A repository for notes on mlrun
- open-source MLOps framework
- abstraction layer to a variety of technology stacks
- Feature and Artifact Store
- Elastic Serverless Runtimes: Kubernetes/Nuclio/Dask/Spark/Horovod
- ML Pipeline Automation: data prep/modeling/real-time pipelines/monitoring
- Central Management: UI/CLI/SDK
- Speed of deployment
- Elastic scaling of batch and real-time jobs
- Feature management system
- Runs anywhere
- Project
- Function
- Run
- Artifact
- Workflow
- UI
- Safari not supported, used Chrome
Must do TWO THINGS before you run tutorial on hosted platform:
- !/User/align_mlrun.sh
- Restart Kernel
-
Locally use Docker Desktop as described in Install MLRun on a Kubernetes Cluster
-
Follow steps described.
-
Install (Make sure you have the latest
pip
). Install on OS X will take several minutes and requires Rust and Cython. -
Create and source a python virtualenv:
python3 -m venv ~/.mlrun-notes && source ~/.mlrun-notes/bin/activate
pip install --upgrade pip && pip install mlrun
Can take 30+ minutes to install and contains many dependency errors.
- install latest Python and Rust):
brew install python
andbrew install rust
ModuleNotFoundError: No module named 'Cython'
RuntimeError: cargo not found in PATH. Please install rust (https://www.rust-lang.org/tools/install) and try again
clang: error: the clang compiler does not support 'faltivec', please use -maltivec and include altivec.h explicitly
ERROR: Could not build wheels for maturin, which is required to install pyproject.toml-based projects
- Install only takes a couple of minutes to install
(.mlrun-notes) ➜ functions git:(main) ✗ mlrun build function.yaml
> 2021-11-26 11:39:57,419 [info] remote deployment started
> 2021-11-26 11:39:57,419 [error] database connection is not configured
> 2021-11-26 11:39:57,419 [info] building image (.mlrun/func-default-remote-git-test-latest)
deploy error, local docker registry is not defined, set DEFAULT_DOCKER_REGISTRY/SECRET env vars
- Replace README.md with official docs link which is up to date and mentions Docker based workflows
- Hello World example in "one line"
- Target environment recommendation: Github Codespaces, AWS Cloudshell, etc
- Hello World using a pre-built Docker pull command
- Separate demos with foolproof "hello world" commands for each architectural component
- More clear link to official docs
- Point to a VM based solution: i.e AWS AMI, etc.