metadata
tags:
- time-series-forecasting
- cloud-computing
- timesfm
- resource-prediction
🌩️ Cloud-Pre (500M)
Cloud-Pre is a specialized time-series foundation model designed for predicting cloud environment resource usage (CPU, RAM, Network).
It is fine-tuned based on Google's highly advanced google/timesfm-2.0-500m-pytorch.
Model Details
- Architecture: Decoder-Only Transformer (TimesFM 2.0 Base)
- Parameters: 500 Million
- Fine-tuning Objective: Cloud CPU/Resource peak and anomaly forecasting to aid with Predictive Auto-scaling.
- Developer: Assem Sabry
How to find the training code?
The complete open-source training pipeline and data engineering scripts can be found on my GitHub Repository.