Experiment flow¶
The flow of an H2O Hydrogen Torch experiment can be described and summarized in the following sequential steps:
In the below sections, each step, in turn, is explained in detail.
Step 1: Import dataset¶
H2O Hydrogen Torch supports diverse problem types in computer vision and natural language. As a requirement, H2O Hydrogen Torch requires the dataset for an experiment to be preprocessed to follow a certain dataset format for the problem type the experiment aims to solve.
-
To learn how to import your own preprocessed dataset, see Data Connectors.
-
To learn more about dataset formats, see Dataset Formats.
-
H2O Hydrogen Torch offers access to an array of preprocessed datasets to highlight all supported problem types, but in particular, they are offered as a reference to understand better the format a dataset needs to follow for a given experiment problem type. To learn how to access one of the preprocessed datasets in H2O Hydrogen Torch, see Preprocessed datasets.
Step 2: Train experiment¶
Each problem type offers several hyperparameters that you can adjust for model training. H2O Hydrogen Torch also provides the ability to enable Grid Search for a given experiment to tune and experiment on specific hyperparameters.
-
To learn about all the supported problem types, see Supported problem types.
-
To learn about the settings available for each supported problem type, see Overview: Experiment settings.
-
To learn how to enable Grid Search, see Grid search.
Through simple interactive graphs, data scientists can understand the impact of selected hyperparameter values during and after the model's training process.
- To learn about available simple and interactive graphs, see Charts.
Step 3: Inspect & Deploy¶
After understanding and inspecting a built model, you can score new data on a built model (experiment) that generates downloadable predictions. You can also deploy a built model easily to any external Python environment or directly to H2O MLOps. You can also use the H2O Hydrogen Torch UI to score new data.
-
To learn how to deploy built models to external Python environments, see Python Environment.
Note
H2O Hydrogen Torch offers a standalone Python Scoring Pipeline running on Linux-based systems.
-
To learn how to deploy built models directly to H2O MLOps, see H2O MLOps.
-
To learn how to score on new data using the H2O Hydrogen Torch UI, see H2O Hydrogen Torch UI (Predict Data).
- Submit and view feedback for this page
- Send feedback about H2O Hydrogen Torch to cloud-feedback@h2o.ai