vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

lookup: complement data from context with general text statistics (#5479)

Browse source 
* lookup: evaluation tools, use corpus/previous gens

* fixup! lookup: evaluation tools, use corpus/previous gens

* fixup! lookup: evaluation tools, use corpus/previous gens

* fixup! lookup: evaluation tools, use corpus/previous gens

* fixup! lookup: evaluation tools, use corpus/previous gens
This commit is contained in:
Johannes Gäßler 2024-03-23 01:24:36 +01:00 • committed by GitHub
parent 56a00f0a2f
commit 50ccaf5eac
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
13 changed files with 774 additions and 63 deletions
Download patch file Download diff file Expand all files Collapse all files

47
examples/lookup/lookup-merge.cpp Normal file
View file

@ -0,0 +1,47 @@
#include "ggml.h"
#include "llama.h"
#include "common.h"
#include "ngram-cache.h"
#include <cstdint>
#include <cstdio>
#include <fstream>
#include <iostream>
#include <string>
#include <unordered_map>
#include <vector>
static void print_usage() {
fprintf(stderr, "Merges multiple lookup cache files into a single one.\n");
fprintf(stderr, "Usage: lookup-merge [--help] lookup_part_1.bin lookup_part_2.bin ... lookup_merged.bin\n");
}
int main(int argc, char ** argv){
if (argc < 3) {
print_usage();
exit(1);
}
std::vector<std::string> args;
args.resize(argc-1);
for (int i = 0; i < argc-1; ++i) {
args[i] = argv[i+1];
if (args[i] == "-h" || args[i] == "--help") {
print_usage();
exit(0);
}
}
fprintf(stderr, "lookup-merge: loading file %s\n", args[0].c_str());
llama_ngram_cache ngram_cache_merged = llama_ngram_cache_load(args[0]);
for (size_t i = 1; i < args.size()-1; ++i) {
fprintf(stderr, "lookup-merge: loading file %s\n", args[i].c_str());
llama_ngram_cache ngram_cache = llama_ngram_cache_load(args[i]);
llama_ngram_cache_merge(ngram_cache_merged, ngram_cache);
}
fprintf(stderr, "lookup-merge: saving file %s\n", args.back().c_str());
llama_ngram_cache_save(ngram_cache_merged, args.back());
}