安装pytorc2.3.1+cu121
步骤 1. 创建并激活一个 conda 环境
conda create -n mmd python=3.8 -y
conda activate mmd
步骤 2. 基于PyTorch 官方说明安装 PyTorch，例如：
pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121
步骤 3. 验证PyTorch安装
[GCC 11.2.0] :: Anaconda, Inc. on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> torch.cuda.is_available()
True
安装waymo-open-dataset-tf-2-6-0
步骤 1. 安装waymo-API
pip install waymo-open-dataset-tf-2-6-0
步骤 2. 处理torch版本冲突
出现如下报错
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
torch 2.3.1+cu121 requires typing-extensions>=4.8.0, but you have typing-extensions 3.7.4.3 which is incompatible.
更新typing-extensions
pip install typing-extensions==4.8.0
步骤 3. 处理tensorflow-2.6.0与protobuf的版本冲突
pip install protobuf==3.20.0
步骤 4. 安装tensorflow-gpu所需cudatoolkit与cudnn
conda install cudatoolkit=11.3.1 cudnn=8.2.1
步骤 5. 验证torch与tensorflow安装
Python 3.8.19 (default, Mar 20 2024, 19:58:24)
[GCC 11.2.0] :: Anaconda, Inc. on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> import tensorflow as tf
>>> torch.cuda.is_available()
True
>>> tf.test.is_gpu_available()
WARNING:tensorflow:From <stdin>:1: is_gpu_available (from tensorflow.python.framework.test_util) is deprecated and will be removed in a future version.
Instructions for updating:
Use `tf.config.list_physical_devices('GPU')` instead.
2024-06-24 10:36:35.105410: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: AVX2 FMA
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
2024-06-24 10:36:35.106672: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2024-06-24 10:36:35.114045: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2024-06-24 10:36:35.114108: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2024-06-24 10:36:35.165378: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2024-06-24 10:36:35.165449: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2024-06-24 10:36:35.165495: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2024-06-24 10:36:35.165545: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1510] Created device /device:GPU:0 with 20170 MB memory: -> device: 0, name: NVIDIA GeForce RTX 4090, pci bus id: 0000:01:00.0, compute capability: 8.9
True
安装MMDetection3D
步骤 1. 使用 MIM 安装 MMEngine，MMCV 和 MMDetection
pip install -U openmim
mim install mmengine
mim install 'mmcv>=2.0.0rc4'
mim install 'mmdet>=3.0.0'
步骤 2. 处理openmim与tensorflow的版本冲突
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
tensorflow 2.6.0 requires numpy~=1.19.2, but you have numpy 1.24.4 which is incompatible.
tensorflow 2.6.0 requires typing-extensions~=3.7.4, but you have typing-extensions 4.8.0 which is incompatible.
参考issues#2858安装numpy==1.23.0
pip install numpy==1.23.0
步骤 2. 安装 MMDetection3D
下载源码备用
git clone https://github.com/open-mmlab/mmdetection3d.git -b dev-1.x
# "-b dev-1.x" 表示切换到 `dev-1.x` 分支。
cd mmdetection3d
将 mmdet3d 作为依赖或第三方 Python 包使用，使用 MIM 安装
mim install "mmdet3d>=1.1.0rc0"
验证安装
步骤 1. 验证mmdet3d
mim download mmdet3d --config pointpillars_hv_secfpn_8xb6-160e_kitti-3d-car --dest .
python demo/pcd_demo.py demo/data/kitti/000008.bin pointpillars_hv_secfpn_8xb6-160e_kitti-3d-car.py hv_pointpillars_secfpn_6x8_160e_kitti-3d-car_20220331_134606-d42d15ed.pth --show
步骤 2. 验证pytorch与tensorflow
Python 3.8.19 (default, Mar 20 2024, 19:58:24)
[GCC 11.2.0] :: Anaconda, Inc. on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> import tensorflow as tf
>>> torch.cuda.is_available()
True
>>> tf.test.is_gpu_available()
WARNING:tensorflow:From <stdin>:1: is_gpu_available (from tensorflow.python.framework.test_util) is deprecated and will be removed in a future version.
Instructions for updating:
Use `tf.config.list_physical_devices('GPU')` instead.
2024-06-24 10:47:24.233284: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: AVX2 FMA
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
2024-06-24 10:47:24.234634: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2024-06-24 10:47:24.237583: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2024-06-24 10:47:24.237644: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2024-06-24 10:47:24.267267: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2024-06-24 10:47:24.267336: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2024-06-24 10:47:24.267382: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2024-06-24 10:47:24.267431: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1510] Created device /device:GPU:0 with 19665 MB memory: -> device: 0, name: NVIDIA GeForce RTX 4090, pci bus id: 0000:01:00.0, compute capability: 8.9
True
>>>