markson14 / PyTorchEngine

A simple yet effective PyTorch-based framework for beginner to familiarize deep learning training pipeline. Plus, this is a good foundation to desigin your own training pipeline from scratch with guided.

PyTorchEngine

** Finished four main step before using this framework**

1. engine/dataset

1. define your own dataset
2. use tools/test_cfg.py to test if it works as expected

2. engine/models

1. create your own model pipeline
2. use tools/test_cfg.py to test if it works as expected

3. engine/loss

1. create your training loss
2. use tools/test_cfg.py to test if it works as expected
3. copy it to engine/engine.py - compute_loss()

4. engine/engine.py

1. After finished the previous steps, you are almost ready
2. setup your config/myconfig.yaml file
3. run train.py to see if it works properly
4. define your evaluation metric
5. use your evaluation metric result to save your model checkpoint