chore: fix new feature

src/concrete/ml/torch/compile.py

@@ -212,7 +212,7 @@ def _compile_torch_or_onnx_model(
 
         # mypy
         assert not isinstance(n_bits_rounding, str) and n_bits_rounding is not None
-        rounding_threshold_bits = {"n_bits": n_bits_rounding, "method": method.name}
+        rounding_threshold_bits = {"n_bits": n_bits_rounding, "method": method}
 
     inputset_as_numpy_tuple = tuple(
         convert_torch_tensor_or_numpy_array_to_numpy_array(val) for val in to_tuple(torch_inputset)

tests/torch/test_compile_torch.py

@@ -1420,3 +1420,35 @@ def test_compile_torch_model_rounding_threshold_bits_errors(
             rounding_threshold_bits=rounding_threshold_bits,
             configuration=default_configuration,
         )
+
+
+@pytest.mark.parametrize(
+    "rounding_method, expected_reinterpret",
+    [
+        ("APPROXIMATE", True),
+        ("EXACT", False),
+    ],
+)
+def test_rounding_mode(rounding_method, expected_reinterpret, default_configuration):
+    """Test that the underlying FHE circuit uses the right rounding method."""
+    model = FCSmall(input_output=5, activation_function=nn.ReLU)
+    torch_inputset = torch.randn(10, 5)
+    configuration = default_configuration
+
+    compiled_module = compile_torch_model(
+        torch_model=model,
+        torch_inputset=torch_inputset,
+        rounding_threshold_bits={"method": rounding_method, "n_bits": 4},
+        configuration=configuration,
+    )
+
+    # Convert compiled module to string to search for patterns
+    mlir = compiled_module.fhe_circuit.mlir
+    if expected_reinterpret:
+        assert (
+            "FHE.reinterpret_precision" in mlir and "FHE.round" not in mlir
+        ), "Expected 'FHE.reinterpret_precision' found but 'FHE.round' should not be present."
+    else:
+        assert (
+            "FHE.reinterpret_precision" not in mlir
+        ), "Unexpected 'FHE.reinterpret_precision' found."

use_case_examples/cifar/cifar_brevitas_training/evaluate_one_example_fhe.py

@@ -138,11 +138,9 @@ def wrapper(*args, **kwargs):
     print(f"Quantization of a single input (image) took {quantization_execution_time} seconds")
     print(f"Size of CLEAR input is {q_x_numpy.nbytes} bytes\n")
 
-    # Use new VL with .simulate() once CP's multi-parameter/precision bug is fixed
-    # FIXME: https://github.com/zama-ai/concrete-ml-internal/issues/3856
     p_error = quantized_numpy_module.fhe_circuit.p_error
     expected_quantized_prediction, clear_inference_time = measure_execution_time(
-        partial(quantized_numpy_module.fhe_circuit.graph, p_error=p_error)
+        partial(quantized_numpy_module.fhe_circuit.simulate, p_error=p_error)
     )(q_x_numpy)
 
     # Encrypt the input