kv-cache : use ggml_set_rows (llama/14285)

* kv-cache : use ggml_set_rows

ggml-ci

* graph : separate k and v indices

ggml-ci

* cont : remove redundant ifs

ggml-ci

* kv-cache : improve find_slot impl

* kv-cache : bounds-check when accessing slot_info indices

* kv-cache : add comments

ggml-ci

* ggml : add TODOs for adding GGML_OP_SET_ROWS support in the backends

ggml-ci

diff --git a/src/ggml-cann/ggml-cann.cpp b/src/ggml-cann/ggml-cann.cpp
index 8a3d2026ed898c459b78d4d9afb6852a7ef4203a..eae575cc040cdfd19f4575ec93d4efd533f99d11 100755 (executable)
--- a/src/ggml-cann/ggml-cann.cpp
+++ b/src/ggml-cann/ggml-cann.cpp
@@ -2086,6 +2086,12 @@ static bool ggml_backend_cann_supports_op(ggml_backend_dev_t dev,
                     return false;
             }
         } break;
+        case GGML_OP_SET_ROWS:
+            {
+                // TODO: add support
+                // ref: https://github.com/ggml-org/llama.cpp/pull/14274
+                return false;
+            } break;
         case GGML_OP_CPY: {
             ggml_tensor *src = op->src[0];
             if ((op->type != GGML_TYPE_F32 && op->type != GGML_TYPE_F16) ||

diff --git a/src/ggml-opencl/ggml-opencl.cpp b/src/ggml-opencl/ggml-opencl.cpp
index 71829c05a5bf626e85fc72848c7fed8534aab5aa..9436e6ea9a08de7801b60fe4e4291dc3d931048b 100644 (file)
--- a/src/ggml-opencl/ggml-opencl.cpp
+++ b/src/ggml-opencl/ggml-opencl.cpp
@@ -2222,6 +2222,12 @@ static bool ggml_opencl_supports_op(ggml_backend_dev_t dev, const struct ggml_te
                 default:
                     return false;
             }
+        case GGML_OP_SET_ROWS:
+            {
+                // TODO: add support
+                // ref: https://github.com/ggml-org/llama.cpp/pull/14274
+                return false;
+            } break;
         case GGML_OP_CPY:
         case GGML_OP_DUP:
         case GGML_OP_CONT:

diff --git a/src/ggml-sycl/ggml-sycl.cpp b/src/ggml-sycl/ggml-sycl.cpp
index 1d41d7a4be8206182bca3c3bbb3f86d48badf644..ee4c8f7b2fa5cb9a1470519dd820cf2ba42b3e0c 100644 (file)
--- a/src/ggml-sycl/ggml-sycl.cpp
+++ b/src/ggml-sycl/ggml-sycl.cpp
@@ -4285,6 +4285,12 @@ static bool ggml_backend_sycl_device_supports_op(ggml_backend_dev_t dev, const g
                         return false;
                 }
             }
+        case GGML_OP_SET_ROWS:
+            {
+                // TODO: add support
+                // ref: https://github.com/ggml-org/llama.cpp/pull/14274
+                return false;
+            } break;
         case GGML_OP_CPY:
             {
                 ggml_type src0_type = op->src[0]->type;

diff --git a/src/ggml-vulkan/ggml-vulkan.cpp b/src/ggml-vulkan/ggml-vulkan.cpp
index ccb0d47b6759b696a4bb902d194afc7a76d56dc7..8114cb42046e47bf201d24e3154b978eb7e7010e 100644 (file)
--- a/src/ggml-vulkan/ggml-vulkan.cpp
+++ b/src/ggml-vulkan/ggml-vulkan.cpp
@@ -10339,6 +10339,12 @@ static bool ggml_backend_vk_device_supports_op(ggml_backend_dev_t dev, const ggm
                         return false;
                 }
             } break;
+        case GGML_OP_SET_ROWS:
+            {
+                // TODO: add support
+                // ref: https://github.com/ggml-org/llama.cpp/pull/14274
+                return false;
+            } break;
         case GGML_OP_CONT:
         case GGML_OP_CPY:
         case GGML_OP_DUP: