Hey guys,
When I run Ansor from an ONNX model, it generates for me a set of results, and each row of those results is a task.
I would like to know what determines the layer I am running in this task, the workload_key or just the “hash” in the task?
For example below for resnet18 the first task I received was this:
{“i”: [["[“ee21c953c075817832b0e81ff3155d5f”, [1, 2, 28, 28, 64], [16, 2, 1, 1, 64, 16], [1, 16, 1, 1, 16], [1, 16, 14, 14, 16]]", "llvm -keys=cpu ", [8, 64, 64, 0, 0, 0, 0, 0], “”, 2, []], [[], [[“SP”, 2, 0, 1, [1, 1, 1], 1], [“SP”, 2, 4, 16, [8, 2, 1], 1], [“SP”, 2, 8, 14, [1, 1, 2], 1], [“SP”, 2, 12, 14, [2, 7, 1], 1], [“SP”, 2, 16, 16, [1, 4, 4], 1], [“SP”, 2, 20, 128, [16], 1], [“SP”, 2, 22, 1, [1], 1], [“SP”, 2, 24, 1, [1], 1], [“RE”, 2, [0, 4, 8, 12, 16, 1, 5, 9, 13, 17, 20, 22, 24, 2, 6, 10, 14, 18, 21, 23, 25, 3, 7, 11, 15, 19]], [“FSP”, 4, 0, 0, 1], [“FSP”, 4, 2, 1, 1], [“FSP”, 4, 4, 2, 1], [“FSP”, 4, 6, 3, 1], [“FSP”, 4, 8, 4, 1], [“RE”, 4, [0, 2, 4, 6, 8, 1, 3, 5, 7, 9]], [“CA”, 2, 4, 4], [“FU”, 4, [0, 1, 2, 3, 4]], [“AN”, 4, 0, 3], [“PR”, 2, 0, “auto_unroll_max_step$16”], [“AN”, 2, 25, 2], [“AN”, 4, 5, 2]]]], “r”: [[0.000183951, 0.000190844, 0.000210957], 0, 0.587965, 1700682778], “v”: “v0.6”}
what determines the layer signature is the workload_key:
“[“ee21c953c075817832b0e81ff3155d5f”, [1, 2, 28, 28, 64], [16, 2, 1, 1, 64, 16], [1, 16, 1, 1, 16], [1, 16, 14, 14, 16]]”, "llvm -keys=cpu ", [8, 64, 64, 0, 0, 0, 0, 0], “”, 2, []
or just this “hash” code which is the name of the workload_key:
ee21c953c075817832b0e81ff3155d5f
Thank you for your attention.
Michael