byte_pair_encoding.hpp
Go to the documentation of this file.
1 /*
2  * SPDX-FileCopyrightText: Copyright (c) 2022-2024, NVIDIA CORPORATION.
3  * SPDX-License-Identifier: Apache-2.0
4  */
5 #pragma once
6 
7 #include <cudf/column/column.hpp>
9 #include <cudf/scalar/scalar.hpp>
12 #include <cudf/utilities/export.hpp>
14 
15 namespace CUDF_EXPORT nvtext {
16 
29  struct bpe_merge_pairs_impl;
30 
38  bpe_merge_pairs(std::unique_ptr<cudf::column>&& input,
41 
52 
53  ~bpe_merge_pairs();
55 
56  private:
57  friend bpe_merge_pairs_impl const* get_bpe_merge_pairs_impl(bpe_merge_pairs const&);
58  bpe_merge_pairs_impl* impl{};
59 };
60 
86 std::unique_ptr<bpe_merge_pairs> load_merge_pairs(
87  cudf::strings_column_view const& merge_pairs,
90 
118 std::unique_ptr<cudf::column> byte_pair_encoding(
119  cudf::strings_column_view const& input,
120  bpe_merge_pairs const& merges_pairs,
121  cudf::string_scalar const& separator = cudf::string_scalar(" "),
124  // end of group
126 } // namespace CUDF_EXPORT nvtext
An owning class to represent a string in device memory.
Definition: scalar.hpp:410
Given a column-view of strings type, an instance of this class provides a wrapper on this compound co...
Class definition for cudf::column.
column view class definitions
rmm::cuda_stream_view const get_default_stream()
Get the current default stream.
rmm::device_async_resource_ref get_current_device_resource_ref()
Get the current device memory resource reference.
detail::cccl_async_resource_ref< cuda::mr::resource_ref< cuda::mr::device_accessible > > device_async_resource_ref
std::unique_ptr< bpe_merge_pairs > load_merge_pairs(cudf::strings_column_view const &merge_pairs, rmm::cuda_stream_view stream=cudf::get_default_stream(), rmm::device_async_resource_ref mr=cudf::get_current_device_resource_ref())
Create a nvtext::bpe_merge_pairs from a strings column.
std::unique_ptr< cudf::column > byte_pair_encoding(cudf::strings_column_view const &input, bpe_merge_pairs const &merges_pairs, cudf::string_scalar const &separator=cudf::string_scalar(" "), rmm::cuda_stream_view stream=cudf::get_default_stream(), rmm::device_async_resource_ref mr=cudf::get_current_device_resource_ref())
Byte pair encode the input strings.
NVText APIs.
Class definitions for cudf::scalar.
Class definition for cudf::strings_column_view.
The table of merge pairs for the BPE encoder.
bpe_merge_pairs(cudf::strings_column_view const &input, rmm::cuda_stream_view stream=cudf::get_default_stream(), rmm::device_async_resource_ref mr=cudf::get_current_device_resource_ref())
Construct a new bpe merge pairs object.
bpe_merge_pairs(std::unique_ptr< cudf::column > &&input, rmm::cuda_stream_view stream=cudf::get_default_stream(), rmm::device_async_resource_ref mr=cudf::get_current_device_resource_ref())
Construct a new bpe merge pairs object.