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
95 lines
3.4 KiB
C++
95 lines
3.4 KiB
C++
#pragma once
|
|
|
|
#include "engine/framework/core/backend_weight_store.h"
|
|
#include "engine/framework/core/execution_context.h"
|
|
#include "engine/framework/modules/conv_modules.h"
|
|
#include "engine/framework/modules/linear_module.h"
|
|
#include "engine/framework/modules/norm_modules.h"
|
|
#include "engine/framework/modules/streaming_conv_modules.h"
|
|
#include "engine/models/index_tts2/assets.h"
|
|
#include "engine/models/index_tts2/semantic_encoder.h"
|
|
|
|
#include "ggml-backend.h"
|
|
|
|
#include <memory>
|
|
#include <vector>
|
|
|
|
namespace engine::models::index_tts2 {
|
|
|
|
struct IndexTTS2VocosConvNeXtBlockWeights {
|
|
engine::modules::DepthwiseConv1dWeights depthwise;
|
|
engine::modules::NormWeights norm;
|
|
engine::modules::LinearWeights pointwise_in;
|
|
engine::modules::LinearWeights pointwise_out;
|
|
engine::core::TensorValue gamma;
|
|
};
|
|
|
|
struct IndexTTS2VocosBackboneWeights {
|
|
engine::modules::Conv1dWeights embed;
|
|
engine::modules::NormWeights norm;
|
|
std::vector<IndexTTS2VocosConvNeXtBlockWeights> blocks;
|
|
engine::modules::NormWeights final_norm;
|
|
};
|
|
|
|
struct IndexTTS2SemanticCodecWeights {
|
|
std::shared_ptr<engine::core::BackendWeightStore> store;
|
|
IndexTTS2VocosBackboneWeights encoder_backbone;
|
|
engine::modules::LinearWeights encoder_projection;
|
|
engine::modules::Conv1dWeights quantizer_in;
|
|
engine::core::TensorValue codebook;
|
|
engine::core::TensorValue normalized_codebook;
|
|
engine::modules::Conv1dWeights quantizer_out;
|
|
IndexTTS2VocosBackboneWeights decoder_backbone;
|
|
engine::modules::LinearWeights decoder_projection;
|
|
// v2.5 only: conv applied after the 2x nearest upsample in the decode path.
|
|
engine::modules::Conv1dWeights up;
|
|
};
|
|
|
|
struct IndexTTS2SemanticCodecOutput {
|
|
std::vector<int32_t> codes;
|
|
std::vector<float> embedding_channel_first;
|
|
int64_t frames = 0;
|
|
int64_t dims = 0;
|
|
};
|
|
|
|
std::shared_ptr<const IndexTTS2SemanticCodecWeights> load_index_tts2_semantic_codec_weights(
|
|
const IndexTTS2Assets & assets,
|
|
ggml_backend_t backend,
|
|
engine::core::BackendType backend_type,
|
|
engine::assets::TensorStorageType matmul_storage_type,
|
|
engine::assets::TensorStorageType conv_storage_type,
|
|
size_t weight_context_bytes);
|
|
|
|
class IndexTTS2SemanticCodecRuntime {
|
|
public:
|
|
IndexTTS2SemanticCodecRuntime(
|
|
std::shared_ptr<const IndexTTS2Assets> assets,
|
|
engine::core::ExecutionContext & execution,
|
|
size_t graph_arena_bytes,
|
|
size_t weight_context_bytes,
|
|
engine::assets::TensorStorageType matmul_storage_type,
|
|
engine::assets::TensorStorageType conv_storage_type);
|
|
~IndexTTS2SemanticCodecRuntime();
|
|
|
|
IndexTTS2SemanticCodecRuntime(const IndexTTS2SemanticCodecRuntime &) = delete;
|
|
IndexTTS2SemanticCodecRuntime & operator=(const IndexTTS2SemanticCodecRuntime &) = delete;
|
|
|
|
void prepare_quantize(int64_t frames);
|
|
void prepare_codes(int64_t frames);
|
|
IndexTTS2SemanticCodecOutput quantize(const IndexTTS2SemanticEmbedding & semantic);
|
|
IndexTTS2SemanticCodecOutput codes_to_embedding(const std::vector<int32_t> & codes, int64_t frames);
|
|
void release_graphs();
|
|
|
|
private:
|
|
class QuantizeGraph;
|
|
class CodesGraph;
|
|
|
|
std::shared_ptr<const IndexTTS2Assets> assets_;
|
|
engine::core::ExecutionContext * execution_ = nullptr;
|
|
size_t graph_arena_bytes_ = 0;
|
|
std::shared_ptr<const IndexTTS2SemanticCodecWeights> weights_;
|
|
std::unique_ptr<QuantizeGraph> quantize_graph_;
|
|
std::unique_ptr<CodesGraph> codes_graph_;
|
|
};
|
|
|
|
} // namespace engine::models::index_tts2
|