[Transform] add apply_repeated, add some docs on adding transforms

src/finn/transformation/__init__.py

+"""
+Guide to writing FINN transformations
+-------------------------------------
+
+* Your transformation should take in an ONNX model, and return a tuple with
+ (transformed_model: ModelProto, model_was_changed: Bool)
+* The original model should not be modified, use e.g. copy.deepcopy() if you
+  want to work on a copy of the graph for modifications.
+* model_was_changed indicates whether your transformation made any changes to
+  the model. If you know your transformation needs to be called only once and
+  repeated calls have no further effect, you can return False even if the model
+  was changed.
+* You MUST return model_was_changed=False at some point when your transformation
+  is called multiple times, otherwise apply_repeated() will loop infinitely.
+* If you cannot guarantee that the transformation will reach a fixed point,
+  you must declare this and return only the transformed model instead of a tuple.
+"""

src/finn/transformation/general.py

@@ -13,6 +13,20 @@ def give_unique_names(model):
     return new_model
 
 
+def apply_repeated(model, transform):
+    """Applies given transform repeatedly until no more changes can be made.
+    Transform must return (transformed_model, model_was_changed)."""
+    transformed_model = model
+    model_was_changed = True
+    while model_was_changed:
+        (transformed_model, model_was_changed) = transform(transformed_model)
+    return transformed_model
+
+
+# TODO consider making a wrapper for ONNX model and make the below functions
+# members - they aren't really proper transformations
+
+
 def get_tensor_shape(model, tensor_name):
     """Returns the shape of tensor with given name, if it has ValueInfoProto."""
     graph = model.graph