update ggml to 2024-10-03

leejet · Nov 20, 2024 · c9c9810 · c9c9810
1 parent ac54e00
commit c9c9810
Show file tree

Hide file tree

Showing 4 changed files with 19 additions and 13 deletions.
diff --git a/ggml b/ggml
diff --git a/ggml_extend.hpp b/ggml_extend.hpp
@@ -100,20 +100,21 @@ __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;
         }
     }
+    //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];
+    //        break;
+    //    }
+    //}
     return res;
 }
 
@@ -1118,7 +1119,7 @@ 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);
             }

diff --git a/stable-diffusion.cpp b/stable-diffusion.cpp
@@ -1,3 +1,4 @@
+#include "ggml.h"
 #include "ggml_extend.hpp"
 
 #include "model.h"
@@ -47,6 +48,8 @@ const char* sampling_methods_str[] = {
     "LCM",
 };
 
+//static_assert((int)SD_TYPE_COUNT == (int)GGML_TYPE_COUNT);
+
 /*================================================== Helper Functions ================================================*/
 
 void calculate_alphas_cumprod(float* alphas_cumprod,

diff --git a/stable-diffusion.h b/stable-diffusion.h
@@ -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,
 };
+89 −1		AUTHORS
+12 −2		CMakeLists.txt
+18 −86		README.md
+4 −0		examples/CMakeLists.txt
+2 −0		examples/common-ggml.cpp
+1 −1		examples/gpt-2/main-alloc.cpp
+0 −6		examples/gpt-2/main-backend.cpp
+1 −6		examples/gpt-2/main-batched.cpp
+1 −3		examples/gpt-2/main-sched.cpp
+1 −1		examples/mnist/CMakeLists.txt
+69 −53		examples/mnist/README.md
+279 −110		examples/mnist/mnist-common.cpp
+61 −34		examples/mnist/mnist-common.h
+24 −16		examples/mnist/mnist-eval.cpp
+4 −6		examples/mnist/mnist-train-cnn.py
+5 −4		examples/mnist/mnist-train-fc.py
+5 −5		examples/mnist/mnist-train.cpp
+7 −0		examples/perf-metal/CMakeLists.txt
+152 −0		examples/perf-metal/perf-metal.cpp
+1 −1		examples/sam/main.cpp
+1 −7		examples/simple/simple-backend.cpp
+1 −1		examples/simple/simple-ctx.cpp
+2 −2		include/ggml-alloc.h
+146 −59		include/ggml-backend.h
+3 −3		include/ggml-blas.h
+15 −8		include/ggml-cann.h
+17 −15		include/ggml-cuda.h
+4 −7		include/ggml-metal.h
+5 −5		include/ggml-rpc.h
+8 −8		include/ggml-sycl.h
+7 −7		include/ggml-vulkan.h
+189 −135		include/ggml.h
+20 −6		scripts/sync-llama-am.sh
+1 −1		scripts/sync-llama.last
+75 −43		src/CMakeLists.txt
+2,127 −1,098		src/ggml-aarch64.c
+6 −0		src/ggml-alloc.c
+155 −70		src/ggml-backend-impl.h
+58 −6		src/ggml-backend.c
+2,496 −0		src/ggml-backend.cpp
+9 −10		src/ggml-blas.cpp
+162 −97		src/ggml-cann.cpp
+1 −0		src/ggml-cann/common.h
+20 −0		src/ggml-common.h
+614 −0		src/ggml-cpu-impl.h
+487 −207		src/ggml-cuda.cu
+79 −0		src/ggml-cuda/argmax.cu
+3 −0		src/ggml-cuda/argmax.cuh
+67 −0		src/ggml-cuda/binbcast.cu
+2 −0		src/ggml-cuda/binbcast.cuh
+19 −0		src/ggml-cuda/common.cuh
+64 −0		src/ggml-cuda/count-equal.cu
+5 −0		src/ggml-cuda/count-equal.cuh
+70 −16		src/ggml-cuda/cpy.cu
+62 −2		src/ggml-cuda/cross-entropy-loss.cu
+2 −0		src/ggml-cuda/cross-entropy-loss.cuh
+1 −1		src/ggml-cuda/fattn-tile-f16.cu
+5 −1		src/ggml-cuda/fattn-tile-f32.cu
+3 −3		src/ggml-cuda/fattn-vec-f16.cuh
+3 −3		src/ggml-cuda/fattn.cu
+0 −1		src/ggml-cuda/im2col.cu
+5 −1		src/ggml-cuda/mmq.cu
+2 −2		src/ggml-cuda/mmq.cuh
+80 −0		src/ggml-cuda/opt-step-adamw.cu
+5 −0		src/ggml-cuda/opt-step-adamw.cuh
+51 −0		src/ggml-cuda/out-prod.cu
+3 −0		src/ggml-cuda/out-prod.cuh
+89 −0		src/ggml-cuda/rwkv-wkv.cu
+5 −0		src/ggml-cuda/rwkv-wkv.cuh
+47 −0		src/ggml-cuda/sum.cu
+5 −0		src/ggml-cuda/sum.cuh
+86 −0		src/ggml-cuda/unary.cu
+9 −0		src/ggml-cuda/unary.cuh
+1 −0		src/ggml-cuda/vendors/hip.h
+2 −39		src/ggml-cuda/vendors/musa.h
+39 −606		src/ggml-impl.h
+4 −24		src/ggml-kompute.cpp
+1,961 −1,839		src/ggml-metal.m
+7 −7		src/ggml-metal.metal
+965 −85		src/ggml-quants.c
+15 −4		src/ggml-quants.h
+38 −36		src/ggml-rpc.cpp
+67 −66		src/ggml-sycl.cpp
+0 −1		src/ggml-sycl/common.hpp
+2 −2		src/ggml-sycl/dmmv.cpp
+344 −335		src/ggml-vulkan.cpp
+2,152 −1,385		src/ggml.c
+4 −6		src/vulkan-shaders/argsort.comp
+5 −0		src/vulkan-shaders/vulkan-shaders-gen.cpp
+1,406 −270		tests/test-backend-ops.cpp
+24 −30		tests/test-conv-transpose-1d.cpp
+5 −11		tests/test-conv1d.cpp
+5 −11		tests/test-conv2d.cpp
+8 −4		tests/test-grad0.cpp
+1 −6		tests/test-mul-mat.cpp
+6 −0		tests/test-quantize-fns.cpp
+2 −2		tests/test1.c