whispercpp 1.3.3 → 1.3.4

data/ext/sources/examples/talk-llama/llama-kv-cache-unified.h

@@ -1,303 +0,0 @@
-#pragma once
-
-#include "llama-batch.h"
-#include "llama-graph.h"
-#include "llama-kv-cells.h"
-#include "llama-memory.h"
-
-#include <unordered_map>
-#include <vector>
-
-struct llama_cparams;
-struct llama_hparams;
-struct llama_model;
-struct llama_context;
-
-//
-// llama_kv_cache_unified
-//
-
-class llama_kv_cache_unified : public llama_memory_i {
-public:
-    static uint32_t get_padding(const llama_cparams & cparams);
-
-    // this callback is used to filter out layers that should not be included in the cache
-    using layer_filter_cb = std::function<bool(int32_t il)>;
-
-    using ubatch_heads = std::vector<uint32_t>;
-
-    struct defrag_info {
-        bool empty() const {
-            return ids.empty();
-        }
-
-        // contains information about which cell moves where:
-        //  - cell i moves to ids[i]
-        //  - if ids[i] == i || ids[i] == ids.size(), then cell i is not moved
-        std::vector<uint32_t> ids;
-    };
-
-    llama_kv_cache_unified(
-            const llama_model &  model,
-              layer_filter_cb && filter,
-                    ggml_type    type_k,
-                    ggml_type    type_v,
-                         bool    v_trans,
-                         bool    offload,
-                     uint32_t    kv_size,
-                     uint32_t    n_seq_max,
-                     uint32_t    n_pad,
-                     uint32_t    n_swa,
-               llama_swa_type    swa_type);
-
-    ~llama_kv_cache_unified() = default;
-
-    //
-    // llama_memory_i
-    //
-
-    llama_memory_context_ptr init_batch(
-            llama_batch_allocr & balloc,
-            uint32_t n_ubatch,
-            bool embd_all) override;
-
-    llama_memory_context_ptr init_full() override;
-
-    llama_memory_context_ptr init_update(llama_context * lctx, bool optimize) override;
-
-    bool get_can_shift() const override;
-
-    void clear(bool data) override;
-
-    bool seq_rm  (llama_seq_id seq_id,                              llama_pos p0, llama_pos p1) override;
-    void seq_cp  (llama_seq_id seq_id_src, llama_seq_id seq_id_dst, llama_pos p0, llama_pos p1) override;
-    void seq_keep(llama_seq_id seq_id)                                                          override;
-    void seq_add (llama_seq_id seq_id,                              llama_pos p0, llama_pos p1, llama_pos shift) override;
-    void seq_div (llama_seq_id seq_id,                              llama_pos p0, llama_pos p1, int d) override;
-
-    llama_pos seq_pos_min(llama_seq_id seq_id) const override;
-    llama_pos seq_pos_max(llama_seq_id seq_id) const override;
-
-    // state write/load
-
-    void state_write(llama_io_write_i & io, llama_seq_id seq_id = -1) const override;
-    void state_read (llama_io_read_i  & io, llama_seq_id seq_id = -1)       override;
-
-    //
-    // llama_kv_cache_unified specific API
-    //
-
-    uint32_t get_size() const;
-
-    bool get_has_shift() const;
-
-    //
-    // graph_build API
-    //
-
-    uint32_t get_n_kv() const;
-
-    // get views of the current state of the cache
-    ggml_tensor * get_k(ggml_context * ctx, int32_t il, uint32_t n_kv) const;
-    ggml_tensor * get_v(ggml_context * ctx, int32_t il, uint32_t n_kv) const;
-
-    // store k_cur and v_cur in the cache based on the provided head location
-    ggml_tensor * cpy_k(ggml_context * ctx, ggml_tensor * k_cur, int32_t il, uint32_t head_cur) const;
-    ggml_tensor * cpy_v(ggml_context * ctx, ggml_tensor * v_cur, int32_t il, uint32_t head_cur) const;
-
-    //
-    // preparation API
-    //
-
-    // find places for the provided ubatches in the cache, returns the head locations
-    // return empty vector on failure
-    ubatch_heads prepare(const std::vector<llama_ubatch> & ubatches);
-
-    bool update(llama_context * lctx, bool do_shift, const defrag_info & dinfo);
-
-    // return the cell position where we can insert the ubatch
-    // return -1 on failure to find a contiguous slot of kv cells
-    int32_t find_slot(const llama_ubatch & ubatch) const;
-
-    // emplace the ubatch context into slot: [head_cur, head_cur + ubatch.n_tokens)
-    void apply_ubatch(uint32_t head_cur, const llama_ubatch & ubatch);
-
-    //
-    // set_input API
-    //
-
-    void set_input_kq_mask   (ggml_tensor * dst, const llama_ubatch * ubatch, bool causal_attn) const;
-    void set_input_k_shift   (ggml_tensor * dst) const;
-    void set_input_pos_bucket(ggml_tensor * dst, const llama_ubatch * ubatch) const;
-
-private:
-    const llama_model & model;
-    const llama_hparams & hparams;
-
-    struct kv_layer {
-        // layer index in the model
-        // note: can be different from the layer index in the KV cache
-        uint32_t il;
-
-        ggml_tensor * k;
-        ggml_tensor * v;
-    };
-
-    bool v_trans = true;  // the value tensor is transposed
-
-    // the current index from where we start searching for a free slot in the ring buffer of KV cells (see find_slot())
-    // note: this is not part of the KV state and it's only used to speed-up the find_slot() method
-    uint32_t head = 0;
-
-    const uint32_t n_seq_max = 1;
-
-    // required padding
-    const uint32_t n_pad = 1;
-
-    // SWA
-    const uint32_t n_swa = 0;
-
-    int debug = 0;
-
-    const llama_swa_type swa_type = LLAMA_SWA_TYPE_NONE;
-
-    std::vector<ggml_context_ptr>        ctxs;
-    std::vector<ggml_backend_buffer_ptr> bufs;
-
-    llama_kv_cells_unified cells;
-
-    std::vector<kv_layer> layers;
-
-    // model layer id -> KV cache layer id
-    std::unordered_map<int32_t, int32_t> map_layer_ids;
-
-    // return non-empty vector if cells have been moved
-    defrag_info defrag_prepare(int32_t n_max_nodes) const;
-
-    size_t total_size() const;
-
-    size_t size_k_bytes() const;
-    size_t size_v_bytes() const;
-
-    bool is_masked_swa(llama_pos p0, llama_pos p1) const;
-
-    ggml_tensor * build_rope_shift(
-            const llama_cparams & cparams,
-                   ggml_context * ctx,
-                    ggml_tensor * cur,
-                    ggml_tensor * shift,
-                    ggml_tensor * factors,
-                          float   freq_base,
-                          float   freq_scale) const;
-
-    llm_graph_result_ptr build_graph_shift(
-            const llama_cparams & cparams,
-                   ggml_context * ctx,
-                    ggml_cgraph * gf) const;
-
-    llm_graph_result_ptr build_graph_defrag(
-            const llama_cparams & cparams,
-                   ggml_context * ctx,
-                    ggml_cgraph * gf,
-              const defrag_info & dinfo) const;
-
-    void state_write_meta(llama_io_write_i & io, const std::vector<std::pair<uint32_t, uint32_t>> & cell_ranges, llama_seq_id seq_id = -1) const;
-    void state_write_data(llama_io_write_i & io, const std::vector<std::pair<uint32_t, uint32_t>> & cell_ranges) const;
-
-    bool state_read_meta(llama_io_read_i & io, uint32_t cell_count, llama_seq_id dest_seq_id = -1);
-    bool state_read_data(llama_io_read_i & io, uint32_t cell_count);
-};
-
-class llama_kv_cache_unified_context : public llama_memory_context_i {
-public:
-    // some shorthands
-    using ubatch_heads = llama_kv_cache_unified::ubatch_heads;
-    using defrag_info  = llama_kv_cache_unified::defrag_info;
-
-    // used for errors
-    llama_kv_cache_unified_context(llama_memory_status status);
-
-    // used to create a full-cache context
-    llama_kv_cache_unified_context(
-            llama_kv_cache_unified * kv);
-
-    // used to create an update context
-    llama_kv_cache_unified_context(
-            llama_kv_cache_unified * kv,
-            llama_context * lctx,
-            bool do_shift,
-            defrag_info dinfo);
-
-    // used to create a batch procesing context from a batch
-    llama_kv_cache_unified_context(
-            llama_kv_cache_unified * kv,
-            ubatch_heads heads,
-            std::vector<llama_ubatch> ubatches);
-
-    virtual ~llama_kv_cache_unified_context();
-
-    //
-    // llama_memory_context_i
-    //
-
-    bool next()  override;
-    bool apply() override;
-
-    llama_memory_status  get_status() const override;
-    const llama_ubatch & get_ubatch() const override;
-
-    //
-    // llama_kv_cache_unified_context specific API
-    //
-
-    uint32_t get_n_kv() const;
-
-    // get views of the current state of the cache
-    ggml_tensor * get_k(ggml_context * ctx, int32_t il) const;
-    ggml_tensor * get_v(ggml_context * ctx, int32_t il) const;
-
-    // store k_cur and v_cur in the cache based on the provided head location
-    ggml_tensor * cpy_k(ggml_context * ctx, ggml_tensor * k_cur, int32_t il) const;
-    ggml_tensor * cpy_v(ggml_context * ctx, ggml_tensor * v_cur, int32_t il) const;
-
-    void set_input_k_shift(ggml_tensor * dst) const;
-
-    void set_input_kq_mask   (ggml_tensor * dst, const llama_ubatch * ubatch, bool causal_attn) const;
-    void set_input_pos_bucket(ggml_tensor * dst, const llama_ubatch * ubatch) const;
-
-private:
-    llama_memory_status status;
-
-    llama_kv_cache_unified * kv;
-    llama_context * lctx;
-
-    //
-    // update context
-    //
-
-    bool do_shift = false;
-
-    defrag_info dinfo;
-
-    //
-    // batch processing context
-    //
-
-    // the index of the next ubatch to process
-    size_t i_next = 0;
-
-    ubatch_heads heads;
-
-    std::vector<llama_ubatch> ubatches;
-
-    //
-    // data needed for building the compute graph for the current ubatch:
-    //
-
-    // a heuristic, to avoid attending the full cache if it is not yet utilized
-    // as the cache gets filled, the benefit from this heuristic disappears
-    int32_t n_kv;
-
-    // the beginning of the current slot in which the ubatch will be inserted
-    int32_t head;
-};

data/ext/sources/ggml/include/ggml-kompute.h

@@ -1,50 +0,0 @@
-#pragma once
-
-#include "ggml.h"
-#include "ggml-backend.h"
-
-#include <stdbool.h>
-#include <stddef.h>
-#include <stdint.h>
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-#define GGML_KOMPUTE_MAX_DEVICES 16
-
-struct ggml_vk_device {
-    int index;
-    int type; // same as VkPhysicalDeviceType
-    size_t heapSize;
-    const char * name;
-    const char * vendor;
-    int subgroupSize;
-    uint64_t bufferAlignment;
-    uint64_t maxAlloc;
-};
-
-struct ggml_vk_device * ggml_vk_available_devices(size_t memoryRequired, size_t * count);
-bool ggml_vk_get_device(struct ggml_vk_device * device, size_t memoryRequired, const char * name);
-bool ggml_vk_has_vulkan(void);
-bool ggml_vk_has_device(void);
-struct ggml_vk_device ggml_vk_current_device(void);
-
-//
-// backend API
-//
-
-// forward declaration
-typedef struct ggml_backend * ggml_backend_t;
-
-GGML_BACKEND_API ggml_backend_t ggml_backend_kompute_init(int device);
-
-GGML_BACKEND_API bool ggml_backend_is_kompute(ggml_backend_t backend);
-
-GGML_BACKEND_API ggml_backend_buffer_type_t ggml_backend_kompute_buffer_type(int device);
-
-GGML_BACKEND_API ggml_backend_reg_t ggml_backend_kompute_reg(void);
-
-#ifdef __cplusplus
-}
-#endif

data/ext/sources/ggml/src/ggml-amx/CMakeLists.txt

@@ -1,107 +0,0 @@
-if (CMAKE_OSX_ARCHITECTURES STREQUAL "x86_64" OR CMAKE_GENERATOR_PLATFORM_LWR MATCHES "^(x86_64|i686|amd64|x64|win32)$" OR
-        (NOT CMAKE_OSX_ARCHITECTURES AND NOT CMAKE_GENERATOR_PLATFORM_LWR AND
-         CMAKE_SYSTEM_PROCESSOR MATCHES "^(x86_64|i686|AMD64)$") AND
-        CMAKE_COMPILER_IS_GNUCC AND CMAKE_CXX_COMPILER_VERSION VERSION_GREATER 11.0)
-    message(STATUS "Using AMX")
-
-    file(GLOB   GGML_HEADERS_AMX "*.h")
-    list(APPEND GGML_HEADERS_AMX "../../include/ggml-amx.h")
-
-    file(GLOB   GGML_SOURCES_AMX "*.cpp")
-
-    add_library(ggml-amx
-                ${GGML_HEADERS_AMX}
-                ${GGML_SOURCES_AMX})
-
-    target_link_libraries(ggml-amx PRIVATE ggml-base)
-    target_include_directories(ggml-amx PRIVATE . ..)
-
-    # this is duplicated from the CPU backend, since the AMX backend also depends on the architecture flags
-    # TODO: integrate AMX backend into the CPU backend
-    if (MSVC)
-        # instruction set detection for MSVC only
-        if (GGML_NATIVE)
-            # TODO: improve, should not reference files from the parent folder
-            include(../ggml-cpu/cmake/FindSIMD.cmake)
-        endif ()
-        if (GGML_AVX512)
-            list(APPEND ARCH_FLAGS /arch:AVX512)
-            # MSVC has no compile-time flags enabling specific
-            # AVX512 extensions, neither it defines the
-            # macros corresponding to the extensions.
-            # Do it manually.
-            if (GGML_AVX512_VBMI)
-                add_compile_definitions($<$<COMPILE_LANGUAGE:C>:__AVX512VBMI__>)
-                add_compile_definitions($<$<COMPILE_LANGUAGE:CXX>:__AVX512VBMI__>)
-            endif()
-            if (GGML_AVX512_VNNI)
-                add_compile_definitions($<$<COMPILE_LANGUAGE:C>:__AVX512VNNI__>)
-                add_compile_definitions($<$<COMPILE_LANGUAGE:CXX>:__AVX512VNNI__>)
-            endif()
-            if (GGML_AVX512_BF16)
-                add_compile_definitions($<$<COMPILE_LANGUAGE:C>:__AVX512BF16__>)
-                add_compile_definitions($<$<COMPILE_LANGUAGE:CXX>:__AVX512BF16__>)
-            endif()
-            if (GGML_AMX_TILE)
-                add_compile_definitions($<$<COMPILE_LANGUAGE:C>:__AMX_TILE__>)
-                add_compile_definitions($<$<COMPILE_LANGUAGE:CXX>:__AMX_TILE__>)
-            endif()
-            if (GGML_AMX_INT8)
-                add_compile_definitions($<$<COMPILE_LANGUAGE:C>:__AMX_INT8__>)
-                add_compile_definitions($<$<COMPILE_LANGUAGE:CXX>:__AMX_INT8__>)
-            endif()
-            if (GGML_AMX_BF16)
-                add_compile_definitions($<$<COMPILE_LANGUAGE:C>:__AMX_BF16__>)
-                add_compile_definitions($<$<COMPILE_LANGUAGE:CXX>:__AMX_BF16__>)
-            endif()
-        elseif (GGML_AVX2)
-            list(APPEND ARCH_FLAGS /arch:AVX2)
-        elseif (GGML_AVX)
-            list(APPEND ARCH_FLAGS /arch:AVX)
-        endif()
-    else()
-        if (GGML_NATIVE)
-            list(APPEND ARCH_FLAGS -march=native)
-        endif()
-        if (GGML_F16C)
-            list(APPEND ARCH_FLAGS -mf16c)
-        endif()
-        if (GGML_FMA)
-            list(APPEND ARCH_FLAGS -mfma)
-        endif()
-        if (GGML_AVX)
-            list(APPEND ARCH_FLAGS -mavx)
-        endif()
-        if (GGML_AVX2)
-            list(APPEND ARCH_FLAGS -mavx2)
-        endif()
-        if (GGML_AVX512)
-            list(APPEND ARCH_FLAGS -mavx512f)
-            list(APPEND ARCH_FLAGS -mavx512dq)
-            list(APPEND ARCH_FLAGS -mavx512bw)
-        endif()
-        if (GGML_AVX512_VBMI)
-            list(APPEND ARCH_FLAGS -mavx512vbmi)
-        endif()
-        if (GGML_AVX512_VNNI)
-            list(APPEND ARCH_FLAGS -mavx512vnni)
-        endif()
-        if (GGML_AVX512_BF16)
-            list(APPEND ARCH_FLAGS -mavx512bf16)
-        endif()
-        if (GGML_AMX_TILE)
-            list(APPEND ARCH_FLAGS -mamx-tile)
-        endif()
-        if (GGML_AMX_INT8)
-            list(APPEND ARCH_FLAGS -mamx-int8)
-        endif()
-        if (GGML_AMX_BF16)
-            list(APPEND ARCH_FLAGS -mamx-bf16)
-        endif()
-    endif()
-
-    target_compile_options(ggml-amx PRIVATE ${ARCH_FLAGS})
-else()
-    set(GGML_AMX OFF PARENT_SCOPE)
-    message(WARNING "AMX requires x86 and gcc version > 11.0. Turning off GGML_AMX.")
-endif()

data/ext/sources/ggml/src/ggml-amx/common.h

@@ -1,94 +0,0 @@
-#pragma once
-
-#include "ggml.h"
-// hack until AMX is moved into the CPU backend
-#include "../ggml-cpu/ggml-cpu-impl.h" // <immintrin.h>
-
-#include <algorithm>
-#include <memory>
-#include <type_traits>
-
-#if defined(_OPENMP)
-#include <omp.h>
-#endif
-
-#define TILE_M 16
-#define TILE_N 16
-#define TILE_K 32
-#define VNNI_BLK 4
-
-#define AMX_BLK_SIZE 32
-
-#define TMM0 0
-#define TMM1 1
-#define TMM2 2
-#define TMM3 3
-#define TMM4 4
-#define TMM5 5
-#define TMM6 6
-#define TMM7 7
-
-// parallel routines
-template <typename T, typename std::enable_if<std::is_integral<T>::value, int>::type = 0>
-inline T div_up(T x, T y) { return (x + y - 1) / y; }
-
-template <typename T>
-inline void balance211(T n, T nth, T ith, T& n_start, T& n_end) {
-#if 0
-    // onednn partition pattern
-    T& n_my = n_end;
-    if (nth <= 1 || n == 0) {
-        n_start = 0;
-        n_my = n;
-    } else {
-        T n1 = div_up(n, nth);
-        T n2 = n1 - 1;
-        T T1 = n - n2 * nth;
-        n_my = ith < T1 ? n1 : n2;
-        n_start = ith <= T1 ? ith*n1 : T1 * n1 + (ith - T1) * n2;
-    }
-    n_end += n_start;
-#else
-    // pytorch aten partition pattern
-    T n_my = div_up(n, nth);
-    n_start = ith * n_my;
-    n_end = std::min(n_start + n_my, n);
-#endif
-}
-
-template <typename func_t>
-inline void parallel_for(int nth, int n, const func_t& f) {
-#if defined(_OPENMP)
-#pragma omp parallel num_threads(nth)
-{
-    //int nth = omp_get_num_threads();
-    int ith = omp_get_thread_num();
-    int tbegin, tend;
-    balance211(n, nth, ith, tbegin, tend);
-    f(tbegin, tend);
-}
-#else
-    f(0, n);
-
-    GGML_UNUSED(nth);
-#endif
-}
-
-// quantized types that have AMX support
-inline bool qtype_has_amx_kernels(const enum ggml_type type) {
-    // TODO: fix padding for vnni format
-    return (type == GGML_TYPE_Q4_0) ||
-        (type == GGML_TYPE_Q4_1);
-        //(type == GGML_TYPE_Q8_0) ||
-        //(type == GGML_TYPE_Q4_K) ||
-        //(type == GGML_TYPE_Q5_K) ||
-        //(type == GGML_TYPE_Q6_K) ||
-        //(type == GGML_TYPE_IQ4_XS);
-}
-
-// ggml backend context
-struct ggml_backend_amx_context {
-    int n_threads = GGML_DEFAULT_N_THREADS;
-    std::unique_ptr<char[]> work_data;
-    size_t work_size = 0;
-};