- Start any of the jupyterlab notebook from the IDE tab.
- Once running, click the jupyterlab icon to launch jupyterlab
- Open terminal in Jupyterlab and run
> wget https://raw.githubusercontent.com/oneconvergence/dkube-examples/tensorflow/clinical_reg/pipeline_withslurm.ipynb
- Open pipeline.ipynb and run cells to generate the tar file and create run.
- Download the tar file by right-clicking on it(optional).
- Upload the tar file into the DKube pipeline UI(optional).
- Go to Model Catalog and from model version click deploy model.
- Give name.
- Serving image: default
- Deployment type: Test
- Select transformer
- Transformer script:
clinical_reg/transformer.py
- Transformer script:
- Deploy using: CPU and Submit.
- Deployed Model will be available in Model Serving.
- Download the csv data file cli_inp.csv and any sample image from images folder from sample_data/images
- open https://{your-dkube-url}/inference,
- In DKube UI, once the pipeline run has completed, navigate to ‘Deployments’ on the left pane
- Copy the ‘Endpoint’ URL in the row using the clipboard icon
- Enter the Endpoint URL into the Model Serving URL field of inference page,
- Copy the token from ‘Developer Settings’ and paste into ‘Authorization Token’ box
- Select Model Type as ‘Regression’ on the next dropdown selection
- Click ‘Upload Image’ to load image from [1], ‘Upload File’ to load csv from [1]
- Click ‘Predict’ to run Inference.
- Go to IDE section
- Create Notebook
- Give a name
- Code: regression
- Framework : Tensorflow
- Framework version : 2.3
- Datasets: - i. clinical Mount point: /opt/dkube/input/clinical - ii. images Mount point: /opt/dkube/input/images - iii. rna Mount Point: /opt/dkube/input/rna i3. Submit
- Open workflow.ipynb from location
workspace/regression/clinical_reg
- Run cells and wait for output (In case of running the notebook second time, restart the kernel)
- Delete if workflow.py is already there and export the workflow notebook as executable.
- Upload it into Juyterlab.
- Make changes in py file, comment/remove the following line numbers: -i. 239-240 -ii. 268 -iii. 435-end
- Save and commit the workflow.py
- Create a model named workflow with source none.
- Create training run using workflow.py
- Give a name
- Code: regression
- Framework : Tensorflow
- Framework version : 2.0.0
- Startup command: python workflow.py
- Datasets:
- i. clinical Mount point: /opt/dkube/input/clinical
- ii. images Mount point: /opt/dkube/input/images
- iii. rna Mount Point: /opt/dkube/input/rna
- Output model: workflow, Mount point : /opt/dkube/output