This adds support for multiple subgraphs in some of the Python MLIR utils, in particular the _remove_tensors_from_model, strip_lcedequantize_ops and _find_int8_quantized_inputs_outputs functions.
How Has This Been Tested?
This has been tested internally in the past with networks with multiple subgraphs (i.e. a while-loop). I did not add new tests for this functionality.
What do these changes do?
This adds support for multiple subgraphs in some of the Python MLIR utils, in particular the
_remove_tensors_from_model
,strip_lcedequantize_ops
and_find_int8_quantized_inputs_outputs
functions.How Has This Been Tested?
This has been tested internally in the past with networks with multiple subgraphs (i.e. a while-loop). I did not add new tests for this functionality.
Benchmark Results
Related issue number
(internal issue only)