Has important links, tools, blogs or concepts that i learnt during EMLO-4.0
Dockers
- gitpod.io
- github code spaces
Low cost cloud GPUs
- jarvislabs
- runpod
- papersapce
- AWS spot instances
Developing GPU scheduling in MLOPS
- Use NVIDIA Time slicing for Low end GPUS
- Use NVIDIA MIG for high end GPUs
LLM
- Use vLLM to serve to large number of users and with batch serving
Tools
| S.NO | Purpose | Package Name |
|---|---|---|
| 1 | Argument management | Hydra |
| 2 | Logging | aim, Comet, MLflow |
| 3 | Data versioning | DVC with Cloud (GCS) |
| 4 | Markdown file generation | Tabulate |
| 5 | Code formatter | Black |
| 6 | Google Drive data download | gdown |
| 7 | GitHub Actions commenting | cml |
| 8 | Unit testing | Pytest |
| 9 | Test coverage reporting | Coverage |
| 10 | AI code assistance | Cursor |
| 11 | Hyper parameter Optimization | Optuna |
| 12 | Multi run parallel | Joblib |
| 13 | Run Github actions locally | act |