Azure OpenAI Service - Microsoft Q&A
This browser is no longer supported. Upgrade to Microsoft Edge to take advantage of the latest features, security updates, and technical support.
Introduction to Azure OpenAI Service
The Azure OpenAI Service is an Azure service that provides access to OpenAI’s GPT-3 models with enterprise capabilities. It allows users to perform various tasks such as uploading batch jobs, creating structured output formats, fine-tuning models, and much more.
Batch Job Upload Issue
Users have reported issues with uploading batch jobs, where the job gets validated but completes very quickly with incomplete outputs. Some users have experienced jobs completing with only a fraction of the expected output lines, without any error notifications or warnings.
API Call Format
Users have encountered challenges when calling the API in a specific format, resulting in a "404 deployment not found" error from the Azure API side. Despite searching for examples, there is a lack of clear documentation on how to properly use or call the API.
Structured Output Format Troubleshooting
Although the structured output format with the batch API has been successful for several weeks, some users have faced issues after updating prompts and attempting to re-run tasks. This has caused discrepancies in the output results.
One user shared an example where the AI model was unable to provide answers accurately, even with detailed prompting.
Fine-Tuning Models and Deployment Quota
Users have encountered difficulties while fine-tuning models, especially when specifying paths in the training/validation files. Some have faced challenges with deploying models due to undefined quotas, despite having ample quota remaining.
Support and Deployment Challenges
Users have sought expedited support to increase their quotas for the Azure OpenAI service, as they plan to scale up their workload across multiple models. Some users have also faced deployment challenges in specific regions, such as limitations in selecting deployment types.
Data Analysis and Document Processing
Some users have expressed the need for user-level data analysis in Azure AI Studio, specifically to track individual users' usage. Others have explored the use of batch models for cost-efficient analysis of large document sets.
Technical Issues and Error Handling
There have been reports of technical issues such as errors while processing documents, parsing PDFs, and fine-tuning models with images. Users have also highlighted bugs and unexpected behaviors when combining different AI components.
Quota Changes and Service Availability
Users have raised concerns about sudden changes in quotas and service availability, leading to confusion and challenges in utilizing Azure OpenAI services effectively.
For further information and support regarding Azure OpenAI services, users are encouraged to refer to Microsoft's official documentation and support channels.
