Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-25 05:48:47 +01:00
Code Issues Packages Projects Releases Wiki Activity

ggml : tag ggml_tensor::backend as deprecated (#7290)

Browse Source
This commit is contained in:
slaren 2024-05-15 15:08:48 +02:00 committed by GitHub
parent 9a17ab914b
commit 344f9126cc
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
4 changed files with 12 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
examples/llava/llava.cpp
View File

@ -88,7 +88,6 @@ static struct clip_image_grid_shape get_anyres_image_grid_shape(const std::pair<
// Take the image segments in a grid configuration and return the embeddings and the number of embeddings into preallocated memory (image_embd_out) // Take the image segments in a grid configuration and return the embeddings and the number of embeddings into preallocated memory (image_embd_out)
static bool clip_llava_handle_patches(clip_ctx * ctx_clip, std::vector<float *> & image_embd_v, struct clip_image_grid_shape grid_shape, float * image_embd_out, int * n_img_pos_out) { static bool clip_llava_handle_patches(clip_ctx * ctx_clip, std::vector<float *> & image_embd_v, struct clip_image_grid_shape grid_shape, float * image_embd_out, int * n_img_pos_out) {
struct { struct {
struct ggml_tensor * newline;
struct ggml_context * ctx; struct ggml_context * ctx;
} model; } model;
@ -150,20 +149,6 @@ static bool clip_llava_handle_patches(clip_ctx * ctx_clip, std::vector<float *>
model.ctx = ggml_init(params); model.ctx = ggml_init(params);
ggml_tensor * newline_tmp = clip_get_newline_tensor(ctx_clip);
model.newline = ggml_new_tensor_1d(model.ctx, GGML_TYPE_F32, newline_tmp->ne[0]);
if (newline_tmp->backend != GGML_BACKEND_TYPE_CPU) {
if (newline_tmp->buffer == NULL) {
LOG_TEE("newline_tmp tensor buffer is NULL\n");
}
ggml_backend_tensor_get(newline_tmp, model.newline->data, 0, ggml_nbytes(newline_tmp));
} else {
model.newline->data = newline_tmp->data;
if (model.newline->data == NULL) {
LOG_TEE("newline_tmp tensor data is NULL\n");
}
}
struct ggml_tensor * image_features = ggml_new_tensor_3d(model.ctx, GGML_TYPE_F32, clip_n_mmproj_embd(ctx_clip), clip_n_patches(ctx_clip), num_images - 1); // example: 4096 x 576 x 4 struct ggml_tensor * image_features = ggml_new_tensor_3d(model.ctx, GGML_TYPE_F32, clip_n_mmproj_embd(ctx_clip), clip_n_patches(ctx_clip), num_images - 1); // example: 4096 x 576 x 4
// ggml_tensor_printf(image_features,"image_features",__LINE__,false,false); // ggml_tensor_printf(image_features,"image_features",__LINE__,false,false);
// fill it with the image embeddings, ignoring the base // fill it with the image embeddings, ignoring the base

1
ggml-backend.c
View File

@ -1895,7 +1895,6 @@ void ggml_backend_view_init(ggml_backend_buffer_t buffer, struct ggml_tensor * t
tensor->buffer = buffer; tensor->buffer = buffer;
tensor->data = (char *)tensor->view_src->data + tensor->view_offs; tensor->data = (char *)tensor->view_src->data + tensor->view_offs;
tensor->backend = tensor->view_src->backend;
ggml_backend_buffer_init_tensor(buffer, tensor); ggml_backend_buffer_init_tensor(buffer, tensor);
} }

10
ggml.c
View File

@ -3178,6 +3178,12 @@ static struct ggml_tensor * ggml_new_tensor_impl(
struct ggml_tensor * const result = (struct ggml_tensor *)((char *)ctx->mem_buffer + obj_new->offs); struct ggml_tensor * const result = (struct ggml_tensor *)((char *)ctx->mem_buffer + obj_new->offs);
#ifdef __clang__
// temporary until ggml_tensor::backend is removed
#pragma clang diagnostic push
#pragma clang diagnostic ignored "-Wdeprecated-declarations"
#endif
*result = (struct ggml_tensor) { *result = (struct ggml_tensor) {
/*.type =*/ type, /*.type =*/ type,
/*.backend =*/ GGML_BACKEND_TYPE_CPU, /*.backend =*/ GGML_BACKEND_TYPE_CPU,
@ -3200,6 +3206,10 @@ static struct ggml_tensor * ggml_new_tensor_impl(
/*.padding =*/ { 0 }, /*.padding =*/ { 0 },
}; };
#ifdef __clang__
#pragma clang diagnostic pop
#endif
// TODO: this should not be needed as long as we don't rely on aligned SIMD loads // TODO: this should not be needed as long as we don't rely on aligned SIMD loads
//ggml_assert_aligned(result->data); //ggml_assert_aligned(result->data);

3
ggml.h
View File

@ -565,7 +565,8 @@ extern "C" {
// n-dimensional tensor // n-dimensional tensor
struct ggml_tensor { struct ggml_tensor {
enum ggml_type type; enum ggml_type type;
enum ggml_backend_type backend;
GGML_DEPRECATED(enum ggml_backend_type backend, "use the buffer type to find the storage location of the tensor");
struct ggml_backend_buffer * buffer; struct ggml_backend_buffer * buffer;