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
80 lines
2.3 KiB
C++
80 lines
2.3 KiB
C++
#pragma once
|
|
|
|
#include "engine/framework/assets/resource_bundle.h"
|
|
#include "engine/framework/assets/tensor_source.h"
|
|
|
|
#include <cstdint>
|
|
#include <memory>
|
|
|
|
namespace engine::models::glm_tts {
|
|
|
|
struct GlmTTSLlamaConfig {
|
|
int64_t bos_token_id = 1;
|
|
int64_t eos_token_id = 2;
|
|
int64_t hidden_size = 0;
|
|
int64_t intermediate_size = 0;
|
|
int64_t max_position_embeddings = 0;
|
|
int64_t num_attention_heads = 0;
|
|
int64_t num_hidden_layers = 0;
|
|
int64_t num_key_value_heads = 0;
|
|
int64_t head_dim = 0;
|
|
int64_t vocab_size = 0;
|
|
float rms_norm_eps = 1.0e-5F;
|
|
float rope_theta = 10000.0F;
|
|
};
|
|
|
|
struct GlmTTSSpeechTokenizerConfig {
|
|
int64_t feature_size = 128;
|
|
int64_t sampling_rate = 16000;
|
|
int64_t d_model = 0;
|
|
int64_t encoder_attention_heads = 0;
|
|
int64_t encoder_ffn_dim = 0;
|
|
int64_t encoder_layers = 0;
|
|
int64_t max_source_positions = 0;
|
|
int64_t pooling_kernel_size = 0;
|
|
int64_t pooling_position = 0;
|
|
int64_t quantize_position = 0;
|
|
int64_t quantize_vocab_size = 0;
|
|
};
|
|
|
|
struct GlmTTSFlowConfig {
|
|
int64_t speech_token_dim = 512;
|
|
int64_t vocab_size = 100000;
|
|
int64_t mel_dim = 80;
|
|
int64_t trans_dim = 768;
|
|
int64_t depth = 18;
|
|
int64_t heads = 12;
|
|
int64_t dim_head = 64;
|
|
int64_t conv_layers = 4;
|
|
int64_t mel_framerate = 50;
|
|
float input_frame_rate = 25.0F;
|
|
float inference_cfg_rate = 0.7F;
|
|
int64_t inference_steps = 10;
|
|
};
|
|
|
|
struct GlmTTSConfig {
|
|
GlmTTSLlamaConfig llama;
|
|
GlmTTSSpeechTokenizerConfig speech_tokenizer;
|
|
GlmTTSFlowConfig flow;
|
|
int64_t sample_rate = 24000;
|
|
int64_t audio_token_start = 0;
|
|
int64_t audio_token_end = 0;
|
|
int64_t begin_audio_token = 0;
|
|
int64_t end_audio_token = 0;
|
|
int64_t pad_token = 0;
|
|
};
|
|
|
|
struct GlmTTSAssets {
|
|
GlmTTSConfig config;
|
|
assets::ResourceBundle resources;
|
|
std::shared_ptr<const assets::TensorSource> llama_weights;
|
|
std::shared_ptr<const assets::TensorSource> speech_tokenizer_weights;
|
|
std::shared_ptr<const assets::TensorSource> flow_weights;
|
|
std::shared_ptr<const assets::TensorSource> hift_weights;
|
|
std::shared_ptr<const assets::TensorSource> campplus_weights;
|
|
};
|
|
|
|
std::shared_ptr<const GlmTTSAssets> load_glm_tts_assets(
|
|
const std::filesystem::path & model_path);
|
|
|
|
} // namespace engine::models::glm_tts
|