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

Removed API get_input_type

This commit is contained in:
Xuejun Zhai 2025-12-04 06:24:38 -08:00 committed by Mustafa Cavus
parent 891a3beb2d
commit 42ca27f714
4 changed files with 6 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
ggml/src/ggml-openvino/ggml-decoder.cpp
View File

@ -747,10 +747,6 @@ std::vector<size_t> GgmlOvDecoder::get_input_stride(int node_idx, const std::str
return get_stride(m_node_info_list[node_idx].node_inputs.at(name)); return get_stride(m_node_info_list[node_idx].node_inputs.at(name));
} }
ov::element::Type GgmlOvDecoder::get_input_type(const std::string & name) const {
return get_ov_type(m_inputs.at(name));
}
ov::element::Type GgmlOvDecoder::get_input_type(int node_idx, const std::string & name) const { ov::element::Type GgmlOvDecoder::get_input_type(int node_idx, const std::string & name) const {
return get_ov_type(m_node_info_list[node_idx].node_inputs.at(name)); return get_ov_type(m_node_info_list[node_idx].node_inputs.at(name));
} }

2
ggml/src/ggml-openvino/ggml-decoder.h
View File

@ -83,8 +83,6 @@ public:
virtual std::vector<size_t> get_input_stride(int node_idx, const std::string & name) const override; virtual std::vector<size_t> get_input_stride(int node_idx, const std::string & name) const override;
virtual ov::element::Type get_input_type(const std::string & name) const override;
virtual ov::element::Type get_input_type(int node_idx, const std::string & name) const override; virtual ov::element::Type get_input_type(int node_idx, const std::string & name) const override;
virtual size_t get_input_size() const override; virtual size_t get_input_size() const override;

2
ggml/src/ggml-openvino/openvino/decoder.hpp
View File

@ -20,8 +20,6 @@ public:
virtual std::vector<size_t> get_input_stride(int node_idx, const std::string& name) const = 0; virtual std::vector<size_t> get_input_stride(int node_idx, const std::string& name) const = 0;
virtual element::Type get_input_type(const std::string& name) const = 0;
virtual element::Type get_input_type(int node_idx, const std::string& name) const = 0; virtual element::Type get_input_type(int node_idx, const std::string& name) const = 0;
virtual size_t get_input_size() const = 0; virtual size_t get_input_size() const = 0;

12
ggml/src/ggml-openvino/utils.cpp
View File

@ -480,9 +480,9 @@ ov::Tensor convert_ggml_input_to_ov(std::shared_ptr<GgmlOvDecoder> ggml_decoder,
// This case is added to make test-backend-ops work // This case is added to make test-backend-ops work
input_shape = ggml_decoder->get_shape(ggml_tensor->view_src); input_shape = ggml_decoder->get_shape(ggml_tensor->view_src);
} else { } else {
input_shape = ggml_decoder->get_input_shape(name).to_shape(); input_shape = ggml_decoder->get_shape(ggml_tensor);
} }
auto input_tensor = ov::Tensor(ggml_decoder->get_input_type(name), input_shape, input_data); auto input_tensor = ov::Tensor(ggml_decoder->get_ov_type(ggml_tensor), input_shape, input_data);
return input_tensor; return input_tensor;
} }
} // namespace } // namespace
@ -506,7 +506,7 @@ ov::Tensor get_ov_input_tensor_static_decode(std::shared_ptr<GgmlOvDecoder> ggml
(op->op == GGML_OP_SET_ROWS && op->src[1] == ggml_tensor)) { (op->op == GGML_OP_SET_ROWS && op->src[1] == ggml_tensor)) {
assert(ggml_tensor->ne[0] == 1); assert(ggml_tensor->ne[0] == 1);
ov::Shape input_shape = {1, 1, 1, 1}; ov::Shape input_shape = {1, 1, 1, 1};
ov::Tensor input_tensor(ggml_decoder->get_input_type(param_name), input_shape); ov::Tensor input_tensor(ggml_decoder->get_ov_type(ggml_tensor), input_shape);
if (ggml_tensor->type == GGML_TYPE_I32) { if (ggml_tensor->type == GGML_TYPE_I32) {
*input_tensor.data<int32_t>() = *((int32_t *) ggml_tensor->data); *input_tensor.data<int32_t>() = *((int32_t *) ggml_tensor->data);
} else if (ggml_tensor->type == GGML_TYPE_I64) { } else if (ggml_tensor->type == GGML_TYPE_I64) {
@ -519,7 +519,7 @@ ov::Tensor get_ov_input_tensor_static_decode(std::shared_ptr<GgmlOvDecoder> ggml
if (param_name == "inp_out_ids") { if (param_name == "inp_out_ids") {
ov::Shape input_shape = {1, 1, 1, 1}; ov::Shape input_shape = {1, 1, 1, 1};
ov::Tensor input_tensor(ggml_decoder->get_input_type(param_name), input_shape); ov::Tensor input_tensor(ggml_decoder->get_ov_type(ggml_tensor), input_shape);
int32_t inp_out_id = *((int32_t *) ggml_tensor->data); int32_t inp_out_id = *((int32_t *) ggml_tensor->data);
assert(ggml_tensor->ne[0] == 1); assert(ggml_tensor->ne[0] == 1);
assert(inp_out_id == 0); assert(inp_out_id == 0);
@ -553,7 +553,7 @@ ov::Tensor get_ov_input_tensor_static_prefill(std::shared_ptr<GgmlOvDecoder> ggm
if (param_name == "inp_pos" || param_name == "inp_tokens" || if (param_name == "inp_pos" || param_name == "inp_tokens" ||
(op->op == GGML_OP_SET_ROWS && op->src[1] == ggml_tensor)) { (op->op == GGML_OP_SET_ROWS && op->src[1] == ggml_tensor)) {
ov::Shape input_shape = {1, 1, 1, chunk_size}; ov::Shape input_shape = {1, 1, 1, chunk_size};
ov::Tensor input_tensor(ggml_decoder->get_input_type(param_name), input_shape); ov::Tensor input_tensor(ggml_decoder->get_ov_type(ggml_tensor), input_shape);
// copy the chunk_index-th chunk from ggml_tensor // copy the chunk_index-th chunk from ggml_tensor
size_t element_size = ggml_type_size(ggml_tensor->type); size_t element_size = ggml_type_size(ggml_tensor->type);
void * input_data = (char *) ggml_tensor->data + chunk_index * chunk_size * element_size; void * input_data = (char *) ggml_tensor->data + chunk_index * chunk_size * element_size;
@ -581,7 +581,7 @@ ov::Tensor get_ov_input_tensor_static_prefill(std::shared_ptr<GgmlOvDecoder> ggm
if (param_name == "inp_out_ids") { if (param_name == "inp_out_ids") {
size_t output_len = ggml_decoder->get_compute_params().output_len; size_t output_len = ggml_decoder->get_compute_params().output_len;
ov::Shape input_shape = {1, 1, 1, output_len}; ov::Shape input_shape = {1, 1, 1, output_len};
ov::Tensor input_tensor(ggml_decoder->get_input_type(param_name), input_shape); ov::Tensor input_tensor(ggml_decoder->get_ov_type(ggml_tensor), input_shape);
if (ggml_tensor->ne[0] == 0) { if (ggml_tensor->ne[0] == 0) {
*input_tensor.data<int32_t>() = 0; *input_tensor.data<int32_t>() = 0;
} else { } else {