Guides

Introduction

To get started with Tensorleap, we provided a few step-by-step guides, describing an integration of datasets and models, in addition to various analyses.

Integrating the Dataset and Model

The first step to getting started with Tensorleap is to integrate your datasets and models. This can be done through the Integration Script, which defines how to read each sample and metadata.

There are multiple ways to integrate your model:

• Pre-Trained - You can import a pre-trained model through the Import Model feature or through the CLI Model Integration. After the model and dataset are imported, you must Evaluate the Model (inference) in order to get the metrics and analyses.

• Built from Scratch - Some users prefer building the model and altering it in the Network View. Then the model is Trained within the platform.

In the full-guides below, we present both cases for each use-case.

Tensorleap is now available for a free trial! For more info, see Try Tensorleap for Free.

Full Guides

Guides are designed to get you started quickly with Tensorleap.

Choose one to get started:

• MNIST - Image Classification - integration, training, analysis and metrics.

Sample's Details

Fetch Similars

Similar Samples Heat Map

• IMDB - Semantic Text classification - integration, training, analysis and metrics.

Heat map for Negative Output

Metrics Dashboard

Dataset Script

Guide

A guide about the Dataset Script can be found here.

Examples

• Wikipedia Toxicity (using Tensorflow Datasets)
• CelebA (using Google Cloud Storage)

The Reference documentation for the Tensorleap user interface (UI) and Command Line Interface (CLI) may prove useful when going through these tutorials.