression Autoencoder
Efficient Diffusion Models with Deep Compression Autoencoder
0:10
Efficient Diffusion Models with Deep Compression Autoencoder
ISCA 2024 Atomique: A Quantum Compiler for Reconfigurable Neutral Atom Arrays by Hanrui Wang
12:01
ISCA 2024 Atomique: A Quantum Compiler for Reconfigurable Neutral Atom Arrays by Hanrui Wang
MLSys'24 Best Paper - AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration
18:57
MLSys'24 Best Paper - AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration
DistriFusion: Distributed Parallel Inferencefor High-Resolution Diffusion Models
5:00
DistriFusion: Distributed Parallel Inferencefor High-Resolution Diffusion Models
Condition-Aware Neural Network for Controlled Image Generation
5:01
Condition-Aware Neural Network for Controlled Image Generation
GazeSAM Demo: Combining EfficientViT-SAM with Gaze Estimation
0:26
GazeSAM Demo: Combining EfficientViT-SAM with Gaze Estimation
TinyChat Computer running Llama2-7B Jetson Orin Nano. Key technique: AWQ 4bit quantization.
0:51
TinyChat Computer running Llama2-7B Jetson Orin Nano. Key technique: AWQ 4bit quantization.
[Demo] On-Device Training Under 256KB Memory
1:48
[Demo] On-Device Training Under 256KB Memory
Ji Lin's PhD Defense, Efficient Deep Learning Computing: From TinyML to Large Language Model. @MIT
56:18
Ji Lin's PhD Defense, Efficient Deep Learning Computing: From TinyML to Large Language Model. @MIT