Fix output_dir of dtype generalization pass

graph_net/test/dtype_gen_test.sh

@@ -43,25 +43,3 @@ EOF
 )
 
 
-# Step 3: Valiation
-SUCCESS_CNT=0
-FAIL_CNT=0
-
-for model_path in "$OUTPUT_DIR"/*; do
-    echo "[VALIDATE] $model_path"
-
-    output=$(python -m graph_net.torch.validate \
-        --model-path "$model_path" 2>&1)
-
-    if echo "$output" | grep -q "Validation success, model_path="; then
-        echo "SUCCESS"
-        ((SUCCESS_CNT++))
-    else
-        echo "FAIL"
-        ((FAIL_CNT++))
-    fi
-done
-
-echo "===================="
-echo "SUCCESS $SUCCESS_CNT"
-echo "FAIL    $FAIL_CNT"

graph_net/torch/sample_pass/dtype_generalizer.py

@@ -42,6 +42,8 @@
 from graph_net.sample_pass.sample_pass import SamplePass
 from graph_net.sample_pass.resumable_sample_pass_mixin import ResumableSamplePassMixin
 
+from graph_net.hash_util import get_sha256_hash
+
 # Weights that must remain float32 for numerical stability
 FLOAT32_PRESERVED_WEIGHTS = {
     "running_mean",
@@ -296,7 +298,7 @@ def sample_handled(self, rel_model_path: str) -> bool:
     def __call__(self, rel_model_path: str):
         self.resumable_handle_sample(rel_model_path)
 
-    def resume(self, model_path: str) -> List[str]:
+    def resume(self, rel_model_path: str) -> List[str]:
         """
         Apply dtype passes to generate new samples.
 
@@ -308,24 +310,36 @@ def resume(self, model_path: str) -> List[str]:
         """
         # Apply model_path_prefix if provided
         if self.model_path_prefix:
-            model_path = str(Path(self.model_path_prefix) / model_path)
+            abs_model_path = str(Path(self.model_path_prefix) / rel_model_path)
 
         # Read pass names from graph_net.json
-        dtype_pass_names = self._read_dtype_pass_names(model_path)
+        dtype_pass_names = self._read_dtype_pass_names(abs_model_path)
 
         if not dtype_pass_names:
-            logging.warning(f"No dtype passes found in {model_path}/graph_net.json")
+            logging.warning(f"No dtype passes found in {abs_model_path}/graph_net.json")
             return []
 
         # Parse the computation graph
-        traced_model = parse_immutable_model_path_into_sole_graph_module(model_path)
+        traced_model = parse_immutable_model_path_into_sole_graph_module(abs_model_path)
+
+        # Copy the originl sample
+        files_copied = [
+            "model.py",
+            "graph_hash.txt",
+            "graph_net.json",
+            "weight_meta.py",
+            "input_meta.py",
+            "input_tensor_constraints.py",
+            "subgraph_sources.json",
+        ]
+        self._copy_sample_files(rel_model_path, "float32", files_copied)
 
         # Generate samples for each pass
         generated_samples = []
         for pass_name in dtype_pass_names:
             try:
                 sample_dir = self._apply_pass_and_generate(
-                    model_path, traced_model, pass_name
+                    rel_model_path, traced_model, pass_name
                 )
                 generated_samples.append(sample_dir)
                 logging.info(f"Generated sample: {sample_dir}")
@@ -388,8 +402,7 @@ def _apply_pass_and_generate(
         gm_modified = dtype_pass.rewrite(gm_copy)
 
         # Generate output directory
-        model_name = Path(model_path).name
-        output_sample_dir = Path(self.output_dir) / f"{model_name}_{dtype}"
+        output_sample_dir = Path(self.output_dir) / dtype / model_path
         output_sample_dir.mkdir(parents=True, exist_ok=True)
 
         # Write modified model.py
@@ -398,11 +411,20 @@ def _apply_pass_and_generate(
         with open(output_sample_dir / "model.py", "w") as f:
             f.write(write_code)
 
+        # Write modified graph_hash.txt
+        model_hash = get_sha256_hash(model_code)
+        with open(output_sample_dir / "graph_hash.txt", "w") as f:
+            f.write(model_hash)
+
         # Copy metadata files
-        for fname in ["graph_net.json", "weight_meta.py", "input_meta.py"]:
-            src = Path(model_path) / fname
-            if src.exists():
-                shutil.copy(src, output_sample_dir / fname)
+        files_copied = [
+            "graph_net.json",
+            "weight_meta.py",
+            "input_meta.py",
+            "input_tensor_constraints.py",
+            "subgraph_sources.json",
+        ]
+        self._copy_sample_files(model_path, dtype, files_copied)
 
         # Update graph_net.json with dtype information
         self._update_sample_metadata(output_sample_dir, dtype)
@@ -429,6 +451,25 @@ def _update_sample_metadata(self, sample_dir: Path, dtype: str) -> None:
         update_json(graph_net_json_path, kDtypeGeneralizationPrecision, dtype)
         update_json(graph_net_json_path, kDtypeGeneralizationGenerated, True)
 
+    def _copy_sample_files(
+        self, rel_model_path: str, dtype: str, files_copied: list
+    ) -> None:
+        """
+        Copy files of sample.
+
+        Args:
+            rel_model_path: relative model path
+        """
+        # Generate output directory
+        output_sample_dir = Path(self.output_dir) / dtype / rel_model_path
+        output_sample_dir.mkdir(parents=True, exist_ok=True)
+
+        # Copy files of original sample
+        for fname in files_copied:
+            src = Path(rel_model_path) / fname
+            if src.exists():
+                shutil.copy(src, output_sample_dir / fname)
+
 
 class MultiDtypeFilter:
     """