vulkan: enable cooperative matrix support and improve shader generation

Author: l3utterfly

common/sampling.cpp

@@ -262,7 +262,7 @@ void common_sampler_reset(struct common_sampler * gsmpl) {
 void common_sampler_reinit_grammar(struct common_sampler * gsmpl, const struct llama_model * model, const char * grammar) {
     llama_sampler_reset(gsmpl->grmr);
 
-    gsmpl->grmr = llama_sampler_init_grammar(model, grammar, "root");
+    gsmpl->grmr = llama_sampler_init_grammar(llama_model_get_vocab(model), grammar, "root");
 }
 
 struct common_sampler * common_sampler_clone(common_sampler * gsmpl) {

examples/llava/clip.cpp

@@ -1111,8 +1111,6 @@ static ggml_cgraph * clip_image_build_graph(clip_ctx * ctx, const clip_image_f32
 
 // read and create ggml_context containing the tensors and their data
 struct clip_ctx * clip_model_load(const char * fname, const int verbosity = 1) {
-    throw new std::runtime_error("Not implemented");
-
     struct ggml_context * meta = NULL;
 
     struct gguf_init_params params = {
@@ -2446,8 +2444,6 @@ bool clip_image_encode(struct clip_ctx * ctx, const int n_threads, clip_image_f3
 }
 
 bool clip_image_batch_encode(clip_ctx * ctx, const int n_threads, const clip_image_f32_batch * imgs, float * vec) {
-    throw new std::runtime_error("Not implemented");
-
     if (!ctx->has_vision_encoder) {
         LOG_ERR("This gguf file seems to have no vision encoder\n");
         return false;

examples/llava/llava.cpp

@@ -101,8 +101,6 @@ 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)
 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) {
-    throw new std::runtime_error("Not implemented");
-
     struct {
         struct ggml_context * ctx;
     } model;