Giskard-AI/community-content

How to upload a Keras ML model with Giskard?

Opened this issue · 3 comments

jmsquare commented
How to upload a Keras ML model with Giskard?
olujerry commented

Outline

Introduction

  • Briefly introduce the concept of model testing and the importance of quality assurance in ML applications.

  • Introduce Giskard as an open-source testing framework for ML models.

  • Mention the focus of the article: step-by-step guide to uploading a Keras ML model with Giskard.

Preparing for Model Upload

  • Install Giskard: Provide instructions on how to install Giskard using pip.

  • Prepare the Keras Model: Explain the requirement of having a trained Keras model saved as a file.

Importing Giskard and Loading the Model

Import Giskard Library: Showcase the import statement for the Giskard library in Python.
Load the Keras Model: Describe how to load a pre-trained Keras model into memory using appropriate methods

Uploading the Model to Giskard

  • Utilize the upload_model() Function: Explain the usage of the `giskard.upload_model()` function to upload the Keras model.
  • Specify Model Details: Discuss any additional parameters that need to be provided, such as model name or version.

Testing and Validation with Giskard

  • Overview of Giskard Features: Highlight the various testing and validation functionalities offered by Giskard.

  • Vulnerability Detection: Explain how Giskard can detect vulnerabilities in the uploaded model.

  • Automatic Test Generation: Discuss the generation of relevant tests based on detected vulnerabilities.

  • Leveraging Giskard Catalog: Explain how to utilize the Giskard catalog for quality assurance best practices.

Conclusion

  • Recap the process: Summarize the steps involved in uploading a Keras ML model with Giskard.
  • Highlight the benefits: Mention the advantages of using Giskard for model testing and validation.
  • Encourage further exploration: Suggest exploring the extensive capabilities of Giskard beyond model upload.
alexcombessie commented

@luca-martial what do you think?

luca-martial commented

@olujerry has registered through the community writing program, we've taken the conversation offline