sync: update ggml

Author: leejet

ggml

@@ -22,6 +22,7 @@
 
 #include "ggml-alloc.h"
 #include "ggml-backend.h"
+#include "ggml-cpu.h"
 #include "ggml.h"
 
 #ifdef SD_USE_CUBLAS
@@ -100,17 +101,11 @@ __STATIC_INLINE__ ggml_fp16_t ggml_tensor_get_f16(const ggml_tensor* tensor, int
 
 static struct ggml_tensor* get_tensor_from_graph(struct ggml_cgraph* gf, const char* name) {
     struct ggml_tensor* res = NULL;
-    for (int i = 0; i < gf->n_nodes; i++) {
-        // printf("%d, %s \n", i, gf->nodes[i]->name);
-        if (strcmp(ggml_get_name(gf->nodes[i]), name) == 0) {
-            res = gf->nodes[i];
-            break;
-        }
-    }
-    for (int i = 0; i < gf->n_leafs; i++) {
-        // printf("%d, %s \n", i, gf->leafs[i]->name);
-        if (strcmp(ggml_get_name(gf->leafs[i]), name) == 0) {
-            res = gf->leafs[i];
+    for (int i = 0; i < ggml_graph_n_nodes(gf); i++) {
+        struct ggml_tensor* node = ggml_graph_node(gf, i);
+        // printf("%d, %s \n", i, ggml_get_name(node));
+        if (strcmp(ggml_get_name(node), name) == 0) {
+            res = node;
             break;
         }
     }
@@ -1129,7 +1124,7 @@ struct GGMLRunner {
         ggml_graph_print(gf);
 #endif
         if (output != NULL) {
-            auto result = gf->nodes[gf->n_nodes - 1];
+            auto result = ggml_graph_node(gf, -1);
             if (*output == NULL && output_ctx != NULL) {
                 *output = ggml_dup_tensor(output_ctx, result);
             }

ggml_extend.hpp

@@ -13,6 +13,7 @@
 
 #include "ggml-alloc.h"
 #include "ggml-backend.h"
+#include "ggml-cpu.h"
 #include "ggml.h"
 
 #include "stable-diffusion.h"
@@ -733,25 +734,25 @@ void convert_tensor(void* src,
         if (src_type == GGML_TYPE_F16) {
             ggml_fp16_to_fp32_row((ggml_fp16_t*)src, (float*)dst, n);
         } else {
-            auto qtype = ggml_internal_get_type_traits(src_type);
-            if (qtype.to_float == NULL) {
+            auto qtype = ggml_get_type_traits(src_type);
+            if (qtype->to_float == NULL) {
                 throw std::runtime_error(format("type %s unsupported for integer quantization: no dequantization available",
                                                 ggml_type_name(src_type)));
             }
-            qtype.to_float(src, (float*)dst, n);
+            qtype->to_float(src, (float*)dst, n);
         }
     } else {
         // src_type == GGML_TYPE_F16 => dst_type is quantized
         // src_type is quantized => dst_type == GGML_TYPE_F16 or dst_type is quantized
-        auto qtype = ggml_internal_get_type_traits(src_type);
-        if (qtype.to_float == NULL) {
+        auto qtype = ggml_get_type_traits(src_type);
+        if (qtype->to_float == NULL) {
             throw std::runtime_error(format("type %s unsupported for integer quantization: no dequantization available",
                                             ggml_type_name(src_type)));
         }
         std::vector<char> buf;
         buf.resize(sizeof(float) * n);
         char* src_data_f32 = buf.data();
-        qtype.to_float(src, (float*)src_data_f32, n);
+        qtype->to_float(src, (float*)src_data_f32, n);
         if (dst_type == GGML_TYPE_F16) {
             ggml_fp32_to_fp16_row((float*)src_data_f32, (ggml_fp16_t*)dst, n);
         } else {

model.cpp

@@ -93,6 +93,8 @@ enum sd_type_t {
     SD_TYPE_Q4_0_4_4 = 31,
     SD_TYPE_Q4_0_4_8 = 32,
     SD_TYPE_Q4_0_8_8 = 33,
+    SD_TYPE_TQ1_0    = 34,
+    SD_TYPE_TQ2_0    = 35,
     SD_TYPE_COUNT,
 };
 

stable-diffusion.h

@@ -22,6 +22,7 @@
 #include <unistd.h>
 #endif
 
+#include "ggml-cpu.h"
 #include "ggml.h"
 #include "stable-diffusion.h"
 
@@ -410,7 +411,6 @@ const char* sd_get_system_info() {
     static char buffer[1024];
     std::stringstream ss;
     ss << "System Info: \n";
-    ss << "    BLAS = " << ggml_cpu_has_blas() << std::endl;
     ss << "    SSE3 = " << ggml_cpu_has_sse3() << std::endl;
     ss << "    AVX = " << ggml_cpu_has_avx() << std::endl;
     ss << "    AVX2 = " << ggml_cpu_has_avx2() << std::endl;