Add interactive optimization mode

Author: ricardoV94

pytensor/configdefaults.py

@@ -515,6 +515,22 @@ def add_compile_configvars():
         in_c_key=False,
     )
 
+    config.add(
+        "optimizer_interactive",
+        "If True, we interrupt after every optimization being applied and display how the graph changed",
+        BoolParam(False),
+        in_c_key=False,
+    )
+
+    config.add(
+        "optimizer_interactive_skip_rewrites",
+        (
+            "Do not interrupt after changes from optimizers with these names. Separate names with ',"
+        ),
+        StrParam(""),
+        in_c_key=False,
+    )
+
     config.add(
         "on_opt_error",
         (

pytensor/graph/features.py

@@ -3,6 +3,7 @@
 import time
 import warnings
 from collections import OrderedDict
+from difflib import Differ
 from functools import partial
 from io import StringIO
 
@@ -563,8 +564,19 @@ def replace_all_validate(
     ):
         chk = fgraph.checkpoint()
 
+        interactive = config.optimizer_interactive
+
         if verbose is None:
-            verbose = config.optimizer_verbose
+            verbose = config.optimizer_verbose or interactive
+
+        if interactive:
+            differ = Differ()
+            bef = pytensor.dprint(
+                fgraph, file="str", print_type=True, id_type="", print_topo_order=False
+            )
+            skip_rewrites = config.optimizer_interactive_skip_rewrites.replace(
+                " ", ""
+            ).split(",")
 
         for r, new_r in replacements:
             try:
@@ -611,6 +623,22 @@ def replace_all_validate(
             print(
                 f"rewriting: rewrite {reason} replaces {r} of {r.owner} with {new_r} of {new_r.owner}"
             )
+            if interactive and str(reason) not in skip_rewrites:
+                aft = pytensor.dprint(
+                    fgraph,
+                    file="str",
+                    print_type=True,
+                    id_type="",
+                    print_topo_order=False,
+                )
+                if bef != aft:
+                    diff = list(
+                        differ.compare(
+                            bef.splitlines(keepends=True), aft.splitlines(keepends=True)
+                        )
+                    )
+                    sys.stdout.writelines(diff)
+                    input("Press any key to continue")
 
         # The return is needed by replace_all_validate_remove
         return chk

pytensor/printing.py

@@ -119,6 +119,7 @@ def debugprint(
     print_destroy_map: bool = False,
     print_view_map: bool = False,
     print_fgraph_inputs: bool = False,
+    print_topo_order: bool = True,
 ) -> Union[str, TextIO]:
     r"""Print a graph as text.
 
@@ -175,6 +176,8 @@ def debugprint(
         Whether to print the `view_map`\s of printed objects
     print_fgraph_inputs
         Print the inputs of `FunctionGraph`\s.
+    print_topo_order
+        Whether to print the toposort ordering of nodes
 
     Returns
     -------
@@ -231,7 +234,10 @@ def debugprint(
             else:
                 storage_maps.extend([None for item in obj.maker.fgraph.outputs])
             topo = obj.maker.fgraph.toposort()
-            topo_orders.extend([topo for item in obj.maker.fgraph.outputs])
+            if print_topo_order:
+                topo_orders.extend([topo for item in obj.maker.fgraph.outputs])
+            else:
+                topo_orders.extend([None for item in obj.maker.fgraph.outputs])
         elif isinstance(obj, FunctionGraph):
             if print_fgraph_inputs:
                 inputs_to_print.extend(obj.inputs)
@@ -241,7 +247,10 @@ def debugprint(
                 [getattr(obj, "storage_map", None) for item in obj.outputs]
             )
             topo = obj.toposort()
-            topo_orders.extend([topo for item in obj.outputs])
+            if print_topo_order:
+                topo_orders.extend([topo for item in obj.outputs])
+            else:
+                topo_orders.extend([None for item in obj.outputs])
         elif isinstance(obj, (int, float, np.ndarray)):
             print(obj, file=_file)
         elif isinstance(obj, (In, Out)):