Open saibaldas opened 11 months ago
Roberta-base fine-tuned for the boolq dataset uses the tokenizer encode_plus to encode both the question and context. When exported to ONNX , the inputs are
name: input_ids tensor: int64[batch_size,sequence_length]
name: attention_mask tensor: int64[batch_size,sequence_length]
Please suggest the steps to run the ONNX for inference in ML.NET . The BertTokenizer can't be used as the tokenizer is BPE Tried to use the https://gist.github.com/luisquintanilla/bc91de8668cfa7c3755b20329fadd027 without much success
Roberta-base fine-tuned for the boolq dataset uses the tokenizer encode_plus to encode both the question and context. When exported to ONNX , the inputs are
name: input_ids tensor: int64[batch_size,sequence_length]
name: attention_mask tensor: int64[batch_size,sequence_length]
Please suggest the steps to run the ONNX for inference in ML.NET . The BertTokenizer can't be used as the tokenizer is BPE Tried to use the https://gist.github.com/luisquintanilla/bc91de8668cfa7c3755b20329fadd027 without much success