Hardware
- On-premises compute (VM or bare metal) 32 GB RAM, 4vCPU
- GPU access, 16Gb or higher
Software
- Ubuntu 22.04 or comparable Linux environment (WSL is also possible)
- Miniconda environments
- Python 3.10 or higher
- Jupyter notebooks installed.
Storage
- 500-1000 GB for combined model cache and PDF dataset storage
- As you experiment, this will grow.
Network
- Access to public internet
Note: Test environment: Dell PowerEdge with Nvidia T4 and Dell Precision Mobile Workstation 7780 with Nvidia RTX5000.