fanre的专栏

用多GPU跑的时候总报这样的错误，感觉就是进程没有关掉导致的，我没有找到解决方案，也捕获不到RuntimeError，就直接把raise注释掉了。About the MultiThreadedAugmenter & SingleThreadedAugmenter · Issue #696 · MIC-DKFZ/nnUNet · GitHub...

一、引言tensorRT是一个高性能深度学习推理库。tensorrt包含推理优化器和运行时。TensorRT能够以更高的吞吐量和更低的延迟运行。本指南包含tensorrt基本的安装、转换、运行时选择以及最佳运用。二、安装tensorrt包含：容器、Debian、pip安装方式。2.1、容器中安装公有云上部署参照：NGC Certified Public Clouds Documentation2.2、Debian安装1、下载tensorRT2、安装os="ubu

def resample_mesh(img, target_spacing ,mode="bilinear",device='cuda:0',is_seg = False): data = sitk.GetArrayFromImage(img) size = img.GetSize()[::-1] origin = img.GetSpacing()[::-1] spacing = img.GetSpacing()[::-1] target_size = np..

https://pytorch.apachecn.org/docs/1.7/02.html最近没有正经看过源码，在回溯一下，以免忘记。pytorch有自动微分机制。张量可以理解为多维数组张量初始化import torchimport numpy as np#****************直接生成张量****************#data = [[1,2],[3,4]]x_data = torch.tensor(data)#***************numpy生成

参考：https://zhuanlan.zhihu.com/p/165152789 https://zhuanlan.zhihu.com/p/176998729 https://pytorch.org/docs/stable/amp.html https://pytorch.org/docs/stable/notes/amp_examples.html#amp-examplespytorch 版本有点旧，更新一下就好了，我直接更新到1.7import torchprint(torch._

tensorflow_addons 里面的tfa.image.transform(input_, transforms, interpolation)做平移很方便，但是pytorch版的平移版本太多了。正好做刚体配准的时候，感觉有一个版本的平移比较好用：def torch_image_translate(input_, tx, ty, interpolation='nearest'): # got these parameters from solving the equations for

import tensorflow as tfimport numpy as npimport tensorflow_addons as tfaimport SimpleITK as sitkprint(tf.__version__)import tensorflow as tfimport matplotlib.pyplot as pltgpus = tf.config.experimental.list_physical_devices('GPU')tf.config.experim.

from torch.nn import functional as Fimport torchimport SimpleITK as sitkimport matplotlib.pyplot as pltimport numpy as npimport mathimg_path = r'xxxdrr.jpg'img_itk = sitk.ReadImage(img_path)img = sitk.GetArrayFromImage(img_itk)plt.close('all')p.

参考：一、https://stackoverflow.com/questions/37572948/extracting-weights-from-caffemodel-without-caffe-installed-in-python/37573293#37573293将caffe.proto转成caffe_pb2.py：protoc --python_out=. caffe.pr...