Developer information from closed issues

Constant propagation pass is done in model floating point precision

• Original issue: https://github.com/ROCm/AMDMIGraphX/issues/2302
• Consider changing constant propagation to be done at a higher precision to ensure accuracy.
  • This is more important for datatypes like fp16.