This is the source code for "Unsupervised Anomaly Detection on Microservice Traces through Graph VAE".
pip3 install -r requirements.txt.- Convert the dataset with
python3 -m tracegnn.cli.data_process preprocess -i [input_path] -o [dataset_path]. The sample dataset is undersample_dataset. (Note: This sample dataset only shows data format and usage, and cannot be used to evaluate model performance. Please replace it with your dataset.) sample:
python3 -m tracegnn.cli.data_process preprocess -i sample_dataset -o sample_dataset
- Train the model with
bash train.sh [dataset_path]:
bash train.sh sample_dataset
- Evaluate the model with
bash teset.sh [model_path] [dataset_path]. The default model path is underresults/train/models/final.pt:
bash test.sh results/train/models/final.pt sample_dataset