diff --git a/docs/guides/performance_improving/paddle_tensorrt_infer.md b/docs/guides/performance_improving/paddle_tensorrt_infer.md index 2890eceb4ab..4590884c5a8 100644 --- a/docs/guides/performance_improving/paddle_tensorrt_infer.md +++ b/docs/guides/performance_improving/paddle_tensorrt_infer.md @@ -60,7 +60,7 @@ config->EnableTensorRtEngine(1 << 20 /* workspace_size*/, ## Paddle-TRT 样例编译测试 1. 下载或编译带有 TensorRT 的 paddle 预测库,参考[安装与编译 C++预测库](../../inference_deployment/inference/build_and_install_lib_cn.html)。 -2. 从[NVIDIA 官网](https://developer.nvidia.com/nvidia-tensorrt-download)下载对应本地环境中 cuda 和 cudnn 版本的 TensorRT,需要登陆 NVIDIA 开发者账号。 +2. 从[NVIDIA 官网](https://developer.nvidia.com/tensorrt)下载对应本地环境中 cuda 和 cudnn 版本的 TensorRT,需要登陆 NVIDIA 开发者账号。 3. 下载[预测样例](https://paddle-inference-dist.bj.bcebos.com/tensorrt_test/paddle_inference_sample_v1.7.tar.gz)并解压,进入`sample/paddle-TRT`目录下。 `paddle-TRT` 文件夹目录结构如下: diff --git a/docs/guides/performance_improving/paddle_tensorrt_infer_en.md b/docs/guides/performance_improving/paddle_tensorrt_infer_en.md index 0acc384ab2a..29bd16445b9 100644 --- a/docs/guides/performance_improving/paddle_tensorrt_infer_en.md +++ b/docs/guides/performance_improving/paddle_tensorrt_infer_en.md @@ -53,7 +53,7 @@ The details of this interface is as following: ## Paddle-TRT example compiling test 1. Download or compile Paddle Inference with TensorRT support, refer to [Install and Compile C++ Inference Library](../../inference_deployment/inference/build_and_install_lib_en.html). -2. Download NVIDIA TensorRT(with consistent version of cuda and cudnn in local environment) from [NVIDIA TensorRT](https://developer.nvidia.com/nvidia-tensorrt-download) with an NVIDIA developer account. +2. Download NVIDIA TensorRT(with consistent version of cuda and cudnn in local environment) from [NVIDIA TensorRT](https://developer.nvidia.com/tensorrt) with an NVIDIA developer account. 3. Download [Paddle Inference sample](https://paddle-inference-dist.bj.bcebos.com/tensorrt_test/paddle_inference_sample_v1.7.tar.gz) and uncompress, and enter `sample/paddle-TRT` directory. `paddle-TRT` directory structure is as following: diff --git a/docs/install/compile/linux-compile.md b/docs/install/compile/linux-compile.md index 4201c19e8f6..12ed39b9fe9 100644 --- a/docs/install/compile/linux-compile.md +++ b/docs/install/compile/linux-compile.md @@ -16,7 +16,7 @@ * 如果您的计算机有 NVIDIA® GPU,请确保满足以下条件以编译 GPU 版 PaddlePaddle - * **CUDA 工具包 10.1 配合 cuDNN 7 (cuDNN 版本>=7.6.5, 如需多卡支持,需配合 NCCL2.7 及更高)** + * **CUDA 工具包 10.1 配合 cuDNN 7 (cuDNN 版本>=7.6.5, 如需多卡支持,需配合 NCCL2.7 及更高;不支持使用 TensorRT)** * **CUDA 工具包 10.2 配合 cuDNN 7 (cuDNN 版本>=7.6.5, 如需多卡支持,需配合 NCCL2.7 及更高;如需使用 PaddleTensorRT 推理,需配合 TensorRT7.0.0.11)** * **CUDA 工具包 11.1 配合 cuDNN v8.1.1(如需多卡支持,需配合 NCCL2.7 及更高;如需使用 PaddleTensorRT 推理,需配合 TensorRT7.2.3.4)** * **CUDA 工具包 11.2 配合 cuDNN v8.1.1(如需多卡支持,需配合 NCCL2.7 及更高;如需使用 PaddleTensorRT 推理,需配合 TensorRT8.0.3.4)** @@ -24,7 +24,7 @@ * **CUDA 工具包 11.7 配合 cuDNN v8.4.1(如需多卡支持,需配合 NCCL2.7 及更高;如需使用 PaddleTensorRT 推理,需配合 TensorRT8.4.2.4)** * **GPU 运算能力超过 3.5 的硬件设备** - 您可参考 NVIDIA 官方文档了解 CUDA、CUDNN 和 TensorRT 的安装流程和配置方法,请见[CUDA](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/),[cuDNN](https://docs.nvidia.com/deeplearning/sdk/cudnn-install/),[TensorRT](https://developer.nvidia.com/nvidia-tensorrt-download) + 您可参考 NVIDIA 官方文档了解 CUDA、CUDNN 和 TensorRT 的安装流程和配置方法,请见[CUDA](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/),[cuDNN](https://docs.nvidia.com/deeplearning/sdk/cudnn-install/),[TensorRT](https://developer.nvidia.com/tensorrt) ## 安装步骤 diff --git a/docs/install/compile/linux-compile_en.md b/docs/install/compile/linux-compile_en.md index 2c7f77d69d8..755cfc70667 100644 --- a/docs/install/compile/linux-compile_en.md +++ b/docs/install/compile/linux-compile_en.md @@ -16,7 +16,7 @@ * If your computer has NVIDIA® GPU, and the following conditions are met,GPU version of PaddlePaddle is recommended. - * **CUDA toolkit 10.1 with cuDNN 7 (cuDNN version>=7.6.5, for multi card support, NCCL2.7 or higher)** + * **CUDA toolkit 10.1 with cuDNN 7 (cuDNN version>=7.6.5, for multi card support, NCCL2.7 or higher;TensorRT is not supported)** * **CUDA toolkit 10.2 with cuDNN 7 (cuDNN version>=7.6.5, for multi card support, NCCL2.7 or higher;for PaddleTensorRT deployment, TensorRT7.0.0.11)** * **CUDA toolkit 11.1 with cuDNN v8.1.1(for multi card support, NCCL2.7 or higher;for PaddleTensorRT deployment, TensorRT7.2.3.4)** * **CUDA toolkit 11.2 with cuDNN v8.1.1(for multi card support, NCCL2.7 or higher;for PaddleTensorRT deployment, TensorRT8.0.3.4)** @@ -24,7 +24,7 @@ * **CUDA toolkit 11.7 with cuDNN v8.4.1(for multi card support, NCCL2.7 or higher;for PaddleTensorRT deployment, TensorRT8.4.2.4)** * **Hardware devices with GPU computing power over 3.5** - You can refer to NVIDIA official documents for installation process and configuration method of CUDA, cuDNN and TensorRT. Please refer to[CUDA](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/),[cuDNN](https://docs.nvidia.com/deeplearning/sdk/cudnn-install/),[TensorRT](https://developer.nvidia.com/nvidia-tensorrt-download) + You can refer to NVIDIA official documents for installation process and configuration method of CUDA, cuDNN and TensorRT. Please refer to[CUDA](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/),[cuDNN](https://docs.nvidia.com/deeplearning/sdk/cudnn-install/),[TensorRT](https://developer.nvidia.com/tensorrt) ## Installation steps diff --git a/docs/install/conda/linux-conda.md b/docs/install/conda/linux-conda.md index e6662874f2f..d4a64bb2ef8 100644 --- a/docs/install/conda/linux-conda.md +++ b/docs/install/conda/linux-conda.md @@ -1,6 +1,6 @@ # Linux 下的 Conda 安装 -[Anaconda](https://www.anaconda.com/)是一个免费开源的 Python 和 R 语言的发行版本,用于计算科学,Anaconda 致力于简化包管理和部署。Anaconda 的包使用软件包管理系统 Conda 进行管理。Conda 是一个开源包管理系统和环境管理系统,可在 Windows、macOS 和 Linux 上运行。 +[Anaconda](https://www.anaconda.com/)是一个免费开源的 Python 和 R 语言的发行版本,用于计算科学,Anaconda 致力于简化包管理和部署。Anaconda 的包使用软件包管理系统 Conda 进行管理。Conda 是一个开源包管理系统和环境管理系统,可在 Windows、macOS 和 Linux 上运行。本文档为你介绍 Anaconda 安装方式,飞桨提供的 Anaconda 安装包支持分布式训练(多机多卡)、TensorRT 推理功能。 ## 一、环境准备 diff --git a/docs/install/conda/windows-conda.md b/docs/install/conda/windows-conda.md index 9ecd6a3343d..e08fda606df 100644 --- a/docs/install/conda/windows-conda.md +++ b/docs/install/conda/windows-conda.md @@ -1,8 +1,6 @@ # Windows 下的 Conda 安装 -[Anaconda](https://www.anaconda.com/)是一个免费开源的 Python 和 R 语言的发行版本,用于计算科学,Anaconda 致力于简化包管理和部署。Anaconda 的包使用软件包管理系统 Conda 进行管理。Conda 是一个开源包管理系统和环境管理系统,可在 Windows、macOS 和 Linux 上运行。 - - +[Anaconda](https://www.anaconda.com/)是一个免费开源的 Python 和 R 语言的发行版本,用于计算科学,Anaconda 致力于简化包管理和部署。Anaconda 的包使用软件包管理系统 Conda 进行管理。Conda 是一个开源包管理系统和环境管理系统,可在 Windows、macOS 和 Linux 上运行。本文档为你介绍 Anaconda 安装方式,飞桨提供的 Anaconda 安装包支持分布式训练(多机多卡)、TensorRT 推理功能。 ## 一、环境准备 diff --git a/docs/install/pip/linux-pip.md b/docs/install/pip/linux-pip.md index e49a51ee75c..d791993c77f 100644 --- a/docs/install/pip/linux-pip.md +++ b/docs/install/pip/linux-pip.md @@ -1,5 +1,7 @@ # Linux 下的 PIP 安装 +The Python Package Index(PyPI)是 Python 的包管理器。本文档为你介绍 PyPI 安装方式,飞桨提供的 PyPI 安装包支持分布式训练(多机多卡)、TensorRT 推理功能;PyPI 下载详见 PyPI 官网(PyPI 官网设置链接:https://pypi.org/)。 + ## 一、环境准备 ### 1.1 目前飞桨支持的环境 @@ -75,7 +77,7 @@ * 如果您的计算机有 NVIDIA® GPU,请确保满足以下条件并且安装[GPU 版 PaddlePaddle](#gpu) - * **CUDA 工具包 10.1 配合 cuDNN 7 (cuDNN 版本>=7.6.5, 如需多卡支持,需配合 NCCL2.7 及更高)** + * **CUDA 工具包 10.1 配合 cuDNN 7 (cuDNN 版本>=7.6.5, 如需多卡支持,需配合 NCCL2.7 及更高; 不支持使用 TensorRT)** * **CUDA 工具包 10.2 配合 cuDNN 7 (cuDNN 版本>=7.6.5, 如需多卡支持,需配合 NCCL2.7 及更高;如需使用 PaddleTensorRT 推理,需配合 TensorRT7.0.0.11)** @@ -89,7 +91,7 @@ * **GPU 运算能力超过 3.5 的硬件设备** - 您可参考 NVIDIA 官方文档了解 CUDA、CUDNN 和 TensorRT 的安装流程和配置方法,请见[CUDA](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/),[cuDNN](https://docs.nvidia.com/deeplearning/sdk/cudnn-install/),[TensorRT](https://developer.nvidia.com/nvidia-tensorrt-download) + 您可参考 NVIDIA 官方文档了解 CUDA、CUDNN 和 TensorRT 的安装流程和配置方法,请见[CUDA](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/),[cuDNN](https://docs.nvidia.com/deeplearning/sdk/cudnn-install/),[TensorRT](https://developer.nvidia.com/tensorrt) * 如果您需要使用多卡环境请确保您已经正确安装 nccl2,或者按照以下指令安装 nccl2(这里提供的是 CUDA10.2,cuDNN7 下 nccl2 的安装指令,更多版本的安装信息请参考 NVIDIA[官方网站](https://developer.nvidia.com/nccl)): @@ -163,11 +165,16 @@ 2.2.4 CUDA11.2 的 PaddlePaddle - + cuDNN8.1.1: ``` python -m pip install paddlepaddle-gpu==2.4.0rc0.post112 -f https://www.paddlepaddle.org.cn/whl/linux/mkl/avx/stable.html ``` + 如果你想使用 PaddleTensorRT 进行推理,cudnn8.2.1 与 TensorRT8.0.3.4 联编的安装包能够获得更优的推理性能,安装命令如下: + ``` + python -m pip install paddlepaddle-gpu==2.4.0rc0.post112 -f https://www.paddlepaddle.org.cn/whl/linux/cuda11.2-cudnn8.2-tensorrt8.html + ``` + 2.2.5 CUDA11.6 的 PaddlePaddle @@ -236,8 +243,6 @@ python -m pip download paddlepaddle==2.4.0rc0 -f https://www.paddlepaddle.org.cn/whl/linux/openblas/avx/stable.html --no-index --no-deps ``` -* 如果你想在`cuda11.2`环境下,获得更好的`PaddleTensorRT`推理性能,需配合`cudnn8.2.1`,并安装联编`tensorrt8.0.3.4`的 Paddle 包,可以参考[下载安装 Linux 预测库](https://paddleinference.paddlepaddle.org.cn/user_guides/download_lib.html)。 - diff --git a/docs/install/pip/linux-pip_en.md b/docs/install/pip/linux-pip_en.md index 9da0d3e6104..221934d1a2b 100644 --- a/docs/install/pip/linux-pip_en.md +++ b/docs/install/pip/linux-pip_en.md @@ -75,7 +75,7 @@ If you installed Python via Homebrew or the Python website, `pip` was installed * If your computer has NVIDIA® GPU, please make sure that the following conditions are met and install [the GPU Version of PaddlePaddle](#gpu) - * **CUDA toolkit 10.1 with cuDNN 7 (cuDNN version>=7.6.5, for multi card support, NCCL2.7 or higher)** + * **CUDA toolkit 10.1 with cuDNN 7 (cuDNN version>=7.6.5, for multi card support, NCCL2.7 or higher; TensorRT is not supported)** * **CUDA toolkit 10.2 with cuDNN 7 (cuDNN version>=7.6.5, for multi card support, NCCL2.7 or higher;for PaddleTensorRT deployment, TensorRT7.0.0.11)** @@ -89,7 +89,7 @@ If you installed Python via Homebrew or the Python website, `pip` was installed * **Hardware devices with GPU computing power over 3.5** - You can refer to NVIDIA official documents for installation process and configuration method of CUDA, cuDNN and TensorRT. Please refer to [CUDA](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/),[cuDNN](https://docs.nvidia.com/deeplearning/sdk/cudnn-install/),[TensorRT](https://developer.nvidia.com/nvidia-tensorrt-download) + You can refer to NVIDIA official documents for installation process and configuration method of CUDA, cuDNN and TensorRT. Please refer to [CUDA](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/),[cuDNN](https://docs.nvidia.com/deeplearning/sdk/cudnn-install/),[TensorRT](https://developer.nvidia.com/tensorrt) * If you need to use a multi-card environment, please make sure that you have installed nccl2 correctly, or install nccl2 according to the following instructions (here are the installation instructions of nccl2 under CUDA10.2 and cuDNN7. For more version installation information, please refer to NVIDIA [Official Website](https://developer.nvidia.com/nccl)): diff --git a/docs/install/pip/macos-pip.md b/docs/install/pip/macos-pip.md index 7d30b0c87e0..b40df5d36ea 100644 --- a/docs/install/pip/macos-pip.md +++ b/docs/install/pip/macos-pip.md @@ -1,5 +1,7 @@ # MacOS 下的 PIP 安装 +The Python Package Index(PyPI)是 Python 的包管理器。本文档为你介绍 PyPI 安装方式。PyPI 下载详见 PyPI 官网(PyPI 官网设置链接:https://pypi.org/)。 + ## 一、环境准备 ### 1.1 目前飞桨支持的环境 diff --git a/docs/install/pip/windows-pip.md b/docs/install/pip/windows-pip.md index f25e72905cd..509c4a87eba 100644 --- a/docs/install/pip/windows-pip.md +++ b/docs/install/pip/windows-pip.md @@ -1,5 +1,7 @@ # Windows 下的 PIP 安装 +The Python Package Index(PyPI)是 Python 的包管理器。本文档为你介绍 PyPI 安装方式,飞桨提供的 PyPI 安装包支持分布式训练(多机多卡)、TensorRT 推理功能;PyPI 下载详见 PyPI 官网(PyPI 官网设置链接:https://pypi.org/)。 + ## 一、环境准备 ### 1.1 目前飞桨支持的环境 @@ -51,7 +53,7 @@ * 如果您的计算机有 NVIDIA® GPU,请确保满足以下条件并且安装 GPU 版 PaddlePaddle - * **CUDA 工具包 10.1 配合 cuDNN v7.6.5** + * **CUDA 工具包 10.1 配合 cuDNN v7.6.5(不支持使用 TensorRT)** * **CUDA 工具包 10.2 配合 cuDNN v7.6.5(如需使用 PaddleTensorRT 推理,需配合 TensorRT7.0.0.11)** @@ -65,7 +67,7 @@ * **GPU 运算能力超过 3.5 的硬件设备** - * 注:目前官方发布的 windows 安装包仅包含 CUDA 10.1/10.2/11.1/11.2/11.6/11.7,如需使用其他 cuda 版本,请通过源码自行编译。您可参考 NVIDIA 官方文档了解 CUDA、CUDNN 和 TensorRT 的安装流程和配置方法,请见[CUDA](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/),[cuDNN](https://docs.nvidia.com/deeplearning/sdk/cudnn-install/),[TensorRT](https://developer.nvidia.com/nvidia-tensorrt-download) + * 注:目前官方发布的 windows 安装包仅包含 CUDA 10.1/10.2/11.1/11.2/11.6/11.7,如需使用其他 cuda 版本,请通过源码自行编译。您可参考 NVIDIA 官方文档了解 CUDA、CUDNN 和 TensorRT 的安装流程和配置方法,请见[CUDA](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/),[cuDNN](https://docs.nvidia.com/deeplearning/sdk/cudnn-install/),[TensorRT](https://developer.nvidia.com/tensorrt) diff --git a/docs/install/pip/windows-pip_en.md b/docs/install/pip/windows-pip_en.md index dd0b0e5cd55..254a537ca64 100644 --- a/docs/install/pip/windows-pip_en.md +++ b/docs/install/pip/windows-pip_en.md @@ -50,7 +50,7 @@ If you installed Python via Homebrew or the Python website, `pip` was installed * If your computer has NVIDIA® GPU, please make sure that the following conditions are met and install [the GPU Version of PaddlePaddle](#gpu) - * **CUDA toolkit 10.1 with cuDNN v7.6.5** + * **CUDA toolkit 10.1 with cuDNN v7.6.5(TensorRT is not supported)** * **CUDA toolkit 10.2 with cuDNN v7.6.5(for PaddleTensorRT deployment, TensorRT7.0.0.11)** @@ -64,7 +64,7 @@ If you installed Python via Homebrew or the Python website, `pip` was installed * **GPU CUDA capability over 3.5** - You can refer to NVIDIA official documents for installation process and configuration method of CUDA, cuDNN and TensorRT. Please refer to [CUDA](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/),[cuDNN](https://docs.nvidia.com/deeplearning/sdk/cudnn-install/),[TensorRT](https://developer.nvidia.com/nvidia-tensorrt-download) + You can refer to NVIDIA official documents for installation process and configuration method of CUDA, cuDNN and TensorRT. Please refer to [CUDA](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/),[cuDNN](https://docs.nvidia.com/deeplearning/sdk/cudnn-install/),[TensorRT](https://developer.nvidia.com/tensorrt) ## Installation Step