These known issues have been identified in Data Labeling.
Prefix matching a file name causes Needs Attention state
- The prefix for a file name is used for "starts with" searches and not for exact matches. If any file has a name that matches the prefix exactly, the record generation fails, and the dataset goes to a state of "Needs Attention" after the Generate Records action finishes.
- Change the prefix or file name in object storage. Upload the new files to the dataset and generate the records again.
List Records only returns 1000 entries
List Recordsreturns no more than 1000 entries in a single API call.
- Workaround 1
- If you need to fetch more than 1000 records, use multiple API calls.
- Workaround 2
- If you are looking for a particular record, include the name of it using the
- Workaround 3
- If you want a count of the records, you can get it by calling
Summarize RecordsAPI for that compartment and dataset.
Labels cannot be seen if you zoom in too much
- For entity extraction classification, if you zoom in too much, the label and text overlap.
- Zoom out until you can clearly see the labels.
Cannot indicate nothing to label with entity extraction or object detection
- When you classify a piece of text using entity extraction, or classify an image using object detection, you cannot indicate that there is nothing to label. By not labeling such a piece of text or image, it is considered an unlabeled record, not a record with nothing to label.
- We're working on a resolution.
Files with a file extension in capital letters are not recognized
- Files with a file extension in capital letters are not recognized.
- Use lower case for the file extensions. For example,