FAQ
How to view the cause of failure after model training fails?
Answer: When model training fails, you can find the model that failed training in Studio-Models-Training Failed, click to enter the training details page, view the task status, and check the logs. After locating the problem, retrain or start new training.
When uploading a file dataset, what are the requirements for the uploaded file?
Answer: Smart Vision currently supports two file formats: txt and json. The file size needs to be within 30M. Please upload the dataset file that meets the requirements on the Studio.
When creating an enterprise knowledge base, why can't I select high-quality indexing mode in Text Preprocessing and Cleaning settings?
Answer: The high-quality indexing mode relies on calling the embedded interface of Microsoft's Azure-GPT-3.5 for processing. You need to complete the deployment of the model in Studio-Models (if the studio does not have the model, you need to add the model from Model Store first) before you can use this indexing method.
When orchestrating Agent applications, why can't the Retrieval-Augmented Generation(RAG) select sentence window search?
Answer: The selectable RAG is related to Text Preprocessing and Cleaning settings rules of the mounted knowledge base. If the mounted enterprise knowledge base selects the sentence window, RAG can select the sentence window retrieval when orchestrating AI applications. In addition, if the mounted enterprise knowledge base selects the AutoMergingTextSplitter splitter, RAG can select automatic merge retrieval when orchestrating AI applications.
When uploading files in Agent application, why doesn't the application summarize like the enhanced application?
Answer: Unlike the enhanced application, which automatically summarizes the understanding after uploading files, the agent application does not automatically use the uploaded files as context. Instead, it needs to be combined with tool calling to call the uploaded files as context for processing.
I added a tool from Tool Store to my Studio, but I still can't use it. What should I do?
Answer: Some tools provided by Tool Store require authorization before use. You can authorize unauthorized tools in the Studio-Applications-Tools, so that they can be used directly when orchestrating AI applications or workflows. In addition, if you have not set up authorization in advance, you can also authorize when adding tools in application/workflow orchestration.
I added a workflow during application orchestration. After I modified the workflow, the workflow in the application orchestration was not updated. What should I do?
Answer: After the workflow is orchestrated and debugged, it needs to be configured before it can be added to the AI application. After the workflow is modified, it needs to be reconfigured to synchronize the modified content to the application to ensure the effect of your application.
