I am trying to export a wav2vec model (cahya/wav2vec2-base-turkish-artificial-cv) to ONNX format with convert_graph_to_onnx.py script provided in transformers repository.
When I try to use these script with this line:
python convert_graph_to_onnx.py --framework pt --model cahya/wav2vec2-base-turkish-artificial-cv exported_model.onnx
I am getting this error:
====== Converting model to ONNX ====== ONNX opset version set to: 11 Loading pipeline (model: cahya/wav2vec2-base-turkish-artificial-cv, tokenizer: cahya/wav2vec2-base-turkish-artificial-cv) Some weights of the model checkpoint at cahya/wav2vec2-base-turkish-artificial-cv were not used when initializing Wav2Vec2Model: ['lm_head.bias', 'lm_head.weight'] - This IS expected if you are initializing Wav2Vec2Model from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model). - This IS NOT expected if you are initializing Wav2Vec2Model from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model). Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. Error while converting the model: __init__() got an unexpected keyword argument 'feature_extractor'
All I understand is that this script is not made for specifically wav2vec models. If I am right, how can I convert the wav2vec model to ONNX format?