Implementation of SCARL: Attentive Reinforcement Learning-Based Scheduling in a Multi-Resource Heterogeneous Cluster

Description

• baselines.py: baseline heuristics.
• networks.py: Model Networks implemented.
• mc_job_dist.py: Job sequence generation.
• mc_env.py: OpenAI Gym compatible cluster simulation environment.
• mc_eval.py: Run model training
• generate-train-data.ipynb: train data generation
• envTest-30-2_6.ipynb: simple evaluation.

Usage example

1. create dataset using generate-train-data.ipynb
2. python mc_eval.py --step_size=15 --save_dir=model-2.0 --input=tr_2.0.pkl

---

I'm sorry that I can't disclose the script for training/running on actual k8s enironment. I can't find things I've done.

Implementation of the environment is mainly brought and extended from env of deeprm