dropt-example

This repo provides examples of parameter tuning with DrOpt.

Directory structure

.
|
+- trials/	# parameter tuning trial examples
|
+- data/	# datasets for trials
|
+- util/	# utility Python scirpt
|
+- log/		# trial logs (create upon running)

Models

Our trial examples include the following models:

• func-eggholder: test function for optimization
• func-rosenbrock: another test function for optimization
• titanic-xgboost: Titanic survival prediction by XGBoost
• mnist-pytorch: handwritten digit recognition with PyTorch framework
• imagenet-pytorch: image classification with PyTorch framework

Run a trial

1. Set up Python enviroment.
   • (Optional) Set up Python virtual enviroment.
2. Go to the folder of a trial example.
3. Install required Python package:

   $ pip install -r requirements.txt

4. Run the trial with DrOpt:

   $ droptctl -t TOKEN create

5. Inspect the trial result on the DrOpt service cloud.

References

1. DrOpt official service cloud: https://dropt.goedge.ai
2. DrOpt documentation: https://dropt-cli.readthedocs.io
3. dropt-cli source repo: https://github.com/GoEdge-ai/dropt-cli