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
55 lines
1.7 KiB
C++
55 lines
1.7 KiB
C++
#pragma once
|
|
|
|
#include "engine/framework/assets/tensor_source.h"
|
|
#include "engine/framework/core/execution_context.h"
|
|
#include "engine/framework/core/attention_fallback.h"
|
|
#include "engine/models/qwen3_asr/assets.h"
|
|
#include "engine/models/qwen3_asr/types.h"
|
|
|
|
#include <cstddef>
|
|
#include <memory>
|
|
#include <optional>
|
|
|
|
namespace engine::models::qwen3_asr {
|
|
|
|
struct Qwen3ASRPerfSettings {
|
|
std::optional<core::AttentionPreference> attention;
|
|
bool packed_projections = false;
|
|
ggml_type kv_cache_type = GGML_TYPE_F32;
|
|
int64_t kv_cache_alignment = 1;
|
|
int64_t offline_batch_size = 1;
|
|
};
|
|
|
|
class Qwen3ASRThinkerRuntime {
|
|
public:
|
|
struct Impl;
|
|
|
|
Qwen3ASRThinkerRuntime(
|
|
std::shared_ptr<const Qwen3ASRAssets> assets,
|
|
core::ExecutionContext & execution,
|
|
size_t prefill_graph_arena_bytes,
|
|
size_t decode_graph_arena_bytes,
|
|
size_t weight_context_bytes,
|
|
assets::TensorStorageType weight_storage_type,
|
|
const Qwen3ASRPerfSettings & perf = {});
|
|
~Qwen3ASRThinkerRuntime();
|
|
|
|
Qwen3ASRGeneratedTokens generate(
|
|
const Qwen3ASRPrompt & prompt,
|
|
const Qwen3ASRAudioEmbeddings & audio_embeddings,
|
|
const Qwen3ASRGenerationOptions & options);
|
|
|
|
std::vector<Qwen3ASRGeneratedTokens> generate_batch(
|
|
const std::vector<Qwen3ASRPrompt> & prompts,
|
|
const std::vector<Qwen3ASRAudioEmbeddings> & embeddings,
|
|
const std::vector<Qwen3ASRGenerationOptions> & options);
|
|
|
|
std::vector<int32_t> classify_prompt(
|
|
const Qwen3ASRPrompt & prompt,
|
|
const Qwen3ASRAudioEmbeddings & audio_embeddings);
|
|
|
|
private:
|
|
std::unique_ptr<Impl> impl_;
|
|
};
|
|
|
|
} // namespace engine::models::qwen3_asr
|