ML developers can reduce cloud expenses by running proof of concept experiments and debug projects locally, using the same containerised environments that will be used to deploy the model.
The Hardware
Motherboard: B550-A PRO MSI B550-A PRO AM4 ATX MOTHERBOARD
DRAM: CMK32GX4M2D3600C18 CORSAIR VENGEANCE LPX 32GB (2 X 16GB) DDR4 DRAM 3600MHZ C18 DESKTOP RAM - BLACK
CPU: 100-100000065BOX AMD RYZEN 5 5600X 4.60GHZ 6C/12T AM4 DESKTOP PROCESSOR
PSU: SST-ST85F-PT SILVERSTONE STRIDER 80+ PLATINUM 850W POWER SUPPLY
GPU: MSI GEFORCE RTX 3060 VENTUS 2X 12G OC
CASE: CC-9011210-WW CORSAIR 5000D AIRFLOW TEMPERED GLASS MID-TOWER CASE - BLACK
STORAGE: MZ-77E1T0BW SAMSUNG 870 EVO 1TB 2.5” INTERNAL SSD, WD Black 2TB
HCI: Keychron Q10 keyboard, Logitech MX ERGO mouse
OS and Software
OS: Ubuntu 22.04.1 LTS 64-bit
NVIDIA driver: 525.85.12
CUDA: 12.0
nvidia-ctk: 1.12.0
Docker: 20.10.23
Docker-compose: 1.29.2
Poetry: 1.3.2 (to assist in project config management)
Text editors: VS Code, Obsidian
Now I can use the template repos in my GitHub account to initialise a project and maintain a stable deployment when the development work is complete.