llama_cpp 0.7.1 → 0.9.0

data/ext/llama_cpp/src/ggml.h

@@ -231,8 +231,9 @@
 #define GGML_EXIT_SUCCESS 0
 #define GGML_EXIT_ABORTED 1
 
-#define GGUF_MAGIC   0x46554747 // "GGUF"
-#define GGUF_VERSION 2
+#define GGUF_MAGIC "GGUF"
+
+#define GGUF_VERSION 3
 
 #define GGUF_DEFAULT_ALIGNMENT 32
 
@@ -400,15 +401,16 @@ extern "C" {
         GGML_OP_ALIBI,
         GGML_OP_CLAMP,
         GGML_OP_CONV_1D,
-        GGML_OP_CONV_2D,
+        GGML_OP_CONV_1D_STAGE_0,  // internal
+        GGML_OP_CONV_1D_STAGE_1,  // internal
         GGML_OP_CONV_TRANSPOSE_1D,
+        GGML_OP_CONV_2D,
+        GGML_OP_CONV_2D_STAGE_0, // internal
+        GGML_OP_CONV_2D_STAGE_1, // internal
         GGML_OP_CONV_TRANSPOSE_2D,
         GGML_OP_POOL_1D,
         GGML_OP_POOL_2D,
 
-        GGML_OP_CONV_1D_STAGE_0,  // internal
-        GGML_OP_CONV_1D_STAGE_1,  // internal
-
         GGML_OP_UPSCALE, // nearest interpolate
 
         GGML_OP_FLASH_ATTN,
@@ -1019,9 +1021,9 @@ extern "C" {
             struct ggml_tensor  * b,
             float                 eps);
 
-    // A: n columns, m rows
-    // B: n columns, p rows  (i.e. we transpose it internally)
-    // result is m columns, p rows
+    // A: k columns, n rows => [ne03, ne02, n, k]
+    // B: k columns, m rows  (i.e. we transpose it internally) => [ne03 * x, ne02 * y, m, k]
+    // result is n columns, m rows => [ne03 * x, ne02 * y, m, n]
     GGML_API struct ggml_tensor * ggml_mul_mat(
             struct ggml_context * ctx,
             struct ggml_tensor  * a,

data/ext/llama_cpp/src/k_quants.c

@@ -46,7 +46,7 @@ inline static int32_t vaddvq_s32(int32x4_t v) {
 #if defined(_MSC_VER) || defined(__MINGW32__)
 #include <intrin.h>
 #else
-#if !defined(__riscv)
+#if !defined(__riscv) && !defined(__s390__)
 #include <immintrin.h>
 #endif
 #endif
@@ -462,12 +462,9 @@ void quantize_row_q2_K(const float * restrict x, void * restrict vy, int k) {
 }
 
 size_t ggml_quantize_q2_K(const float * restrict src, void * restrict dst, int n, int k, int64_t * restrict hist) {
-    const int nb = k / QK_K;
-
-    // TODO - collect histograms - although, at a second thought, I don't really care about them
-    (void)hist;
+    (void)hist; // TODO: collect histograms
 
-    for (int j = 0; j < nb; j += k) {
+    for (int j = 0; j < n; j += k) {
         block_q2_K * restrict y = (block_q2_K *)dst + j/QK_K;
         quantize_row_q2_K_reference(src + j, y, k);
     }
@@ -678,12 +675,9 @@ void quantize_row_q3_K(const float * restrict x, void * restrict vy, int k) {
 }
 
 size_t ggml_quantize_q3_K(const float * restrict src, void * restrict dst, int n, int k, int64_t * restrict hist) {
-    const int nb = k / QK_K;
-
-    // TODO - collect histograms - although, at a second thought, I don't really care about them
-    (void)hist;
+    (void)hist; // TODO: collect histograms
 
-    for (int j = 0; j < nb; j += k) {
+    for (int j = 0; j < n; j += k) {
         block_q3_K * restrict y = (block_q3_K *)dst + j/QK_K;
         quantize_row_q3_K_reference(src + j, y, k);
     }
@@ -846,9 +840,9 @@ void quantize_row_q4_K(const float * restrict x, void * restrict vy, int k) {
 
 size_t ggml_quantize_q4_K(const float * restrict src, void * restrict dst, int n, int k, int64_t * restrict hist) {
     assert(k % QK_K == 0);
-    const int nb = k / QK_K;
     (void)hist; // TODO: collect histograms
-    for (int j = 0; j < nb; j += k) {
+
+    for (int j = 0; j < n; j += k) {
         block_q4_K * restrict y = (block_q4_K *)dst + j/QK_K;
         quantize_row_q4_K_reference(src + j, y, k);
     }
@@ -1052,9 +1046,9 @@ void quantize_row_q5_K(const float * restrict x, void * restrict vy, int k) {
 
 size_t ggml_quantize_q5_K(const float * restrict src, void * restrict dst, int n, int k, int64_t * restrict hist) {
     assert(k % QK_K == 0);
-    const int nb = k / QK_K;
-    (void)hist;
-    for (int j = 0; j < nb; j += k) {
+    (void)hist; // TODO: collect histograms
+
+    for (int j = 0; j < n; j += k) {
         block_q5_K * restrict y = (block_q5_K *)dst + j/QK_K;
         quantize_row_q5_K_reference(src + j, y, k);
     }
@@ -1200,11 +1194,9 @@ void quantize_row_q6_K(const float * restrict x, void * restrict vy, int k) {
 
 size_t ggml_quantize_q6_K(const float * src, void * dst, int n, int k, int64_t * hist) {
     assert(k % QK_K == 0);
-    const int nb = k / QK_K;
-
-    (void)hist; // TODO
+    (void)hist; // TODO: collect histograms
 
-    for (int j = 0; j < nb; j += k) {
+    for (int j = 0; j < n; j += k) {
         block_q6_K * restrict y = (block_q6_K *)dst + j/QK_K;
         quantize_row_q6_K_reference(src + j, y, k);
     }