clip operator fails when min or max values are provided as tensors (ONNX compatibility issue)

What commit version of aidge do you use

aidge_backend_cuda 0.5.1
aidge_core 0.5.1 aidge_onnx 0.4.1 aidge_quantization 0.3.1

Problem description

The clip operator currently only accepts scalar values for the min and max parameters. However, in some cases (e.g., when working with ONNX graphs), these values might be provided as tensors, see a model onnxgoogle_mobilenet_v1_1.0_224_sim.onnx. This causes the operator to fail, as it does not handle tensor inputs for min and max.

Expected Behavior

The clip operator should accept tensor inputs for min and max and handle them appropriately (e.g., by treating them as scalars if they are 0D or 1D tensors with a single value).

@mszczep I don't have this issue with my model and the latest version of aidge.

Do you still have the issue?

yes, I do have the same issue with

aidge_backend_cpu            0.5.0
aidge_core                   0.5.1

for the same model, I provide more details to make sure I'm doing it right.

scheduler_aidge.forward(True, [img_tensor_batch_aidge])

where

img_tensor_batch_aidge : 150528
img_tensor_batch_aidge : [1, 3, 224, 224]
img_tensor_batch_aidge: <class 'aidge_core.aidge_core.Tensor'>

**[FATAL] - Expected Input#1 and Input#2 to be scalar (Tensors of empty shapes)**
  0%|                                                                                                              | 0/50000 [00:00<?, ?batch/s]
Traceback (most recent call last):
  File "/.../test_PTQ_model_classif.py", line 92, in <module>
    valider_float.eval(modelQuantized=False)
  File "/../AidgeClassificationHandler.py", line 220, in eval
    prediction = self.run_inference(input_data)
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/.../AidgeClassificationHandler.py", line 161, in run_inference
    self.scheduler_aidge.forward(True, [img_tensor_batch_aidge])
RuntimeError: Expected Input#1 and Input#2 to be scalar (Tensors of empty shapes)```

What version on aidge_onnx do you have?

Also this maybe an issue on how you set your inputs can you provide your script?

Yes, my script can be find here. Please let me know if you reproduce the same error. I can clarify if it is needed because it is a compilation of several files. test_model_classif.py I use aidge_core 0.5.1 thanks for help

added Bug 🐛 label

mentioned in commit 2b1fbc89

Should be fixed on dev, please re-open if there is still an issue.

closed