/my-pipelines

Primary LanguageJupyter Notebook

EZ Pipelines

Writing lines of code for data and models preparations can take hours to prepare if ones start everything from scratch once again for every new projects coming in. Moreover, even though there are several tutorials out there online, the data preparation and formating can be somewhat different across different tutorials. Here I present my collections of simple pipelines for multiple common machine learning tasks with clear explaination and command line interface. All you need to do is to prepare your data into a super-simple format, run few lines of codes, and start your training rightaway. EZ

(Actually, these are my pipelines from my previous work and I just want to store them together for easy usage)

Current Available Pipelines

1. Image Classification

  • Utilize state-of-the-art pre-trained models from torchvision for fast fine-tuning and inference
  • Feature learning rate finder to identify most suitable learning rate initialization
  • Error analysis by confusion matrix
  • Visualization model's vision

Working-In-Progress Pipelines

2. Image Segmentation

3. Object Detection

4. NLP Token Classification

5. Machine Translation

6. Forecasting with Regression