Point cloud computation is important for AR/VR and ADAS. It involves sparse and irregular computation patterns, requiring specialized high-performance kernels. Existing GPU libraries offer two dataflow types for sparse point cloud convolution. The gather-GEMM-scatter dataflow is easy to implement but not optimal in performance, while the dataflows with overlapped computation and memory access (e.g. implicit GEMM) are highly performant but have very high engineering costs. In this work we introduce TorchSparse++, a new GPU library that achieves the best of both worlds. We create a highly efficient Sparse Kernel Generator that generates performant sparse point cloud convolution kernels at less than one-tenth of the engineering cost of the current state-of-the-art system. On top of this, we design the Sparse Autotuner, which extends the design space of existing point cloud libraries and searches for the best dataflow configurations for training and inference workloads. Consequently, TorchSparse++ achieves 2.9x, 3.3x, 2.2x and 1.7x measured end-to-end speedup on an NVIDIA A100 GPU over state-of-the-art MinkowskiEngine, SpConv 1.2, TorchSparse and SpConv v2 in inference; and is 1.2-1.3x faster than SpConv v2 in mixed precision training.
TorchSparse++ is a high-performance computing library for efficient 3D sparse convolution. It offers significant performance improvement over TorchSparse++ by overlapping computation with memory access. It also searches for the best execution strategy for sparse workloads within a large design space through auto-tuning.
On-device learning and efficient fine-tuning enable continuous and privacy-preserving customization (e.g., locally fine-tuning large language models on personalized data). However, existing training frameworks are designed for cloud servers with powerful accelerators (e.g., GPUs, TPUs) and lack the optimizations for learning on the edge, which faces challenges of resource limitations and edge hardware diversity. In this paper, we introduce PockEngine: a tiny, sparse and efficient engine to enable fine-tuning on various edge devices. PockEngine supports sparse backpropagation: it prunes the backward graph and sparsely updates the model with measured memory saving and latency reduction while maintaining the model quality. Secondly, PockEngine is compilation first: the entire training graph (including forward, backward and optimization steps) is derived at compile-time, which reduces the runtime overhead and brings opportunities for graph transformations. PockEngine also integrates a rich set of training graph optimizations, thus can further accelerate the training cost, including operator reordering and backend switching. PockEngine supports diverse applications, frontends and hardware backends: it flexibly compiles and tunes models defined in PyTorch/TensorFlow/Jax and deploys binaries to mobile CPU/GPU/DSPs. We evaluated PockEngine on both vision models and large language models. PockEngine achieves up to 15x speedup over off-the-shelf TensorFlow (Raspberry Pi), 5.6x memory saving back-propagation (Jetson Orin). Remarkably, PockEngine enables fine-tuning LLaMA2-7B on NVIDIA Jetson Orin at 550 tokens/s, 7.9x faster than the PyTorch. PockEngine feautres Sparse backpropagation
This project introduce PockEngine: a tiny, sparse and efficient engine to enable fine-tuning on various edge devices. PockEngine supports sparse backpropagation: it prunes the backward graph and sparsely updates the model with measured memory saving and latency reduction while maintaining the model quality.
High-resolution dense prediction enables many appealing real-world applications, such as computational photography, autonomous driving, etc. However, the vast computational cost makes deploying state-of-the-art high-resolution dense prediction models on hardware devices difficult. This work presents EfficientViT, a new family of high-resolution vision models with novel lightweight multi-scale attention. Unlike prior high-resolution dense prediction models that rely on heavy self-attention, hardware-inefficient large-kernel convolution, or complicated topology structure to obtain good performances, our lightweight multi-scale attention achieves a global receptive field and multi-scale learning (two critical features for high-resolution dense prediction) with only lightweight and hardware-efficient operations. As such, EfficientViT delivers remarkable performance gains over previous state-of-the-art high-resolution dense prediction models with significant speedup on diverse hardware platforms, including mobile CPU, edge GPU, and cloud GPU. Without performance loss on Cityscapes, our EfficientViT provides up to 8.8x and 3.8x GPU latency reduction over SegFormer and SegNeXt, respectively. For super-resolution, EfficientViT provides up to 6.4x speedup over Restormer while providing 0.11dB gain in PSNR.
EfficientViT is a new family of vision models for high-resolution dense prediction. It achieves global receptive field and multi-scale learning with only hardware-efficient operations. EfficientViT delivers remarkable performance gains over previous models with speedup on diverse hardware platforms, including mobile CPU, edge GPU, and cloud GPU.
Deploying Large Language Models (LLMs) in streaming applications such as multi-round dialogue, where long interactions are expected, is urgently needed but poses two major challenges. Firstly, during the decoding stage, caching previous tokens' Key and Value states (KV) consumes extensive memory. Secondly, popular LLMs cannot generalize to longer texts than the training sequence length. Window attention, where only the most recent KVs are cached, is a natural approach --- but we show that it fails when the text length surpasses the cache size. We observe an interesting phenomenon, namely attention sink, that keeping the KV of initial tokens will largely recover the performance of window attention. In this paper, we first demonstrate that the emergence of attention sink is due to the strong attention scores towards initial tokens as a ``sink'' even if they are not semantically important. Based on the above analysis, we introduce StreamingLLM, an efficient framework that enables LLMs trained with a finite length attention window to generalize to infinite sequence length without any fine-tuning. We show that StreamingLLM can enable Llama-2, MPT, Falcon, and Pythia to perform stable and efficient language modeling with up to 4 million tokens and more. In addition, we discover that adding a placeholder token as a dedicated attention sink during pre-training can further improve streaming deployment. In streaming settings, StreamingLLM outperforms the sliding window recomputation baseline by up to 22.2x speedup.
We enable LLMs to work on infinite-length texts without compromising efficiency and performance.
