The TT-Forge FE is a graph compiler designed to optimize and transform computational graphs for deep learning models, enhancing their performance and efficiency.
In order to generate unique op tests, we need to have knowledge about input shapes for each op during codegen.
Previously, we only needed to know details about generated op names,
inputs, params, etc. However, these are not enough in order to support
logic for splitting models across different op tests
Fix #592