Getting Started

Environment

Python 3.7.13, PyTorch 1.10.0, scikit-learn 1.0.2, fastText 0.9.2, and DGL 0.9.2 are suggested.

Dataset

D1: https://github.com/CloudWise-OpenSource/GAIA-DataSet

D1 contains two datasets: MicroSS and Companion Data. We use MicroSS, for it provides trace, log, and metric at the same time.

Demo

We provide a demo. Please run:

python main.py --config gaia_config.yaml

Parameter Description in the Demo

fastText & Instance Embedding

• vector_dim: The dimension of event embedding vectors. (default: 100)
• sample_count: The number of samples per type after data augmentation. (default: 1000)
• edit_count: The number of events modified per sample during data augmentation. (default: 1)
• minCount: The minimum number of occurrences of the event (events that occur less than this number are ignored). (default: 1)

DGL

• epoch: Training rounds. (default: 6000)
• batch_size: The number of samples contained in a batch of data. (default: 1000)
• win_size: The length of the judgment window for ending training early. (default: 10)
• win_threshole: The thresh for ending training early. (default: 0.0001)
• lr: The learning rate. (default: 0.001)