Python库 | kp-0.6.0.tar.gz

     <table> <tr> <td width="20%"> <img src="https://raw.githubusercontent.com/EthicalML/vulkan-kompute/master/docs/images/kompute.jpg"> </td> <td> <h1>Vulkan Kompute</h1> <h3>The general purpose GPU compute framework for cross vendor graphics cards (AMD, Qualcomm, NVIDIA & friends).</h3> </td> </tr> </table> <h4>Blazing fast, mobile-enabled, asynchronous, and optimized for advanced GPU processing usecases.</h4> 🔋 [Documentation](https://kompute.cc) 💻 [Blog Post](https://medium.com/@AxSaucedo/machine-learning-and-data-processing-in-the-gpu-with-vulkan-kompute-c9350e5e5d3a) ⌨ [Examples](#more-examples) 💾 ## Principles & Features * Low level extensible [C++ SDK](#your-first-kompute-c) with high level optimized [Python module](#your-first-kompute-python) * [Asynchronous & parallel processing](#asynchronous-and-parallel-operations) support through GPU family queues * [Mobile enabled](#mobile-enabled) with examples via Android NDK across several architectures * BYOV: [Bring-your-own-Vulkan design](#motivations) to play nice with existing Vulkan applications * Explicit relationships for GPU and host [memory ownership and memory management](https://kompute.cc/overview/memory-management.html) * [Hands on examples](#simple-examples) showing the core features * Longer tutorials for [machine learning 🤖](https://towardsdatascience.com/machine-learning-and-data-processing-in-the-gpu-with-vulkan-kompute-c9350e5e5d3a), [mobile development 📱](https://towardsdatascience.com/gpu-accelerated-machine-learning-in-your-mobile-applications-using-the-android-ndk-vulkan-kompute-1e9da37b7617) and [game development 🎮](https://towardsdatascience.com/supercharging-game-development-with-gpu-accelerated-ml-using-vulkan-kompute-the-godot-game-engine-4e75a84ea9f0).  ## Getting Started Below you can find a GPU multiplication example using the C++ and Python Kompute interfaces. ### Your First Kompute (C++) The C++ interface provides low level access to the native components of Kompute and Vulkan, enabling for [advanced optimizations](https://kompute.cc/overview/async-parallel.html) as well as [extension of components](https://kompute.cc/overview/reference.html). ```c++ int main() { // 1. Create Kompute Manager with default settings (device 0 and first compute compatible queue) kp::Manager mgr; // 2. Create and initialise Kompute Tensors through manager auto tensorInA = mgr.buildTensor({ 2., 2., 2. }); auto tensorInB = mgr.buildTensor({ 1., 2., 3. }); auto tensorOut = mgr.buildTensor({ 0., 0., 0. }); // 3. Specify "multiply shader" code (can also be raw string, spir-v bytes or file path) std::string shaderString = (R"( #version 450 layout (local_size_x = 1) in; // The input tensors bind index is relative to index in parameter passed layout(set = 0, binding = 0) buffer bina { float tina[]; }; layout(set = 0, binding = 1) buffer binb { float tinb[]; }; layout(set = 0, binding = 2) buffer bout { float tout[]; }; void main() { uint index = gl_GlobalInvocationID.x; tout[index] = tina[index] * tinb[index]; } )"); // 3. Run operation with string shader synchronously mgr.evalOpDefault<kp::OpAlgoBase>( { tensorInA, tensorInB, tensorOut }, std::vector<char>(shaderString.begin(), shaderString.end())); // 4. Map results back from GPU memory to print the results mgr.evalOpDefault<kp::OpTensorSyncLocal>({ tensorInA, tensorInB, tensorOut }); // Prints the output which is Output: { 2, 4, 6 } for (const float& elem : tensorOut->data()) std::cout << elem << " "; } ``` ### Your First Kompute (Python) The [Python package](https://kompute.cc/overview/python-package.html) provides a [high level interactive interface](https://kompute.cc/overview/python-reference.html) that enables for experimentation whilst ensuring high performance and fast development workflows. ```python # 1. Create Kompute Manager with default settings (device 0 and first compute compatible queue) mgr = Manager() # 2. Create and initialise Kompute Tensors (can be initialized with List[] or np.Array) tensor_in_a = Tensor([2, 2, 2]) tensor_in_b = Tensor([1, 2, 3]) tensor_out = Tensor([0, 0, 0]) mgr.eval_tensor_create_def([tensor_in_a, tensor_in_b, tensor_out]) # 3. Specify "multiply shader" code (can also be raw string, spir-v bytes or file path) @python2shader def compute_shader_multiply(index=("input", "GlobalInvocationId", ivec3), data1=("buffer", 0, Array(f32)), data2=("buffer", 1, Array(f32)), data3=("buffer", 2, Array(f32))): i = index.x data3[i] = data1[i] * data2[i] # 4. Run multiplication operation synchronously mgr.eval_algo_data_def( [tensor_in_a, tensor_in_b, tensor_out], compute_shader_multiply.to_spirv()) # 5. Map results back from GPU memory to print the results mgr.eval_tensor_sync_local_def([tensor_out]) # Prints [2.0, 4.0, 6.0] print(tensor_out.data()) ``` ### Interactive Notebooks & Hands on Videos You are able to try out the interactive Colab Notebooks which allow you to use a free GPU. The available examples are the Python and C++ examples below: <table> <tr> <td width="50%"> <h5>Try the interactive <a href="https://colab.research.google.com/drive/1l3hNSq2AcJ5j2E3YIw__jKy5n6M615GP?authuser=1#scrollTo=1BipBsO-fQRD">C++ Colab</a> from <a href="https://towardsdatascience.com/machine-learning-and-data-processing-in-the-gpu-with-vulkan-kompute-c9350e5e5d3a">Blog Post</a></h5> </td> <td> <h5>Try the interactive <a href="https://colab.research.google.com/drive/15uQ7qMZuOyk8JcXF-3SB2R5yNFW21I4P">Python Colab</a> from <a href="https://towardsdatascience.com/beyond-cuda-gpu-accelerated-python-for-machine-learning-in-cross-vendor-graphics-cards-made-simple-6cc828a45cc3">Blog Post</a></h5> </td> </tr> <tr> <td width="50%"> <a href="https://colab.research.google.com/drive/1l3hNSq2AcJ5j2E3YIw__jKy5n6M615GP?authuser=1#scrollTo=1BipBsO-fQRD"> <img src="https://raw.githubusercontent.com/EthicalML/vulkan-kompute/master/docs/images/binder-cpp.jpg"> </a> </td> <td> <a href="https://colab.research.google.com/drive/15uQ7qMZuOyk8JcXF-3SB2R5yNFW21I4P"> <img src="https://raw.githubusercontent.com/EthicalML/vulkan-kompute/master/docs/images/binder-python.jpg"> </a> </td> </tr> </table> You can also check out the two following talks presented at the FOSDEM 2021 conference. Both videos have timestamps which will allow you to skip to the most relevant section for you - the intro & motivations for both is almost the same so you can skip to the more specific content. <table> <tr> <td width="50%"> <h5>Watch the video for <a href="https://www.youtube.com/watch?v=Xz4fiQNmGSA">C++ & Vulkan SDK</a> Enthusiasts</h5> </td> <td> <h5>Watch the video for <a href="https://www.youtube.com/watch?v=AJRyZ09IUdg">Python & Machine Learning</a> Enthusiasts</h5> </td> </tr> <tr> <td width="50%"> <a href="https://www.youtube.com/watch?v=Xz4fiQNmGSA"> <img src="https://raw.githubusercontent.com/EthicalML/vulkan-kompute/master/docs/images/kompute-cpp-video.png"> </a> </td> <td> <a href="https://www.youtube.com/watch?v=AJRyZ09IUdg"> <img src="https://raw.githubusercontent.com/EthicalML/vulkan-kompute/master/docs/images/kompute-python-video.png"> </a> </td> </tr> </table> ## Architectural Overview The core architecture of Kompute includes the following: * [Kompute Manager](https://kompute.cc/overview/reference.html#manager) - Bas