diff options
Diffstat (limited to 'common/common.h')
| -rw-r--r-- | common/common.h | 27 |
1 files changed, 23 insertions, 4 deletions
diff --git a/common/common.h b/common/common.h index 979762e1..50035897 100644 --- a/common/common.h +++ b/common/common.h @@ -33,6 +33,15 @@ #define DEFAULT_MODEL_PATH "models/7B/ggml-model-f16.gguf" +struct llama_lora_adapter_info { + std::string path; + float scale; +}; + +struct llama_lora_adapter_container : llama_lora_adapter_info { + struct llama_lora_adapter * adapter; +}; + // build info extern int LLAMA_BUILD_NUMBER; extern char const * LLAMA_COMMIT; @@ -126,8 +135,8 @@ struct gpt_params { std::vector<std::string> antiprompt; // strings upon which more user input is prompted (a.k.a. reverse prompts) std::vector<llama_model_kv_override> kv_overrides; - // TODO: avoid tuple, use struct - std::vector<std::tuple<std::string, float>> lora_adapter; // lora adapter path with user defined scale + bool lora_init_without_apply = false; // only load lora to memory, but do not apply it to ctx (user can manually apply lora later using llama_lora_adapter_apply) + std::vector<llama_lora_adapter_info> lora_adapters; // lora adapter path with user defined scale std::vector<llama_control_vector_load_info> control_vectors; // control vector with user defined scale @@ -278,6 +287,8 @@ std::vector<std::string> string_split(std::string input, char separator); std::string string_strip(const std::string & str); std::string string_get_sortable_timestamp(); +void string_replace_all(std::string & s, const std::string & search, const std::string & replace); + template<class T> static std::vector<T> string_split(const std::string & str, char delim) { std::vector<T> values; @@ -309,8 +320,13 @@ std::string fs_get_cache_file(const std::string & filename); // Model utils // -// TODO: avoid tuplue, use struct -std::tuple<struct llama_model *, struct llama_context *> llama_init_from_gpt_params(gpt_params & params); +struct llama_init_result { + struct llama_model * model = nullptr; + struct llama_context * context = nullptr; + std::vector<llama_lora_adapter_container> lora_adapters; +}; + +struct llama_init_result llama_init_from_gpt_params(gpt_params & params); struct llama_model_params llama_model_params_from_gpt_params (const gpt_params & params); struct llama_context_params llama_context_params_from_gpt_params(const gpt_params & params); @@ -318,6 +334,9 @@ struct llama_context_params llama_context_params_from_gpt_params(const gpt_param struct llama_model * llama_load_model_from_url(const char * model_url, const char * path_model, const char * hf_token, const struct llama_model_params & params); struct llama_model * llama_load_model_from_hf(const char * repo, const char * file, const char * path_model, const char * hf_token, const struct llama_model_params & params); +// clear LoRA adapters from context, then apply new list of adapters +void llama_lora_adapters_apply(struct llama_context * ctx, std::vector<llama_lora_adapter_container> & lora_adapters); + // Batch utils void llama_batch_clear(struct llama_batch & batch); |