Summary:
Sometimes we only want to generate a replacement for a matched pattern
once we know some information about the nodes in the pattern.
So far, we have found this the most useful to do matches based on specific
shapes of tensors flowing into functions.
Use a callback function similar to `match_filters`. By default this isn't used.
Had to make `replacement` a None-able parameter because Callable was
already used to detect a case where a graph needed to be traced.
Differential Revision: D62412628
Pull Request resolved: https://github.com/pytorch/pytorch/pull/135553
Approved by: https://github.com/SherlockNoMad
For cases where the pattern graph matches on x number of arguments, but the matching graph omits some of these arguments (by using the default values instead), right now SubgraphMatcher fails because these graphs have a different number of arguments. So instead in the case where we see the pattern/replacement nodes have different number of arguments, we will add the default values onto whichever argument set is lacking arguments.
Note this support is only for when we are matching targets that are instances of OpOverload, which have a schema and default values tied to them.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/99431
Approved by: https://github.com/jerryzh168
Fixes#98974
When `torch.fx.subgraph_rewriter._replace_pattern` is used to remove nodes from a graph, if there are two adjacent matches then after the first removal, the nodes in `InternalMatch.nodes_map` and `placeholder_nodes` become outdated because they contain nodes that were just removed from the graph.
This fix is to update the `match.nodes_map` and `match.placeholder_nodes` using the node changes stored in `match_changed_node`.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/99039
Approved by: https://github.com/angelayi
Pattern replacement behaves incorrectly when the replacement pattern maps inputs to outputs (such a pattern can be used to replace redundant code). However, current code in `torch.fx.subgraph_rewriter._replace_pattern` causes the list of replacement nodes to include the entire graph before that node, resulting in an exponential slowdown due to recursive calls traversing the entire graph multiple times.
The proposed fix is to add a check in `_replace_pattern` to prevent the call to `get_replacement_nodes`:
```python
for ret_node in copied_returning_nodes:
if ret_node in match.placeholder_nodes:
replacement_nodes.append(ret_node)
else:
get_replacement_nodes(ret_node)
```
Fixes#97817
Pull Request resolved: https://github.com/pytorch/pytorch/pull/97903
Approved by: https://github.com/angelayi
Summary: Modified replace_pattern in the subgraph rewriter to return a list of pairs of matches along with their corresponding replacement nodes in the modified graph (`List[Tuple[Match, List[Node]]]`). This allows us to easily modify the replaced nodes, including setting the metadata.
Test Plan: CI
Differential Revision: D41737056
Pull Request resolved: https://github.com/pytorch/pytorch/pull/90244
Approved by: https://github.com/SherlockNoMad
Summary:
att, this is experimental api so not marking it as bc-breaking.
The match will be accepted only if all the filters in the list passes.
Changing the filter arg to be list also allows us to pass in empty list that means no filter, which makes user code cleaner.
Test Plan:
python test/test_fx.py -k test_replace_pattern_with_filters
Reviewers:
Subscribers:
Tasks:
Tags:
Differential Revision: [D40810943](https://our.internmc.facebook.com/intern/diff/D40810943)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/87998
Approved by: https://github.com/SherlockNoMad
Summary:
att, this is experimental api so not marking it as bc-breaking.
The match will be accepted only if all the filters in the list passes.
Changing the filter arg to be list also allows us to pass in empty list that means no filter, which makes user code cleaner.
Test Plan:
python test/test_fx.py -k test_replace_pattern_with_filters
Reviewers:
Subscribers:
Tasks:
Tags:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/87257
Approved by: https://github.com/SherlockNoMad
This PR introduces an interface for user defined function that filters the matches in SubgraphRewriter. The function will have the following signature.
callable(match: InternalMatch, original_graph: Graph, pattern_graph: Graph) -> bool
This filter is applied after SubgraphMatcher returns the matches, and before replacement takes place.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/86430
Approved by: https://github.com/jerryzh168
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/51532
- Change output of `replace_pattern` to `List[Match]` reflecting the
pattern(s) matched in the original graph
- Ensure that all Callables (not just FunctionType objects) work with
the rewriter
- Fix incorrect matching in degenerate case (`test_subgraph_rewriter_correct_output_replacement`)
- Verify that pattern matching works when pattern and original graph are
the same
Test Plan: Imported from OSS
Reviewed By: jamesr66a
Differential Revision: D26193082
Pulled By: ansley
fbshipit-source-id: 7f40c3862012a44adb88f403ade7afc37e50417f
