You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Because PatternTransformation, i.e. most of the data flow transformations, including MapFusion are called through the apply() function, that lacks the pipeline_results argument that apply_pass() has.
To make this argument available to such passes, PatternTransformation has this member _pipeline_results, that is set to the value of pipeline_results argument that was passed to the apply_pass() method.
However, this member is only set during the execution of apply(), but when can_be_applied() is run the value is None.
I do not see a reason why it has to be that way and would propose to change this, such that _pipeline_results is also set during can_be_applied().
The text was updated successfully, but these errors were encountered:
Because
PatternTransformation
, i.e. most of the data flow transformations, includingMapFusion
are called through theapply()
function, that lacks thepipeline_results
argument thatapply_pass()
has.To make this argument available to such passes,
PatternTransformation
has this member_pipeline_results
, that is set to the value ofpipeline_results
argument that was passed to theapply_pass()
method.However, this member is only set during the execution of
apply()
, but whencan_be_applied()
is run the value isNone
.I do not see a reason why it has to be that way and would propose to change this, such that
_pipeline_results
is also set duringcan_be_applied()
.The text was updated successfully, but these errors were encountered: