voice_audio_multi/include/engine/models/niagara_asr/assets.h
Fabio c36510c21e
Some checks failed
Linux Build / Linux cpu (push) Has been cancelled
Linux Build / Linux cpu-clang (push) Has been cancelled
Linux Build / Linux vulkan (push) Has been cancelled
mac-build / macOS CPU (push) Has been cancelled
Nix Build / Nix (cpu) (push) Has been cancelled
Nix Build / Nix (vulkan) (push) Has been cancelled
windows-build / Windows CPU (push) Has been cancelled
Initial import: audio.cpp with Orange Pi documentation
2026-11-04 19:05:22 +01:00

52 lines
1.3 KiB
C++

#pragma once
#include "engine/framework/assets/tensor_source.h"
#include "engine/framework/model_spec/package.h"
#include "engine/framework/tokenizers/sentencepiece.h"
#include <cstdint>
#include <filesystem>
#include <memory>
#include <string>
#include <vector>
namespace engine::models::niagara_asr {
struct NiagaraFrontendConfig {
int64_t sample_rate = 16000;
int64_t n_fft = 400;
int64_t win_length = 400;
int64_t hop_length = 160;
int64_t n_mels = 80;
};
struct NiagaraEncoderConfig {
int64_t hidden_size = 0;
int64_t intermediate_size = 0;
int64_t num_layers = 0;
int64_t state_size = 0;
int64_t state_channels = 0;
int64_t dense_input_features = 0;
int64_t subsampling_kernel_height = 4;
int64_t subsampling_kernel_width = 3;
};
struct NiagaraAsrConfig {
std::string model_type = "niagara_asr";
int64_t vocab_size = 256;
int64_t blank_id = 256;
NiagaraFrontendConfig frontend;
NiagaraEncoderConfig encoder;
};
struct NiagaraAsrAssets {
assets::ResourceBundle resources;
std::shared_ptr<const assets::TensorSource> source;
NiagaraAsrConfig config;
std::vector<tokenizers::SentencePiecePiece> tokenizer_pieces;
};
std::shared_ptr<const NiagaraAsrAssets> load_niagara_asr_assets(
const std::filesystem::path & model_path);
} // namespace engine::models::niagara_asr