microsoft/onnxruntime-extensions
Publicmirrored fromhttps://github.com/microsoft/onnxruntime-extensionsAvailable
operators/tokenizer/gpt2_tokenizer.hpp
27lines · modecode
| 1 | #include <list> |
| 2 | #include "ocos.h" |
| 3 | #include "ustring.h" |
| 4 | |
| 5 | |
| 6 | class VocabData; |
| 7 | |
| 8 | struct KernelBpeTokenizer : BaseKernel { |
| 9 | KernelBpeTokenizer(OrtApi api, const OrtKernelInfo* info); |
| 10 | void Compute(OrtKernelContext* context); |
| 11 | |
| 12 | private: |
| 13 | std::vector<int64_t> Tokenize(const ustring& input, int64_t max_length); |
| 14 | |
| 15 | int64_t padding_length_; |
| 16 | std::list<int> byte_list_; |
| 17 | std::shared_ptr<VocabData> bbpe_tokenizer_; |
| 18 | }; |
| 19 | |
| 20 | struct CustomOpBpeTokenizer : Ort::CustomOpBase<CustomOpBpeTokenizer, KernelBpeTokenizer> { |
| 21 | void* CreateKernel(OrtApi api, const OrtKernelInfo* info) const; |
| 22 | const char* GetName() const; |
| 23 | size_t GetInputTypeCount() const; |
| 24 | ONNXTensorElementDataType GetInputType(size_t index) const; |
| 25 | size_t GetOutputTypeCount() const; |
| 26 | ONNXTensorElementDataType GetOutputType(size_t index) const; |
| 27 | }; |
| 28 | |