Commit graph

64 commits

Author SHA1 Message Date
sandyiscool
37e257c48e
make : clean *.so files (#1857) 2023-06-15 20:36:06 +03:00
Kerfuffle
74d4cfa343
Allow "quantizing" to f16 and f32 (#1787)
* Allow "quantizing" to f16 and f32

Fix an issue where quantizing didn't respect LLAMA_NO_K_QUANTS

Add brief help to the list of quantization types in the quantize tool

Ignore case for quantization type arguments in the quantize tool
2023-06-13 04:23:23 -06:00
rankaiyx
555275a693
make : add SSSE3 compilation use case (#1659) 2023-06-10 09:41:59 +03:00
Georgi Gerganov
5c64a0952e
k-quants : allow to optionally disable at compile time (#1734)
* k-quants : put behind optional compile flag LLAMA_K_QUANTS

* build : enable k-quants by default
2023-06-07 10:59:52 +03:00
Georgi Gerganov
2d43387daf
ggml : fix builds, add ggml-quants-k.o (close #1712, close #1710) 2023-06-06 10:18:03 +03:00
Kawrakow
99009e72f8
ggml : add SOTA 2,3,4,5,6 bit k-quantizations (#1684)
* Starting to add k-quantization to ggml

I think it is better to have quantization separate from
ggml. For now just adding the k-quants there, but it would be
better to also factor out the existing ggml quantizations.

* Adding Q3_K and Q8_K (de)-quantization

* Q3_K now working on CUDA and AVX2/scalar

CUDA is not ideal - ~50% slower than Q4_0 for
single token prediction, about the same in batch
mode (perplexity). CPU single token is ~55 ms
(on Ryzen 7950X).

* Some improvement for Q3_K on CUDA

It is now ~22.5 ms/token on my GPU, so ~30% slower than Q4_0.

* Some more CUDA optimizations for Q3_K

Single token is now 20.5 ms/token (~20% slower than Q4_0).
Perplexity is on par with Q4_0.

* Adding Q4_K - scalar, AVX2, CUDA

Performance is the same or perhaps very slightly better than Q4_0 on the CPU.
On the GPU, single token prediction is ~10% better than Q4_0,
batch mode (perplexity is about the same).

* Adding Q6_K - scalar, AVX2, CUDA

Performance is ~40% lower compared to Q4_K on the CPU.
This is to be expected, considering that we are memory bound
on the CPU and the 6-bit model is ~44% larger than the 4-bit.
On the GPU, single token prediction is ~6% lower than Q4_0,
batch mode (perplexity) is even closer (but still slower).

* Adding Q5_K - scalar, AVX2, CUDA

Performance is ~20% lower compared to Q4_K on the CPU.
This is to be expected, considering that we are memory bound
on the CPU and the 5-bit model is ~22% larger than the 4-bit.
On the GPU, single token prediction is about the same as Q4_0
for both, single token and batch prediction.

* Per convention, all QX_K quantizations use Q5_K for output.weight

* Adding quantization mixes

* Quantization mixes: didn't quite get what I wanted in the last commit

* Q4_K dot product for ARM_NEON

* Q6_K dot product for ARM_NEON

* Q5_K dot product for ARM_NEON

* Adding Q3_K dot for ARM_NEON

It is 22% slower than Q4_K, despite the smaller model size.
On x86_64, where we are memory bound, the Q3_K model is
quite a bit faster than Q4_K.

* A very slightly faster ARM_NEON Q3_K dot

* Adding Q2_K - just CUDA for now

Token prediction is pretty good - about 15.5 ms on a RTX 4080.
Perplexity is about the same as Q4_K.

* Adding scalar and AVX2 Q2_K dot

* Adding ARM_NEON Q2_K dot

About the same performance as Q4_K.

* A slightly faster ARM_NEON Q2_K dot

Single token prediction is now ~36 ms on M2 Max.
The code is much simpler too.

* Fixed bug in Q2_K CUDA dot product kernel

Stranegly enough, for the few prompts I tried with the 7B model
the responses looked perfectly reasonable. Only realized something
is not quite right when I tried the larger models and started getting
nonse back.

In any case, Q2_K single token evaluation time on an RTX 4080 in a Ryzen7950X
box iusing CUDA and model fully loaded on the GPU are
  ~15.5 ms for 7B, ~25.4 ms for 13B, and ~55.8 ms for 30B.
The max number of layers that fit in VRAM for The 65B is 32.
With that, we get ~330 ms per token, which is not that much faster
than just running on the CPU (~470 ms per token).

* Don't print zeros/NaNs when no count histogram has been collected

* A 10% faster CUDA vector dot kernel for Q3_K

Q3_K is now running at ~18.5 ms / token on CUDA,
so the gap to Q4_0 is only 10%.
It seems memory acccess pattern is more important for
performance than the amount of computation the kernel
does.

* A slightly daster Q4_K AVX2 dot product

For perplexity, where we are less memory bound, time per
pass drops by ~5%. Barely measurable difference for single
token prediction.

* A slightly faster ARM_NEON A4_K dot product

* Minor

* Fix quantization error test

We cannot possibly be expecting rmse < 0.002 for 2- and 3-bit
quantization variants.

* Fix docker build

I have been sloppy with vector reinterpret casts on ARM_NEON.
It seems clang is very forgiving in that regard.

* Added forgotten ggml.o dependence on k_quants.h to the Makefile

* Had unintentionally committed the Makefile with -Ofast enabled

* ggml : rename k_quants -> ggml-quants-k, use lowercase in code

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-05 22:56:18 +03:00
Georgi Gerganov
ecb217db4f
llama : Metal inference (#1642)
* mtl : export the LLaMA computation graph

* ci : disable temporary

* mtl : adapt the MNIST example as starter

* mtl : no need for mtl-export tool, add cli arg for main instead

* mtl : export just a small part of the graph for now to make it easier

* mtl : move MSL code into separate file for easy editing

* mtl : initial get_rows_q4_0 kernel

* mtl : confirmed get_rows_q4_0 is working correctly

* mtl : add rms_norm kernel + confirm working

* mtl : add mul kernel + confirm working

* mtl : initial mul_mat Q4 kernel (wrong results)

* mtl : mul_mat fixes (still wrong)

* mtl : another mul_mat Q4 (still does not work)

* mtl : working mul_mat q4

* ggml : fix handling of "view" ops in ggml_graph_import()

* mtl : add rope kernel

* mtl : add reshape and transpose handling

* ggml : store offset as opt arg for ggml_view_xd() operators

* mtl : add cpy kernel + handle view ops

* mtl : confirm f16 x f32 attention mul mat

* mtl : add scale kernel

* mtl : add diag_mask_inf kernel

* mtl : fix soft_max kernel

* ggml : update ggml_nbytes() to handle non-contiguous tensors

* mtl : verify V tensor contents

* mtl : add f32 -> f32 cpy kernel

* mtl : add silu kernel

* mtl : add non-broadcast mul kernel

* mtl : full GPU inference of the computation graph

* mtl : optimize rms_norm and soft_max kernels

* mtl : add f16 mat x f32 vec multiplication kernel

* mtl : fix bug in f16 x f32 mul mat + speed-up computation

* mtl : faster mul_mat_q4_0_f32 kernel

* mtl : fix kernel signature + roll inner loop

* mtl : more threads for rms_norm + better timing

* mtl : remove printfs from inner loop

* mtl : simplify implementation

* mtl : add save/load vocab to ggml file

* mtl : plug Metal inference into llama.cpp (very quick-n-dirty)

* mtl : make it work with main example

Lots of hacks but at least now it generates text

* mtl : preparing for merge

* mtl : clean-up ggml mtl interface + suport scratch / inplace

* mtl : remove temp / debug code

* metal : final refactoring and simplification

* Revert "ci : disable temporary"

This reverts commit 98c267fc77fe811082f672538fc91bcfc9072d63.

* metal : add comments

* metal : clean-up stuff, fix typos

* readme : add Metal instructions

* readme : add example for main
2023-06-04 23:34:30 +03:00
Johannes Gäßler
3b126f654f
LLAMA_DEBUG adds debug symbols (#1617) 2023-05-28 21:01:02 +02:00
Kerfuffle
0df7d63e5b
Include server in releases + other build system cleanups (#1610)
Set `LLAMA_BUILD_SERVER` in workflow so the `server` example gets build. This currently only applies to Windows builds because it seems like only Windows binary artifacts are included in releases.

Add `server` example target to `Makefile` (still uses `LLAMA_BUILD_SERVER` define and does not build by default)

Fix issue where `vdot` binary wasn't removed when running `make clean`.

Fix compile warnings in `server` example.

Add `.hpp` files to trigger workflow (the server example has one).
2023-05-27 11:04:14 -06:00
Johannes Gäßler
1fcdcc28b1
cuda : performance optimizations (#1530)
* xor hack

* block y dim

* loop unrolling

* Fixed cmake LLAMA_CUDA_BY option

* Removed hipblas compatibility code

* Define GGML_CUDA_DMMV_BLOCK_Y if not defined

* Fewer iters, more ops per iter

* Renamed DMMV X/Y compilation options
2023-05-26 00:07:29 +03:00
0cc4m
2e6cd4b025
OpenCL Token Generation Acceleration (#1459)
* Move back to C++ for OpenCL

* Refactor OpenCL code to work more like the CUDA code, add missing functions

* Deduplicate dequant kernels

* Add OpenCL compile options

* Use compile args for preprocessing constants

* Restore default platform + device selection by id behavior

---------

Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
Co-authored-by: Henri Vasserman <henv@hot.ee>
2023-05-23 00:33:24 +03:00
Stefan Sydow
7780e4f479
make : .PHONY clean (#1553) 2023-05-21 17:03:44 +03:00
Zenix
b8ee340abe
feature : support blis and other blas implementation (#1536)
* feature: add blis support

* feature: allow all BLA_VENDOR to be assigned in cmake arguments. align with whisper.cpp pr 927

* fix: version detection for BLA_SIZEOF_INTEGER, recover min version of cmake

* Fix typo in INTEGER

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

* Fix: blas changes on ci

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-05-20 17:58:31 +03:00
Georgi Gerganov
ea600071cb
Revert "feature : add blis and other BLAS implementation support (#1502)"
This reverts commit 07e9ace0f9.
2023-05-20 12:03:48 +03:00
Zenix
07e9ace0f9
feature : add blis and other BLAS implementation support (#1502)
* feature: add blis support

* feature: allow all BLA_VENDOR to be assigned in cmake arguments. align with whisper.cpp pr 927

* fix: version detection for BLA_SIZEOF_INTEGER, recover min version of cmake

* Fix typo in INTEGER

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-05-20 12:02:48 +03:00
sandyiscool
2a5ee023ad
Add alternate include path for openblas (#1476)
In some linux distributions (fedora, for example), the include path for openblas is located at '/usr/local/include'
2023-05-16 10:30:15 +02:00
Georgi Gerganov
bda4d7c215 make : fix PERF build with cuBLAS 2023-05-13 17:25:09 +03:00
DaniAndTheWeb
173d0e6419
makefile: automatic Arch Linux detection (#1332)
This commit is a port of a detection method used in koboldcpp's Makefile in order to automatically set the -lcblas option on Arch Linux
2023-05-05 23:57:14 +02:00
Ionoclast Laboratories
2d13786e91
Fix for OpenCL / clbast builds on macOS. (#1329) 2023-05-05 14:18:21 +02:00
DannyDaemonic
55bc5f0900
Call sh on build-info.sh (#1294) 2023-05-02 17:52:35 -07:00
DannyDaemonic
f4cef87edf
Add git-based build information for better issue tracking (#1232)
* Add git-based build information for better issue tracking

* macOS fix

* "build (hash)" and "CMAKE_SOURCE_DIR" changes

* Redo "CMAKE_CURRENT_SOURCE_DIR" and clearer build messages

* Fix conditional dependency on missing target

* Broke out build-info.cmake, added find_package fallback, and added build into to all examples, added dependencies to Makefile

* 4 space indenting for cmake, attempt to clean up my mess in Makefile

* Short hash, less fancy Makefile, and don't modify build-info.h if it wouldn't change it
2023-05-01 18:23:47 +02:00
Pavol Rusnak
6f79699286
build: add armv{6,7,8} support to cmake (#1251)
- flags copied from Makefile
- updated comments in both CMakeLists.txt and Makefile to match reality
2023-04-30 20:48:38 +02:00
Stephan Walter
f0d70f147d
Various fixes to mat_mul benchmark (#1253) 2023-04-30 12:32:37 +00:00
Georgi Gerganov
214b6a3570
ggml : adjust mul_mat_f16 work memory (#1226)
* llama : minor - remove explicity int64_t cast

* ggml : reduce memory buffer for F16 mul_mat when not using cuBLAS

* ggml : add asserts to guard for incorrect wsize
2023-04-29 18:43:28 +03:00
Georgi Gerganov
305eb5afd5
build : fix reference to old llama_util.h 2023-04-29 13:53:12 +03:00
slaren
7fc50c051a
cuBLAS: use host pinned memory and dequantize while copying (#1207)
* cuBLAS: dequantize simultaneously while copying memory

* cuBLAS: use host pinned memory

* cuBLAS: improve ggml_compute_forward_mul_mat_f16_f32 with pinned memory

* cuBLAS: also pin kv cache

* fix rebase
2023-04-29 02:04:18 +02:00
0cc4m
7296c961d9
ggml : add CLBlast support (#1164)
* Allow use of OpenCL GPU-based BLAS using ClBlast instead of OpenBLAS for context processing

* Improve ClBlast implementation, avoid recreating buffers, remove redundant transfers

* Finish merge of ClBlast support

* Move CLBlast implementation to separate file

Add buffer reuse code (adapted from slaren's cuda implementation)

* Add q4_2 and q4_3 CLBlast support, improve code

* Double CLBlast speed by disabling OpenBLAS thread workaround

Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com>
Co-authored-by: slaren <2141330+slaren@users.noreply.github.com>

* Fix device selection env variable names

* Fix cast in opencl kernels

* Add CLBlast to CMakeLists.txt

* Replace buffer pool with static buffers a, b, qb, c

Fix compile warnings

* Fix typos, use GGML_TYPE defines, improve code

* Improve btype dequant kernel selection code, add error if type is unsupported

* Improve code quality

* Move internal stuff out of header
* Use internal enums instead of CLBlast enums
* Remove leftover C++ includes and defines
* Make event use easier to read

Co-authored-by: Henri Vasserman <henv@hot.ee>

* Use c compiler for opencl files

* Simplify code, fix include

* First check error, then release event

* Make globals static, fix indentation

* Rename dequant kernels file to conform with other file names

* Fix import cl file name

---------

Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com>
Co-authored-by: slaren <2141330+slaren@users.noreply.github.com>
Co-authored-by: Henri Vasserman <henv@hot.ee>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-04-28 17:57:16 +03:00
Johannes Gäßler
92a6e13a31
Add Manjaro CUDA include and lib dirs to Makefile (#1212) 2023-04-28 15:40:32 +02:00
slaren
e4cf982e0d
Fix cuda compilation (#1128)
* Fix: Issue with CUBLAS compilation error due to missing -fPIC flag

---------

Co-authored-by: B1gM8c <89020353+B1gM8c@users.noreply.github.com>
2023-04-24 17:29:58 +02:00
Georgi Gerganov
e4422e299c
ggml : better PERF prints + support "LLAMA_PERF=1 make" 2023-04-23 18:15:39 +03:00
Georgi Gerganov
872c365a91 ggml : fix AVX build + update to new Q8_0 format 2023-04-22 11:08:12 +03:00
slaren
50cb666b8a
Improve cuBLAS performance by using a memory pool (#1094)
* Improve cuBLAS performance by using a memory pool

* Move cuda specific definitions to ggml-cuda.h/cu

* Add CXX flags to nvcc

* Change memory pool synchronization mechanism to a spin lock
General code cleanup
2023-04-21 21:59:17 +02:00
slaren
2005469ea1
Add Q4_3 support to cuBLAS (#1086) 2023-04-20 20:49:53 +02:00
源文雨
5addcb120c
fix: LLAMA_CUBLAS=1 undefined reference 'shm_open' (#1080) 2023-04-20 15:28:43 +02:00
slaren
02d6988121
Improve cuBLAS performance by dequantizing on the GPU (#1065) 2023-04-20 03:14:14 +02:00
Stephan Walter
f3d4edf504
ggml : Q4 cleanup - remove 4-bit dot product code (#1061)
* Q4 cleanup

* Remove unused AVX512 Q4_0 code
2023-04-19 19:06:37 +03:00
slaren
8944a13296
Add NVIDIA cuBLAS support (#1044) 2023-04-19 11:22:45 +02:00
Kawrakow
5ecff35151
Adding a simple program to measure speed of dot products (#1041)
On my Mac, the direct Q4_1 product is marginally slower
(~69 vs ~55 us for Q4_0). The SIMD-ified ggml version
is now almost 2X slower (~121 us).

On a Ryzen 7950X CPU, the direct product for Q4_1 quantization
is faster than the AVX2 implementation (~60 vs ~62 us).

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2023-04-18 19:00:14 +00:00
Georgi Gerganov
e95b6554b4
ggml : add Q8_0 quantization for intermediate results (#951)
* ggml : add Q8_0 quantization for intermediate results

* quantize-stats : fix test + add it to Makefile default

* Q8: use int8_t, AVX/AVX2 optimizations

* ggml : fix quantize_row_q8_0() ARM_NEON rounding

* minor : updates after rebase to latest master

* quantize-stats : delete obsolete strings

* ggml : fix q4_1 dot func

---------

Co-authored-by: Stephan Walter <stephan@walter.name>
2023-04-15 17:53:22 +03:00
Stephan Walter
93265e988a
make : fix dependencies, use auto variables (#983) 2023-04-14 22:39:48 +03:00
Georgi Gerganov
9190e8eac8
llama : merge llama_internal.h into llama.h
Hide it behind an #ifdef
2023-04-13 18:04:45 +03:00
CRD716
8cda5c981d
fix whitespace (#944) 2023-04-13 16:03:57 +02:00
SebastianApel
95ea26f6e9
benchmark : add tool for timing q4_0 matrix multiplication (#653)
* Initial version of q4_0 matrix multiplication benchmark

* Bugfix: Added dependency to ggml.o to benchmark

* Reviewer requests: added parameter for threads, switched to ggml_time_us()

* Reviewer input: removed rtsc, use epsilon for check

* Review comment: Removed set_locale

* Feature: Param for numer of iterations, Bugfix for use of parameter threads

* Reviewer suggestion: Moved to examples

* Reviewer feedback: Updated clean: and benchmark: sections

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-04-13 15:46:23 +03:00
comex
f963b63afa Rewrite loading code to try to satisfy everyone:
- Support all three formats (ggml, ggmf, ggjt).  (However, I didn't
  include the hack needed to support GPT4All files without conversion.
  Those can still be used after converting them with convert.py from my
  other PR.)

- Support both mmap and read (mmap is used by default, but can be
  disabled with `--no-mmap`, and is automatically disabled for pre-ggjt
  files or on platforms where mmap is not supported).

- Support multi-file models like before, but automatically determine the
  number of parts rather than requiring `--n_parts`.

- Improve validation and error checking.

- Stop using the per-file type field (f16) entirely in favor of just
  relying on the per-tensor type/size fields.  This has no immediate
  benefit, but makes it easier to experiment with different formats, and
  should make it easier to support the new GPTQ-for-LLaMa models in the
  future (I have some work in progress on that front).

- Support VirtualLock on Windows (using the same `--mlock` option as on
  Unix).

    - Indicate loading progress when using mmap + mlock.  (Which led me
      to the interesting observation that on my Linux machine, with a
      warm file cache, mlock actually takes some time, whereas mmap
      without mlock starts almost instantly...)

      - To help implement this, move mlock support from ggml to the
        loading code.

- madvise/PrefetchVirtualMemory support (based on #740)

- Switch from ifstream to the `fopen` family of functions to avoid
  unnecessary copying and, when mmap is enabled, allow reusing the same
  file descriptor for both metadata reads and mmap (whereas the existing
  implementation opens the file a second time to mmap).

- Quantization now produces a single-file output even with multi-file
  inputs (not really a feature as much as 'it was easier this way').

Implementation notes:

I tried to factor the code into more discrete pieces than before.

Regarding code style: I tried to follow the code style, but I'm naughty
and used a few advanced C++ features repeatedly:

- Destructors to make it easier to ensure everything gets cleaned up.

- Exceptions.  I don't even usually use exceptions when writing C++, and
  I can remove them if desired... but here they make the loading code
  much more succinct while still properly handling a variety of errors,
  ranging from API calls failing to integer overflow and allocation
  failure.  The exceptions are converted to error codes at the
  API boundary.)

Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-10 01:10:46 +02:00
unbounded
62cfc54f77
Add quantize-stats command for testing quantization (#728)
Command that calculates some statistics over the errors introduced by
quantization, like mean square error, max error and some percentile errors for layer
weights. Should be useful for testing quantization improvements.

Exposes some internal state from ggml and llama for testing
2023-04-08 00:09:18 +02:00
bhubbb
698f7b5d63
make : add libllama.so target for llama-cpp-python (#797)
I was able to get llama-cpp-python working but only when I build libllama.so with make.
2023-04-07 19:11:58 +03:00
Ivan Stepanov
0c44427df1
make : missing host optimizations in CXXFLAGS (#763) 2023-04-05 17:38:37 +03:00
Fabian
c4f89d8d73
make : use -march=native -mtune=native on x86 (#609) 2023-04-02 10:17:05 +03:00
david raistrick
1f0414feec
make : fix darwin f16c flags check (#615)
...there was no check.  ported upstream from https://github.com/zanussbaum/gpt4all.cpp/pull/2 (I dont see any clean path for upstream patches)
2023-03-30 20:34:45 +03:00
Stephan Walter
436e561931
all : be more strict about converting float to double (#458)
* Be more strict about converting float to double

* Test equivalence of round, SILU implementations

Test module is commented out in CMakeLists.txt because the tests may
take a long time, depending on how much the compiler optimizes.

* Fix softmax in perplexity.cpp

* all : prefer float over double where appropriate

* perplexity : add <cmath>

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-03-28 19:48:20 +03:00