Adding Facebook/Encodec support for ONNX

[coolyashas]

Fixes #27602

[fxmarty]

Thank you @coolyashas, feel free if you need any help.

[shoh27]

Hi,

I tried exporting the model by configuring it and incorporating 'encodec' into the TasksManager class within the tasks.py file. I tested this solution in both the source and stable versions. main.py model_checkpoint = "facebook/encodec_24khz" ; base_model = AutoModel.from_pretrained(model_checkpoint); onnx_path = Path(save_directory,"model_encodec.onnx") ; onnx_config_constructor = TasksManager.get_exporter_config_constructor("onnx", base_model); onnx_config = onnx_config_constructor(base_model.config); onnx_inputs, onnx_outputs = export(base_model, onnx_config, onnx_path, onnx_config.DEFAULT_ONNX_OPSET);

tasks.py "encodec": supported_tasks_mapping( "default", "feature-extraction", onnx="EncodecOnnxConfig", ),

Regrettably, the proposed solution isn't effective. While following the example provided in the HuggingFace documentation on ONNX exporters https://huggingface.co/docs/optimum/main/en/exporters/onnx/usage_guides/contribute](onnx), I encountered an error within self._create_dummy_input_generator_classes(**kwargs).

/home/user/anaconda3/envs/onnx_from_src/lib/python3.8/site-packages/torch/nn/utils/weight_norm.py:28: UserWarning: torch.nn.utils.weight_norm is deprecated in favor of torch.nn.utils.parametrizations.weight_norm. warnings.warn("torch.nn.utils.weight_norm is deprecated in favor of torch.nn.utils.parametrizations.weight_norm.") Passing the argument library_nametoget_supported_tasks_for_model_typeis required, but got library_name=None. Defaulting totransformers. An error will be raised in a future version of Optimum if library_nameis not provided. Using framework PyTorch: 2.2.1+cu121 Traceback (most recent call last): File "/home/user/anaconda3/envs/onnx_from_src/lib/python3.8/site-packages/optimum/exporters/onnx/convert.py", line 547, in export_pytorch dummy_inputs = config.generate_dummy_inputs(framework="pt", **input_shapes) File "/home/user/anaconda3/envs/onnx_from_src/lib/python3.8/site-packages/optimum/exporters/onnx/base.py", line 463, in generate_dummy_inputs dummy_inputs_generators = self._create_dummy_input_generator_classes(**kwargs) File "/home/user/anaconda3/envs/onnx_from_src/lib/python3.8/site-packages/optimum/exporters/onnx/model_configs.py", line 1917, in _create_dummy_input_generator_classes DummyAudioInputGenerator(**kwargs) TypeError: __init__() missing 2 required positional arguments: 'task' and 'normalized_config'