Brunobkr/llama.cpp_AlgMor24_github

1#pragma once2 3//4// GGML Tensor Library5//6// This documentation is still a work in progress.7// If you wish some specific topics to be covered, feel free to drop a comment:8//9//   https://github.com/ggml-org/whisper.cpp/issues/4010//11// ## Overview12//13// This library implements:14//15//  - a set of tensor operations16//  - automatic differentiation17//  - basic optimization algorithms18//19// The aim of this library is to provide a minimalistic approach for various machine learning tasks. This includes,20// but is not limited to, the following:21//22//  - linear regression23//  - support vector machines24//  - neural networks25//26// The library allows the user to define a certain function using the available tensor operations. This function27// definition is represented internally via a computation graph. Each tensor operation in the function definition28// corresponds to a node in the graph. Having the computation graph defined, the user can choose to compute the29// function's value and/or its gradient with respect to the input variables. Optionally, the function can be optimized30// using one of the available optimization algorithms.31//32// For example, here we define the function: f(x) = a*x^2 + b33//34//   {35//       struct ggml_init_params params = {36//           .mem_size   = 16*1024*1024,37//           .mem_buffer = NULL,38//       };39//40//       // memory allocation happens here41//       struct ggml_context * ctx = ggml_init(params);42//43//       struct ggml_tensor * x = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, 1);44//45//       ggml_set_param(ctx, x); // x is an input variable46//47//       struct ggml_tensor * a  = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, 1);48//       struct ggml_tensor * b  = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, 1);49//       struct ggml_tensor * x2 = ggml_mul(ctx, x, x);50//       struct ggml_tensor * f  = ggml_add(ctx, ggml_mul(ctx, a, x2), b);51//52//       ...53//   }54//55// Notice that the function definition above does not involve any actual computation. The computation is performed only56// when the user explicitly requests it. For example, to compute the function's value at x = 2.0:57//58//   {59//       ...60//61//       struct ggml_cgraph * gf = ggml_new_graph(ctx);62//       ggml_build_forward_expand(gf, f);63//64//       // set the input variable and parameter values65//       ggml_set_f32(x, 2.0f);66//       ggml_set_f32(a, 3.0f);67//       ggml_set_f32(b, 4.0f);68//69//       ggml_graph_compute_with_ctx(ctx, &gf, n_threads);70//71//       printf("f = %f\n", ggml_get_f32_1d(f, 0));72//73//       ...74//   }75//76// The actual computation is performed in the ggml_graph_compute() function.77//78// The ggml_new_tensor_...() functions create new tensors. They are allocated in the memory buffer provided to the79// ggml_init() function. You have to be careful not to exceed the memory buffer size. Therefore, you have to know80// in advance how much memory you need for your computation. Alternatively, you can allocate a large enough memory81// and after defining the computation graph, call the ggml_used_mem() function to find out how much memory was82// actually needed.83//84// The ggml_set_param() function marks a tensor as an input variable. This is used by the automatic85// differentiation and optimization algorithms.86//87// The described approach allows to define the function graph once and then compute its forward or backward graphs88// multiple times. All computations will use the same memory buffer allocated in the ggml_init() function. This way89// the user can avoid the memory allocation overhead at runtime.90//91// The library supports multi-dimensional tensors - up to 4 dimensions. The FP16 and FP32 data types are first class92// citizens, but in theory the library can be extended to support FP8 and integer data types.93//94// Each tensor operation produces a new tensor. Initially the library was envisioned to support only the use of unary95// and binary operations. Most of the available operations fall into one of these two categories. With time, it became96// clear that the library needs to support more complex operations. The way to support these operations is not clear97// yet, but a few examples are demonstrated in the following operations:98//99//   - ggml_permute()100//   - ggml_conv_1d_1s()101//   - ggml_conv_1d_2s()102//103// For each tensor operator, the library implements a forward and backward computation function. The forward function104// computes the output tensor value given the input tensor values. The backward function computes the adjoint of the105// input tensors given the adjoint of the output tensor. For a detailed explanation of what this means, take a106// calculus class, or watch the following video:107//108//   What is Automatic Differentiation?109//   https://www.youtube.com/watch?v=wG_nF1awSSY110//111//112// ## Tensor data (struct ggml_tensor)113//114// The tensors are stored in memory via the ggml_tensor struct. The structure provides information about the size of115// the tensor, the data type, and the memory buffer where the tensor data is stored. Additionally, it contains116// pointers to the "source" tensors - i.e. the tensors that were used to compute the current tensor. For example:117//118//   {119//       struct ggml_tensor * c = ggml_add(ctx, a, b);120//121//       assert(c->src[0] == a);122//       assert(c->src[1] == b);123//   }124//125// The multi-dimensional tensors are stored in row-major order. The ggml_tensor struct contains fields for the126// number of elements in each dimension ("ne") as well as the number of bytes ("nb", a.k.a. stride). This allows127// to store tensors that are not contiguous in memory, which is useful for operations such as transposition and128// permutation. All tensor operations have to take the stride into account and not assume that the tensor is129// contiguous in memory.130//131// The data of the tensor is accessed via the "data" pointer. For example:132//133//   {134//       const int nx = 2;135//       const int ny = 3;136//137//       struct ggml_tensor * a = ggml_new_tensor_2d(ctx, GGML_TYPE_F32, nx, ny);138//139//       for (int y = 0; y < ny; y++) {140//           for (int x = 0; x < nx; x++) {141//               *(float *) ((char *) a->data + y*a->nb[1] + x*a->nb[0]) = x + y;142//           }143//       }144//145//       ...146//   }147//148// Alternatively, there are helper functions, such as ggml_get_f32_1d() and ggml_set_f32_1d() that can be used.149//150// ## The matrix multiplication operator (ggml_mul_mat)151//152// TODO153//154//155// ## Multi-threading156//157// TODO158//159//160// ## Overview of ggml.c161//162// TODO163//164//165// ## SIMD optimizations166//167// TODO168//169//170// ## Debugging ggml171//172// TODO173//174//175 176#ifdef GGML_SHARED177#    if defined(_WIN32) && !defined(__MINGW32__)178#        ifdef GGML_BUILD179#            define GGML_API __declspec(dllexport) extern180#        else181#            define GGML_API __declspec(dllimport) extern182#        endif183#    else184#        define GGML_API __attribute__ ((visibility ("default"))) extern185#    endif186#else187#    define GGML_API extern188#endif189 190// TODO: support for clang191#ifdef __GNUC__192#    define GGML_DEPRECATED(func, hint) func __attribute__((deprecated(hint)))193#elif defined(_MSC_VER)194#    define GGML_DEPRECATED(func, hint) __declspec(deprecated(hint)) func195#else196#    define GGML_DEPRECATED(func, hint) func197#endif198 199#ifndef __GNUC__200#    define GGML_ATTRIBUTE_FORMAT(...)201#elif defined(__MINGW32__) && !defined(__clang__)202#    define GGML_ATTRIBUTE_FORMAT(...) __attribute__((format(gnu_printf, __VA_ARGS__)))203#else204#    define GGML_ATTRIBUTE_FORMAT(...) __attribute__((format(printf, __VA_ARGS__)))205#endif206 207#if defined(_WIN32) && !defined(_WIN32_WINNT)208#    define _WIN32_WINNT 0x0A00209#endif210 211#include <stdbool.h>212#include <stddef.h>213#include <stdint.h>214#include <stdio.h>215 216#define GGML_FILE_MAGIC   0x67676d6c // "ggml"217#define GGML_FILE_VERSION 2218 219#define GGML_QNT_VERSION        2    // bump this on quantization format changes220#define GGML_QNT_VERSION_FACTOR 1000 // do not change this221 222#define GGML_MAX_DIMS           4223#define GGML_MAX_PARAMS         2048224#define GGML_MAX_SRC            10225#define GGML_MAX_N_THREADS      512226#define GGML_MAX_OP_PARAMS      64227 228#ifndef GGML_MAX_NAME229#   define GGML_MAX_NAME        64230#endif231 232#define GGML_DEFAULT_N_THREADS  4233#define GGML_DEFAULT_GRAPH_SIZE 2048234 235#if UINTPTR_MAX == 0xFFFFFFFF236    #define GGML_MEM_ALIGN 4237#elif defined(__EMSCRIPTEN__)238// emscripten uses max_align_t == 8, so we need GGML_MEM_ALIGN == 8 for 64-bit wasm.239// (for 32-bit wasm, the first conditional is true and GGML_MEM_ALIGN stays 4.)240// ref: https://github.com/ggml-org/llama.cpp/pull/18628241    #define GGML_MEM_ALIGN 8242#else243    #define GGML_MEM_ALIGN 16244#endif245 246#define GGML_EXIT_SUCCESS 0247#define GGML_EXIT_ABORTED 1248 249// TODO: convert to enum https://github.com/ggml-org/llama.cpp/pull/16187#discussion_r2388538726250#define GGML_ROPE_TYPE_NORMAL 0251#define GGML_ROPE_TYPE_NEOX   2252#define GGML_ROPE_TYPE_MROPE  8253#define GGML_ROPE_TYPE_VISION 24254#define GGML_ROPE_TYPE_IMROPE 40 // binary: 101000255 256#define GGML_MROPE_SECTIONS   4257 258#define GGML_UNUSED(x) (void)(x)259#ifdef __CUDACC__260template<typename... Args>261__host__ __device__ constexpr inline void ggml_unused_vars_impl(Args&&...) noexcept {}262#define GGML_UNUSED_VARS(...) ggml_unused_vars_impl(__VA_ARGS__)263#else264#define GGML_UNUSED_VARS(...) do { (void)sizeof((__VA_ARGS__, 0)); } while(0)265#endif // __CUDACC__266 267#define GGML_PAD(x, n) (((x) + (n) - 1) & ~((n) - 1))268 269#ifndef NDEBUG270#   define GGML_UNREACHABLE() do { fprintf(stderr, "statement should be unreachable\n"); abort(); } while(0)271#elif defined(__GNUC__)272#   define GGML_UNREACHABLE() __builtin_unreachable()273#elif defined(_MSC_VER)274#   define GGML_UNREACHABLE() __assume(0)275#else276#   define GGML_UNREACHABLE() ((void) 0)277#endif278 279#ifdef __cplusplus280#   define GGML_NORETURN [[noreturn]]281#elif defined(_MSC_VER)282#   define GGML_NORETURN __declspec(noreturn)283#else284#   define GGML_NORETURN _Noreturn285#endif286 287#define GGML_ABORT(...) ggml_abort(__FILE__, __LINE__, __VA_ARGS__)288#define GGML_ASSERT(x) if (!(x)) GGML_ABORT("GGML_ASSERT(%s) failed", #x)289 290// used to copy the number of elements and stride in bytes of tensors into local variables.291// main purpose is to reduce code duplication and improve readability.292//293// example:294//295//    GGML_TENSOR_LOCALS(int64_t, ne1, src1, ne);296//    GGML_TENSOR_LOCALS(size_t,  nb1, src1, nb);297//298#define GGML_TENSOR_LOCALS_1(type, prefix, pointer, array) \299    const type prefix##0 = (pointer) ? (pointer)->array[0] : 0; \300    GGML_UNUSED(prefix##0);301#define GGML_TENSOR_LOCALS_2(type, prefix, pointer, array) \302    GGML_TENSOR_LOCALS_1    (type, prefix, pointer, array) \303    const type prefix##1 = (pointer) ? (pointer)->array[1] : 0; \304    GGML_UNUSED(prefix##1);305#define GGML_TENSOR_LOCALS_3(type, prefix, pointer, array) \306    GGML_TENSOR_LOCALS_2    (type, prefix, pointer, array) \307    const type prefix##2 = (pointer) ? (pointer)->array[2] : 0; \308    GGML_UNUSED(prefix##2);309#define GGML_TENSOR_LOCALS(type, prefix, pointer, array) \310    GGML_TENSOR_LOCALS_3  (type, prefix, pointer, array) \311    const type prefix##3 = (pointer) ? (pointer)->array[3] : 0; \312    GGML_UNUSED(prefix##3);313 314#define GGML_TENSOR_UNARY_OP_LOCALS \315    GGML_TENSOR_LOCALS(int64_t, ne0, src0, ne) \316    GGML_TENSOR_LOCALS(size_t,  nb0, src0, nb) \317    GGML_TENSOR_LOCALS(int64_t, ne,  dst,  ne) \318    GGML_TENSOR_LOCALS(size_t,  nb,  dst,  nb)319 320#define GGML_TENSOR_BINARY_OP_LOCALS \321    GGML_TENSOR_LOCALS(int64_t, ne0, src0, ne) \322    GGML_TENSOR_LOCALS(size_t,  nb0, src0, nb) \323    GGML_TENSOR_LOCALS(int64_t, ne1, src1, ne) \324    GGML_TENSOR_LOCALS(size_t,  nb1, src1, nb) \325    GGML_TENSOR_LOCALS(int64_t, ne,  dst,  ne) \326    GGML_TENSOR_LOCALS(size_t,  nb,  dst,  nb)327 328#define GGML_TENSOR_TERNARY_OP_LOCALS \329    GGML_TENSOR_LOCALS(int64_t, ne0, src0, ne) \330    GGML_TENSOR_LOCALS(size_t,  nb0, src0, nb) \331    GGML_TENSOR_LOCALS(int64_t, ne1, src1, ne) \332    GGML_TENSOR_LOCALS(size_t,  nb1, src1, nb) \333    GGML_TENSOR_LOCALS(int64_t, ne2, src2, ne) \334    GGML_TENSOR_LOCALS(size_t,  nb2, src2, nb) \335    GGML_TENSOR_LOCALS(int64_t, ne,  dst,  ne) \336    GGML_TENSOR_LOCALS(size_t,  nb,  dst,  nb)337 338#define GGML_TENSOR_BINARY_OP_LOCALS01 \339    GGML_TENSOR_LOCALS(int64_t, ne0, src0, ne) \340    GGML_TENSOR_LOCALS(size_t,  nb0, src0, nb) \341    GGML_TENSOR_LOCALS(int64_t, ne1, src1, ne) \342    GGML_TENSOR_LOCALS(size_t,  nb1, src1, nb)343 344#ifdef  __cplusplus345extern "C" {346#endif347 348    // Function type used in fatal error callbacks349    typedef void (*ggml_abort_callback_t)(const char * error_message);350 351    // Set the abort callback (passing null will restore original abort functionality: printing a message to stdout)352    // Returns the old callback for chaining353    GGML_API ggml_abort_callback_t ggml_set_abort_callback(ggml_abort_callback_t callback);354 355    GGML_NORETURN GGML_ATTRIBUTE_FORMAT(3, 4)356    GGML_API void ggml_abort(const char * file, int line, const char * fmt, ...);357 358    enum ggml_status {359        GGML_STATUS_ALLOC_FAILED = -2,360        GGML_STATUS_FAILED = -1,361        GGML_STATUS_SUCCESS = 0,362        GGML_STATUS_ABORTED = 1,363    };364 365    // get ggml_status name string366    GGML_API const char * ggml_status_to_string(enum ggml_status status);367 368    // ieee 754-2008 half-precision float16369    // todo: make this not an integral type370    typedef uint16_t ggml_fp16_t;371    GGML_API float       ggml_fp16_to_fp32(ggml_fp16_t);372    GGML_API ggml_fp16_t ggml_fp32_to_fp16(float);373    GGML_API void        ggml_fp16_to_fp32_row(const ggml_fp16_t *, float *, int64_t);374    GGML_API void        ggml_fp32_to_fp16_row(const float *, ggml_fp16_t *, int64_t);375 376    // google brain half-precision bfloat16377    typedef struct { uint16_t bits; } ggml_bf16_t;378    GGML_API ggml_bf16_t ggml_fp32_to_bf16(float);379    GGML_API float       ggml_bf16_to_fp32(ggml_bf16_t);  // consider just doing << 16380    GGML_API void        ggml_bf16_to_fp32_row(const ggml_bf16_t *, float *, int64_t);381    GGML_API void        ggml_fp32_to_bf16_row_ref(const float *, ggml_bf16_t *, int64_t);382    GGML_API void        ggml_fp32_to_bf16_row(const float *, ggml_bf16_t *, int64_t);383 384    struct ggml_object;385    struct ggml_context;386    struct ggml_cgraph;387 388    // NOTE: always add types at the end of the enum to keep backward compatibility389    enum ggml_type {390        GGML_TYPE_F32     = 0,391        GGML_TYPE_F16     = 1,392        GGML_TYPE_Q4_0    = 2,393        GGML_TYPE_Q4_1    = 3,394        // GGML_TYPE_Q4_2 = 4, support has been removed395        // GGML_TYPE_Q4_3 = 5, support has been removed396        GGML_TYPE_Q5_0    = 6,397        GGML_TYPE_Q5_1    = 7,398        GGML_TYPE_Q8_0    = 8,399        GGML_TYPE_Q8_1    = 9,400        GGML_TYPE_Q2_K    = 10,401        GGML_TYPE_Q3_K    = 11,402        GGML_TYPE_Q4_K    = 12,403        GGML_TYPE_Q5_K    = 13,404        GGML_TYPE_Q6_K    = 14,405        GGML_TYPE_Q8_K    = 15,406        GGML_TYPE_IQ2_XXS = 16,407        GGML_TYPE_IQ2_XS  = 17,408        GGML_TYPE_IQ3_XXS = 18,409        GGML_TYPE_IQ1_S   = 19,410        GGML_TYPE_IQ4_NL  = 20,411        GGML_TYPE_IQ3_S   = 21,412        GGML_TYPE_IQ2_S   = 22,413        GGML_TYPE_IQ4_XS  = 23,414        GGML_TYPE_I8      = 24,415        GGML_TYPE_I16     = 25,416        GGML_TYPE_I32     = 26,417        GGML_TYPE_I64     = 27,418        GGML_TYPE_F64     = 28,419        GGML_TYPE_IQ1_M   = 29,420        GGML_TYPE_BF16    = 30,421        // GGML_TYPE_Q4_0_4_4 = 31, support has been removed from gguf files422        // GGML_TYPE_Q4_0_4_8 = 32,423        // GGML_TYPE_Q4_0_8_8 = 33,424        GGML_TYPE_TQ1_0   = 34,425        GGML_TYPE_TQ2_0   = 35,426        // GGML_TYPE_IQ4_NL_4_4 = 36,427        // GGML_TYPE_IQ4_NL_4_8 = 37,428        // GGML_TYPE_IQ4_NL_8_8 = 38,429        GGML_TYPE_MXFP4   = 39, // MXFP4 (1 block)430        GGML_TYPE_NVFP4   = 40, // NVFP4 (4 blocks, E4M3 scale)431        GGML_TYPE_Q1_0    = 41,432        GGML_TYPE_Q2_0    = 42,433        GGML_TYPE_COUNT   = 43,434    };435 436    // precision437    enum ggml_prec {438        GGML_PREC_DEFAULT =  0, // stored as ggml_tensor.op_params, 0 by default439        GGML_PREC_F32     = 10,440    };441 442    // op hint443    enum ggml_op_hint {444        GGML_HINT_NONE             = 0,445        GGML_HINT_SRC0_IS_HADAMARD = 1,446    };447 448    // model file types449    enum ggml_ftype {450        GGML_FTYPE_UNKNOWN        = -1,451        GGML_FTYPE_ALL_F32        = 0,452        GGML_FTYPE_MOSTLY_F16     = 1,  // except 1d tensors453        GGML_FTYPE_MOSTLY_Q4_0    = 2,  // except 1d tensors454        GGML_FTYPE_MOSTLY_Q4_1    = 3,  // except 1d tensors455        GGML_FTYPE_MOSTLY_Q4_1_SOME_F16 = 4, // tok_embeddings.weight and output.weight are F16456        GGML_FTYPE_MOSTLY_Q8_0    = 7,  // except 1d tensors457        GGML_FTYPE_MOSTLY_Q5_0    = 8,  // except 1d tensors458        GGML_FTYPE_MOSTLY_Q5_1    = 9,  // except 1d tensors459        GGML_FTYPE_MOSTLY_Q2_K    = 10, // except 1d tensors460        GGML_FTYPE_MOSTLY_Q3_K    = 11, // except 1d tensors461        GGML_FTYPE_MOSTLY_Q4_K    = 12, // except 1d tensors462        GGML_FTYPE_MOSTLY_Q5_K    = 13, // except 1d tensors463        GGML_FTYPE_MOSTLY_Q6_K    = 14, // except 1d tensors464        GGML_FTYPE_MOSTLY_IQ2_XXS = 15, // except 1d tensors465        GGML_FTYPE_MOSTLY_IQ2_XS  = 16, // except 1d tensors466        GGML_FTYPE_MOSTLY_IQ3_XXS = 17, // except 1d tensors467        GGML_FTYPE_MOSTLY_IQ1_S   = 18, // except 1d tensors468        GGML_FTYPE_MOSTLY_IQ4_NL  = 19, // except 1d tensors469        GGML_FTYPE_MOSTLY_IQ3_S   = 20, // except 1d tensors470        GGML_FTYPE_MOSTLY_IQ2_S   = 21, // except 1d tensors471        GGML_FTYPE_MOSTLY_IQ4_XS  = 22, // except 1d tensors472        GGML_FTYPE_MOSTLY_IQ1_M   = 23, // except 1d tensors473        GGML_FTYPE_MOSTLY_BF16    = 24, // except 1d tensors474        GGML_FTYPE_MOSTLY_MXFP4   = 25, // except 1d tensors475        GGML_FTYPE_MOSTLY_NVFP4   = 26, // except 1d tensors476        GGML_FTYPE_MOSTLY_Q1_0    = 27, // except 1d tensors477        GGML_FTYPE_MOSTLY_Q2_0    = 28, // except 1d tensors478    };479 480    // available tensor operations:481    enum ggml_op {482        GGML_OP_NONE = 0,483 484        GGML_OP_DUP,485        GGML_OP_ADD,486        GGML_OP_ADD_ID,487        GGML_OP_ADD1,488        GGML_OP_ACC,489        GGML_OP_SUB,490        GGML_OP_MUL,491        GGML_OP_DIV,492        GGML_OP_SQR,493        GGML_OP_SQRT,494        GGML_OP_LOG,495        GGML_OP_SIN,496        GGML_OP_COS,497        GGML_OP_SUM,498        GGML_OP_SUM_ROWS,499        GGML_OP_CUMSUM,500        GGML_OP_MEAN,501        GGML_OP_ARGMAX,502        GGML_OP_COUNT_EQUAL,503        GGML_OP_REPEAT,504        GGML_OP_REPEAT_BACK,505        GGML_OP_CONCAT,506        GGML_OP_SILU_BACK,507        GGML_OP_NORM, // normalize508        GGML_OP_RMS_NORM,509        GGML_OP_RMS_NORM_BACK,510        GGML_OP_GROUP_NORM,511        GGML_OP_L2_NORM,512 513        GGML_OP_MUL_MAT,514        GGML_OP_MUL_MAT_ID,515        GGML_OP_OUT_PROD,516 517        GGML_OP_SCALE,518        GGML_OP_SET,519        GGML_OP_CPY,520        GGML_OP_CONT,521        GGML_OP_RESHAPE,522        GGML_OP_VIEW,523        GGML_OP_PERMUTE,524        GGML_OP_TRANSPOSE,525        GGML_OP_GET_ROWS,526        GGML_OP_GET_ROWS_BACK,527        GGML_OP_SET_ROWS,528        GGML_OP_DIAG,529        GGML_OP_DIAG_MASK_INF,530        GGML_OP_DIAG_MASK_ZERO,531        GGML_OP_SOFT_MAX,532        GGML_OP_SOFT_MAX_BACK,533        GGML_OP_ROPE,534        GGML_OP_ROPE_BACK,535        GGML_OP_CLAMP,536        GGML_OP_CONV_TRANSPOSE_1D,537        GGML_OP_IM2COL,538        GGML_OP_IM2COL_BACK,539        GGML_OP_IM2COL_3D,540        GGML_OP_COL2IM_1D,541        GGML_OP_CONV_2D,542        GGML_OP_CONV_3D,543        GGML_OP_CONV_2D_DW,544        GGML_OP_CONV_TRANSPOSE_2D,545        GGML_OP_POOL_1D,546        GGML_OP_POOL_2D,547        GGML_OP_POOL_2D_BACK,548        GGML_OP_UPSCALE,549        GGML_OP_PAD,550        GGML_OP_PAD_REFLECT_1D,551        GGML_OP_ROLL,552        GGML_OP_ARANGE,553        GGML_OP_TIMESTEP_EMBEDDING,554        GGML_OP_ARGSORT,555        GGML_OP_TOP_K,556        GGML_OP_LEAKY_RELU,557        GGML_OP_TRI,558        GGML_OP_FILL,559 560        GGML_OP_FLASH_ATTN_EXT,561        GGML_OP_FLASH_ATTN_BACK,562        GGML_OP_SSM_CONV,563        GGML_OP_SSM_SCAN,564        GGML_OP_WIN_PART,565        GGML_OP_WIN_UNPART,566        GGML_OP_GET_REL_POS,567        GGML_OP_ADD_REL_POS,568        GGML_OP_RWKV_WKV6,569        GGML_OP_GATED_LINEAR_ATTN,570        GGML_OP_RWKV_WKV7,571        GGML_OP_SOLVE_TRI,572        GGML_OP_GATED_DELTA_NET,573        GGML_OP_LIGHTNING_INDEXER,574        GGML_OP_DSV4_HC_COMB,575        GGML_OP_DSV4_HC_PRE,576        GGML_OP_DSV4_HC_POST,577 578        GGML_OP_UNARY,579 580        GGML_OP_MAP_CUSTOM1,581        GGML_OP_MAP_CUSTOM2,582        GGML_OP_MAP_CUSTOM3,583 584        GGML_OP_CUSTOM,585 586        GGML_OP_CROSS_ENTROPY_LOSS,587        GGML_OP_CROSS_ENTROPY_LOSS_BACK,588        GGML_OP_OPT_STEP_ADAMW,589        GGML_OP_OPT_STEP_SGD,590 591        GGML_OP_GLU,592 593        GGML_OP_COUNT,594    };595 596    enum ggml_unary_op {597        GGML_UNARY_OP_ABS,598        GGML_UNARY_OP_SGN,599        GGML_UNARY_OP_NEG,600        GGML_UNARY_OP_STEP,601        GGML_UNARY_OP_TANH,602        GGML_UNARY_OP_ELU,603        GGML_UNARY_OP_RELU,604        GGML_UNARY_OP_SIGMOID,605        GGML_UNARY_OP_GELU,606        GGML_UNARY_OP_GELU_QUICK,607        GGML_UNARY_OP_SILU,608        GGML_UNARY_OP_HARDSWISH,609        GGML_UNARY_OP_HARDSIGMOID,610        GGML_UNARY_OP_EXP,611        GGML_UNARY_OP_EXPM1,612        GGML_UNARY_OP_SOFTPLUS,613        GGML_UNARY_OP_GELU_ERF,614        GGML_UNARY_OP_XIELU,615        GGML_UNARY_OP_FLOOR,616        GGML_UNARY_OP_CEIL,617        GGML_UNARY_OP_ROUND,618        GGML_UNARY_OP_TRUNC,619 620        GGML_UNARY_OP_COUNT,621    };622 623    enum ggml_glu_op {624        GGML_GLU_OP_REGLU,625        GGML_GLU_OP_GEGLU,626        GGML_GLU_OP_SWIGLU,627        GGML_GLU_OP_SWIGLU_OAI,628        GGML_GLU_OP_GEGLU_ERF,629        GGML_GLU_OP_GEGLU_QUICK,630 631        GGML_GLU_OP_COUNT,632    };633 634    enum ggml_object_type {635        GGML_OBJECT_TYPE_TENSOR,636        GGML_OBJECT_TYPE_GRAPH,637        GGML_OBJECT_TYPE_WORK_BUFFER638    };639 640    enum ggml_log_level {641        GGML_LOG_LEVEL_NONE  = 0,642        GGML_LOG_LEVEL_DEBUG = 1,643        GGML_LOG_LEVEL_INFO  = 2,644        GGML_LOG_LEVEL_WARN  = 3,645        GGML_LOG_LEVEL_ERROR = 4,646        GGML_LOG_LEVEL_CONT  = 5, // continue previous log647    };648 649    // this tensor...650    enum ggml_tensor_flag {651        GGML_TENSOR_FLAG_INPUT   =  1, // ...is an input for the GGML compute graph652        GGML_TENSOR_FLAG_OUTPUT  =  2, // ...is an output for the GGML compute graph653        GGML_TENSOR_FLAG_PARAM   =  4, // ...contains trainable parameters654        GGML_TENSOR_FLAG_LOSS    =  8, // ...defines loss for numerical optimization (multiple loss tensors add up)655        GGML_TENSOR_FLAG_COMPUTE = 16, // ...must be computed656    };657 658    enum ggml_tri_type {659        GGML_TRI_TYPE_UPPER_DIAG = 0,660        GGML_TRI_TYPE_UPPER      = 1,661        GGML_TRI_TYPE_LOWER_DIAG = 2,662        GGML_TRI_TYPE_LOWER      = 3663    };664 665    struct ggml_init_params {666        // memory pool667        size_t mem_size;   // bytes668        void * mem_buffer; // if NULL, memory will be allocated internally669        bool   no_alloc;   // don't allocate memory for the tensor data670    };671 672    // n-dimensional tensor673    struct ggml_tensor {674        enum ggml_type type;675 676        struct ggml_backend_buffer * buffer;677 678        int64_t ne[GGML_MAX_DIMS]; // number of elements679        size_t  nb[GGML_MAX_DIMS]; // stride in bytes:680                                   // nb[0] = ggml_type_size(type)681                                   // nb[1] = nb[0]   * (ne[0] / ggml_blck_size(type)) + padding682                                   // nb[i] = nb[i-1] * ne[i-1]683 684        // compute data685        enum ggml_op op;686 687        // op params - allocated as int32_t for alignment688        int32_t op_params[GGML_MAX_OP_PARAMS / sizeof(int32_t)];689 690        int32_t flags;691 692        struct ggml_tensor * src[GGML_MAX_SRC];693 694        // source tensor and offset for views695        struct ggml_tensor * view_src;696        size_t               view_offs;697 698        void * data;699 700        char name[GGML_MAX_NAME];701 702        void * extra; // extra things e.g. for ggml-cuda.cu703 704        char padding[8];705    };706 707    static const size_t GGML_TENSOR_SIZE = sizeof(struct ggml_tensor);708 709    // Abort callback710    // If not NULL, called before ggml computation711    // If it returns true, the computation is aborted712    typedef bool (*ggml_abort_callback)(void * data);713 714 715    //716    // GUID717    //718 719    // GUID types720    typedef uint8_t ggml_guid[16];721    typedef ggml_guid * ggml_guid_t;722 723    GGML_API bool ggml_guid_matches(ggml_guid_t guid_a, ggml_guid_t guid_b);724 725    // misc726 727    GGML_API const char * ggml_version(void);728    GGML_API const char * ggml_commit(void);729 730    GGML_API void    ggml_time_init(void); // call this once at the beginning of the program731    GGML_API int64_t ggml_time_ms(void);732    GGML_API int64_t ggml_time_us(void);733    GGML_API int64_t ggml_cycles(void);734    GGML_API int64_t ggml_cycles_per_ms(void);735 736    // accepts a UTF-8 path, even on Windows737    GGML_API FILE *  ggml_fopen(const char * fname, const char * mode);738 739    GGML_API void    ggml_print_object (const struct ggml_object * obj);740    GGML_API void    ggml_print_objects(const struct ggml_context * ctx);741 742    GGML_API int64_t ggml_nelements (const struct ggml_tensor * tensor);743    GGML_API int64_t ggml_nrows     (const struct ggml_tensor * tensor);744    GGML_API size_t  ggml_nbytes    (const struct ggml_tensor * tensor);745    GGML_API size_t  ggml_nbytes_pad(const struct ggml_tensor * tensor); // same as ggml_nbytes() but padded to GGML_MEM_ALIGN746 747    GGML_API int64_t ggml_blck_size(enum ggml_type type);748    GGML_API size_t  ggml_type_size(enum ggml_type type);             // size in bytes for all elements in a block749    GGML_API size_t  ggml_row_size (enum ggml_type type, int64_t ne); // size in bytes for all elements in a row750 751    GGML_DEPRECATED(752    GGML_API double ggml_type_sizef(enum ggml_type type), // ggml_type_size()/ggml_blck_size() as float753    "use ggml_row_size() instead");754 755    GGML_API const char * ggml_type_name(enum ggml_type type);756    GGML_API const char * ggml_op_name  (enum ggml_op   op);757    GGML_API const char * ggml_op_symbol(enum ggml_op   op);758 759    GGML_API const char * ggml_unary_op_name(enum ggml_unary_op op);760    GGML_API const char * ggml_glu_op_name(enum ggml_glu_op op);761    GGML_API const char * ggml_op_desc(const struct ggml_tensor * t); // unary or op name762 763    GGML_API size_t  ggml_element_size(const struct ggml_tensor * tensor);764 765    GGML_API bool    ggml_is_quantized(enum ggml_type type);766 767    // TODO: temporary until model loading of ggml examples is refactored768    GGML_API enum ggml_type ggml_ftype_to_ggml_type(enum ggml_ftype ftype);769 770    GGML_API bool ggml_is_transposed(const struct ggml_tensor * tensor);771    GGML_API bool ggml_is_permuted  (const struct ggml_tensor * tensor);772    GGML_API bool ggml_is_empty     (const struct ggml_tensor * tensor);773    GGML_API bool ggml_is_view      (const struct ggml_tensor * tensor);774    GGML_API bool ggml_is_scalar    (const struct ggml_tensor * tensor);775    GGML_API bool ggml_is_vector    (const struct ggml_tensor * tensor);776    GGML_API bool ggml_is_matrix    (const struct ggml_tensor * tensor);777    GGML_API bool ggml_is_3d        (const struct ggml_tensor * tensor);778    GGML_API int  ggml_n_dims       (const struct ggml_tensor * tensor); // returns 1 for scalars779 780    // returns whether the tensor elements can be iterated over with a flattened index (no gaps, no permutation)781    GGML_API bool ggml_is_contiguous  (const struct ggml_tensor * tensor);782    GGML_API bool ggml_is_contiguous_0(const struct ggml_tensor * tensor); // same as ggml_is_contiguous()783    GGML_API bool ggml_is_contiguous_1(const struct ggml_tensor * tensor); // contiguous for dims >= 1784    GGML_API bool ggml_is_contiguous_2(const struct ggml_tensor * tensor); // contiguous for dims >= 2785 786    GGML_API bool ggml_is_contiguous_to_1(const struct ggml_tensor * tensor); // contiguous for dims < 1787    GGML_API bool ggml_is_contiguous_to_2(const struct ggml_tensor * tensor); // contiguous for dims < 2788    GGML_API bool ggml_is_contiguous_to_3(const struct ggml_tensor * tensor); // contiguous for dims < 3789 790    // returns whether the tensor elements are allocated as one contiguous block of memory (no gaps, but permutation ok)791    GGML_API bool ggml_is_contiguously_allocated(const struct ggml_tensor * tensor);792 793    // true for tensor that is stored in memory as CxWxHxN and has been permuted to WxHxCxN794    GGML_API bool ggml_is_contiguous_channels(const struct ggml_tensor * tensor);795 796    // true if the elements in dimension 0 are contiguous, or there is just 1 block of elements797    GGML_API bool ggml_is_contiguous_rows(const struct ggml_tensor * tensor);798 799    GGML_API bool ggml_are_same_shape (const struct ggml_tensor * t0, const struct ggml_tensor * t1);800    GGML_API bool ggml_are_same_stride(const struct ggml_tensor * t0, const struct ggml_tensor * t1);801 802    GGML_API bool ggml_can_repeat(const struct ggml_tensor * t0, const struct ggml_tensor * t1);803 804    // use this to compute the memory overhead of a tensor805    GGML_API size_t ggml_tensor_overhead(void);806 807    GGML_API bool ggml_validate_row_data(enum ggml_type type, const void * data, size_t nbytes);808 809    // main810 811    GGML_API struct ggml_context * ggml_init (struct ggml_init_params params);812    GGML_API void                  ggml_reset(struct ggml_context * ctx);813    GGML_API void                  ggml_free (struct ggml_context * ctx);814 815    GGML_API size_t  ggml_used_mem(const struct ggml_context * ctx);816 817    GGML_API bool    ggml_get_no_alloc(struct ggml_context * ctx);818    GGML_API void    ggml_set_no_alloc(struct ggml_context * ctx, bool no_alloc);819 820    GGML_API void *  ggml_get_mem_buffer     (const struct ggml_context * ctx);821    GGML_API size_t  ggml_get_mem_size       (const struct ggml_context * ctx);822    GGML_API size_t  ggml_get_max_tensor_size(const struct ggml_context * ctx);823 824    GGML_API struct ggml_tensor * ggml_new_tensor(825            struct ggml_context * ctx,826            enum   ggml_type type,827            int    n_dims,828            const int64_t *ne);829 830    GGML_API struct ggml_tensor * ggml_new_tensor_1d(831            struct ggml_context * ctx,832            enum   ggml_type type,833            int64_t ne0);834 835    GGML_API struct ggml_tensor * ggml_new_tensor_2d(836            struct ggml_context * ctx,837            enum   ggml_type type,838            int64_t ne0,839            int64_t ne1);840 841    GGML_API struct ggml_tensor * ggml_new_tensor_3d(842            struct ggml_context * ctx,843            enum   ggml_type type,844            int64_t ne0,845            int64_t ne1,846            int64_t ne2);847 848    GGML_API struct ggml_tensor * ggml_new_tensor_4d(849            struct ggml_context * ctx,850            enum   ggml_type type,851            int64_t ne0,852            int64_t ne1,853            int64_t ne2,854            int64_t ne3);855 856    GGML_API void * ggml_new_buffer(struct ggml_context * ctx, size_t nbytes);857 858    GGML_API struct ggml_tensor * ggml_dup_tensor (struct ggml_context * ctx, const struct ggml_tensor * src);859    GGML_API struct ggml_tensor * ggml_view_tensor(struct ggml_context * ctx, struct ggml_tensor * src);860 861    // Context tensor enumeration and lookup862    GGML_API struct ggml_tensor * ggml_get_first_tensor(const struct ggml_context * ctx);863    GGML_API struct ggml_tensor * ggml_get_next_tensor (const struct ggml_context * ctx, struct ggml_tensor * tensor);864    GGML_API struct ggml_tensor * ggml_get_tensor(struct ggml_context * ctx, const char * name);865 866    // Converts a flat index into coordinates867    GGML_API void ggml_unravel_index(const struct ggml_tensor * tensor, int64_t i, int64_t * i0, int64_t * i1, int64_t * i2, int64_t * i3);868 869    GGML_API enum ggml_unary_op ggml_get_unary_op(const struct ggml_tensor * tensor);870    GGML_API enum ggml_glu_op ggml_get_glu_op(const struct ggml_tensor * tensor);871 872    GGML_API void *  ggml_get_data    (const struct ggml_tensor * tensor);873    GGML_API float * ggml_get_data_f32(const struct ggml_tensor * tensor);874 875    GGML_API const char *         ggml_get_name   (const struct ggml_tensor * tensor);876    GGML_API struct ggml_tensor * ggml_set_name   (      struct ggml_tensor * tensor, const char * name);877    GGML_ATTRIBUTE_FORMAT(2, 3)878    GGML_API struct ggml_tensor * ggml_format_name(      struct ggml_tensor * tensor, const char * fmt, ...);879 880    // Tensor flags881    GGML_API void ggml_set_input(struct ggml_tensor * tensor);882    GGML_API void ggml_set_output(struct ggml_tensor * tensor);883    GGML_API void ggml_set_param(struct ggml_tensor * tensor);884    GGML_API void ggml_set_loss(struct ggml_tensor * tensor);885 886    //887    // operations on tensors with backpropagation888    //889 890    GGML_API struct ggml_tensor * ggml_dup(891            struct ggml_context * ctx,892            struct ggml_tensor  * a);893 894    // in-place, returns view(a)895    GGML_API struct ggml_tensor * ggml_dup_inplace(896            struct ggml_context * ctx,897            struct ggml_tensor  * a);898 899    GGML_API struct ggml_tensor * ggml_add(900            struct ggml_context * ctx,901            struct ggml_tensor  * a,902            struct ggml_tensor  * b);903 904    GGML_API struct ggml_tensor * ggml_add_inplace(905            struct ggml_context * ctx,906            struct ggml_tensor  * a,907            struct ggml_tensor  * b);908 909    GGML_API struct ggml_tensor * ggml_add_cast(910            struct ggml_context * ctx,911            struct ggml_tensor  * a,912            struct ggml_tensor  * b,913            enum   ggml_type      type);914 915    // dst[i0, i1, i2] = a[i0, i1, i2] + b[i0, ids[i1, i2]]916    GGML_API struct ggml_tensor * ggml_add_id(917            struct ggml_context * ctx,918            struct ggml_tensor  * a,919            struct ggml_tensor  * b,920            struct ggml_tensor  * ids);921 922    GGML_DEPRECATED(GGML_API struct ggml_tensor * ggml_add1(923            struct ggml_context * ctx,924            struct ggml_tensor  * a,925            struct ggml_tensor  * b),926        "use ggml_add instead");927 928    GGML_DEPRECATED(GGML_API struct ggml_tensor * ggml_add1_inplace(929            struct ggml_context * ctx,930            struct ggml_tensor  * a,931            struct ggml_tensor  * b),932        "use ggml_add_inplace instead");933 934    // dst = a935    // view(dst, nb1, nb2, nb3, offset) += b936    // return dst937    GGML_API struct ggml_tensor * ggml_acc(938            struct ggml_context * ctx,939            struct ggml_tensor  * a,940            struct ggml_tensor  * b,941            size_t                nb1,942            size_t                nb2,943            size_t                nb3,944            size_t                offset);945 946    GGML_API struct ggml_tensor * ggml_acc_inplace(947            struct ggml_context * ctx,948            struct ggml_tensor  * a,949            struct ggml_tensor  * b,950            size_t                nb1,951            size_t                nb2,952            size_t                nb3,953            size_t                offset);954 955    GGML_API struct ggml_tensor * ggml_sub(956            struct ggml_context * ctx,957            struct ggml_tensor  * a,958            struct ggml_tensor  * b);959 960    GGML_API struct ggml_tensor * ggml_sub_inplace(961            struct ggml_context * ctx,962            struct ggml_tensor  * a,963            struct ggml_tensor  * b);964 965    GGML_API struct ggml_tensor * ggml_mul(966            struct ggml_context * ctx,967            struct ggml_tensor  * a,968            struct ggml_tensor  * b);969 970    GGML_API struct ggml_tensor * ggml_mul_inplace(971            struct ggml_context * ctx,972            struct ggml_tensor  * a,973            struct ggml_tensor  * b);974 975    GGML_API struct ggml_tensor * ggml_div(976            struct ggml_context * ctx,977            struct ggml_tensor  * a,978            struct ggml_tensor  * b);979 980    GGML_API struct ggml_tensor * ggml_div_inplace(981            struct ggml_context * ctx,982            struct ggml_tensor  * a,983            struct ggml_tensor  * b);984 985    GGML_API struct ggml_tensor * ggml_sqr(986            struct ggml_context * ctx,987            struct ggml_tensor  * a);988 989    GGML_API struct ggml_tensor * ggml_sqr_inplace(990            struct ggml_context * ctx,991            struct ggml_tensor  * a);992 993    GGML_API struct ggml_tensor * ggml_sqrt(994            struct ggml_context * ctx,995            struct ggml_tensor  * a);996 997    GGML_API struct ggml_tensor * ggml_sqrt_inplace(998            struct ggml_context * ctx,999            struct ggml_tensor  * a);1000 1001    GGML_API struct ggml_tensor * ggml_log(1002            struct ggml_context * ctx,1003            struct ggml_tensor  * a);1004 1005    GGML_API struct ggml_tensor * ggml_log_inplace(1006            struct ggml_context * ctx,1007            struct ggml_tensor  * a);1008 1009    GGML_API struct ggml_tensor * ggml_expm1(1010            struct ggml_context * ctx,1011            struct ggml_tensor  * a);1012 1013    GGML_API struct ggml_tensor * ggml_expm1_inplace(1014            struct ggml_context * ctx,1015            struct ggml_tensor  * a);1016 1017    GGML_API struct ggml_tensor * ggml_softplus(1018            struct ggml_context * ctx,1019            struct ggml_tensor  * a);1020 1021    GGML_API struct ggml_tensor * ggml_softplus_inplace(1022            struct ggml_context * ctx,1023            struct ggml_tensor  * a);1024 1025    GGML_API struct ggml_tensor * ggml_sin(1026            struct ggml_context * ctx,1027            struct ggml_tensor  * a);1028 1029    GGML_API struct ggml_tensor * ggml_sin_inplace(1030            struct ggml_context * ctx,1031            struct ggml_tensor  * a);1032 1033    GGML_API struct ggml_tensor * ggml_cos(1034            struct ggml_context * ctx,1035            struct ggml_tensor  * a);1036 1037    GGML_API struct ggml_tensor * ggml_cos_inplace(1038            struct ggml_context * ctx,1039            struct ggml_tensor  * a);1040 1041    // return scalar1042    GGML_API struct ggml_tensor * ggml_sum(1043            struct ggml_context * ctx,1044            struct ggml_tensor  * a);1045 1046    // sums along rows, with input shape [a,b,c,d] return shape [1,b,c,d]1047    GGML_API struct ggml_tensor * ggml_sum_rows(1048            struct ggml_context * ctx,1049            struct ggml_tensor  * a);1050 1051    GGML_API struct ggml_tensor * ggml_cumsum(1052        struct ggml_context * ctx,1053        struct ggml_tensor  * a);1054 1055    // mean along rows1056    GGML_API struct ggml_tensor * ggml_mean(1057            struct ggml_context * ctx,1058            struct ggml_tensor  * a);1059 1060    // argmax along rows1061    GGML_API struct ggml_tensor * ggml_argmax(1062            struct ggml_context * ctx,1063            struct ggml_tensor  * a);1064 1065    // count number of equal elements in a and b1066    GGML_API struct ggml_tensor * ggml_count_equal(1067            struct ggml_context * ctx,1068            struct ggml_tensor  * a,1069            struct ggml_tensor  * b);1070 1071    // if a is the same shape as b, and a is not parameter, return a1072    // otherwise, return a new tensor: repeat(a) to fit in b1073    GGML_API struct ggml_tensor * ggml_repeat(1074            struct ggml_context * ctx,1075            struct ggml_tensor  * a,1076            struct ggml_tensor  * b);1077 1078    // repeat a to the specified shape1079    GGML_API struct ggml_tensor * ggml_repeat_4d(1080            struct ggml_context * ctx,1081            struct ggml_tensor  * a,1082                       int64_t    ne0,1083                       int64_t    ne1,1084                       int64_t    ne2,1085                       int64_t    ne3);1086 1087    // sums repetitions in a into shape of b1088    GGML_API struct ggml_tensor * ggml_repeat_back(1089            struct ggml_context * ctx,1090            struct ggml_tensor  * a,1091            struct ggml_tensor  * b); // sum up values that are adjacent in dims > 0 instead of repeated with same stride1092 1093    // concat a and b along dim1094    // used in stable-diffusion1095    GGML_API struct ggml_tensor * ggml_concat(1096            struct ggml_context * ctx,1097            struct ggml_tensor  * a,1098            struct ggml_tensor  * b,1099            int                   dim);1100 1101    GGML_API struct ggml_tensor * ggml_abs(1102            struct ggml_context * ctx,1103            struct ggml_tensor  * a);1104 1105    GGML_API struct ggml_tensor * ggml_abs_inplace(1106            struct ggml_context * ctx,1107            struct ggml_tensor  * a);1108 1109    GGML_API struct ggml_tensor * ggml_sgn(1110            struct ggml_context * ctx,1111            struct ggml_tensor  * a);1112 1113    GGML_API struct ggml_tensor * ggml_sgn_inplace(1114            struct ggml_context * ctx,1115            struct ggml_tensor  * a);1116 1117    GGML_API struct ggml_tensor * ggml_neg(1118            struct ggml_context * ctx,1119            struct ggml_tensor  * a);1120 1121    GGML_API struct ggml_tensor * ggml_neg_inplace(1122            struct ggml_context * ctx,1123            struct ggml_tensor  * a);1124 1125    GGML_API struct ggml_tensor * ggml_step(1126            struct ggml_context * ctx,1127            struct ggml_tensor  * a);1128 1129    GGML_API struct ggml_tensor * ggml_step_inplace(1130            struct ggml_context * ctx,1131            struct ggml_tensor  * a);1132 1133    GGML_API struct ggml_tensor * ggml_tanh(1134            struct ggml_context * ctx,1135            struct ggml_tensor  * a);1136 1137    GGML_API struct ggml_tensor * ggml_tanh_inplace(1138            struct ggml_context * ctx,1139            struct ggml_tensor  * a);1140 1141    GGML_API struct ggml_tensor * ggml_elu(1142            struct ggml_context * ctx,1143            struct ggml_tensor  * a);1144 1145    GGML_API struct ggml_tensor * ggml_elu_inplace(1146            struct ggml_context * ctx,1147            struct ggml_tensor  * a);1148 1149    GGML_API struct ggml_tensor * ggml_relu(1150            struct ggml_context * ctx,1151            struct ggml_tensor  * a);1152 1153    GGML_API struct ggml_tensor * ggml_leaky_relu(1154            struct ggml_context * ctx,1155            struct ggml_tensor  * a, float negative_slope, bool inplace);1156 1157    GGML_API struct ggml_tensor * ggml_relu_inplace(1158            struct ggml_context * ctx,1159            struct ggml_tensor  * a);1160 1161    GGML_API struct ggml_tensor * ggml_sigmoid(1162            struct ggml_context * ctx,1163            struct ggml_tensor  * a);1164 1165    GGML_API struct ggml_tensor * ggml_sigmoid_inplace(1166            struct ggml_context * ctx,1167            struct ggml_tensor  * a);1168 1169    GGML_API struct ggml_tensor * ggml_gelu(1170            struct ggml_context * ctx,1171            struct ggml_tensor  * a);1172 1173    GGML_API struct ggml_tensor * ggml_gelu_inplace(1174            struct ggml_context * ctx,1175            struct ggml_tensor  * a);1176 1177    // GELU using erf (error function) when possible1178    // some backends may fallback to approximation based on Abramowitz and Stegun formula1179    GGML_API struct ggml_tensor * ggml_gelu_erf(1180            struct ggml_context * ctx,1181            struct ggml_tensor  * a);1182 1183    GGML_API struct ggml_tensor * ggml_gelu_erf_inplace(1184            struct ggml_context * ctx,1185            struct ggml_tensor  * a);1186 1187    GGML_API struct ggml_tensor * ggml_gelu_quick(1188            struct ggml_context * ctx,1189            struct ggml_tensor  * a);1190 1191    GGML_API struct ggml_tensor * ggml_gelu_quick_inplace(1192            struct ggml_context * ctx,1193            struct ggml_tensor  * a);1194 1195    GGML_API struct ggml_tensor * ggml_silu(1196            struct ggml_context * ctx,1197            struct ggml_tensor  * a);1198 1199    GGML_API struct ggml_tensor * ggml_silu_inplace(1200            struct ggml_context * ctx,

Showing the first 1,200 of 2937 lines. Download the file for the rest.