Split AugAssign

CHANGELOGS.rst

@@ -4,6 +4,8 @@ Change Logs
 0.5.0
 +++++
 
+* :pr:`83`, :pr:`85`: improves the automated rewriting of control flow (test)
+
 0.4.4
 +++++
 

_unittests/ut_torch_export_patches/test_patch_module.py

@@ -349,6 +349,21 @@ def forward(self, x, y):
         self.assertEqualAny(expected, ep.module()(x, y))
         self.assertEqualAny(expected_, ep.module()(-x, y))
 
+    def test_rewrite_PLBartEncoderLayer(self):
+        from transformers.models.plbart.modeling_plbart import PLBartEncoderLayer
+
+        rewritten = transform_method(PLBartEncoderLayer.forward, verbose=self.verbose)
+        self.assertIn(
+            (
+                "torch.cond(hidden_states.dtype == torch.float16 and "
+                "(torch.isinf(hidden_states).any() or torch.isnan(hidden_states).any()), "
+                "branch_cond_then_1, branch_cond_else_1, [hidden_states])"
+            ),
+            rewritten.code,
+        )
+        print()
+        print(rewritten.code)
+
 
 if __name__ == "__main__":
     unittest.main(verbosity=2)

onnx_diagnostic/helpers/config_helper.py

@@ -2,7 +2,7 @@
 import importlib
 import inspect
 import re
-from typing import Any, Dict, Optional, Tuple, Union
+from typing import Any, Callable, Dict, Optional, Tuple, Union
 import transformers
 
 
@@ -42,8 +42,16 @@ def update_config(config: Any, mkwargs: Dict[str, Any]):
         setattr(config, k, v)
 
 
-def _pick(config, *atts):
+def _pick(config, *atts, exceptions: Optional[Dict[str, Callable]] = None):
     """Returns the first value found in the configuration."""
+    if (
+        exceptions
+        and hasattr(config, "architectures")
+        and len(config.architectures) == 1
+        and config.architectures[0] in exceptions
+    ):
+        excs = exceptions[config.architectures[0]]
+        return excs(config)
     for a in atts:
         if isinstance(a, str):
             if hasattr(config, a):

onnx_diagnostic/tasks/text2text_generation.py

@@ -164,6 +164,11 @@ def random_input_kwargs(config: Any) -> Tuple[Dict[str, Any], Callable]:
                 ("decoder_attention_heads", "encoder_attention_heads"),
             ),
         )
+    # exceptions = {
+    #     "PLBartForConditionalGeneration": (
+    #         lambda c: c.encoder_attention_heads + c.decoder_attention_heads
+    #    )
+    # }
     kwargs = dict(
         batch_size=2,
         sequence_length=30,
@@ -181,6 +186,7 @@ def random_input_kwargs(config: Any) -> Tuple[Dict[str, Any], Callable]:
                 "num_key_value_heads",
                 "num_heads",
                 (sum, "encoder_attention_heads", "decoder_attention_heads"),
+                # exceptions=exceptions,
             )
         ),
         encoder_dim=512 if config is None else _pick(config, "n_positions", "d_model"),

onnx_diagnostic/torch_export_patches/patch_module.py

@@ -1,4 +1,5 @@
 import ast
+import copy
 import inspect
 import types
 import textwrap
@@ -341,11 +342,34 @@ def visit(self, node):
         return node
 
 
+class _SelectiveAssignNormalizer(ast.NodeTransformer):
+    def visit_If(self, node):
+        self.generic_visit(node)
+        node.body = [self._transform_if_needed(stmt) for stmt in node.body]
+        node.orelse = [self._transform_if_needed(stmt) for stmt in node.orelse]
+        return node
+
+    def _transform_if_needed(self, stmt):
+        if isinstance(stmt, ast.AugAssign):
+            return ast.Assign(
+                targets=[stmt.target],
+                value=ast.BinOp(left=copy.deepcopy(stmt.target), op=stmt.op, right=stmt.value),
+            )
+        if isinstance(stmt, ast.AnnAssign) and stmt.value is not None:
+            return ast.Assign(targets=[stmt.target], value=stmt.value)
+        return self.visit(stmt)
+
+
 def inplace_add_parent(tree: "ast.Node"):
     """Adds parents to an AST tree."""
     _AddParentTransformer().visit(tree)
 
 
+def normalize_assignment_in_test(tree: "ast.Node"):
+    """Split AugAssign into BinOp and Assign to simplify whatever comes after."""
+    _SelectiveAssignNormalizer().visit(tree)
+
+
 def transform_method(
     func: Callable,
     prefix: str = "branch_cond",
@@ -451,6 +475,7 @@ def forward(self, x, y):
         skip_objects=modules,
         args_names=set(sig.parameters),
     )
+    normalize_assignment_in_test(tree)
     inplace_add_parent(tree)
     new_tree = transformer.visit(tree)
     if verbose > 1:

onnx_diagnostic/torch_models/hghub/hub_data.py

@@ -97,6 +97,7 @@
     Phi3ForCausalLM,text-generation
     PhiForCausalLM,text-generation
     Pix2StructForConditionalGeneration,image-to-text
+    PLBartForConditionalGeneration,text2text-generation
     PoolFormerModel,image-feature-extraction
     PvtForImageClassification,image-classification
     Qwen2ForCausalLM,text-generation