github/gpt4all
1
0
Fork 0
mirror of https://github.com/nomic-ai/gpt4all.git synced 2025-11-02 23:13:29 +00:00
Code Issues Packages Projects Releases Wiki Activity

backend: rebase llama.cpp on upstream as of Sep 26th (#2998)

Browse Source 
Signed-off-by: Jared Van Bortel <jared@nomic.ai>
This commit is contained in:
Jared Van Bortel
2024-09-27 12:05:59 -04:00
committed by GitHub
parent 8bd937eb68
commit f9d6be8afb
16 changed files with 165 additions and 600 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
gpt4all-backend/include/gpt4all-backend/llmodel.h
View File

@@ -7,6 +7,7 @@
#include <cstdint>
#include <functional>
#include <optional>
#include <span>
#include <stdexcept>
#include <string>
#include <string_view>
@@ -149,9 +150,9 @@ public:
virtual bool isEmbeddingModel(const std::string &modelPath) const { (void)modelPath; return false; }
virtual bool isModelLoaded() const = 0;
virtual size_t requiredMem(const std::string &modelPath, int n_ctx, int ngl) = 0;
virtual size_t stateSize() const { return 0; }
virtual size_t saveState(uint8_t *dest) const { (void)dest; return 0; }
virtual size_t restoreState(const uint8_t *src) { (void)src; return 0; }
virtual size_t stateSize() const = 0;
virtual size_t saveState(std::span<uint8_t> dest) const = 0;
virtual size_t restoreState(std::span<const uint8_t> src) = 0;
// This method requires the model to return true from supportsCompletion otherwise it will throw
// an error
@@ -215,7 +216,8 @@ protected:
virtual std::vector<Token> tokenize(PromptContext &ctx, std::string_view str, bool special = false) = 0;
virtual bool isSpecialToken(Token id) const = 0;
virtual std::string tokenToString(Token id) const = 0;
virtual Token sampleToken(PromptContext &ctx) const = 0;
virtual void initSampler(PromptContext &ctx) = 0;
virtual Token sampleToken() const = 0;
virtual bool evalTokens(PromptContext &ctx, const std::vector<int32_t> &tokens) const = 0;
virtual void shiftContext(PromptContext &promptCtx) = 0;
virtual int32_t contextLength() const = 0;

12
gpt4all-backend/include/gpt4all-backend/llmodel_c.h
View File

@@ -148,18 +148,20 @@ uint64_t llmodel_get_state_size(llmodel_model model);
* NOTE: This state data is specific to the type of model you have created.
* @param model A pointer to the llmodel_model instance.
* @param dest A pointer to the destination.
* @return the number of bytes copied
* @param size The size of the destination buffer.
* @return the number of bytes copied, or zero on error.
*/
uint64_t llmodel_save_state_data(llmodel_model model, uint8_t *dest);
uint64_t llmodel_save_state_data(llmodel_model model, uint8_t *dest, uint64_t size);
/**
* Restores the internal state of the model using data from the specified address.
* NOTE: This state data is specific to the type of model you have created.
* @param model A pointer to the llmodel_model instance.
* @param src A pointer to the src.
* @return the number of bytes read
* @param src A pointer to the state data.
* @param size The size of the source data.
* @return The number of bytes read, or zero on error.
*/
uint64_t llmodel_restore_state_data(llmodel_model model, const uint8_t *src);
uint64_t llmodel_restore_state_data(llmodel_model model, const uint8_t *src, size_t size);
/**
* Generate a response using the model.