Unit Test failing for Concrete ML when running locally

Zama Concrete ML
1

When I try to run the tests locally for Concrete-ML using make pytest, I get the following 12 failures -

FAILED tests/onnx/test_onnx_utils.py::test_check_onnx_model_large - AssertionError: Regex pattern did not match.
FAILED tests/pandas/test_pandas.py::test_merge[index-left-True] - RuntimeError: Tried to transform transport value with incompatible type info.
FAILED tests/pandas/test_pandas.py::test_merge[index-left-False] - RuntimeError: Tried to transform transport value with incompatible type info.
FAILED tests/pandas/test_pandas.py::test_merge[index-right-True] - RuntimeError: Tried to transform transport value with incompatible type info.
FAILED tests/pandas/test_pandas.py::test_merge[index-right-False] - RuntimeError: Tried to transform transport value with incompatible type info.
FAILED tests/pandas/test_pandas.py::test_merge[None-left-True] - RuntimeError: Tried to transform transport value with incompatible type info.
FAILED tests/pandas/test_pandas.py::test_merge[None-left-False] - RuntimeError: Tried to transform transport value with incompatible type info.
FAILED tests/pandas/test_pandas.py::test_merge[None-right-True] - RuntimeError: Tried to transform transport value with incompatible type info.
FAILED tests/pandas/test_pandas.py::test_merge[None-right-False] - RuntimeError: Tried to transform transport value with incompatible type info.
FAILED tests/pandas/test_pandas.py::test_parameter_sets - AssertionError: The new generated client file is not equal to the one stored in source.
FAILED tests/pandas/test_pandas.py::test_schema_input - RuntimeError: Tried to transform transport value with incompatible type info.
FAILED tests/torch/test_compile_torch.py::test_compile_torch_or_onnx_networks[get_and_compile-is_onnx-FHE_simulation-model1-5-relu] - AssertionError

Out of the above 12 failures, the first failure -

E           AssertionError: Regex pattern did not match.
E           Regex: 'Message onnx.ModelProto exceeds maximum protobuf size of 2GB:'
E           Input: 'This protobuf of onnx model is too large (>2GB). Call check_model with model path instead.'

I think the above failure is occurring because of wrong regex present in the unit test of the repo. The input message seems to match from the onnx repo - https://github.com/onnx/onnx/blob/99e1fd352c05c3176770080824fd7a8c474c97c0/onnx/checker.py#L177.
For the rest of the 11 failures, I found this PR - https://github.com/zama-ai/concrete-ml/pull/798#issuecomment-2239288027 that has the exact same errors. I need some help to fix these failures so that I can set up Concrete-ML repo. Running these unit test on Mac M1 Sanoma 14.6.

2

Hello @Samanvaya_Panda ,
This error generally occurs if you have changed our Concrete Python dependency version locally ! Is that the case ?

  • if so and if this is expected, then you can simply run make update_encrypted_dataframe locally and that should fix the issue. However, is there a particular reason why you would want to use a different Concrete Python version ? Concrete ML is build and delivered on fixed CP versions and we cannot guarantee full compatibility with other ones, so be sure to know what you are doing ! If you needed new features from Concrete, please tell us and we could try to provide a new release based on a newer Concrete Python version
  • if not, could you update the dependencies ? We recommend to use a virtual environment and then run make setup_env (or make sync_env if you already have one)

Hope that helps !