ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-02-24 23:24:13 +00:00
Code Issues Packages Projects Releases Wiki Activity

Enable CUDA graphs for MoE models + GPT-OSS support (#689)

Browse Source 
* gmp-oss: common

* gpt-oss: attnetion sinks, swiglu_oai

* gpt-oss: WIP llama

Model loads and runs (CPU only), but PPL is much to high
(~1500 for 1st batch vs ~200 in mainline).
Is it because of SWA, because of vocab, or did I introduce a bug somewhere?

* gpt-oss: CPU seems to be working

It was the SWA thta was missing in the previous commit.

There are issues with EOG tokens, so this still needs to be added.

* CUDA: ADD_ID

Just a copy from mainline

* gpt-oss: Seems to be working on CUDA

* gpt-oss: add sinks to the attn-vec kernels

* CUDA: add head size of 64 to new mma

Haven't turned it on yet, but observe slightly better PP and slightly
worse TG performance with that.

* gpt-oss: add ability to use -fmoe (only CUDA for now)

* Move row sums to the write place

* Add sinks to iqk flash attention

* gpt_oss: Implement -fmoe on the CPU

* Simdify swiglu_oai

Turning it off for now as performance becomes more variable,
so perhaps I'm running into thermal trottling imore often
because of making the CPU work too hard.

* llama: factor out model loader

* Builds successfully

* It runs, but mmap does not work

* Fix llama_mmap so mmap works

* Minor

* Fix CUDA after latest changes

* Attempt to use CUDA graphs with MoE models - not working

* CUDA graphs WIP - still not working

* CUDA graphs - seems to be working

Likely not all MLA variants are working.
I no longer remember why I added the q8_0 cpy that
transposes the tensor, but if really needed, this is now
missing. Also missing is q6_0.

* Make q8_0 cache work for DeepSeek models with CUDA graphs

* cuda: cpy for q6_0

* Fix llama_mmap on non-Linux platforms

* Adding forgotten file

* Iterating on Windows build failures

* cuda: re-add q8_0 -> q8_0 transpose

so mla = 2 can be used with CUDA graphs and q8_0 cache.

* Disable graphs without -fmoe

* Minor

* Turn graphs on by default

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
This commit is contained in:
Kawrakow
2025-08-15 09:18:07 +03:00
committed by GitHub
parent e082df47f2
commit fc06bc9d27
56 changed files with 8720 additions and 5115 deletions
Download Patch File Download Diff File Expand all files Collapse all files

19
src/unicode.h
View File

@@ -4,9 +4,7 @@
#include <string>
#include <vector>
// TODO: prefix all symbols with "llama_"
struct codepoint_flags {
struct unicode_cpt_flags {
enum {
UNDEFINED = 0x0001,
NUMBER = 0x0002, // regex: \p{N}
@@ -35,7 +33,7 @@ struct codepoint_flags {
uint16_t is_nfd : 1;
// decode from uint16
inline codepoint_flags(const uint16_t flags=0) {
inline unicode_cpt_flags(const uint16_t flags = 0) {
*reinterpret_cast<uint16_t*>(this) = flags;
}
@@ -50,19 +48,20 @@ struct codepoint_flags {
size_t unicode_len_utf8(char src);
std::string unicode_cpt_to_utf8(uint32_t cp);
uint32_t unicode_cpt_from_utf8(const std::string & utf8, size_t & offset);
std::string unicode_cpt_to_utf8 (uint32_t cpt);
uint32_t unicode_cpt_from_utf8(const std::string & utf8, size_t & offset);
std::vector<uint32_t> unicode_cpts_from_utf8(const std::string & utf8);
std::vector<uint32_t> unicode_cpts_normalize_nfd(const std::vector<uint32_t> & cpts);
codepoint_flags unicode_cpt_flags(const uint32_t cp);
codepoint_flags unicode_cpt_flags(const std::string & utf8);
unicode_cpt_flags unicode_cpt_flags_from_cpt (uint32_t cpt);
unicode_cpt_flags unicode_cpt_flags_from_utf8(const std::string & utf8);
std::string unicode_byte_to_utf8(uint8_t byte);
uint8_t unicode_utf8_to_byte(const std::string & utf8);
uint8_t unicode_utf8_to_byte(const std::string & utf8);
uint32_t unicode_tolower(uint32_t cp);
uint32_t unicode_tolower(uint32_t cpt);
bool unicode_cpt_is_han(uint32_t cpt);