Update the news for TorchSparse (#212)

* [Major] Update README for v2.1.0

* [Minor] Update README.md

* [Minor] Update README.md

* [Minor] Update README.md

* [Minor] Add installation.

* [Minor] Fix.

* [Minor] Update README.md

---------

Co-authored-by: ys-2020 <shangy@mit.edu>
Co-authored-by: Haotian Tang <haotian@k5.mit.edu>

README.md

@@ -9,6 +9,16 @@ TorchSparse is a high-performance neural network library for point cloud process
 
 Point cloud computation has become an increasingly more important workload for autonomous driving and other applications. Unlike dense 2D computation, point cloud convolution has **sparse** and **irregular** computation patterns and thus requires dedicated inference system support with specialized high-performance kernels. While existing point cloud deep learning libraries have developed different dataflows for convolution on point clouds, they assume a single dataflow throughout the execution of the entire model. In this work, we systematically analyze and improve existing dataflows. Our resulting system, TorchSparse, achieves **2.9x**, **3.3x**, **2.2x** and **1.7x** measured end-to-end speedup on an NVIDIA A100 GPU over the state-of-the-art MinkowskiEngine, SpConv 1.2, TorchSparse (MLSys) and SpConv v2 in inference respectively. 
 
+## News
+
+**\[2023/6/18\]** TorchSparse++ has been released and presented at CVPR 2023 workshops on autonomous driving. It achieves 1.7-2.9x inference speedup over previous state-of-the-art systems.
+
+**\[2022/8/29\]** TorchSparse is presented at MLSys 2022. Talk video is available [here](https://www.youtube.com/watch?v=IIh4EwmcLUs).
+
+**\[2022/1/15\]** TorchSparse has been accepted to MLSys 2022, featuring adaptive matrix multiplication grouping and locality-aware memory access.
+
+**\[2021/6/24\]** TorchSparse v1.4 has been released.
+
 ## Installation
 
 We provide pre-built torchsparse v2.1.0 packages (recommended) with different PyTorch and CUDA versions to simplify the building for the Linux system.
@@ -49,7 +59,7 @@ TorchSparse-MLsys on cloud GPUs. It also improves the latency of SpConv 2.3.5 by
 
 ![train_benchmark.png](./docs/figs/train_benchmark.png)
 
-TorchSparse achieves superior mixed-precision training speed compared with MinkowskiEngine, TorchSparse-MLSys and SpConv 2.3.5. Specifically, it is **1.16x** faster on Tesla A100, **1.27x** faster on RTX 2080 Ti than state-of-the-art SpConv 2.3.5. It also significantly outperforms MinkowskiEngine by **4.6-4.8x*** across seven benchmarks on A100 and 2080 Ti. Measured with batch size = 2.
+TorchSparse achieves superior mixed-precision training speed compared with MinkowskiEngine, TorchSparse-MLSys and SpConv 2.3.5. Specifically, it is **1.16x** faster on Tesla A100, **1.27x** faster on RTX 2080 Ti than state-of-the-art SpConv 2.3.5. It also significantly outperforms MinkowskiEngine by **4.6-4.8x** across seven benchmarks on A100 and 2080 Ti. Measured with batch size = 2.
 
 
 ## Team