matt/ollama37
1
0
Fork 0
mirror of https://github.com/dogkeeper886/ollama37.git synced 2025-12-12 00:37:04 +00:00
Code Issues Packages Projects Releases Wiki Activity

llama: update vendored code to commit 46e3556 (#8308)

Browse Source
This commit is contained in:
Jeffrey Morgan
2025-01-08 11:22:01 -08:00
committed by GitHub
parent 57f038ec7b
commit 1deafd8254
305 changed files with 16048 additions and 12926 deletions
Download Patch File Download Diff File Expand all files Collapse all files

45
llama/llama.h vendored
View File

@@ -1,5 +1,5 @@
/**
* llama.cpp - commit ba1cb19cdd0d92e012e0f6e009e0620f854b6afd - do not edit this file
* llama.cpp - commit 46e3556e01b824e52395fb050b29804b6cff2a7c - do not edit this file
*
* MIT License
*
@@ -131,6 +131,7 @@ extern "C" {
LLAMA_VOCAB_PRE_TYPE_EXAONE = 25,
LLAMA_VOCAB_PRE_TYPE_CHAMELEON = 26,
LLAMA_VOCAB_PRE_TYPE_MINERVA = 27,
LLAMA_VOCAB_PRE_TYPE_DEEPSEEK3_LLM = 28,
};
enum llama_rope_type {
@@ -373,6 +374,7 @@ extern "C" {
bool offload_kqv; // whether to offload the KQV ops (including the KV cache) to GPU
bool flash_attn; // whether to use flash attention [EXPERIMENTAL]
bool no_perf; // whether to measure performance timings
bool cross_attn; // whether to use cross attention
// Abort callback
// if it returns true, execution of llama_decode() will be aborted
@@ -412,6 +414,7 @@ extern "C" {
} llama_chat_message;
// lora adapter
// TODO: rename to llama_adapter_lora
struct llama_lora_adapter;
// Helpers for getting default parameters
@@ -443,6 +446,7 @@ extern "C" {
const char * path_model,
struct llama_model_params params);
// TODO: rename to llama_model_free
LLAMA_API void llama_free_model(struct llama_model * model);
// TODO: rename to llama_init_from_model
@@ -513,9 +517,6 @@ extern "C" {
// Returns the total number of parameters in the model
LLAMA_API uint64_t llama_model_n_params(const struct llama_model * model);
// Get a llama model tensor
LLAMA_API struct ggml_tensor * llama_get_model_tensor(struct llama_model * model, const char * name);
// Returns true if the model contains an encoder that requires llama_encode() call
LLAMA_API bool llama_model_has_encoder(const struct llama_model * model);
@@ -535,14 +536,19 @@ extern "C" {
const char * fname_out,
const llama_model_quantize_params * params);
//
// Adapters
//
// Load a LoRA adapter from file
// The loaded adapter will be associated to the given model, and will be free when the model is deleted
// TODO: rename to llama_adapter_lora_init
LLAMA_API struct llama_lora_adapter * llama_lora_adapter_init(
struct llama_model * model,
const char * path_lora);
// Add a loaded LoRA adapter to given context
// This will not modify model's weight
// TODO: rename to llama_set_adapter_lora
LLAMA_API int32_t llama_lora_adapter_set(
struct llama_context * ctx,
struct llama_lora_adapter * adapter,
@@ -550,16 +556,18 @@ extern "C" {
// Remove a specific LoRA adapter from given context
// Return -1 if the adapter is not present in the context
// TODO: rename to llama_rm_adapter_lora
LLAMA_API int32_t llama_lora_adapter_remove(
struct llama_context * ctx,
struct llama_lora_adapter * adapter);
// Remove all LoRA adapters from given context
LLAMA_API void llama_lora_adapter_clear(
struct llama_context * ctx);
// TODO: rename to llama_clear_adapter_lora
LLAMA_API void llama_lora_adapter_clear(struct llama_context * ctx);
// Manually free a LoRA adapter
// Note: loaded adapters will be free when the associated model is deleted
// TODO: rename to llama_adapter_lora_free
LLAMA_API void llama_lora_adapter_free(struct llama_lora_adapter * adapter);
// Apply a loaded control vector to a llama_context, or if data is NULL, clear
@@ -568,6 +576,7 @@ extern "C" {
// to an n_embd x n_layers buffer starting from layer 1.
// il_start and il_end are the layer range the vector should apply to (both inclusive)
// See llama_control_vector_load in common to load a control vector.
// TODO: rename to llama_adapter_cvec_apply
LLAMA_API int32_t llama_control_vector_apply(
struct llama_context * lctx,
const float * data,
@@ -580,6 +589,8 @@ extern "C" {
// KV cache
//
// TODO: remove llama_kv_cache_view_* API
// Information associated with an individual cell in the KV cache view.
struct llama_kv_cache_view_cell {
// The position for this cell. Takes KV cache shifts into account.
@@ -626,8 +637,11 @@ extern "C" {
LLAMA_API void llama_kv_cache_view_free(struct llama_kv_cache_view * view);
// Update the KV cache view structure with the current state of the KV cache. (use only for debugging purposes)
// TODO: change signature to llama_kv_cache_view_update(struct llama_kv_cache_view * view, const struct llama_context * ctx)
LLAMA_API void llama_kv_cache_view_update(const struct llama_context * ctx, struct llama_kv_cache_view * view);
///
// Returns the number of tokens in the KV cache (slow, use only for debug)
// If a KV cell has multiple sequences assigned to it, it will be counted multiple times
LLAMA_API int32_t llama_get_kv_cache_token_count(const struct llama_context * ctx);
@@ -697,6 +711,9 @@ extern "C" {
struct llama_context * ctx,
llama_seq_id seq_id);
// TODO: the llama_kv_cache_defrag and llama_kv_cache_update API tightly couples llama_context with llama_kv_cache
// how to avoid this?
// Defragment the KV cache
// This will be applied:
// - lazily on next llama_decode()
@@ -1170,16 +1187,12 @@ extern "C" {
const char * grammar_str,
const char * grammar_root);
/// NOTE: Avoid using on the full vocabulary as searching for repeated tokens can become slow. For example, apply top-k or top-p sampling first.
LLAMA_API struct llama_sampler * llama_sampler_init_penalties(
int32_t n_vocab, // llama_n_vocab()
llama_token special_eos_id, // llama_token_eos()
llama_token linefeed_id, // llama_token_nl()
int32_t penalty_last_n, // last n tokens to penalize (0 = disable penalty, -1 = context size)
float penalty_repeat, // 1.0 = disabled
float penalty_freq, // 0.0 = disabled
float penalty_present, // 0.0 = disabled
bool penalize_nl, // consider newlines as a repeatable token
bool ignore_eos); // ignore the end-of-sequence token
int32_t penalty_last_n, // last n tokens to penalize (0 = disable penalty, -1 = context size)
float penalty_repeat, // 1.0 = disabled
float penalty_freq, // 0.0 = disabled
float penalty_present); // 0.0 = disabled
/// @details DRY sampler, designed by p-e-w, as described in: https://github.com/oobabooga/text-generation-webui/pull/5677, porting Koboldcpp implementation authored by pi6am: https://github.com/LostRuins/koboldcpp/pull/982
LLAMA_API struct llama_sampler * llama_sampler_init_dry(