mirror of
https://github.com/ggml-org/llama.cpp.git
synced 2025-11-12 10:47:01 +00:00
0a1b3982cd
* add conv3d support * add ggml_pad_ext for cpu & cuda backend * cuda/cpu: add im2col_3d support * cuda: make im2col a little faster * fix cuda pad/scale/im2col3d * make im2col_3d faster * gguf: support loading tensors which n_dims > GGML_MAX_DIMS * fix cuda get_rows * avoid ggml_conv_3d conflict * correct GGML_OP_COUNT assertion * avoid build failure * avoid build failure on MacOS * cuda: remove unnecessary MIN define * fix cpu im2col_3d * adjust the code style * cuda: use simpler loop in get_rows * add test_im2col_3d to test-backend-ops * test-backend-ops.cpp: remove trailing whitespace * cpu: im2col_3d support non continuous src Co-authored-by: Jeff Bolz <jbolz@nvidia.com> * fix test_im2col_3d * remove unused variables * cuda: get_rows: dfloat2 -> float2 * add test_pad_ext to test-backend-ops.cpp * add gguf_init_from_file_ext impl * Revert "gguf: support loading tensors which n_dims > GGML_MAX_DIMS" This reverts commitd8377a0a37. * Revert "add gguf_init_from_file_ext impl" This reverts commitd9f1d13208. * update ggml_backend_vk_device_supports_op * fix ggml_backend_vk_device_supports_op * update other backend supports op for ggml_pad_ext * metal/opencl/sycl/vulkan: fix GGML_OP_PAD check in supports_op --------- Co-authored-by: Jeff Bolz <jbolz@nvidia.com>
116 lines
8.9 KiB
C++
116 lines
8.9 KiB
C++
#pragma once
|
|
|
|
#include "ggml.h"
|
|
|
|
//
|
|
// cache line
|
|
//
|
|
|
|
#if defined(__cpp_lib_hardware_interference_size)
|
|
#define CACHE_LINE_SIZE std::hardware_destructive_interference_size
|
|
#else
|
|
#if defined(__POWER9_VECTOR__)
|
|
#define CACHE_LINE_SIZE 128
|
|
#elif defined(__VXE__) || defined(__VXE2__)
|
|
#define CACHE_LINE_SIZE 256
|
|
#else
|
|
#define CACHE_LINE_SIZE 64
|
|
#endif
|
|
#endif
|
|
|
|
static const size_t CACHE_LINE_SIZE_F32 = CACHE_LINE_SIZE/sizeof(float);
|
|
|
|
// Work buffer size for im2col operations in CONV2D
|
|
#define GGML_IM2COL_WORK_SIZE (16 * 1024 * 1024)
|
|
|
|
#ifdef __cplusplus
|
|
extern "C" {
|
|
#endif
|
|
|
|
void ggml_compute_forward_dup(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_add(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_add_id(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_add1(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_acc(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_sum(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_sum_rows(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_mean(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_argmax(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_count_equal(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_repeat(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_repeat_back(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_concat(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_silu_back(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_norm(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_rms_norm(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_rms_norm_back(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_group_norm(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_l2_norm(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_out_prod(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_scale(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_set(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_cpy(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_cont(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_reshape(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_view(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_permute(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_transpose(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_get_rows(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_get_rows_back(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_set_rows(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_diag(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_diag_mask_inf(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_diag_mask_zero(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_soft_max(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_soft_max_ext_back(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_rope(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_rope_back(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_clamp(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_conv_transpose_1d(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_im2col(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_im2col_back_f32(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_im2col_3d(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_conv_2d(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_conv_3d(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_conv_transpose_2d(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_conv_2d_dw(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_pool_1d(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_pool_2d(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_pool_2d_back(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_upscale(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_pad(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_pad_reflect_1d(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_roll(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_arange(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_timestep_embedding(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_argsort(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_leaky_relu(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_flash_attn_ext(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_flash_attn_back(
|
|
const struct ggml_compute_params * params,
|
|
const bool masked,
|
|
struct ggml_tensor * dst);
|
|
void ggml_compute_forward_ssm_conv(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_ssm_scan(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_win_part(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_win_unpart(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_unary(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_glu(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_get_rel_pos(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_add_rel_pos(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_rwkv_wkv6(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_rwkv_wkv7(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_gla(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_map_custom1(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_map_custom2(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_map_custom3(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_custom(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_cross_entropy_loss(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_cross_entropy_loss_back(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_opt_step_adamw(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_mul_mat(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
void ggml_compute_forward_opt_step_sgd(const struct ggml_compute_params * params, struct ggml_tensor * dst);
|
|
#ifdef __cplusplus
|
|
}
|
|
#endif
|