LogoLogo
  • Tensorleap
  • Examples
    • Semantic Segmentation
    • Image Analysis
    • Sentiment Analysis
    • MNIST Project Walkthrough
    • IMDB Project Walkthrough
  • Quickstart using CLI
  • Guides
    • Full Guides
      • MNIST Guide
        • Dataset Integration
        • Model Integration
        • Model Perception Analysis
        • Advanced Metrics
      • IMDB Guide
        • Dataset Integration
        • Model Integration
        • Model Perception Analysis
        • Advanced Metrics
    • Integration Script
      • Preprocess Function
      • Input Encoder
      • Ground Truth Encoder
      • Metadata Function
      • Visualizer Function
      • Prediction
      • Custom Metrics
      • Custom Loss Function
      • Custom Layers
      • Unlabeled Data
      • Examples
        • CelebA Object Detection (YoloV7)
        • Wikipedia Toxicity (using Tensorflow Datasets)
        • Confusion Matrix
        • CelebA Classification (using GCS)
  • Platform
    • Resources Management
    • Project
    • Dataset
    • Secret Manager
    • Network
      • Dataset Node
      • Layers
      • Loss and Optimizer
      • Visualizers
      • Import Model
      • Metrics
    • Evaluate / Train Model
    • Metrics Dashboard
    • Versions
    • Issues
    • Tests
    • Analysis
      • helpers
        • detection
          • YOLO
    • Team management
    • Insights
  • API
    • code_loader
      • leap_binder
        • add_custom_metric
        • set_preprocess
        • set_unlabeled_data_preprocess
        • set_input
        • set_ground_truth
        • set_metadata
        • add_prediction
        • add_custom_loss
        • set_visualizer
      • enums
        • DatasetMetadataType
        • LeapDataType
      • datasetclasses
        • PreprocessResponse
      • visualizer_classes
        • LeapImage
        • LeapImageWithBBox
        • LeapGraph
        • LeapText
        • LeapHorizontalBar
        • LeapImageMask
        • LeapTextMask
  • Tips & Tricks
    • Import External Code
  • Legal
    • Terms of Use
    • Privacy Policy
Powered by GitBook
On this page
  • Evaluate Model
  • Evaluate Model Properties
  • Train Model
  • Train Model Properties

Was this helpful?

  1. Platform

Evaluate / Train Model

Learn how to evaluate and train models

PreviousMetricsNextMetrics Dashboard

Last updated 3 years ago

Was this helpful?

Once the model, dataset, loss, and optimizer are set, we can perform operations that fetch the data throughout the model. These operations are Evaluate Model and Train Model.

The difference between these two are that the Evaluate Model operation does not change the state of the model (its weights and parameters), while the Train Model operation does.

Evaluate Model

The Evaluate Model operation inferences the dataset throughout the model, and collects performance metrics and metadata for each sample.

This operation is performed on models that have a state (post train weights), and used to collect metrics on the performance and metadata of the dataset. After it is finished, the collected data is presented on the Dashboard.

More than one evaluation can be performed on a model. For example, different data slices (training/validation/test) on different datasets. In this case, the metrics are added to the Dashboard.

After the model is all set, follow this procedure to evaluate the current dataset:

  1. Click the button at the top to open the Evaluate Model dialog. More info about the properties below.

  2. Select a source model from the right panel.

  3. Set the Evaluated Epoch and Subset.

  4. Set the Batch Size.

  5. Click .

Evaluate Model Properties

  • Selected Subset - Training / Validation / Test

  • Batch Size - the amount of samples to be fetched per iteration

  • Add to Dashboard - make the model visible on the current dashboard

  • Evaluated Network - the Revision Name of the network version. This is based on the model selected for evaluation (see below).

  • Evaluated Model - the Model selected for evaluation (set from the right panel)

  • Evaluated Epoch - shows the current epoch of the selected model

Train Model

If created from scratch, a model needs to be trained.

To train a model:

  2. Select the Training Model.

  3. Set the Batch Size and Number of Epochs.

  4. (Optional) Check the Early Stop with Main KPI metric and Patience Period.

  5. Enter a New Model Name.

Train Model Properties

  • Training Model - Train from Scratch / Copy Initial Weights / Continue Training.

  • Batch Size - the amount of samples to be fetched per iteration

  • Number of Epochs - the number of complete passes through the training data

  • Early Stop - Stops the training when the main KPI stops improving

  • Main KPI - Loss / Accuracy / Mean

  • Patience Period -

  • Add to Dashboard - make the model visible on the current dashboard . Checked by default.

  • Source Network - the Revision Name of the network version (set from the right panel)

  • New Model Name - the resulting Model's Name (set from the right panel)

Tensorleap can import trained and untrained models. More info at .

For more information on importing trained and untrained models into Tensorleap, see .

Click the button at the top to open the Train Model dialog. More info about the properties below.

Click .

Import Model
Import Model
Evaluate Model Dialog
Train Model Dialog