Go to Studio Training. Create a new Project with a Training type.
- Upload the dataset.
Please ensure that the minimal size of a dataset for Microsoft is 12000 unique test segments. We are trimming 10% of the amount of the segment for testing.
3. Select the Microsoft connected account in the providers' list.
If a user is not logged in to the Microsoft account, they will see the following warning. It means a user must pass Microsoft Single-Sign-On flow to get needed auth_tokens.
4. Click on the icon and input the needed credentials saved during the steps above.
client_id - ID of the registered app
tenant_id - ID of the azure scope
translator_id - ID of the custom translator app
5. The user will be redirected to the Microsoft Single-Sign-On Authorization page.
6. After that, the user will see the guard icon. It means that the user can make Microsoft authenticated requests and work with Microsoft Custom Translator API
7. Then, if the training was started successfully, the models' list will be available
The user must be logged in the Microsoft to get the actual status of models training. If the user doesn’t have access to Microsoft, they will get an authentication warning as above.
8. When the training is succeeded for the model, the status of the model will be changed on the Deploying
9. When the model is deployed, the status will be changed to Completed. After that, the user can translate texts with a custom model.