Added tests and made changes to gs_patching.py to improve coverage

Signed-off-by: Hrishith Thadicherla <[email protected]>

Author: hthadicherla

modelopt/onnx/quantization/gs_patching.py

@@ -70,10 +70,8 @@ def _export_tensor_proto(tensor: gs.Constant) -> onnx.TensorProto:
         vals = tensor.values
         if _onnx_supports_int4() and dtype in [onnx.TensorProto.INT4, onnx.TensorProto.UINT4]:
             signed = dtype == onnx.TensorProto.INT4
-            if signed:
-                vals = pack_float32_to_4bit_cpp_based(tensor.values, signed=signed).astype(np.int8)
-            else:
-                vals = pack_float32_to_4bit_cpp_based(tensor.values, signed=signed).astype(np.uint8)
+            packed_dtype = np.int8 if signed else np.uint8
+            vals = pack_float32_to_4bit_cpp_based(tensor.values, signed=signed).astype(packed_dtype)
 
         onnx_tensor = onnx.helper.make_tensor(
             tensor.name,

tests/gpu/onnx/test_quantize_onnx_torch_int4_awq.py

@@ -38,6 +38,55 @@
 #       For that, we need to investigate failure in 'pytest tests/gpu/onnx'.
 #       test_qdq_utils_fp8.py::test_fused_q[bf16,fp16] fails if this script runs after the int4 test, but not before.
 
+def test_safe_cupy_array_all_paths(monkeypatch):
+    """Test safe_cupy_array covering all code paths including ml_dtypes handling"""
+    # Test 1: When ml_dtypes import fails (covers ImportError path)
+    # Temporarily remove ml_dtypes from sys.modules
+    import sys
+    if 'ml_dtypes' in sys.modules:
+        ml_dtypes_backup = sys.modules['ml_dtypes']
+        monkeypatch.delitem(sys.modules, 'ml_dtypes')
+    else:
+        ml_dtypes_backup = None
+    
+    tensor = np.array([1, 2, 3, 4], dtype=np.int8)
+    result = int4.safe_cupy_array(tensor)
+    assert isinstance(result, np.ndarray)  # Should return numpy array
+    
+    # Restore ml_dtypes if it existed
+    if ml_dtypes_backup:
+        sys.modules['ml_dtypes'] = ml_dtypes_backup
+    
+    # Test 2: When ml_dtypes exists and tensor has ml_dtypes.int4 dtype
+    try:
+        import ml_dtypes
+        # Create a mock tensor with int4 dtype
+        class MockInt4Tensor:
+            def __init__(self, data):
+                self.data = data
+                self.dtype = ml_dtypes.int4
+                self.shape = data.shape
+            
+            def astype(self, dtype):
+                return self.data.astype(dtype)
+            
+            def __array__(self):
+                return self.data
+        
+        mock_tensor = MockInt4Tensor(np.array([1, 2, 3, 4], dtype=np.int8))
+        print(mock_tensor.dtype)
+        result = int4.safe_cupy_array(mock_tensor)
+        assert isinstance(result, np.ndarray)
+        assert result.dtype == np.int8
+    except ImportError:
+        # ml_dtypes not available, skip this part
+        pass
+    
+    # Test 3: Normal case with regular numpy array
+    tensor = np.array([1, 2, 3, 4], dtype=np.int8)
+    result = int4.safe_cupy_array(tensor)
+    # Should work normally
+    assert isinstance(result, (np.ndarray, type(tensor)))
 
 def test_int4_awq(tmp_path):
     def _forward_loop(model, dataloader):