In progress experiments & conference submission

• Running inference for smpadhy/smart-ml-hpc-hug0 : https://colab.research.google.com/drive/1Nxai8OnLU5LUJC6tUn23Ukm-KVipOrXd?usp=sharing

  • Note: stuck on model.predict() as it gives error
    • The feature names should match those that were passed during fit. Feature names must be in the same order as they were in fit. when input is formatted as pandas dataframe
    • X does not have valid feature names, but StandardScaler was fitted with feature names when input is formatted as numpy array
    • Need to know how model.fit() was done and maybe explore model formatting other than .pkl

• Trial of OpenClip model usage for inference:

  Google Colab

• Experiments using HF optimum to use ONNX model for inference: https://colab.research.google.com/drive/1YdktJw0CT_oUW2YBDsVSMxNuLWNBmyC5?usp=sharing

  • HF docs: https://huggingface.co/docs/optimum/onnxruntime/usage_guides/pipelines

• Demo submission for Science Gateways 2024:

  Overleaf, Online LaTeX Editor

ML Hub API testing

Running ML Hub’s hub API

Choose how you want to have a copy of ML Hub’s models API docker image to run locally:

• Option 1: Build the image locally (for those developing the API + UI)
• Option 2: Pull the latest image from docker hub (recommended for UI developer)
• Deploying ML Hub

Postman files for ML Hub API testing

• Example API calls to ML Hub models , datasets , and inference APIs - exported in JSON format. Prior to using the endpoints on the API, hit the /auth endpoint to authenticate.

• Collections of endpoints for ML Hub deployed in pods

  MLhub-demo.postman_collection.json

• Collection of endpoints for ML Hub models and datasets API for local testing

  hub-test.postman_collection.json

• Collection of endpoints for ML Hub inference API for local testing

  inference-test.postman_collection.json

Postman Workspace for Tapis Machine Learning Hub API guide

(Postman workspace has been shared with Nathan and Alex):

1. hub-test is used for local deployment/testing of Models & Datasets API
2. inference-test is used for local deployment/testing of Inference API
3. MLhub-demo is used for Models, Datasets and Inference API that is deployed on Pods service