github/gpt4all
1
0
Fork 0
mirror of https://github.com/nomic-ai/gpt4all.git synced 2025-09-06 11:00:48 +00:00
Code Issues Packages Projects Releases Wiki Activity

implement local Nomic Embed via llama.cpp (#2086)

Browse Source 
Signed-off-by: Jared Van Bortel <jared@nomic.ai>
This commit is contained in:
Jared Van Bortel
2024-03-13 18:09:24 -04:00
committed by GitHub
parent 171f4e488e
commit 406e88b59a
23 changed files with 799 additions and 1198 deletions
Download Patch File Download Diff File Expand all files Collapse all files

25
gpt4all-backend/llmodel.h
View File

@@ -1,13 +1,14 @@
#ifndef LLMODEL_H
#define LLMODEL_H
#include <string>
#include <functional>
#include <vector>
#include <string_view>
#include <fstream>
#include <cstdint>
#include <fstream>
#include <functional>
#include <limits>
#include <optional>
#include <string>
#include <string_view>
#include <vector>
#define LLMODEL_MAX_PROMPT_BATCH 128
@@ -44,6 +45,7 @@ public:
static std::vector<GPUDevice> availableGPUDevices();
static int32_t maxContextLength(const std::string &modelPath);
static int32_t layerCount(const std::string &modelPath);
static bool isEmbeddingModel(const std::string &modelPath);
static void setImplementationsSearchPath(const std::string &path);
static const std::string &implementationsSearchPath();
@@ -83,7 +85,8 @@ public:
virtual bool supportsEmbedding() const = 0;
virtual bool supportsCompletion() const = 0;
virtual bool loadModel(const std::string &modelPath, int n_ctx, int ngl) = 0;
virtual bool isModelBlacklisted(const std::string &modelPath) { (void)modelPath; return false; };
virtual bool isModelBlacklisted(const std::string &modelPath) const { (void)modelPath; return false; };
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; }
@@ -101,7 +104,15 @@ public:
bool special = false,
std::string *fakeReply = nullptr);
virtual std::vector<float> embedding(const std::string &text);
virtual size_t embeddingSize() const {
throw std::logic_error(std::string(implementation().modelType()) + " does not support embeddings");
}
// user-specified prefix
virtual void embed(const std::vector<std::string> &texts, float *embeddings, std::optional<std::string> prefix,
int dimensionality = -1, bool doMean = true, bool atlas = false);
// automatic prefix
virtual void embed(const std::vector<std::string> &texts, float *embeddings, bool isRetrieval,
int dimensionality = -1, bool doMean = true, bool atlas = false);
virtual void setThreadCount(int32_t n_threads) { (void)n_threads; }
virtual int32_t threadCount() const { return 1; }