Upload evaluation/eval.py with huggingface_hub

evaluation/eval.py

@@ -2,25 +2,51 @@ import argparse
 import os
 import sys
 import subprocess
+import glob
+import shutil
 
-# Ensure we import the fresh source package instead of any pre-built compiled extension.
-# Add the evaluation directory to sys.path so `import utils` loads evaluation/utils/*.py
-eval_dir = os.path.dirname(__file__)
-if eval_dir not in sys.path:
-    sys.path.insert(0, eval_dir)
-# Invalidate import caches and prefer source .py files over compiled extensions
-import importlib, importlib.util, pkgutil
-# If a compiled extension is present, remove it from sys.modules so Python will import the source package
-for mod in list(sys.modules.keys()):
-    if mod.startswith('utils'):
-        del sys.modules[mod]
-
-# Force import from source files in evaluation/utils
-spec = importlib.util.spec_from_file_location('utils', os.path.join(eval_dir, 'utils', '__init__.py'))
-utils = importlib.util.module_from_spec(spec)
-spec.loader.exec_module(utils)
-import sys as _sys
-_sys.modules['utils'] = utils
+# Ensure evaluation directory is on path so 'utils' package resolves to evaluation/utils
+sys.path.insert(0, os.path.dirname(__file__))
+
+def rebuild_utils_package():
+    """Deterministically rebuild the evaluation.utils package from sources.
+
+    - Remove any existing compiled artifacts (.so, .cpython-*.so, .c) in evaluation/utils
+    - Run setup.py build_ext --inplace inside evaluation/
+    """
+    eval_dir = os.path.dirname(os.path.abspath(__file__))
+    utils_dir = os.path.join(eval_dir, "utils")
+
+    # Remove stale compiled artifacts that can cause non-deterministic imports
+    patterns = ["*.so", "*.cpython-*.so", "*.c", "__pycache__"]
+    for pat in patterns:
+        for path in glob.glob(os.path.join(utils_dir, pat)):
+            try:
+                if os.path.isdir(path):
+                    shutil.rmtree(path)
+                else:
+                    os.remove(path)
+            except Exception:
+                # best-effort; don't fail here so we can attempt a rebuild
+                pass
+
+    # Run deterministic rebuild
+    try:
+        subprocess.run(
+            [sys.executable, "setup.py", "build_ext", "--inplace"],
+            cwd=eval_dir,
+            check=True,
+            capture_output=True,
+            text=True
+        )
+    except subprocess.CalledProcessError as e:
+        print("Error: Failed to rebuild evaluation utils package:", file=sys.stderr)
+        print(e.stdout, file=sys.stderr)
+        print(e.stderr, file=sys.stderr)
+        raise
+
+# Rebuild before importing compiled modules to avoid stale artifacts
+rebuild_utils_package()
 
 from utils.benchmark_utils import BENCHMARK_CALCULATORS