Welcome to MIT HAN Lab! We focus on making AI faster, smarter, and more efficient. Our research covers a broad spectrum, including generative AI (e.g., LLMs and diffusion models), TinyML, system optimization and hardware design. By integrating algorithm and hardware expertise, we strive to push the frontiers of AI efficiency and performance.
Graduated PhD students: Ji Lin (OpenAI), Hanrui Wang (assistant professor @UCLA), Zhijian Liu (assistant professor @UCSD), Han Cai (NVIDIA Research), Haotian Tang (Google DeepMind), Yujun Lin (NVIDIA Research).
Accelerating LLM and Generative AI [slides]:
HART has been highlighted by MIT news: AI tool generates high-quality images faster than state-of-the-art approaches!
🔥⚡ We release TinyChat 2.0, the latest version with significant advancements in prefilling speed of Edge LLMs and VLMs, 1.5-1.7x faster than the previous version of TinyChat. Please refer to our blog for more details.
DistriFusion is integrated in NVIDIA's TensorRT-LLM for distributed inference on high-resolution image generation.
🔥 NVIDIA TensorRT-LLM, AMD, Google Vertex AI, Amazon Sagemaker, Intel Neural Compressor, FastChat, vLLM, HuggingFace TGI, and LMDeploy adopt AWQ to improve LLM serving efficiency. Our AWQ models on HuggingFace has received over 6 million downloads.
Congrats on graduation! Cheers on the next move: Zhijian Liu: assistant professor at UCSD, Hanrui Wang: assistant professor at UCLA, Ji Lin: OpenAI, Han Cai: NVIDIA Research, Wei-Chen Wang (postdoc): Amazon, Wei-Ming Chen (postdoc): NVIDIA.
We show SmoothQuant can enable W8A8 quantization for Llama-1/2, Falcon, Mistral, and Mixtral models with negligible loss.
We supported VILA Vision Languague Models in AWQ & TinyChat! Check our latest demos with multi-image inputs!
StreamingLLM is integrated by HPC-AI Tech SwiftInfer to support infinite input length for LLM inference.
StreamingLLM is integrated by CMU, UW, and OctoAI, enabling endless and efficient LLM generation on iPhone!
Congrats Ji Lin completed and defended his PhD thesis: "Efficient Deep Learning Computing: From TinyML to Large Language Model". Ji joined OpenAI after graduation.
AWQ is integrate by NVIDIA TensorRT-LLM, can fit Falcon-180B on a single H200GPU with INT4 AWQ, and 6.7x faster Llama-70B over A100.
🔥 AWQ is now integrated natively in Hugging Face transformers through from_pretrained. You can either load quantized models from the Hub or your own HF quantized models.
Attention Sinks, an library from community enables StreamingLLM on more Huggingface LLMs. blog.
Data-driven, automatic design space exploration of neural accelerator architecture is desirable for specialization and productivity. Previous frameworks focus on sizing the numerical architectural hyper-parameters while neglect searching the PE connectivities and compiler mappings. We push beyond searching only hardware hyper-parameters and propose the Neural Accelerator Architecture Search (NAAS), which fully exploits the hardware design space and compiler mapping strategies at the same time. Unlike prior work which formulate the hardware parameter search as a pure sizing optimization, NAAS models the co-search as a two-level optimization problem, where each level is a combination of indexing, ordering and sizing optimization. To tackle such challenges, we propose an encoding method which is able to encode the non-numerical parameters such as loop order and parallel dimension chosen as numerical parameters for optimization. Thanks to the low search cost, NAAS can be easily integrated with hardware-aware NAS algorithm by adding another optimization level, achieving the joint searching for neural network architecture, accelerator architecture and compiler mapping. Thus NAAS composes highly matched architectures together with efficient mapping. As a data-driven approach, NAAS rivals the human design Eyeriss by 4.4x EDP reduction with 2.7% accuracy improvement on ImageNet under the same computation resource, and offers 1.4x to 3.5x EDP reduction than only sizing the architectural hyper-parameters.
As a data-driven approach, NAAS holistically composes highly matched accelerator and neural architectures together with efficient compiler mapping.
Federated Learning is an emerging direction in distributed machine learning that enables jointly training a model without sharing the data. Since the data is distributed across many edge devices through wireless/long-distance connections, federated learning suffers from inevitable high communication latency. However, the latency issues are undermined in the current literature [15] and existing approaches suchas FedAvg [27] become less efficient when the latency increases. To overcome the problem, we propose Delayed Gradient Averaging (DGA), which delays the averaging step to improve efficiency and allows local computation in parallel to communication. We theoretically prove that DGA attains a similar convergence rate as FedAvg, and empirically show that our algorithm can tolerate high network latency without compromising accuracy. Specifically, we benchmark the training speed on various vision (CIFAR, ImageNet) and language tasks (Shakespeare),with both IID and non-IID partitions, and show DGA can bring 2.55× to 4.07× speedup. Moreover, we built a 16-node Raspberry Pi cluster and show that DGA can consistently speed up real-world federated learning applications
We propose Delayed Gradient Averaging (DGA), which delays the averaging step to improve efficiency and allows local computation in parallel to communication.
Tiny deep learning on microcontroller units (MCUs) is challenging due to the limited memory size. We find that the memory bottleneck is due to the imbalanced memory distribution in convolutional neural network (CNN) designs: the first several blocks have an order of magnitude larger memory usage than the rest of the network. To alleviate this issue, we propose a generic patch-by-patch inference scheduling, which operates only on a small spatial region of the feature map and significantly cuts down the peak memory. However, naive implementation brings overlapping patches and computation overhead. We further propose network redistribution to shift the receptive field and FLOPs to the later stage and reduce the computation overhead. Manually redistributing the receptive field is difficult. We automate the process with neural architecture search to jointly optimize the neural architecture and inference scheduling, leading to MCUNetV2. Patch-based inference effectively reduces the peak memory usage of existing networks by 4-8x. Co-designed with neural networks, MCUNetV2 sets a record ImageNet accuracy on MCU (71.8%) and achieves >90% accuracy on the visual wake words dataset under only 32kB SRAM. MCUNetV2 also unblocks object detection on tiny devices, achieving 16.9% higher mAP on Pascal VOC compared to the state-of-the-art result. Our study largely addressed the memory bottleneck in tinyML and paved the way for various vision applications beyond image classification.
In MCUNetV2, we propose a generic patch-by-patch inference scheduling, which operates only on a small spatial region of the feature map and significantly cuts down the peak memory. We further propose network redistribution to shift the receptive field and FLOPs to the later stage and reduce the computation overhead.
Deep learning on point clouds plays a vital role in a wide range of applications such as autonomous driving and AR/VR. These applications interact with people in real time on edge devices and thus require low latency and low energy. Compared to projecting the point cloud to 2D space, directly processing 3D point cloud yields higher accuracy and lower #MACs. However, the extremely sparse nature of point cloud poses challenges to hardware acceleration. For example, we need to explicitly determine the nonzero outputs and search for the nonzero neighbors (mapping operation), which is unsupported in existing accelerators. Furthermore, explicit gather and scatter of sparse features are required, resulting in large data movement overhead. In this paper, we comprehensively analyze the performance bottleneck of modern point cloud networks on CPU/GPU/TPU. To address the challenges, we then present PointAcc, a novel point cloud deep learning accelerator. PointAcc maps diverse mapping operations onto one versatile ranking-based kernel, streams the sparse computation with configurable caching, and temporally fuses consecutive dense layers to reduce the memory footprint. Evaluated on 8 point cloud models across 4 applications, PointAcc achieves 3.7X speedup and 22X energy savings over RTX 2080Ti GPU. Codesigned with light-weight neural networks, PointAcc rivals the prior accelerator Mesorasi by 100X speedup with 9.1% higher accuracy running segmentation on the S3DIS dataset. PointAcc paves the way for efficient point cloud recognition.
PointAcc is a novel point cloud deep learning accelerator. It introduces a configurable sorting-based mapping unit that efficiently supports diverse operations in point cloud networks. PointAcc further exploits simplified caching and layer fusion specialized for point cloud models, effectively reducing the DRAM access.
We actively collaborate with industry partners on efficient AI, model compression and acceleration. Our research has influenced and landed in many industrial products: Intel OpenVino, Intel Neural Network Distiller, Intel Neural Compressor, Apple Neural Engine, NVIDIA Sparse Tensor Core, NVIDIA TensorRT LLM, AMD-Xilinx Vitis AI, Qualcomm AI Model Efficiency Toolkit (AIMET), Amazon AutoGluon, Facebook PyTorch, Microsoft NNI, SONY Neural Architecture Search Library, SONY Model Compression Toolkit, ADI MAX78000/MAX78002 Model Training and Synthesis Tool.
