LLM inference in C/C++
C++ ggml
spec : fix naming, spacing (#25410)
Georgi Gerganovc198af4dc24f
diff --git a/common/speculative.cpp b/common/speculative.cppindex 1d2977e424..580728a200 100644--- a/common/speculative.cpp+++ b/common/speculative.cpp@@ -2246,7 +2246,7 @@ common_params common_base_params_to_speculative(const common_params & params) {return result;}-struct common_init_speculative_result::impl {+struct common_speculative_init_result::impl {impl() = default;~impl() = default;@@ -2255,7 +2255,7 @@ struct common_init_speculative_result::impl {llama_context_ptr context;};-common_init_speculative_result::common_init_speculative_result(+common_speculative_init_result::common_speculative_init_result(common_params & params,llama_model * model_tgt,llama_context * ctx_tgt) :@@ -2301,8 +2301,7 @@ common_init_speculative_result::common_init_speculative_result(} else if (spec_mtp) {model_path = params.model.path;- LOG_TRC("%s: creating MTP draft context against the target model '%s'\n",- __func__, model_path.c_str());+ LOG_TRC("%s: creating MTP draft context against the target model '%s'\n", __func__, model_path.c_str());llama_context * ctx_dft = llama_init_from_model(model_tgt, cparams);if (ctx_dft == nullptr) {@@ -2314,18 +2313,18 @@ common_init_speculative_result::common_init_speculative_result(}}-common_init_speculative_result::~common_init_speculative_result() = default;+common_speculative_init_result::~common_speculative_init_result() = default;-llama_model * common_init_speculative_result::model() {+llama_model * common_speculative_init_result::model() {return pimpl->model.get();}-llama_context * common_init_speculative_result::context() {+llama_context * common_speculative_init_result::context() {return pimpl->context.get();}-common_init_speculative_result_ptr common_init_speculative_from_params(common_params & params, llama_model * model_tgt, llama_context * ctx_tgt) {- return std::make_unique<common_init_speculative_result>(params, model_tgt, ctx_tgt);+common_speculative_init_result_ptr common_speculative_init_from_params(common_params & params, llama_model * model_tgt, llama_context * ctx_tgt) {+ return std::make_unique<common_speculative_init_result>(params, model_tgt, ctx_tgt);}// initialization of the speculative decoding systemdiff --git a/common/speculative.h b/common/speculative.hindex 2f6d225b6f..062bf20931 100644--- a/common/speculative.h+++ b/common/speculative.h@@ -83,9 +83,9 @@ struct common_speculative_deleter {typedef std::unique_ptr<common_speculative, common_speculative_deleter> common_speculative_ptr;-struct common_init_speculative_result {- common_init_speculative_result(common_params & params, llama_model * model_tgt, llama_context * ctx_tgt);- ~common_init_speculative_result();+struct common_speculative_init_result {+ common_speculative_init_result(common_params & params, llama_model * model_tgt, llama_context * ctx_tgt);+ ~common_speculative_init_result();llama_model * model();llama_context * context();@@ -95,6 +95,6 @@ private:std::unique_ptr<impl> pimpl;};-using common_init_speculative_result_ptr = std::unique_ptr<common_init_speculative_result>;+using common_speculative_init_result_ptr = std::unique_ptr<common_speculative_init_result>;-common_init_speculative_result_ptr common_init_speculative_from_params(common_params & params, llama_model * model_tgt, llama_context * ctx_tgt);+common_speculative_init_result_ptr common_speculative_init_from_params(common_params & params, llama_model * model_tgt, llama_context * ctx_tgt);diff --git a/tools/server/server-context.cpp b/tools/server/server-context.cppindex 3890fe5f05..df56665815 100644--- a/tools/server/server-context.cpp+++ b/tools/server/server-context.cpp@@ -900,7 +900,7 @@ private:llama_model * model_dft = nullptr;llama_context * ctx_dft = nullptr;- common_init_speculative_result_ptr spec_init;+ common_speculative_init_result_ptr spec_init;common_context_seq_rm_type ctx_tgt_seq_rm_type = COMMON_CONTEXT_SEQ_RM_TYPE_NO;common_context_seq_rm_type ctx_dft_seq_rm_type = COMMON_CONTEXT_SEQ_RM_TYPE_NO;@@ -1171,11 +1171,12 @@ private:{common_params params_dft = common_base_params_to_speculative(params_base);+// progress callbackparams_dft.load_progress_callback = load_progress_callback;params_dft.load_progress_callback_user_data = &load_progress_spec;- spec_init = common_init_speculative_from_params(params_dft, model_tgt, ctx_tgt);+ spec_init = common_speculative_init_from_params(params_dft, model_tgt, ctx_tgt);model_dft = spec_init->model();ctx_dft = spec_init->context();@@ -2306,8 +2307,8 @@ private:// this is not true for SWA models: https://github.com/ggml-org/llama.cpp/pull/24411#issuecomment-4677983225cur.update_pos(slot.prompt.n_tokens() - n_tokens_cur, pos_min, pos_max);- cur.update_tgt(ctx_tgt, slot.id, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY);- cur.update_dft(ctx_dft, slot.id, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY);+ cur.update_tgt(ctx_tgt, slot.id, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY);+ cur.update_dft(ctx_dft, slot.id, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY);// stash the draft's speculative state with the checkpointcommon_speculative_get_state(spec.get(), slot.id, cur.data_spec);@@ -3264,8 +3265,8 @@ private:if (!do_reset) {// restore the context checkpoint- it->load_tgt(ctx_tgt, slot.id, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY);- it->load_dft(ctx_dft, slot.id, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY);+ it->load_tgt(ctx_tgt, slot.id, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY);+ it->load_dft(ctx_dft, slot.id, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY);// restore the draft's speculative statecommon_speculative_set_state(spec.get(), slot.id, it->data_spec);