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

llama : save and restore kv cache for single seq id (#6341)

Browse source 
* llama : save and restore kv cache for single seq id

* remove trailing whitespace

* respond error in case there's no space in the kv cache

* add kv seq save restore to test case

* add --slot-save-path arg to enable save restore and restrict save location

* Returning 0 for some cases, instead of asserting.

* cleanup error cases

* rename sequence state functions

* rename state get set functions

* add previous function names back in with DEPRECATED notice

* update doc

* adjust endpoints to preferred style

* fix restoring zero cell count

* handle seq rm return value

* unused param

* keep in the size check

* fix return types

* add server test case for slot save restore

* cleanup

* add cake

* cleanup style

* add special

* removing a whole sequence never fails

* move sequence state file functionality from server to llama to match session api and add version tags

* catch exceptions on save as well

* error log messages

* check types for stricter restore

* update server doc

* readme : update API changes date

* strict filename validation

* move include, reject bom as well

* also reject empty filename

* reject whitespace and trailing dot

---------

Co-authored-by: Martin Evans <martindevans@gmail.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit is contained in:
Jan Boon 2024-04-08 20:43:30 +08:00 • committed by GitHub
parent 87fb5b4234
commit beea6e1b16
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
11 changed files with 1086 additions and 31 deletions
Download patch file Download diff file Expand all files Collapse all files

73
common/common.cpp
View file

@ -16,6 +16,7 @@
#include <unordered_set>
#include <vector>
#include <cinttypes>
#include <codecvt>
#if defined(__APPLE__) && defined(__MACH__)
#include <sys/types.h>
@ -27,7 +28,6 @@
#ifndef NOMINMAX
# define NOMINMAX
#endif
#include <codecvt>
#include <locale>
#include <windows.h>
#include <fcntl.h>
@ -1500,6 +1500,77 @@ std::string gpt_random_prompt(std::mt19937 & rng) {
GGML_UNREACHABLE();
}
// Validate if a filename is safe to use
// To validate a full path, split the path by the OS-specific path separator, and validate each part with this function
bool validate_file_name(const std::string & filename) {
if (!filename.length()) {
// Empty filename invalid
return false;
}
if (filename.length() > 255) {
// Limit at common largest possible filename on Linux filesystems
// to avoid unnecessary further validation
// (On systems with smaller limits it will be caught by the OS)
return false;
}
std::u32string filename_utf32;
try {
std::wstring_convert<std::codecvt_utf8<char32_t>, char32_t> converter;
filename_utf32 = converter.from_bytes(filename);
// If the reverse conversion mismatches, it means overlong UTF-8 sequences were used,
// or invalid encodings were encountered. Reject such attempts
std::string filename_reencoded = converter.to_bytes(filename_utf32);
if (filename_reencoded != filename) {
return false;
}
} catch (const std::exception &) {
return false;
}
// Check for forbidden codepoints:
// - Control characters
// - Unicode equivalents of illegal characters
// - UTF-16 surrogate pairs
// - UTF-8 replacement character
// - Byte order mark (BOM)
// - Illegal characters: / \ : * ? " < > |
for (char32_t c : filename_utf32) {
if (c <= 0x1F // Control characters (C0)
|| c == 0x7F // Control characters (DEL)
|| (c >= 0x80 && c <= 0x9F) // Control characters (C1)
|| c == 0xFF0E // Fullwidth Full Stop (period equivalent)
|| c == 0x2215 // Division Slash (forward slash equivalent)
|| c == 0x2216 // Set Minus (backslash equivalent)
|| (c >= 0xD800 && c <= 0xDFFF) // UTF-16 surrogate pairs
|| c == 0xFFFD // Replacement Character (UTF-8)
|| c == 0xFEFF // Byte Order Mark (BOM)
|| c == '/' || c == '\\' || c == ':' || c == '*' // Illegal characters
|| c == '?' || c == '"' || c == '<' || c == '>' || c == '|') {
return false;
}
}
// Reject any leading or trailing ' ', or any trailing '.', these are stripped on Windows and will cause a different filename
// Unicode and other whitespace is not affected, only 0x20 space
if (filename.front() == ' ' || filename.back() == ' ' || filename.back() == '.') {
return false;
}
// Reject any ".." (currently stricter than necessary, it should be fine to just check for == ".." instead)
if (filename.find("..") != std::string::npos) {
return false;
}
// Reject "."
if (filename == ".") {
return false;
}
return true;
}
//
// String utils
//