add graph_net/torch/unittest/test_ast_renamer.py

Author: roll-away

graph_net/sample_pass/ast_graph_variable_renamer.py

@@ -234,7 +234,7 @@ def _filter_delete_target(self, target):
         if isinstance(target, ast.Tuple):  # del (a, b)
             kept_elts = [e for e in target.elts if not self._is_input_or_weight_var(e)]
             return ast.Tuple(elts=kept_elts, ctx=ast.Del()) if kept_elts else None
-        elif not self._is_input_or_weight_var(target):
+        elif not self._is_input_or_weight_var(target):  # del a
             return target
         else:
             pass

graph_net/tools/generate_subgraph_dataset.sh

@@ -113,10 +113,11 @@ function rename_decomposed_subgraph() {
         --model-path-list ${range_decomposed_subgraph_list} \
         --handler-config=$(base64 -w 0 <<EOF
 {
-    "handler_path": "$GRAPH_NET_ROOT/graph_net/torch/graph_variable_renamer.py",
-    "handler_class_name": "GraphVariableRenamer",
+    "handler_path": "$GRAPH_NET_ROOT/graph_net/sample_pass/ast_graph_variable_renamer.py",
+    "handler_class_name": "AstGraphVariableRenamer",
     "handler_config": {
         "device": "cuda",
+        "try_run": false,
         "resume": ${RESUME},
         "model_path_prefix": "${RANGE_DECOMPOSE_OUTPUT_DIR}",
         "data_input_predicator_filepath": "$GRAPH_NET_ROOT/graph_net/torch/constraint_util.py",
@@ -221,10 +222,11 @@ function rename_fusible_subgraph() {
         --model-path-list ${fusible_subgraph_list} \
         --handler-config=$(base64 -w 0 <<EOF
 {
-    "handler_path": "$GRAPH_NET_ROOT/graph_net/torch/graph_variable_renamer.py",
-    "handler_class_name": "GraphVariableRenamer",
+    "handler_path": "$GRAPH_NET_ROOT/graph_net/sample_pass/ast_graph_variable_renamer.py",
+    "handler_class_name": "AstGraphVariableRenamer",
     "handler_config": {
         "device": "cuda",
+        "try_run": false,
         "resume": ${RESUME},
         "model_path_prefix": "${FUSIBLE_SUBGRAPH_SAMPLES_DIR}",
         "data_input_predicator_filepath": "$GRAPH_NET_ROOT/graph_net/torch/constraint_util.py",