ggml: update include paths in examples

Signed-off-by: Daniel Bevenius <[email protected]>

fundamentals/ggml/src/backend.c

@@ -1,8 +1,8 @@
 #include <stdio.h>
 
-#include "ggml/ggml.h"
-#include "ggml/ggml-alloc.h"
-#include "ggml/ggml-backend.h"
+#include "ggml.h"
+#include "ggml-alloc.h"
+#include "ggml-backend.h"
 
 void print_backend_info(ggml_backend_buffer_t buffer, struct ggml_context* ctx) {
       printf("------- backend info -------\n");

fundamentals/ggml/src/cont.c

@@ -1,8 +1,8 @@
 #include <stdio.h>
 
-#include "ggml/ggml.h"
-#include "ggml/ggml-alloc.h"
-#include "ggml/ggml-backend.h"
+#include "ggml.h"
+#include "ggml-alloc.h"
+#include "ggml-backend.h"
 
 int main(int argc, char **argv) {
   printf("GGML cont tensor examples\n");

fundamentals/ggml/src/dup.c

@@ -1,8 +1,8 @@
 #include <stdio.h>
 
-#include "ggml/ggml.h"
-#include "ggml/ggml-alloc.h"
-#include "ggml/ggml-backend.h"
+#include "ggml.h"
+#include "ggml-alloc.h"
+#include "ggml-backend.h"
 
 int main(int argc, char **argv) {
   printf("GGML dup example\n");

fundamentals/ggml/src/get_rows.c

@@ -1,8 +1,8 @@
 #include <stdio.h>
 
-#include "ggml/ggml.h"
-#include "ggml/ggml-alloc.h"
-#include "ggml/ggml-backend.h"
+#include "ggml.h"
+#include "ggml-alloc.h"
+#include "ggml-backend.h"
 
 /*
  * This is an example to help understand what ggml_get_rows does.

fundamentals/ggml/src/ggml-quants.c

@@ -1,21 +1,26 @@
 #include <stdio.h>
 
-#include "ggml/ggml.h"
+#include "ggml.h"
 #include "ggml-quants.h"
 #include "ggml-impl.h"
 
+/*
 #ifdef GGML_COMMON_DECL_C
 #warning "[danbev] GGML_COMMON_DECL_C is defined"
 #endif
 
 #ifdef GGML_COMMON_DECL_CUDA
 #warning "[danbev] GGML_COMMON_DECL_CUDA is defined"
 #endif
+*/
 
 int main(int argc, char **argv) {
   printf("GGML Quantization examples\n");
 
-  float data[] = {0.2, 0.3, 0.4, 0.5};
+  float data[] = {0.2, 0.3, 0.4, 0.5, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
+                  0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
+                  0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
+                  0.0, 0.0};
 
   float d = 0.5 / 15.0;
   block_q4_0 block_q4_0 = {
@@ -34,10 +39,18 @@ int main(int argc, char **argv) {
     printf("data[%d]: %f\n", i, block_q4_0.qs[i] * ggml_compute_fp16_to_fp32(block_q4_0.d));
   }
 
+  // Get the type_trait for Q4_0
   ggml_type_traits_t q4_0 = ggml_internal_get_type_traits(GGML_TYPE_Q4_0);
   printf("ggml type trait name: %s\n", q4_0.type_name);
   printf("ggml type trait block size: %d\n", q4_0.blck_size);
   printf("ggml type trait is_quantized: %s\n", q4_0.is_quantized ? "true" : "false");
+  // The type trait contains function pointers to the quantize and dequantize
+  // functions
+  q4_0.from_float(data, &block_q4_0, 32);
+  for (int i = 0; i < QK4_0/2; i++) {
+    printf("block_q4_0.qs[%d]: %d\n", i, block_q4_0.qs[i]);
+  }
+
 
   return 0;
 }

fundamentals/ggml/src/graph.c

@@ -1,8 +1,8 @@
 #include <stdio.h>
 
-#include "ggml/ggml.h"
-#include "ggml/ggml-alloc.h"
-#include "ggml/ggml-backend.h"
+#include "ggml.h"
+#include "ggml-alloc.h"
+#include "ggml-backend.h"
 
 int main(int argc, char **argv) {
   printf("GGML compute graph example\n");

fundamentals/ggml/src/matrix-mul.c

@@ -1,8 +1,8 @@
 #include <stdio.h>
 
-#include "ggml/ggml.h"
-#include "ggml/ggml-alloc.h"
-#include "ggml/ggml-backend.h"
+#include "ggml.h"
+#include "ggml-alloc.h"
+#include "ggml-backend.h"
 
 int main(int argc, char **argv) {
   printf("GGML matrix multiplication example\n");

fundamentals/ggml/src/permute.c

@@ -1,8 +1,8 @@
 #include <stdio.h>
 
-#include "ggml/ggml.h"
-#include "ggml/ggml-alloc.h"
-#include "ggml/ggml-backend.h"
+#include "ggml.h"
+#include "ggml-alloc.h"
+#include "ggml-backend.h"
 
 int main(int argc, char **argv) {
   printf("GGML permute tensor examples\n");

fundamentals/ggml/src/reshape.c

@@ -1,8 +1,8 @@
 #include <stdio.h>
 
-#include "ggml/ggml.h"
-#include "ggml/ggml-alloc.h"
-#include "ggml/ggml-backend.h"
+#include "ggml.h"
+#include "ggml-alloc.h"
+#include "ggml-backend.h"
 
 int main(int argc, char **argv) {
   printf("GGML reshape tensor examples\n");

fundamentals/ggml/src/tensor.c

@@ -1,8 +1,8 @@
 #include <stdio.h>
 
-#include "ggml/ggml.h"
-#include "ggml/ggml-alloc.h"
-#include "ggml/ggml-backend.h"
+#include "ggml.h"
+#include "ggml-alloc.h"
+#include "ggml-backend.h"
 
 int main(int argc, char **argv) {
   printf("GGML tensor example\n");
@@ -67,6 +67,7 @@ int main(int argc, char **argv) {
   for (int y = 0; y < ny; y++) {
       for (int x = 0; x < nx; x++) {
           float* value = (float *)( (char *) matrix->data + (y * matrix->nb[1]) + (x * matrix->nb[0]));
+          printf("value: %p\n", value);
           // Since we want to perform byte-level operations and a char is 1 byte.
           // If we don't do this the additions would be done of the size of the
           // type, so 4 bytes for a float. And after we are done we need to

fundamentals/ggml/src/transpose.c

@@ -1,8 +1,8 @@
 #include <stdio.h>
 
-#include "ggml/ggml.h"
-#include "ggml/ggml-alloc.h"
-#include "ggml/ggml-backend.h"
+#include "ggml.h"
+#include "ggml-alloc.h"
+#include "ggml-backend.h"
 
 int main(int argc, char **argv) {
   printf("GGML reshape tensor examples\n");

fundamentals/ggml/src/view.c

@@ -1,8 +1,8 @@
 #include <stdio.h>
 
-#include "ggml/ggml.h"
-#include "ggml/ggml-alloc.h"
-#include "ggml/ggml-backend.h"
+#include "ggml.h"
+#include "ggml-alloc.h"
+#include "ggml-backend.h"
 
 int main(int argc, char **argv) {
   printf("GGML view examples\n");