Tim Blog

Himmel oder Hölle

pytorch lightning

Its core concept is to seperate model and the engineering code, to make user concentrate on model research. installation pip install pytorch-lightning usage import torch from torch import nn ...

Posted by neverset on October 24, 2022

geo visualization

cartopy cartopy is based on matplotlib, can visualize geo information combining with proj, numpy and shapely. import cartopy.crs as ccrs from cartopy.mpl.ticker import LongitudeFormatter, Latitude...

Posted by neverset on October 15, 2022

dataset distillation

dataset distillation is to extract smaller dataset to train neural network to get better performance https://github.com/Guang000/Awesome-Dataset-Distillation

Posted by neverset on October 15, 2022

push message

Posted by neverset on October 15, 2022

celery

celery celery is a third party python module for task scheduling asynchronously.

Posted by neverset on September 24, 2022

dazel

Run Google’s bazel inside a docker container via a seamless proxy. bazel

Posted by neverset on September 18, 2022

image segmentation

PaddleSeg paddleseg is a real tiem semantic segmentation model. detail: github.com/paddlePaddle/PaddleSeg

Posted by neverset on July 16, 2022

causal analysis

causal-learn #install causal-learn pip install causal-learn #causal discovery G = pc(data, alpha, indep_test, stable, uc_rule, uc_priority, mvpc, correction_name, background_knowledge) #visualizat...

Posted by neverset on December 10, 2021

feature engineering

data transformation 数据转换成图像元数据泄露当处理过的特征在没有应用任何机器学习的情况下，可以非常完美地解释目标时，这可能发生了数据泄露表征学习特征法直接从训练数据中捕捉最显著的特征，无需其他特征工程。均值编码数据分析中经常会遇到类别属性，比如日期、性别、街区编号、IP地址等。绝大部分数据分析算法是无法直接处理这类变量的，需要先把它们先处理成数值型...

Posted by neverset on September 19, 2021

tqdm

tqdm is used to show progress percentage in python app. usage 1) import module import sys if hasattr(sys.modules["__main__"], "get_ipython"): from tqdm import notebook as tqdm else: impor...

Posted by neverset on April 17, 2021