Most reporting features were moved from the ttnn/database.py and ttnn/decorators.py files into the ttnn/graph_report.py tool, but the tensor comparison feature was not updated. Parts of the reporting TTNN report feature were left in those files because of the comparison feature, which is a Python level feature. But there is now duplicated or substantially similar report code to create the database, and to create and populate the tables, since it's done two different ways, depending on whether tensor comparison is being used on not.
We need to refactor the comparison feature in TTNN so that it saves its data to a JSON file, and leave creating and populating the database to the the ttnn.graph_report tool. It will need to be updated to read the tensor comparison data from the JSON file, if present.
Most reporting features were moved from the
ttnn/database.pyandttnn/decorators.pyfiles into thettnn/graph_report.pytool, but the tensor comparison feature was not updated. Parts of the reporting TTNN report feature were left in those files because of the comparison feature, which is a Python level feature. But there is now duplicated or substantially similar report code to create the database, and to create and populate the tables, since it's done two different ways, depending on whether tensor comparison is being used on not.We need to refactor the comparison feature in TTNN so that it saves its data to a JSON file, and leave creating and populating the database to the the
ttnn.graph_reporttool. It will need to be updated to read the tensor comparison data from the JSON file, if present.