transformers.convert_graph_to_onnx.quantize fails in unit tests

[rachthree]

System Info

• transformers version: 4.22.0.dev0
• Platform: Linux-5.10.60.1-microsoft-standard-WSL2-x86_64-with-glibc2.29
  • Used tensorflow/tensorflow:latest Docker image for this environment, then used pip install -e '.[dev,onnx]'
• Python version: 3.8.10
• Huggingface_hub version: 0.8.1
• PyTorch version (GPU?): 1.12.1+cu116 (True)
• Tensorflow version (GPU?): 2.9.1 (True)
• Flax version (CPU?/GPU?/TPU?): 0.5.3 (cpu)
• Jax version: 0.3.6
• JaxLib version: 0.3.5
• Using GPU in script?: Yes
• Using distributed or parallel set-up in script?: No

Other:

• onnxruntime version: 1.12.1

Who can help?

No response

Information

• The official example scripts
• My own modified scripts

Tasks

• An officially supported task in the examples folder (such as GLUE/SQuAD, ...)
• My own task or dataset (give details below)

Reproduction

Run RUN_SLOW=true pytest tests/onnx/test_onnx.py

Get failures:

FAILED tests/onnx/test_onnx.py::OnnxExportTestCase::test_quantize_pytorch - TypeError: 'module' object is not callable
FAILED tests/onnx/test_onnx.py::OnnxExportTestCase::test_quantize_tf - TypeError: 'module' object is not callable

Expected behavior

The unit tests should pass.

I believe this failure is due to onnxruntime.quantization.quantize which is a module that contains functions quantize_static and quantize_dynamic. The API may have changed since the unit test was written. I'm not sure which is the one to use for the unit tests. Even after fixing, not sure how transformers should handle the different versions of onnxruntime or should the required version change in setup.py.

See https://github.com/microsoft/onnxruntime/blob/main/onnxruntime/python/tools/quantization/quantize.py

[LysandreJik]

cc @lewtun

[severinsimmler]

I have fixed this in #18336, but still waiting for a review.

[rachthree]

Looks like this has been fixed! Closing this.