Deploy and Export a Model
Validate a trained model, test it in the playground with real inputs, try a pretrained Hugging Face model, and export a single-file standalone playground you can host anywhere.
What you will do
Take a model you have trained, put it through validation, try it with your own inputs and export a page that runs it in any browser.
Before you start
- A trained model in the open project. Train one in the Model Builder, or follow From Template to Trained Model.
- A current Chrome or Edge
Step 1: Choose the model
Open Deployment. On the Models You Trained tab, select your model from the list. If the list is empty, check that the right project is open.
Step 2: Check the modality
DLWAY reads the model's input and output type: tabular, text, image or audio in; classification, regression or text out. If it could not work them out, a selector lets you set them by hand. Make sure they match what the model really expects.
Step 3: Validate and go live
Press Validate. DLWAY loads the model and tests it with a real input before it calls the deployment live. A failure reports why, for example a shape mismatch, and the playground stays closed until it is fixed.
Step 4: Try it in the playground
Enter values that match the model:
- Tabular: one number per feature, labelled with the training column names
- Text: type a sentence
- Image: upload a picture
Press Run Inference. A classifier shows class probabilities; a regressor shows a number. Change a value and run again to see how the output moves. A good habit is to try a row from your test set whose label you know.
Step 5: Export a standalone playground
Press Export Standalone Playground. You receive one HTML file with the model embedded. Double-click it to run the same model locally, or put it on any static host. It needs no server and no build step; TensorFlow.js is loaded from a CDN when the page opens.
Step 6: Try a pretrained model
Switch to Pretrained Hugging Face Model. Paste a model ID such as Xenova/distilbert-base-uncased-finetuned-sst-2-english, choose the task (here, text classification) and load it. The first time, the model files download and are cached. Validate, then test it with a sentence. Export it too: the exported file fetches the model from the Hub when first opened.
Step 7: Packages from pipelines
If a pipeline has a Build Package step, its packages appear under Pipeline packages. Each is a browser demo and/or a model file with a hashed manifest, produced only for a model the quality gate approved. Download the zip, or open the packaged model back in the Deployment view.
Remember
A package or export is a ready-to-use file set, not a running endpoint. You decide where it is hosted.