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
75 lines
2.6 KiB
C++
75 lines
2.6 KiB
C++
#pragma once
|
|
|
|
#include "engine/models/midashenglm_gen/assets.h"
|
|
#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/framework/runtime/session.h"
|
|
|
|
#include <memory>
|
|
#include <vector>
|
|
|
|
namespace engine::models::midashenglm_gen {
|
|
|
|
struct MiDashengLmGenAudioTokenizerInput {
|
|
std::vector<float> latents;
|
|
int64_t batch = 0;
|
|
int64_t frames = 0;
|
|
int64_t dims = 0;
|
|
};
|
|
|
|
struct MiDashengLmGenAudioTokenizerOutput {
|
|
std::vector<engine::runtime::AudioBuffer> audio;
|
|
};
|
|
|
|
struct MiDashengLmGenVocosBlockWeights {
|
|
engine::modules::DepthwiseConv1dWeights depthwise;
|
|
engine::modules::NormWeights norm;
|
|
engine::modules::LinearWeights pointwise_in;
|
|
engine::modules::LinearWeights pointwise_out;
|
|
engine::core::TensorValue gamma;
|
|
};
|
|
|
|
struct MiDashengLmGenAudioTokenizerWeights {
|
|
std::shared_ptr<engine::core::BackendWeightStore> store;
|
|
engine::modules::ConvTranspose1dWeights upsampler;
|
|
engine::modules::Conv1dWeights embed;
|
|
engine::modules::NormWeights norm;
|
|
std::vector<MiDashengLmGenVocosBlockWeights> blocks;
|
|
engine::modules::NormWeights final_norm;
|
|
engine::modules::LinearWeights head;
|
|
std::vector<float> istft_window;
|
|
};
|
|
|
|
class MiDashengLmGenAudioTokenizerRuntime {
|
|
public:
|
|
MiDashengLmGenAudioTokenizerRuntime(
|
|
std::shared_ptr<const MiDashengLmGenAssets> 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);
|
|
~MiDashengLmGenAudioTokenizerRuntime();
|
|
|
|
MiDashengLmGenAudioTokenizerRuntime(const MiDashengLmGenAudioTokenizerRuntime &) = delete;
|
|
MiDashengLmGenAudioTokenizerRuntime & operator=(const MiDashengLmGenAudioTokenizerRuntime &) = delete;
|
|
|
|
void prepare_decode(int64_t batch, int64_t frames);
|
|
MiDashengLmGenAudioTokenizerOutput decode(const MiDashengLmGenAudioTokenizerInput & input);
|
|
void release_graphs();
|
|
|
|
private:
|
|
class DecodeGraph;
|
|
|
|
std::shared_ptr<const MiDashengLmGenAssets> assets_;
|
|
engine::core::ExecutionContext * execution_ = nullptr;
|
|
size_t graph_arena_bytes_ = 0;
|
|
std::shared_ptr<const MiDashengLmGenAudioTokenizerWeights> weights_;
|
|
std::unique_ptr<DecodeGraph> decode_graph_;
|
|
};
|
|
|
|
} // namespace engine::models::midashenglm_gen
|