riteshkarval / mlflow-demo

dkube-examples

Modified conda.yaml to have correct version of protobuf.

Training

1. Install Conda v4.9 or latest
2. Clone this repo and update the conda.yaml file
3. conda env create -f conda.yaml
4. conda activate insurance-demo
5. python model_run.py

Download model

1. Download the model to local directory

mlflow artifacts  download -r <run-id> -d <local-path>
eg: mlflow artifacts  download -r c263bdaa-9505-4dd5-81fa-f9dbf40190fc -d ./output

Build image

1. Update the conda.yaml file in the downloaded path and add protobuf==3.19.4 in pip dependenicies
2. Run the below command to build the image

mlflow models build-docker -n <image-name> -m <local-path>/sgd-regressor
eg: mlflow models build-docker -n lucifer001/mlflow-insurnace-demo:demo1 -m output/sgd-regressor

3.Push the image

Deployment

1. Select serving image which was build in the previous step.
2. Serving Port: 8000
3. Serving Url Prefix: /invocations
4. Min CPU/Max CPU: 1
5. Min Memory/Max Memory: 5G

Prediction

1. Copy the curl command from the deployment page
2. Change the data section to

-d '{"data": [[31, 1, 22.2, 2, 1, 2]]}'