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
120 lines
4.6 KiB
C++
120 lines
4.6 KiB
C++
#pragma once
|
|
|
|
#include "engine/framework/runtime/cache_slots.h"
|
|
#include "engine/framework/runtime/session_base.h"
|
|
#include "engine/models/index_tts2/assets.h"
|
|
#include "engine/models/index_tts2/audio_features.h"
|
|
#include "engine/models/index_tts2/gpt.h"
|
|
#include "engine/models/index_tts2/qwen_emotion.h"
|
|
#include "engine/models/index_tts2/request.h"
|
|
#include "engine/models/index_tts2/s2mel.h"
|
|
#include "engine/models/index_tts2/semantic_codec.h"
|
|
#include "engine/models/index_tts2/semantic_encoder.h"
|
|
#include "engine/models/index_tts2/style_encoder.h"
|
|
#include "engine/models/index_tts2/tokenizer_text.h"
|
|
#include "engine/models/index_tts2/vocoder.h"
|
|
|
|
#include <cstddef>
|
|
#include <cstdint>
|
|
#include <memory>
|
|
#include <optional>
|
|
#include <string>
|
|
#include <vector>
|
|
|
|
namespace engine::models::index_tts2 {
|
|
|
|
struct IndexTTS2AudioIdentity {
|
|
int sample_rate = 0;
|
|
int channels = 0;
|
|
uint64_t sample_count = 0;
|
|
uint64_t sample_hash = 0;
|
|
};
|
|
|
|
class IndexTTS2Session final
|
|
: public runtime::RuntimeSessionBase
|
|
, public runtime::IOfflineVoiceTaskSession {
|
|
public:
|
|
IndexTTS2Session(
|
|
runtime::TaskSpec task,
|
|
runtime::SessionOptions options,
|
|
std::shared_ptr<const IndexTTS2Assets> assets);
|
|
|
|
std::string family() const override;
|
|
runtime::VoiceTaskKind task_kind() const override;
|
|
runtime::RunMode run_mode() const override;
|
|
void prepare(const runtime::SessionPreparationRequest & request) override;
|
|
runtime::TaskResult run(const runtime::TaskRequest & request) override;
|
|
|
|
private:
|
|
struct SpeakerState {
|
|
IndexTTS2AudioIdentity identity;
|
|
IndexTTS2SemanticEmbedding semantic;
|
|
IndexTTS2MelOutput reference_mel;
|
|
IndexTTS2StyleEmbedding style;
|
|
IndexTTS2S2MelSequence prompt_condition;
|
|
};
|
|
|
|
struct EmotionState {
|
|
IndexTTS2AudioIdentity identity;
|
|
IndexTTS2SemanticEmbedding semantic;
|
|
};
|
|
|
|
struct AudioIdentityEqual {
|
|
bool operator()(
|
|
const IndexTTS2AudioIdentity & lhs,
|
|
const IndexTTS2AudioIdentity & rhs) const;
|
|
};
|
|
|
|
const SpeakerState & resolve_speaker_state(const runtime::AudioBuffer & audio);
|
|
const EmotionState & resolve_emotion_state(const runtime::AudioBuffer & audio);
|
|
std::vector<float> resolve_emotion_vector(
|
|
const IndexTTS2Request & request,
|
|
const SpeakerState & speaker,
|
|
const EmotionState & emotion);
|
|
runtime::AudioBuffer synthesize_segment(
|
|
const std::vector<int32_t> & text_tokens,
|
|
int32_t lang_id,
|
|
const SpeakerState & speaker,
|
|
const EmotionState & emotion,
|
|
const std::vector<float> & emotion_vector,
|
|
const IndexTTS2GenerationOptions & options,
|
|
uint32_t segment_seed);
|
|
|
|
std::vector<float> explicit_emotion_matrix_vector(
|
|
const std::vector<float> & emotion_weights,
|
|
const IndexTTS2StyleEmbedding & style,
|
|
bool use_random,
|
|
uint32_t seed) const;
|
|
|
|
runtime::TaskSpec task_;
|
|
std::shared_ptr<const IndexTTS2Assets> assets_;
|
|
size_t gpt_graph_arena_bytes_ = 32ull * 1024ull * 1024ull;
|
|
size_t s2mel_graph_arena_bytes_ = 32ull * 1024ull * 1024ull;
|
|
size_t reference_graph_arena_bytes_ = 512ull * 1024ull * 1024ull;
|
|
size_t emotion_text_prefill_graph_arena_bytes_ = 2048ull * 1024ull * 1024ull;
|
|
size_t emotion_text_decode_graph_arena_bytes_ = 512ull * 1024ull * 1024ull;
|
|
size_t weight_context_bytes_ = 32ull * 1024ull * 1024ull;
|
|
int64_t emotion_text_max_tokens_ = 256;
|
|
engine::assets::TensorStorageType matmul_weight_storage_type_ = engine::assets::TensorStorageType::Native;
|
|
engine::assets::TensorStorageType conv_weight_storage_type_ = engine::assets::TensorStorageType::Native;
|
|
bool mem_saver_ = false;
|
|
|
|
IndexTTS2TextTokenizer tokenizer_;
|
|
std::unique_ptr<IndexTTS2Wav2Vec2BertRuntime> semantic_encoder_;
|
|
std::unique_ptr<IndexTTS2SemanticCodecRuntime> semantic_codec_;
|
|
std::unique_ptr<IndexTTS2CampplusStyleEncoder> style_encoder_;
|
|
std::unique_ptr<IndexTTS2GptRuntime> gpt_;
|
|
std::unique_ptr<IndexTTS2S2MelRuntime> s2mel_;
|
|
std::unique_ptr<IndexTTS2BigVganVocoder> vocoder_;
|
|
std::unique_ptr<IndexTTS2Qwen3EmotionRuntime> qwen_emotion_;
|
|
|
|
std::vector<float> speaker_matrix_;
|
|
std::vector<float> emotion_matrix_;
|
|
runtime::CacheSlots<IndexTTS2AudioIdentity, SpeakerState, AudioIdentityEqual> speaker_cache_;
|
|
runtime::CacheSlots<IndexTTS2AudioIdentity, EmotionState, AudioIdentityEqual> emotion_cache_;
|
|
runtime::CacheSlots<std::string, std::vector<float>> emotion_text_weights_cache_;
|
|
std::optional<SpeakerState> uncached_speaker_state_;
|
|
std::optional<EmotionState> uncached_emotion_state_;
|
|
};
|
|
|
|
} // namespace engine::models::index_tts2
|