github/gpt4all
1
0
Fork 0
mirror of https://github.com/nomic-ai/gpt4all.git synced 2025-09-07 03:20:26 +00:00
Code Issues Packages Projects Releases Wiki Activity

Trying to shrink the copy+paste code and do more code sharing between backend model impl.

Browse Source
This commit is contained in:
Adam Treat
2023-06-01 23:15:58 -04:00
committed by AT
parent 031d7149a7
commit a41bd6ac0a
9 changed files with 41 additions and 96 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
gpt4all-backend/llmodel.cpp
View File

@@ -1,6 +1,7 @@
#include "llmodel.h"
#include "dlhandle.h"
#include <iostream>
#include <string>
#include <vector>
#include <fstream>
@@ -95,6 +96,28 @@ const LLModel::Implementation* LLModel::implementation(std::ifstream& f, const s
return nullptr;
}
void LLModel::recalculateContext(PromptContext &promptCtx, std::function<bool(bool)> recalculate) {
size_t i = 0;
promptCtx.n_past = 0;
while (i < promptCtx.tokens.size()) {
size_t batch_end = std::min(i + promptCtx.n_batch, promptCtx.tokens.size());
std::vector<int32_t> batch(promptCtx.tokens.begin() + i, promptCtx.tokens.begin() + batch_end);
assert(promptCtx.n_past + int32_t(batch.size()) <= promptCtx.n_ctx);
if (!evalTokens(promptCtx, batch)) {
std::cerr << "LLModel ERROR: Failed to process prompt\n";
goto stop_generating;
}
promptCtx.n_past += batch.size();
if (!recalculate(true))
goto stop_generating;
i = batch_end;
}
assert(promptCtx.n_past == int32_t(promptCtx.tokens.size()));
stop_generating:
recalculate(false);
}
LLModel *LLModel::construct(const std::string &modelPath, std::string buildVariant) {
//TODO: Auto-detect CUDA/OpenCL
if (buildVariant == "auto") {