happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

2nd+ token correct by fix CPY in OV, remove single op backend compute code

This commit is contained in:
Yu, Zijun 2025-04-15 19:43:29 +08:00 committed by Mustafa Cavus
parent 91d2a195b5
commit 8d263bd6a5
1 changed files with 7 additions and 22 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

29
ggml/src/ggml-openvino.cpp
View File

@ -1,18 +1,14 @@
#include "ggml-backend-impl.h" #include "ggml-backend-impl.h"
#include "ggml-cpu-impl.h"
#include "ggml-impl.h" #include "ggml-impl.h"
#include "ggml-openvino.h" #include "ggml-openvino.h"
#include "ggml-openvino/utils.h" #include "ggml-openvino/utils.h"
#include "ggml.h" #include "ggml.h"
#include <string>
#include <mutex> #include <mutex>
#include <vector>
#include <openvino/openvino.hpp> #include <openvino/openvino.hpp>
#include <openvino/op/op.hpp> #include <set>
#include <openvino/op/add.hpp> #include <string>
#include <openvino/op/subtract.hpp> #include <vector>
#include <openvino/opsets/opset1.hpp>
#define GGML_OPENVINO_MAX_STREAMS 8 #define GGML_OPENVINO_MAX_STREAMS 8
@ -55,10 +51,10 @@ static ggml_backend_buffer_type_t ggml_backend_openvino_get_default_buffer_type(
GGML_UNUSED(backend); GGML_UNUSED(backend);
} }
static void ggml_backend_openvino_add_forward(ggml_tensor * dst) { static enum ggml_status
// Step 1: get the input tensor src0 和 src1 ggml_backend_openvino_graph_compute(ggml_backend_t backend, struct ggml_cgraph *cgraph) {
const struct ggml_tensor *src0 = dst->src[0]; int end_node = cgraph->n_nodes - 1;
const struct ggml_tensor *src1 = dst->src[1]; openvino_frontend_compute(backend, cgraph, 0, end_node);
ov::Core core; ov::Core core;
@ -1267,17 +1263,6 @@ static ggml_backend_buffer_t ggml_backend_openvino_device_buffer_from_host_ptr(g
return nullptr; return nullptr;
} }
std::set<std::string> get_openvino_available_opsets() {
ov::Core core;
std::set<std::string> unique_ops;
for (const auto& opset : ov::get_available_opsets()) {
for (const auto& op : opset.second().get_type_info_set()) {
unique_ops.insert(op.name);
}
}
return unique_ops;
}
static bool ggml_backend_openvino_device_supports_op(ggml_backend_dev_t dev, const ggml_tensor * op) { static bool ggml_backend_openvino_device_supports_op(ggml_backend_dev_t dev, const ggml_tensor * op) {
GGML_ASSERT(dev->reg != nullptr); GGML_ASSERT(dev->reg != nullptr);