TraceVAE

This is the source code for "Unsupervised Anomaly Detection on Microservice Traces through Graph VAE".

Usage

1. pip3 install -r requirements.txt.
2. Convert the dataset with python3 -m tracegnn.cli.data_process preprocess -i [input_path] -o [dataset_path]. The sample dataset is under sample_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

3. Train the model with bash train.sh [dataset_path]:

bash train.sh sample_dataset

4. Evaluate the model with bash teset.sh [model_path] [dataset_path]. The default model path is under results/train/models/final.pt:

bash test.sh results/train/models/final.pt sample_dataset