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
72 lines
2.4 KiB
C++
72 lines
2.4 KiB
C++
#pragma once
|
|
|
|
#include "engine/framework/core/backend_weight_store.h"
|
|
#include "engine/framework/modules/feed_forward_modules.h"
|
|
#include "engine/framework/modules/attention/types.h"
|
|
#include "engine/framework/modules/conv_modules.h"
|
|
#include "engine/framework/modules/linear_module.h"
|
|
#include "engine/framework/modules/norm_modules.h"
|
|
#include "engine/models/moonshine_asr/assets.h"
|
|
|
|
#include <memory>
|
|
#include <optional>
|
|
#include <vector>
|
|
|
|
namespace engine::models::moonshine_asr {
|
|
|
|
struct MoonshineFrontendWeights {
|
|
float log_k = 0.0F;
|
|
engine::modules::LinearWeights linear;
|
|
engine::modules::Conv1dWeights conv1;
|
|
engine::modules::Conv1dWeights conv2;
|
|
};
|
|
|
|
struct MoonshineEncoderLayerWeights {
|
|
engine::modules::NormWeights input_norm;
|
|
engine::modules::AttentionWeights self_attn;
|
|
engine::modules::NormWeights post_attention_norm;
|
|
engine::modules::FeedForwardWeights feed_forward;
|
|
};
|
|
|
|
struct MoonshineEncoderWeights {
|
|
MoonshineFrontendWeights frontend;
|
|
std::vector<MoonshineEncoderLayerWeights> layers;
|
|
engine::modules::NormWeights final_norm;
|
|
};
|
|
|
|
struct MoonshineDecoderLayerWeights {
|
|
engine::modules::NormWeights input_norm;
|
|
engine::modules::AttentionWeights self_attn;
|
|
engine::modules::NormWeights post_attention_norm;
|
|
engine::modules::AttentionWeights cross_attn;
|
|
engine::modules::NormWeights final_norm;
|
|
engine::modules::LinearWeights mlp_fc1;
|
|
engine::modules::LinearWeights mlp_fc2;
|
|
};
|
|
|
|
struct MoonshineDecoderWeights {
|
|
engine::core::TensorValue token_embedding;
|
|
engine::core::TensorValue position_embedding;
|
|
std::optional<engine::modules::LinearWeights> adapter_proj;
|
|
std::vector<MoonshineDecoderLayerWeights> layers;
|
|
engine::modules::NormWeights norm;
|
|
engine::modules::LinearWeights output_projection;
|
|
};
|
|
|
|
struct MoonshineWeights {
|
|
std::shared_ptr<engine::core::BackendWeightStore> store;
|
|
MoonshineEncoderWeights encoder;
|
|
MoonshineDecoderWeights decoder;
|
|
};
|
|
|
|
std::shared_ptr<const MoonshineWeights> load_moonshine_asr_weights(
|
|
const MoonshineAssets & assets,
|
|
ggml_backend_t backend,
|
|
engine::core::BackendType backend_type,
|
|
engine::assets::TensorStorageType encoder_storage_type,
|
|
engine::assets::TensorStorageType decoder_storage_type,
|
|
engine::assets::TensorStorageType conv_storage_type,
|
|
size_t weight_context_bytes,
|
|
bool include_decoder = true);
|
|
|
|
} // namespace engine::models::moonshine_asr
|