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
57 lines
1.4 KiB
C++
57 lines
1.4 KiB
C++
#pragma once
|
|
|
|
#include "engine/models/heartmula/assets.h"
|
|
|
|
#include <cstdint>
|
|
#include <memory>
|
|
#include <string>
|
|
#include <vector>
|
|
|
|
namespace engine::models::heartmula {
|
|
|
|
struct HeartMuLaGenerationOptions {
|
|
float duration_seconds = 120.0F;
|
|
float temperature = 1.0F;
|
|
int64_t top_k = 50;
|
|
float guidance_scale = 1.5F;
|
|
float codec_duration = 29.76F;
|
|
int64_t num_inference_steps = 10;
|
|
float codec_guidance_scale = 1.25F;
|
|
bool infinite_mode = false;
|
|
int64_t text_chunk_size = 4096;
|
|
int64_t infinite_chunk_audio_length_ms = 240000;
|
|
};
|
|
|
|
struct HeartMuLaPromptRequest {
|
|
std::string tags;
|
|
std::string lyrics;
|
|
HeartMuLaGenerationOptions options;
|
|
};
|
|
|
|
struct HeartMuLaPromptEncoding {
|
|
int64_t batch_size = 0;
|
|
int64_t prompt_len = 0;
|
|
int64_t parallel_number = 0;
|
|
std::vector<int64_t> tokens;
|
|
std::vector<uint8_t> tokens_mask;
|
|
std::vector<float> muq_embed;
|
|
std::vector<int64_t> muq_idx;
|
|
std::vector<int64_t> pos;
|
|
std::vector<int32_t> tags_ids;
|
|
std::vector<int32_t> lyrics_ids;
|
|
};
|
|
|
|
class HeartMuLaTextTokenizer {
|
|
public:
|
|
struct Impl;
|
|
|
|
explicit HeartMuLaTextTokenizer(std::shared_ptr<const HeartMuLaAssets> assets);
|
|
|
|
std::vector<int32_t> encode(const std::string & text) const;
|
|
HeartMuLaPromptEncoding encode_prompt(const HeartMuLaPromptRequest & request) const;
|
|
|
|
private:
|
|
std::shared_ptr<const Impl> impl_;
|
|
};
|
|
|
|
} // namespace engine::models::heartmula
|