vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

expose llava_image_embed_make_with_clip_img

Browse source
This commit is contained in:
CJ Pais 2024-02-19 07:30:36 -08:00
parent fb16cc9e9f
commit 0ffba498a1
3 changed files with 9 additions and 23 deletions
Download patch file Download diff file Expand all files Collapse all files

2
examples/llava/llava.cpp
View file

@ -311,7 +311,7 @@ bool llava_validate_embed_size(const llama_context * ctx_llama, const clip_ctx *
return true;
}
static bool llava_image_embed_make_with_clip_img(clip_ctx * ctx_clip, int n_threads, const clip_image_u8 * img, float ** image_embd_out, int * n_img_pos_out) {
bool llava_image_embed_make_with_clip_img(clip_ctx * ctx_clip, int n_threads, const clip_image_u8 * img, float ** image_embd_out, int * n_img_pos_out) {
float * image_embd = (float *)malloc(clip_embd_nbytes(ctx_clip)*6); // TODO: base on gridsize/llava model
if (!image_embd) {
fprintf(stderr, "Unable to allocate memory for image embeddings\n");

2
examples/llava/llava.h
View file

@ -31,6 +31,8 @@ struct llava_image_embed {
/** sanity check for clip <-> llava embed size match */
LLAVA_API bool llava_validate_embed_size(const llama_context * ctx_llama, const clip_ctx * ctx_clip);
LLAVA_API bool llava_image_embed_make_with_clip_img(clip_ctx * ctx_clip, int n_threads, const clip_image_u8 * img, float ** image_embd_out, int * n_img_pos_out);
/** build an image embed from image file bytes */
LLAVA_API struct llava_image_embed * llava_image_embed_make_with_bytes(struct clip_ctx * ctx_clip, int n_threads, const unsigned char * image_bytes, int image_bytes_length);
/** build an image embed from a path to an image filename */

28
examples/server/server.cpp
View file

@ -43,23 +43,6 @@ struct server_params
int32_t write_timeout = 600;
};
// RGB uint8 image
struct clip_image_u8 {
int nx;
int ny;
std::vector<uint8_t> buf;
};
// RGB float32 image (NHWC)
// Memory layout: RGBRGBRGB...
struct clip_image_f32 {
int nx;
int ny;
std::vector<float> buf;
};
bool server_verbose = false;
static size_t common_part(const std::vector<llama_token> &a, const std::vector<llama_token> &b)
@ -720,7 +703,11 @@ struct llama_server_context
slot_image img_sl;
img_sl.id = img.count("id") != 0 ? img["id"].get<int>() : slot->images.size();
img_sl.img_data = clip_image_u8_init();
img_sl.img_data->buf = image_buffer;
if (!clip_image_load_from_bytes(image_buffer.data(), image_buffer.size(), img_sl.img_data))
{
LOG_TEE("slot %i - failed to load image [id: %i]\n", slot->id, img_sl.id);
return false;
}
LOG_TEE("slot %i - loaded image\n", slot->id);
img_sl.request_encode_image = true;
slot->images.push_back(img_sl);
@ -998,14 +985,11 @@ struct llama_server_context
continue;
}
llava_image_embed * embed = llava_image_embed_make_with_bytes(clp_ctx, params.n_threads, img.img_data->buf.data(), img.img_data->buf.size());
if (!embed) {
if (!llava_image_embed_make_with_clip_img(clp_ctx, params.n_threads, img.img_data, &img.image_embedding, &img.image_tokens)) {
LOG_TEE("Error processing the given image");
return false;
}
img.image_embedding = embed->embed;
img.image_tokens = embed->n_image_pos;
img.request_encode_image = false;
}