Sheldon,
It seems you've already taken some steps to troubleshoot the issue with integrating your custom language model into the translation service. Here are a few additional steps you might consider to resolve the "The category parameter is invalid" error:
- Verify Category ID: Double-check the category ID you're using to ensure it's exactly correct. It should match the custom category ID provided by the translation service when you created and trained your custom model.
- Access Permissions: Ensure that the user executing the flow has the necessary permissions to use the custom language model. You mentioned sharing the model with the user, but also confirm that the user has the appropriate role and access rights within the translation service.
- API Endpoint: Verify that you're using the correct API endpoint for the translation service that corresponds to the region where your custom model is deployed. Even if you've published the model to multiple regions, you need to use the endpoint specific to the region you're targeting in your flow.
- API Version: Check if the translation service requires a specific API version to use custom models. The API version is typically part of the endpoint URL, and using an outdated or incorrect version can lead to errors.
- Authentication: Confirm that the authentication method and credentials used in the flow are correct. This might involve checking API keys or tokens, and ensuring they have not expired.
- Subscription Key: Make sure you're using the correct subscription key associated with the service where your custom model is hosted. Sometimes, different services or resources within the same cloud provider require different keys.
- Model Deployment Status: Check the status of your custom model to ensure it's fully trained and deployed. Sometimes, a model might still be in a pending or incomplete state, which could cause errors when trying to use it.
- Service Quotas: Verify that you haven't exceeded any service quotas or limits that might prevent the custom model from being invoked.
- Error Logs: Look at detailed error logs if available. Sometimes, cloud service providers offer more detailed logs through their dashboard or a separate logging service.
If all else fails, please share the error logs to us if there is no confidential information. There may be specific steps or configurations required by the service that are not immediately apparent.
Remember to also test the custom model directly using a simple REST client or the service's test interface to ensure the model is working independently of the Power Automate flow. This can help isolate the issue to either the model itself or the way it's being called in the flow.
I hope this helps.
Regards,
Yutong