This PR fixes an issue where runtime values expressed in scientific notation were not being rounded correctly.
Changes
Updated the regex pattern in the extract_single function to properly match scientific notation values.
Testing
Tested with various runtime values, including those in scientific notation, to ensure correct rounding and display in the output table.
Additional Notes
Checking the data from the MIL file, for example: "operation: "fp64",
runtimes: ["mps_graph": 7.0123111500459555e-06, "classic_cpu": 8.414773757700023e-06],
selectedBackend: "classic_cpu",
name: nil,
validationMessages: ["ane": "Tensor buffer live-in not created: Unsupported tensor data type: int32\n"]"
In these cases, the values were displayed as:
GPU: 7.0123
CPU: 8.4147
After the fix, the values are correctly rounded to:
FL33TW00D
commented
4 weeks ago
Summary
This PR fixes an issue where runtime values expressed in scientific notation were not being rounded correctly.
Changes
extract_singlefunction to properly match scientific notation values.Testing
Tested with various runtime values, including those in scientific notation, to ensure correct rounding and display in the output table.
Additional Notes
Checking the data from the MIL file, for example: "operation: "fp64", runtimes: ["mps_graph": 7.0123111500459555e-06, "classic_cpu": 8.414773757700023e-06], selectedBackend: "classic_cpu", name: nil, validationMessages: ["ane": "Tensor buffer live-in not created: Unsupported tensor data type: int32\n"]"
In these cases, the values were displayed as:
After the fix, the values are correctly rounded to: