Commit Graph

41 Commits

Author SHA1 Message Date
Brian
f7cab35ef9
gguf-hash: model wide and per tensor hashing using xxhash and sha1 (#8048)
CLI to hash GGUF files to detect difference on a per model and per tensor level

The hash type we support is:

- `--xxh64`: use xhash 64bit hash mode (default)
- `--sha1`: use sha1
- `--uuid`: use uuid
- `--sha256`: use sha256

While most POSIX systems already have hash checking programs like sha256sum, it
is designed to check entire files. This is not ideal for our purpose if we want
to check for consistency of the tensor data even if the metadata content of the
gguf KV store has been updated.

This program is designed to hash a gguf tensor payload on a 'per tensor layer'
in addition to a 'entire tensor model' hash. The intent is that the entire
tensor layer can be checked first but if there is any detected inconsistencies,
then the per tensor hash can be used to narrow down the specific tensor layer
that has inconsistencies.

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-07-07 22:58:43 +10:00
Georgi Gerganov
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
* scripts : update sync [no ci]

* files : relocate [no ci]

* ci : disable kompute build [no ci]

* cmake : fixes [no ci]

* server : fix mingw build

ggml-ci

* cmake : minor [no ci]

* cmake : link math library [no ci]

* cmake : build normal ggml library (not object library) [no ci]

* cmake : fix kompute build

ggml-ci

* make,cmake : fix LLAMA_CUDA + replace GGML_CDEF_PRIVATE

ggml-ci

* move public backend headers to the public include directory (#8122)

* move public backend headers to the public include directory

* nix test

* spm : fix metal header

---------

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

* scripts : fix sync paths [no ci]

* scripts : sync ggml-blas.h [no ci]

---------

Co-authored-by: slaren <slarengh@gmail.com>
2024-06-26 18:33:02 +03:00
Xuan Son Nguyen
0c7b3595b9
Add cvector-generator example (#7514)
* add control-vector-generator

* calc diff

* add comments

* proof-of-concept stdlib implementation

Implements PCA and file writing using mostly standard libraries. The output is recognized as a functional control vector, but outputs gibberish.

* param parsing, refactor, comments

Added basic command-line parameters for outfile and one each positive/negative prompt.

Refactored some messy code in PCA computation and GGUF exporting.

Left a bunch of comments regarding further work needed.

* example template completions

Implements an example template set built from the positive/negative prompts like the control vector Python implementation.

* add multi prompts, multi-thread for PCA

* fix mem error

* add debugs

* fix matrix transpose multiplication

you have got to be kidding me

* preliminary template/multiprompt support

model is running out of context and that ought to be fixed (segfaulting) but other than that it looks goodish

* fix zero output & param parsing, functional templating

fixed a bug where the output file had no tensor data/was all zero

fixed a bug where single hyphen flags were not being correctly parsed

implements creation of templated prompts from input (still need to adapt based on model)

* fix square_diff matmul index range and CRLF->LF line endings

fixed a logic error where square_diff would not multiply all rows

fixed a formatting error where the provided completions.txt had CRLF line endings

* add command-line args for num threads, num completions file lines, always reload model

refactored a few things and did what the commit message says on the tin

* code aestheticization

* fix compiler warnings

* in-series multithreading for prompt embedding?

added commented-out code to attempt to start implementing mutlithreading for embedding in main

* remove unnecessary multithreading

* interim fix memory leak

* translated everything but PCA (I think)

* tentatively translate the rest

* fix ggml errors and make new ones

at least it compiles and runs

* fix cb_eval

* temporary commit while I move dev environments

it finally outputs a functioning control vector - "functioning" in the sense that it can be loaded and it clearly has the right idea, but makes the model incoherent

* update debug statements

* pre-tokenize so we can allocate correct memory to ctx_diffs_wrapped

* update comments

* (wip) refactor

* clean up PCA ggml implementation

* fix shape of v_diff_original

* add n_batch for pca

* working version

* remember to copy back the last_eigenvector

* fix n_completions

* bring back n_completions

* default n_pca_batch to 20

* fix macos build

* add to makefile all targets

* use ggml_format_name

* add readme

* fix .editorconfig

* use ggml_backend_tensor_copy

* attemp to fix compile problem on mac

* fix compile warn

* reuse allocr

* move param parser to common

* better error handling

* clean up a bit

* add print_usage

* shorten help msg

* beautify help msg

* escape prompt by default

* change compile target to llama-cvector-generator

* typo

* disable GPU for PCA

* code style

---------

Co-authored-by: Christian Zhou-Zheng <christianzhouzheng@gmail.com>
2024-06-15 18:53:40 +02:00
Olivier Chafik
1c641e6aac
build: rename main → llama-cli, server → llama-server, llava-cli → llama-llava-cli, etc... (#7809)
* `main`/`server`: rename to `llama` / `llama-server` for consistency w/ homebrew

* server: update refs -> llama-server

gitignore llama-server

* server: simplify nix package

* main: update refs -> llama

fix examples/main ref

* main/server: fix targets

* update more names

* Update build.yml

* rm accidentally checked in bins

* update straggling refs

* Update .gitignore

* Update server-llm.sh

* main: target name -> llama-cli

* Prefix all example bins w/ llama-

* fix main refs

* rename {main->llama}-cmake-pkg binary

* prefix more cmake targets w/ llama-

* add/fix gbnf-validator subfolder to cmake

* sort cmake example subdirs

* rm bin files

* fix llama-lookup-* Makefile rules

* gitignore /llama-*

* rename Dockerfiles

* rename llama|main -> llama-cli; consistent RPM bin prefixes

* fix some missing -cli suffixes

* rename dockerfile w/ llama-cli

* rename(make): llama-baby-llama

* update dockerfile refs

* more llama-cli(.exe)

* fix test-eval-callback

* rename: llama-cli-cmake-pkg(.exe)

* address gbnf-validator unused fread warning (switched to C++ / ifstream)

* add two missing llama- prefixes

* Updating docs for eval-callback binary to use new `llama-` prefix.

* Updating a few lingering doc references for rename of main to llama-cli

* Updating `run-with-preset.py` to use new binary names.
Updating docs around `perplexity` binary rename.

* Updating documentation references for lookup-merge and export-lora

* Updating two small `main` references missed earlier in the finetune docs.

* Update apps.nix

* update grammar/README.md w/ new llama-* names

* update llama-rpc-server bin name + doc

* Revert "update llama-rpc-server bin name + doc"

This reverts commit e474ef1df4.

* add hot topic notice to README.md

* Update README.md

* Update README.md

* rename gguf-split & quantize bins refs in **/tests.sh

---------

Co-authored-by: HanClinto <hanclinto@gmail.com>
2024-06-13 00:41:52 +01:00
Georgi Gerganov
0cd6bd3483
llama : remove beam search (#7736) 2024-06-04 21:23:05 +03:00
Radoslav Gerganov
5e31828d3e
ggml : add RPC backend (#6829)
* ggml : add RPC backend

The RPC backend proxies all operations to a remote server which runs a
regular backend (CPU, CUDA, Metal, etc).

* set TCP_NODELAY

* add CI workflows

* Address review comments

* fix warning

* implement llama_max_devices() for RPC

* Address review comments

* Address review comments

* wrap sockfd into a struct

* implement get_alignment and get_max_size

* add get_device_memory

* fix warning

* win32 support

* add README

* readme : trim trailing whitespace

* Address review comments

* win32 fix

* Address review comments

* fix compile warnings on macos
2024-05-14 14:27:19 +03:00
Pierrick Hymbert
b804b1ef77
eval-callback: Example how to use eval callback for debugging (#6576)
* gguf-debug: Example how to use ggml callback for debugging

* gguf-debug: no mutex, verify type, fix stride.

* llama: cv eval: move cb eval field in common gpt_params

* ggml_debug: use common gpt_params to pass cb eval.
Fix get tensor SIGV random.

* ggml_debug: ci: add tests

* ggml_debug: EOL in CMakeLists.txt

* ggml_debug: Remove unused param n_batch, no batching here

* ggml_debug: fix trailing spaces

* ggml_debug: fix trailing spaces

* common: fix cb_eval and user data not initialized

* ci: build revert label

* ggml_debug: add main test label

* doc: add a model: add a link to ggml-debug

* ggml-debug: add to make toolchain

* ggml-debug: tests add the main label

* ggml-debug: ci add test curl label

* common: allow the warmup to be disabled in llama_init_from_gpt_params

* ci: add curl test

* ggml-debug: better tensor type support

* gitignore : ggml-debug

* ggml-debug: printing also the sum of each tensor

* ggml-debug: remove block size

* eval-callback: renamed from ggml-debug

* eval-callback: fix make toolchain

---------

Co-authored-by: slaren <slarengh@gmail.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-11 14:51:07 +02:00
Minsoo Cheong
64e7b47c69
examples : add "retrieval" (#6193)
* add `retrieval` example

* add README

* minor fixes

* cast filepos on print

* remove use of variable sized array

* store similarities in separate vector

* print error on insufficient batch size

* fix error message printing

* assign n_batch value to n_ubatch

* fix param definitions

* define retrieval-only parameters in retrieval.cpp

* fix `--context-file` option to be provided multiple times for multiple files

* use vector for `query_emb`

* add usage description in README

* fix merge conflict

* fix usage printing

* remove seed setting

* fix lint

* increase file read buffer size

* retrieval : minor

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-25 09:38:22 +02:00
Pierrick Hymbert
d0d5de42e5
gguf-split: split and merge gguf per batch of tensors (#6135)
* gguf-split: split and merge gguf files per tensor

* gguf-split: build with make toolchain

* gguf-split: rename `--split-tensors-size` to `--split-max-tensors`. Set general.split_count KV to all split

* split : minor style + fix compile warnings

* gguf-split: remove --upload not implemented

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-19 12:05:44 +01:00
DAN™
bcebd7dbf6
llama : add support for GritLM (#5959)
* add gritlm example

* gritlm results match

* tabs to spaces

* comment out debug printing

* rebase to new embed

* gritlm embeddings are back babeee

* add to gitignore

* allow to toggle embedding mode

* Clean-up GritLM sample code.

* Fix types.

* Flush stdout and output ending newline if streaming.

* mostly style fixes; correct KQ_mask comment

* add causal_attn flag to llama_cparams

* gritml : minor

* llama : minor

---------

Co-authored-by: Douglas Hanley <thesecretaryofwar@gmail.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-10 17:56:30 +02:00
John
6c00a06692
gguf : add python reader example (#5216)
* Update CMakeLists.txt

* Create reader.py

* Update reader.py

* Update reader.py

another whitespace :|

* Update reader.py

* lintlintlint
2024-02-13 19:56:38 +02:00
Abhilash Majumder
0f648573dd
ggml : add unified SYCL backend for Intel GPUs (#2690)
* first update for migration

* update init_cublas

* add debug functio, commit all help code

* step 1

* step 2

* step3 add fp16, slower 31->28

* add GGML_LIST_DEVICE function

* step 5 format device and print

* step6, enhance error check, remove CUDA macro, enhance device id to fix none-zero id issue

* support main device is non-zero

* step7 add debug for code path, rm log

* step 8, rename all macro & func from cuda by sycl

* fix error of select non-zero device, format device list

* ren ggml-sycl.hpp -> ggml-sycl.h

* clear CMAKE to rm unused lib and options

* correct queue: rm dtct:get_queue

* add print tensor function to debug

* fix error: wrong result in 658746bb26702e50f2c59c0e4ada8e9da6010481

* summary dpct definition in one header file to replace folder:dpct

* refactor device log

* mv dpct definition from folder dpct to ggml-sycl.h

* update readme, refactor build script

* fix build with sycl

* set nthread=1 when sycl, increase performance

* add run script, comment debug code

* add ls-sycl-device tool

* add ls-sycl-device, rm unused files

* rm rear space

* dos2unix

* Update README_sycl.md

* fix return type

* remove sycl version from include path

* restore rm code to fix hang issue

* add syc and link for sycl readme

* rm original sycl code before refactor

* fix code err

* add know issue for pvc hang issue

* enable SYCL_F16 support

* align pr4766

* check for sycl blas, better performance

* cleanup 1

* remove extra endif

* add build&run script, clean CMakefile, update guide by review comments

* rename macro to intel hardware

* editor config format

* format fixes

* format fixes

* editor format fix

* Remove unused headers

* skip build sycl tool for other code path

* replace tab by space

* fix blas matmul function

* fix mac build

* restore hip dependency

* fix conflict

* ren as review comments

* mv internal function to .cpp file

* export funciton print_sycl_devices(), mv class dpct definition to source file

* update CI/action for sycl code, fix CI error of repeat/dup

* fix action ID format issue

* rm unused strategy

* enable llama_f16 in ci

* fix conflict

* fix build break on MacOS, due to CI of MacOS depend on external ggml, instead of internal ggml

* fix ci cases for unsupported data type

* revert unrelated changed in cuda cmake
remove useless nommq
fix typo of GGML_USE_CLBLAS_SYCL

* revert hip cmake changes

* fix indent

* add prefix in func name

* revert no mmq

* rm cpu blas duplicate

* fix no_new_line

* fix src1->type==F16 bug.

* pass batch offset for F16 src1

* fix batch error

* fix wrong code

* revert sycl checking in test-sampling

* pass void as arguments of ggml_backend_sycl_print_sycl_devices

* remove extra blank line in test-sampling

* revert setting n_threads in sycl

* implement std::isinf for icpx with fast math.

* Update ci/run.sh

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

* Update examples/sycl/run-llama2.sh

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

* Update examples/sycl/run-llama2.sh

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

* Update CMakeLists.txt

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

* Update CMakeLists.txt

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

* Update CMakeLists.txt

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

* Update CMakeLists.txt

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

* add copyright and MIT license declare

* update the cmd example

---------

Co-authored-by: jianyuzh <jianyu.zhang@intel.com>
Co-authored-by: luoyu-intel <yu.luo@intel.com>
Co-authored-by: Meng, Hengyu <hengyu.meng@intel.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 17:56:23 +02:00
Georgi Gerganov
4be5ef556d
metal : remove old API (#4919)
ggml-ci
2024-01-13 20:45:45 +02:00
Kawrakow
326b418b59
Importance Matrix calculation (#4861)
* imatrix: 1st version

* imatrix: WIP

* Cleanup

* Update examples/imatrix/imatrix.cpp

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

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-12 06:59:57 +01:00
Georgi Gerganov
b0034d93ce
examples : add passkey test (#3856)
* examples : add passkey test

* passkey : better prints

* passkey : select pass key pos from CLI

* passkey : simplify n_past logic

* make : add passkey target

* passkey : add "self-extend"-like context extension (#4810)

* llama : "self-extend"-like context extension

* passkey : add comment

* passkey : add readme
2024-01-08 11:14:04 +02:00
LeonEricsson
7082d24cec
lookup : add prompt lookup decoding example (#4484)
* initial commit, going through initializations

* main loop finished, starting to debug

* BUG: generates gibberish/repeating tokens after a while

* kv_cache management

* Added colors to distinguish drafted tokens (--color). Updated README

* lookup : fix token positions in the draft batch

* lookup : use n_draft from CLI params

* lookup : final touches

---------

Co-authored-by: Leon Ericsson <leon.ericsson@icloud.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-22 18:05:56 +02:00
Georgi Gerganov
922754a8d6
lookahead : add example for lookahead decoding (#4207)
* lookahead : init

* lookahead : generate and store n-grams

* lookahead : use loop instead recursion to generate n-grams

* lookahead : initial working implementation

* lookahead : filter repeating n-grams

* lookahead : use deterministic init

* lookahead : add to Makefile

* lookahead : fix a bug in the seq_id of the lookahead tokens

* lookahead : add comments

---------

Co-authored-by: slaren <slarengh@gmail.com>
2023-11-26 20:33:07 +02:00
zakkor
2fa02b4b3d
examples : add tokenize (#4039) 2023-11-17 17:36:44 +02:00
Georgi Gerganov
d1031cf49c
sampling : refactor init to use llama_sampling_params (#3696)
* sampling : refactor init to use llama_sampling_params

* llama : combine repetition, frequency and presence penalties in 1 call

* examples : remove embd-input and gptneox-wip

* sampling : rename penalty params + reduce size of "prev" vector

* sampling : add llama_sampling_print helper

* sampling : hide prev behind API and apply #3661

ggml-ci
2023-10-20 21:07:23 +03:00
M. Yusuf Sarıgöz
370359e5ba
examples: support LLaVA v1.5 (multimodal model) (#3436)
* WIP: start implementing LLaVA

* rm scratch buf for now, will revert after cleanup

* LLaVA image encoder is working. will combine with llama

* Add llava inference code, but it's buggy. debugging

* LLaVA is working e2e, needs to optimize memory allocation + cleanup

* Use ggml_allocr + rm unnecessary code

* fix: crlf -> lf

* fix: new line at EoF

* fix: trailing whitespace

* Add readme

* Update readme

* Some cleanup

* Are you happy editorconfig?

* rm unused batch image preprocessing

* rm unused import

* fix: rm designated initializers

* introduce pad-to-square mode for non-square images

* are you happy editorconfig?

* gitignore /llava

* Handle cases where image file does not exist

* add llava target to Makefile

* add support for 13b model variant

* Maybe seed is unlucky?

* Check if apples are compared to apples

* are you happy editorconfig?

* Use temperature = 0.1 by default

* command line: use gpt_params_parse()

* minor

* handle default n_predict

* fix typo

* llava : code formatting, rename files, fix compile warnings

* do not use Wno-cast-qual for MSVC

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-12 18:23:18 +03:00
Georgi Gerganov
8c70a5ff25
batched : add bench tool (#3545)
* batched : add bench tool

* batched : minor fix table

* batched-bench : add readme + n_kv_max is now configurable

* batched-bench : init warm-up batch

* batched-bench : pass custom set of PP, TG and PL

* batched-bench : add mmq CLI arg
2023-10-11 21:25:33 +03:00
xaedes
0e76a8992c
train : finetune LORA (#2632)
* fix track_max_mem in forward_batch_wo_cache_flash_attn_train

* remove unnecessary Adam(W) optimizer tensors.

reduces optimizer memory overhead from 7*modelsize to 2*modelsize.

additionally allows to optimize models with more than 2^31 parameters by replacing int with int64_t.

bumps training checkpoint file version, but old checkpoints can still be read.
new version with less tensors is saved.

* add gradient clipping to AdamW

* Fix reset of unused g->nodes and g->grads to NULL

* implement gradient checkpointing for training

reduces memory overhead from O(n_layer) to O(sqrt(n_layer))

as explained in readme of https://github.com/cybertronai/gradient-checkpointing

* remove unused compute buffer 3

* add and use function ggml_build_backward_expand to avoid stack overflows with large maximum number of nodes

GGML_API void ggml_build_backward_expand(struct ggml_context * ctx, struct ggml_cgraph * gf, struct ggml_cgraph * gb, bool keep);

* change AdamW decay parameter to work like the torch AdamW decay parameter

It is now relative to Adam learning rate `alpha*sched`.
Before that it was relative to `sched` only.

`alpha` being the maximum learning rate and `sched` being a scaling parameter in [0..1]

* change default AdamW weight decay parameter used in training to 0.1 as used in nanoGPT

* change default AdamW weight decay parameter defined in ggml to 0.0, making Adam default instead of AdamW

btw: the default weight decay parameter for torch.optim.AdamW is 0.01

* bug fixes for cross entropy loss

ggml_cross_entropy_loss: sums where not correctly added in workload of each thread
ggml_cross_entropy_loss_back: simplify backward process, reducing numerical issues

guard usage of exp f16 lookup in cross entropy by #define GGML_CROSS_ENTROPY_EXP_FP16

cross entropy loss is only used once during training, but it is quite sensitive to numerical errors introduced by exp-f16-lookup.
so exp-f16-lookup for cross entropy loss is disabled by default, trading better gradients for very slightly worse runtime performance.

* fix test-grad0 for cross_entropy_loss

the second argument to cross_entropy_loss must sum up to 1 for each row

* fix test-grad0 for soft_max

dont use only sum as aggregation, because sum of softmax is always 1 -> finite differences should not work
instead use sum(log(soft_max()*(1-eps)+eps)); use eps to avoid log(0)

* improve finite differences of test-grad0 by using double instead of float

* change cross_entropy_loss to output average over all rows

this helps keeping the loss and gradients in a sane range

* improve gradient checkpointing

sqrt(n_layers) is only the best checkpoint step when mem size of checkpoints and mem size of layers are equal.
since layers require more memory than the single-tensor-checkpoint we use, the optimal values are compute different:

```
  given: n, u, v
  objective: minimize(a*u+b*v) where a*b=n, a>0, b>0
  b=n/a
  minimize(a*u+v*n/a)
  diff(a*u+v*n/a, a) = u - (v*n/a)/a
  diff(a*u+v*n/a, a) == 0
  u - (v*n/a)/a == 0
  u == v*n/(a*a)
  u*a*a = v*n
  a*a = v*n/u
  a = sqrt(n*v/u)
```

this change results in more checkpoints, requiring less layers to store between checkpoints, overall improving memory usage.

* disable gradient checkpointing debug output

* llama : fix rope usage in train-text-from-scratch after ChatGLM change

* add more training parameters:

--enable-restart N         Only for Adam optimizer. Enable restarts of cos-decay
--disable-restart N        Only for Adam optimizer. Disable restarts of cos-decay
--opt-past N               Number of optimization iterations to track for delta convergence test. Disabled when zero.
--opt-delta N              Maximum delta for delta convergence test. Disabled when <= zero.
--opt-max-no-improvement N Maximum number of optimization iterations with no improvement. Disabled when <= zero.
--adam-epsf N              AdamW epsilon for convergence test. Disabled when <= zero.
--adam-min-alpha N         Adam minimum learning rate alpha, usually 0.1 * alpha

* replace memcpy with reshape operation so that the graph is not cut at the input

this makes it possible to store other values into the input tensor and then simply recompute the graph without rebuilding it

* remove unused function argument from get_example_targets_batch

* measure and print total training time

* add optimization callback to ggml_opt_resume_g

this callback is called before each iteration with custom data and pointer to learning schedule parameter (only used in Adam(W)).

can be used for dynamic learning schedule and setting input data for batches before each iteration

* use optimization callback in training

allows dynamic learning schedule and different batch data for each iteration without relying on low n_iter and high n_examples parameters

reduces runtime by avoiding restart of optimization function and improves training convergence by providing a different batch for each iteration

* add minimum number of tensor dimensions to apply weight decay (default 2)

this allows to not apply weight decay to bias parameters

* rename training parameter cos-decay-alpha to cos-decay-min and clarify that adam-min-alpha also applies to warmup

* fix increase of model.train_samples and model.train_tokens

now that each optimizer iteration gets its own batch we need to multiply by number of opt iterations

* change sampling parameters for prediction after training to defaults of common.h

and clarify what is context for prediction and what are generated tokens

* tighten abs error bounds for cross_entropy_loss in test-grad0

* add conditional compilation of using F16 exp in flash attention

uncomment `// #define GGML_FLASH_ATTN_EXP_FP16` to enable usage of f16 exp in flash attention

* tighten abs error bounds for flash_attn in test-grad0

* tighten abs error bounds for sqrt in test-grad0

* remove out-commented vectorized code of opt_adam

the vectorized code might be bit faster for low number of parameters, but it had a big memory usage overhead

* ggml : update ggml_rms_norm_back with configurable eps

* llama training : fix ggml_rms_norm_back calls to pass configurable eps

* remove trailing whitespace

* add train function using automatic gradient checkpointing backward pass and allocator

* in train function replace add_inplace by regular add

because using add_inplace seems to result in different gradients

* don't use allocate hash_map on context

because the context has no_alloc=True when using memory allocator resulting in NULL data pointers

* correctly clone reshape and permute operations by also cloning tensor->nb values

* fix variable name and add missing type cast

* terminate recursive tensor cloning when reaching tensor without src tensors

* correctly clone view tensors by setting data pointers

without this the checkpointing would only work when being used together with memory allocator

* fix variable names

* swap arguments to commutative ops to be the same as in `forward_batch_wo_cache_flash_attn`

* add input tensors as checkpoints

so that recursive tensor cloning of gradient checkpointing terminates on input tensors

* fix variable name and add missing boolean negation

* make sure some tensors are not reallocated by inserting new temporary nodes depending on them:

output and parameter gradient tensors need to be available at the end of the graph execution

parameter gradient tensors also need to be available before the graph execution because they are set to zero before each optimizer iteration

checkpoint tensors are allocated all together to reduce memory allocator fragmentation

afterwards, in addition to the temporary nodes, we also need to reset the temporary leafs

* fix ASSERT to work with zero layers

* add training options whether to use allocator and/or unified training function

* integrate unified training function which may use memory allocator

the unified training function also supports arguments whether to use flash attention and/or gradient checkpointing

* format name of cloned tensors with " (clone)" suffix

* set names for tensors in unified train function for easier debugging

* allocate graph on context using ggml_new_graph

* remove handwritten training functions

* remove unused training parameters "use_scratch" and "use_unified"

* remove trailing whitespace

* remove unused train params: mem_compute1_gb & mem_compute2_gb

mem_compute_gb is used for compute when automatic memory allocator is not enabled, otherwise it can be very small to only hold the tensor definitions
mem_compute0_gb is used for automatic memory allocator (as long as measurement of max required size is not implemented)

* remove unused forward_batch function

* add debug asserts in ggml_allocr_alloc to some common pitfalls when using this function directly

* only use ggml_allocr_alloc when tensor has NULL data and is no view

* fix test when to create temporary backward graph

temporary backward graph is only necessary when using checkpointing

* fix memory "leak" in optimizers

each iteration a new cplan with new memory for work data was allocated.
now cplan creation only happens at the start of optimization, with each iteration reusing the cplan and its work data.

* reverse order of for loop in ggml_build_backward_expand to save memory when using gradient checkpointing and allocator

with this loop order gradient checkpointing with allocator on 16 layer model saves 13% memory; 2 layer memory it saves 2% memory.

the computation results are the same

* add API functions to access llama model tensors

* add stub example for finetuning, based on train-text-from-scratch

* move and remove code

* add API functions to access remaining model parameters:

mult, head and rot

* first draft for LORA finetune training

* remove const model and layer arguments in API functions for accessing model tensors

* bug fixes to make finetune compile

automatic allocator does not work yet

* add debug prints for training memory improvements

* fix names of lora tensors

* avoid stack overflow resulting from big ggml_cgraph

replace stack allocation and ggml_build_forward by ggml_new_graph in combination with ggml_build_forward_expand

* replace llama API functions to get model tensors by one function to get model tensor by name

LLAMA_API struct ggml_tensor * llama_get_model_tensor(struct llama_model * model, const char * name);

* remove unused call to not existing llama_get_layer_from_model

* implement ggml_compute_forward_out_prod_q_f32

* remove trailing whitespace

* add lora finetune support on quantized base model tensors

* add ggml_add_cast API function

this function works like ggml_add, but accepts a data type for the resulting tensor.
only supported for quantized src0 input.

* use ggml_add_cast in finetuning

lora-applied weights will now have data type F32, which improves gradients when finetuning quantized base models

* bug fix: actually use result type passed to ggml_add_cast

* make sure base model tensors data cannot be used in viewable operations

memory allocator would try to make lora application inplace on base model tensors.
since those are memory mapped this will result in memory access violations

* fix bug in ggml_out_prod which resulted in wrong n_dims of result tensors

* avoid keeping in memory ALL of the gradients

The problem here stems from ggml_graph_reset. This function is called in the optimization function, before each graph computation, to reset the gradients to zero. This required a unique memory slot for each gradient: allocating memory from a previosly freed memory location might lead to non-zero input gradients.

During ggml_compute_backward the gradients are build stepwise by adding or substracting new values, starting from a OP_NONE tensor which needs to contain zero-values. This requires the graph reset.

To avoid this I now remember in ggml_build_backward_expand the original OP_NONE gradient tensors in a hash table, which is passed to ggml_compute_backward. There instead of using add (or sub or similar) I test whether the existing gradient to be changed is a zero-valued-tensor by looking up its existence in the hash table. When it is such a zero-tensor it will not be modified, but replaced by the value to be added, otherwise the regular add (not inplace, allocator will take care of this) will be used. This way none of those zero-tensor values will be necessary in the final backward graph and more importantly they won't need a unique memory slot, just to make them zero.

* remove trailing whitespace

* remove debug prints and function to compute tensor data hash

* improve optimization iteration prints

* adjust maximal values to support finetuning 3B models

* change default finetune params lora_r and lora_alpha to match the n_rank parameters of 4

* bug fix: make sure finetune input gradient is allocated at begin and kept until end

* remove unnecessary src tensor from ggml_get_rows_back

we don't need data of src[2] for computation, only to setup the correct output shape.
remove dependency on src[2], so that allocator can work more freely.

the computational graph is still completely determined, because the output shape is naturally included.
this is similar to how ggml_reshape does it.

* remove unnecessary src tensor from ggml_repeat & ggml_repeat_back

we don't need data of src[1] for computation, only to setup the correct output shape.
remove dependency on src[1], so that allocator can work more freely.

the computational graph is still completely determined, because the output shape is naturally included

* resolve todo

allocator will only make it inplace when they are of the same type

* mixing multiple LORA adapters is now possible

pass more than one '--lora FNAME' argument to apply more than one LORA.
use '--lora-scaled FNAME S' when you want to specify a user-defined scale for an adapter.

* add option to save finetune output every N iterations

* also save latest finetune output with ITERATION="LATEST" and print where files are saved

saving with LATEST makes it easier to resume training from the latest checkpoint
the string "LATEST" can be configured with command line option "--fn-latest STR"

* update checkpoint train stats before saving via "--save-every"

* add command line option `--rank-wo N` for rank of wo tensor

* update finetune README

* fix dump_non_result_info_yaml to output multiple lora adapters

* bug fix: replace GGML_TYPE_SIZE[t] by ggml_type_size(t)

* replace llama_n_mult by llama_n_ff

* finetune bug fixes to compile with merged in code from master

* remove prediction related code to reduce duplicated code with main

use main instead

* reduce large memory overhead in train-text-from-scratch

all gradients had to be pinned so that graph_reset works correctly.
this is no longer necessary with the changes to ggml_compute_backward introduced in this PR.

* add comment explaining why finetune checkpoints are allocated in one block

* make default value of float member a float literal

* handle rms_norm and rope parameters the same as in train-text-from-scratch

* remove unused code

* remove vocab related code as it is unnecessary

* add LLM_KV_TRAINING_TYPE to train-text-from-scratch checkpoints

so that they can be differentiated from lora finetune checkpoints

* add gguf constants and load/save functions from train-text-from-scratch

* add load & save lora finetune checkpoints via gguf

* add python script to convert old finetune checkpoint files to gguf

* remove old checkpoint save & load code

* remove code to print data checksums which was used to verify correctness of new gguf code

* omit tokenization when training is disabled, only save llama lora adapter

training can be disabled by passing '-n 0' to finetune

* remove trailing whitespace

* update README.md

* implement ggml_compute_forward_repeat_f16

* avoid stack overflow of large cgraphs in test-grad0

* add ggml API functions ggml_unravel_index, ggml_get_i32_nd and its analogs for set and for f32

ggml_get_i32_1d, ggml_set_i32_1d, ggml_get_f32_1d, ggml_set_f32_1d now support non-contiguous tensors.
in case of non-contiguous tensor, the 1d index is unraveled into a multi index using ggml_unravel_index to be passed to '_nd' function equivalent.

this fixes a bug in test-grad0 which happens due to ggml_build_backward not building purely contiguous tensors anymore

* increase test-grad0 context mem size to accommodate for bigger cgraph

* add sanity check to ggml_compute_backward, asserting the correct shape of gradients

* fix ggml_acc_or_set to return tensor of correct shape

* remove unused 'inplace' argument from ggml_compute_backward function

inplace operations to add gradients are no longer created by ggml_compute_backward
use allocator to automatically make inplace operations

* add missing argument 'int i0' to ggml_get_i32_nd & ggml_set_i32_nd header declarations

* fix error message in ggml_allocr_alloc to display actual max_avail

* fix check_gradient

ggml_build_backward_expand was previously replaced by ggml_build_backward, but the assignment of forward graph to backward graph missing

* use tensor->view_src instead of ggml_is_view and get_view_source

* move gradient checkpointing code into ggml, new API function:

// build gradient checkpointing backward graph gb for gf using provided checkpoints
// gb_tmp will contain original backward graph with rewritten backward process nodes,
// but without the second forward pass nodes.
GGML_API void ggml_build_backward_gradient_checkpointing(
        struct ggml_context   * ctx,
        struct ggml_cgraph    * gf,
        struct ggml_cgraph    * gb,
        struct ggml_cgraph    * gb_tmp,
        struct ggml_tensor  * * checkpoints,
        int                     n_checkpoints);

* replace custom data getters and setters by ggml functions

* train-text-from-scratch can train (full finetune) gguf models

just pass the gguf model via `--checkpoint-in FN`.
after this, to continue training, pass the generated checkpoint instead of the original gguf model.

tested with smaller models, bigger models may exceed available memory.
use (LORA) finetune for those.

* remove trailing whitespace

* add option to save train-text-from-scratch output every N iterations

* update README.md

* fix warnings

* fix warnings

* remove finetune option to disable allocator

the allocator should always be used.
by making sure that it is always used it gets easier to implement automatic memory requirements computation

* add tensor checkpoints only when gradient checkpointing is enabled

* initialize opt ggml context if none was provided

* add ggml-alloc API function 'ggml_allocr_max_size' to get max size of alloc

GGML_API size_t ggml_allocr_max_size(struct ggml_allocr * alloc);

* finetune: automatically allocate all memory and changes to command line options

remove '--n_examples N' parameter, as it no longer makes sense to call optimization process multiple times in a loop.
add '--only_write_lora' command line option: will skip tokenization and training, to only write a llama.cpp comptabile LORA adapter.
remove memory buffer related command line options.
improve iteration console output.

* add finetune to Makefile

* update README.md

* print time per iteration and estimate remaining time

* increase measured alloc size by tensor_alignment

ggml_allocr_reset will reduce the given size by up to tensor_alignment-1

* fix README.md

* add some more allocator debug prints

* bug fix, probably solves the 'ggml_allocr_alloc: not enough space in the buffer' issue

* revert last commit

"bug fix, probably solves the 'ggml_allocr_alloc: not enough space in the buffer' issue"

"alloc was freeing an externally allocated tensor, because it calculated the end of allocator memory as alloc->data + alloc->max_size instead of alloc->data + alloc->size."

This is intentional to reduce the risk of freeing external tensors when measuring. Unless max_size is not properly calculated, I don't see why this is an issue.

* remove unnecessary "0x" before "%p" output

* move measurement memory segment to upper region of the address space

* update README.md

* fix printf format warnings

* add missing gguf_free in load_checkpoint_lora_file

* load default rms_norm and rope parameters from base model

* add gradient accumulation

specify number accumulation steps with '--grad-acc N'.
this will simulate a bigger batch size of grad_acc*batch.

* fix tracking of train_samples and train_tokens

* build : fix compile warnings

* ggml : fix L-BFGS linesearch loop

* improve finetune time measurement

fix printf warnings on system where int64_t is (long int).
change time datatypes to double because values get big with long training times.
exclude file saving from time measurement.
converge faster to actual time per iteration by removing very small first duration before first iteration was performed.
fix bug in output of total training time, the reported value was 1000 times to small.

* specify default lora rank with '--lora-r N'

'--lora-r N' will specify default rank for all tensors
'--rank-wq N', etc. will override this default rank for specific tensor types.

* fix gradient accumulation bug where the same batch was used for each microstep

* fix gradient accumulation bug where the same batch was used for each microstep

* support grouped-query-attention in ggml_flash_attn and ggml_flash_attn_back

k and v can now be repeated in q along ne[2]

in forward pass just use modulo to compute k and v indices, like ik2 = iq2 % nek2.

in backard pass this won't work as easy, because multiple threads will compete to accumulate to the same k->grad[:,ik1,ik2,ik3] and v->grad[:,iv1,iv2,iv3].
so we change the parallelization over q rows to be over k rows. this ensures non-overlapping (ik2,ik3) across threads.
in each thread we then iterate over the number of repetitions of k/v in q to compute iq2 as iq2 = ik2 + irep*nek2.

since ne2 is not the same for q,k and v we also change how the gradients are concatenated into the result tensor.
additionally the offsets of gradq, gradk and gradv in the result tensor are now memory aligned.

we also simplify the compute_backward part of flash_attn to use ggml_reshape instead of switching over the number of dimensions.
this needs a small change to ggml_reshape, removing the assertion of second argument to be contiguous.
since only the shape (ne) of the second reshape argument is of relevance, its memory layout (nb) is irrelevant -> it can very well be non-contiguous.

change test-grad0 to also test for repeated k/v in q.

this changes the rng and now results in small gradient differences in softmax. these solely come from using f16 exp table lookup in forward softmax: when temporarily changing softmax to use actual exp function, the reported gradient differences go away. gradient differences coming solely from f16 table lookup are acceptable.
added a note to explain this.

* add llama API functions to get grouped-query-attention n_head parameter 'n_head_kv'.

* fix finetune to support grouped-query-attention (using flash-attention)

note: ggml changes to ggml_out_prod are necessary to support grouped-query-attention without flash-attention.

* support broadcastable a in out_prod(a, b) and backward pass of broadcasting mul_mat(a, b)

* test broadcasting mul_mat backward pass

* decouple random number generator of each operation test

when changing one test the rng of others tests is not influenced anymore

* add comment briefly describing what ggml_repeat_back does

* simplify broadcasting mul_mat backward using ggml_repeat_back

* add cgraph evaluation order member and corresponding enum type

this controls in which order ggml_build_forward visits source nodes.
by default the nodes are visited left to right, i.e. src[0] first.
in some cases it is beneficial for ggml-alloc to visit in a different order.
two possible orders are supported: left-to-right (src[0] first) and right-to-left (src[0] last).

* measure max compute size for each cgraph eval order and use best order

this can bring huge memory savings:
e.g. codellama-34b with n_ctx=64, n_batch=1 goes from 92927.8mb down to 4627.6 MB

* remove unused command line options

* add sample start patterns and options to force new or by default resume last shuffling

* update shuffle rng state on reshuffle

* exclude known zero values from computations in flash_attn_f32 & flash_attn_back_f32

* remove probably unnecessary exception type flags from stringstream

* pass correct max number of tokens to llama_tokenize

* account for possible leading whitespace that will be added by tokenizer
e.g. '\t' will be tokenized by llama spm tokenizer to [29871, 12]

* use unrolled vec_mad in out_prod

y is vec_mad result vec.
x is vec_mad input vec.
v is vec_mad input scalar.

ggml_vec_mad_f32_unroll will internally loop over x and v with same y.

GGML_VEC_MAD_UNROLL is by default defined to 32.

This value is empirical optimized using performance test runs of out-prod in openllama-3b finetune with 256 context length and batch size 1. It gives 23% performance boost for out_prod.

Full measurements of out-prod runtime in ms:
	unroll_xv	unroll_yv
1	67014.643	87826.469
2	77117.552	89077.656
4	72091.311	109121.657
8	61077.543	88678.334
16	56914.67	79514.947
24	59024.595	84350.254
28	55952.446	83368.73
32	51476.658	85177.745
36	55973.792	84659.92
40	55139.616	93844.738
48	60736.392	93330.267
64	99856.878	116994.99

Second column is when unrollying yv instead of xv

* set lora_alpha to value of lora_r if it is not set via command line

otherwise only changing lora_r will change scaling of lora adapter used in prediction

* reshuffle original sample order instead of the previous shuffled order

otherwise resumed reshuffle will not result in same sample order

* block tiling for out-prod inspired by mul-mat

block sizes are empirically optimized

roughly doubles the flops of out-prod

* exclude some more known zero values from computations in flash_attn_f32 & flash_attn_back_f32

* add static keywords

* remove outcommented old code

* update train-text-from-scratch with tokenization, sample selection and shuffling from finetune

* remove lbfgs related train parameters

* move common train functions into common/train.[h|cpp]

* move train state into struct train_state

* move train data saving code into callback to unify code of opt_callback

train_params are still different in finetune and train-text-from-scratch, so it can't yet be moved to train.h|cpp

* move common train params into common/train

* move common opt_callback into common/train

* fix consume_common_train_arg

* save and load head_count_kv in lora checkpoints

* increase train_samples by used_samples instead of number of batches

on batch can contain more than one sample when option "fill_with_next_samples" is used

* fix usage of llama_tokenize

* remove static from process_escape since we need it exposed in header

* fix code formating of long function declarations

* fix condition in load_train_state_gguf

* use die("msg") instead of replace GGML_ASSERT(!"msg") or throw std::runtime_error("msg")

* fix saving and loading of training type

* remove terminating '\0' from tokenization

(llama_tokenize is now passed the string length instead of relying on terminating '\0')

* fix compile warnings

* fix compile warnings

* use new/delete for train_state instead of malloc/free

using malloc may result in seg faults when trying to assign string fields

* assert that sample_count > 0, avoiding division by zero

* fix frand to return value in interval [0,1)

* add train option "--sample-random-offsets"

Use samples beginning at random offsets.
The offset is only applied to the first sample in each batch context window.
Together with "--fill-with-next-samples" this may help for training endless text generation.

For example given a dataset containing samples "abcd", "ABCD", "0123".
With context size of 8 and options "--fill-with-next-samples", "--no-separate-with-eos", "--no-separate-with-bos",
the context windows of batches could only be filled with "abcdABCD", "ABCDabcd", "0123abcd", etc.

With "--sample-random-offsets" it can also be filled with "23abcdAB", "bcd0123A", etc.

* deduplicate code into function

* remove n_rot hparam, as it must always be hparam.n_embd_head()

* align code

* assert correct base model tensor shapes

* move some params from lora hparams into model hparams and load model params from gguf

this equalizes the model definition in finetune and text-from-scratch and removes the need for additional llama api functions to get model parameters

* remove now unnecessary llama API functions to get model params that where added by this PR

* train-text-from-scratch: automatically allocate model tensors, remove option '--mem-model N'

* train-text-from-scratch: automatically allocate opt context

* train-text-from-scratch: automatically allocate input tensors

* train-text-from-scratch: automatically allocate compute memory

* remove unused options and equalize train-text-from-scratch with finetune

* initialize opt->loss_after with zero

* add export-lora program

* remove trailing whitespace

* add export-lora build in Makefile

* remove unused struct tensor_info from export-lora

* add export-lora build dependency to llama

because it depends on common, which depends on llama

* update finetune README.md

* cancel optimization when specified number of epochs is completed

* improve handling of export-lora arguments

print errors and warnings when files could not be read or created

* Fix export-lora.cpp "not enough space in the context's memory pool" (#1)

* Fix export-lora.cpp "not enough space in the context's memory pool"

Without this patch, export-lora would sometimes error with "not enough space in the context's memory pool (needed 656784, available 656800)".

* increase required context size by 5*GGML_MEM_ALIGN instead of plain 16

---------

Co-authored-by: xaedes <xaedes@gmail.com>

* improve handling of not yet supported tensor types

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Co-authored-by: meatbag-18a <145869052+meatbag-18a@users.noreply.github.com>
2023-09-28 21:40:11 +03:00
Georgi Gerganov
ec893798b7
llama : custom attention mask + parallel decoding + no context swaps (#3228)
* tests : verify that RoPE is "additive"

* llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask)

* ggml : ggml_rope now takes a vector with positions instead of n_past

* metal : add rope_f16 kernel + optimize cpy kernels

* llama : unified KV cache + batch inference API

* llama : add new llama_decode() API that works with llama_batch

* llama : add cell_max heuristic for more efficient kv_cache

* llama : extend llama_kv_cache API

* llama : more robust cell_max heuristic + wip shift

* metal : disable concurrency optimization

* llama : add llama_kv_cache_shift_seq + no more context swaps

* llama : apply K-cache roping for Falcon and Baichuan

* speculative : fix KV cache management

* parallel : example for serving multiple users in parallel

* parallel : disable hot-plug to avoid cache fragmentation

* fixes : speculative KV cache + llama worst-case graph

* llama : extend batch API to select which logits to output

* llama : fix worst case graph build

* ggml-cuda : update rope implementation for parallel decoding (#3254)

* ggml-cuda : update rope implementation for parallel decoding

* better solution for p0 computation

* fix rope

* simpler rope implementation

---------

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

* make : add parallel to build + fix static functions in llama.cpp

* simple : fix token counting

* parallel : various improvements

* llama : fix cell_max logic + rename functions

* parallel : try smaller batches when the KV cache is fragmented

* parallel : fix sequence termination criteria

* llama : silence errors KV cache errors

* parallel : remove new line from prompt

* parallel : process system prompt once + configurable paramters + llama API

* parallel : remove question with short answers

* parallel : count cache misses

* parallel : print misses on each request

* parallel : minor

* llama : fix n_kv to never become 0

* parallel : rename hot-plug to continuous-batching

* llama : improve llama_batch API + simplify parallel example

* simple : add parallel decoding support

* simple : improve comments + free batch

* ggml-cuda : add rope f16, restore performance with parallel decoding (#3272)

* ggml-cuda : add rope f16, restore performance

* offload KQ_mask with all models

* fix rope shift

---------

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

* llama : disable MPI for now

ggml-ci

* train : make KQ_pos memory buffer permanent via dummy scale op

* ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275)

ggml-ci

* parallel : fix bug (extra BOS) + smaller token_prev array

* parallel : fix cases where the input prompts can overflow the batch

* parallel : add disabled experimental batch chunking in powers of two

* llama : llama.h formatting + comments

* simple : add README.md

* llama : fix kv cache heuristic when context is less than 32

* parallel : fix crash when `-n -1`

* llama : simplify returns if/else branches

* metal : use mm kernels for batch size > 2

* examples : utilize new llama_get_logits_ith()

* examples : add example for batched decoding

* examples : do not eval prompt 2 times (close #3348)

* server : clear the KV cache beyond n_past before llama_decode

* server : avoid context swaps by shifting the KV cache

---------

Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 19:04:36 +03:00
Georgi Gerganov
47068e5170
speculative : PoC for speeding-up inference via speculative sampling (#2926)
* speculative : initial example

* speculative : print encoding speed

* speculative : add --draft CLI arg
2023-09-03 15:12:08 +03:00
Georgi Gerganov
c90d135eb4
examples : fix underscore in beam-search + .gitignore (close #2900) 2023-08-30 12:53:24 +03:00
Matt Pulver
c82742ac9c
llama : add llama_beam_search() (#2267)
* Add llama_beam_search().

* Add '// Beam search' heading to llama.{h,cpp} after llama_grammar_accept_token().

* Add space around * pointers and & references.

* Add spaces around comparison and assignment operators.

* Prefer west const.

* Use llama_ prefix for structs in global namespace.

* Delete obsolete comment from an earlier revision.

* Change eos to eob in llama_beam and llama_beam_view structs.
2023-08-25 18:18:48 +03:00
Georgi Gerganov
6381d4e110
gguf : new file format with flexible meta data (beta) (#2398)
* gguf : first API pass

* gguf : read header + meta data

* gguf : read tensor info

* gguf : initial model loading - not tested

* gguf : add gguf_get_tensor_name()

* gguf : do not support passing existing ggml_context to gguf_init

* gguf : simplify gguf_get_val

* gguf : gguf.c is now part of ggml.c

* gguf : read / write sample models

* gguf : add comments

* refactor : reduce code duplication and better API (#2415)

* gguf : expose the gguf_type enum through the API for now

* gguf : add array support

* gguf.py : some code style changes

* convert.py : start a new simplified implementation by removing old stuff

* convert.py : remove GGML vocab + other obsolete stuff

* GGUF : write tensor (#2426)

* WIP: Write tensor

* GGUF : Support writing tensors in Python

* refactor : rm unused import and upd todos

* fix : fix errors upd writing example

* rm example.gguf

* gitignore *.gguf

* undo formatting

* gguf : add gguf_find_key (#2438)

* gguf.cpp : find key example

* ggml.h : add gguf_find_key

* ggml.c : add gguf_find_key

* gguf : fix writing tensors

* gguf : do not hardcode tensor names to read

* gguf : write sample tensors to read

* gguf : add tokenization constants

* quick and dirty conversion example

* gguf : fix writing gguf arrays

* gguf : write tensors one by one and code reuse

* gguf : fix writing gguf arrays

* gguf : write tensors one by one

* gguf : write tensors one by one

* gguf : write tokenizer data

* gguf : upd gguf conversion script

* Update convert-llama-h5-to-gguf.py

* gguf : handle already encoded string

* ggml.h : get array str and f32

* ggml.c : get arr str and f32

* gguf.py : support any type

* Update convert-llama-h5-to-gguf.py

* gguf : fix set is not subscriptable

* gguf : update convert-llama-h5-to-gguf.py

* constants.py : add layer norm eps

* gguf.py : add layer norm eps and merges

* ggml.h : increase GGML_MAX_NAME to 64

* ggml.c : add gguf_get_arr_n

* Update convert-llama-h5-to-gguf.py

* add gptneox gguf example

* Makefile : add gptneox gguf example

* Update convert-llama-h5-to-gguf.py

* add gptneox gguf example

* Update convert-llama-h5-to-gguf.py

* Update convert-gptneox-h5-to-gguf.py

* Update convert-gptneox-h5-to-gguf.py

* Update convert-llama-h5-to-gguf.py

* gguf : support custom alignment value

* gguf : fix typo in function call

* gguf : mmap tensor data example

* fix : update convert-llama-h5-to-gguf.py

* Update convert-llama-h5-to-gguf.py

* convert-gptneox-h5-to-gguf.py : Special tokens

* gptneox-main.cpp : special tokens

* Update gptneox-main.cpp

* constants.py : special tokens

* gguf.py : accumulate kv and tensor info data + special tokens

* convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens

* gguf : gguf counterpart of llama-util.h

* gguf-util.h : update note

* convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens

* convert-llama-h5-to-gguf.py : special tokens

* Delete gptneox-common.cpp

* Delete gptneox-common.h

* convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer

* gptneox-main.cpp : gpt2 bpe tokenizer

* gpt2 bpe tokenizer (handles merges and unicode)

* Makefile : remove gptneox-common

* gguf.py : bytesarray for gpt2bpe tokenizer

* cmpnct_gpt2bpe.hpp : comments

* gguf.py : use custom alignment if present

* gguf : minor stuff

* Update gptneox-main.cpp

* map tensor names

* convert-gptneox-h5-to-gguf.py : map tensor names

* convert-llama-h5-to-gguf.py : map tensor names

* gptneox-main.cpp : map tensor names

* gguf : start implementing libllama in GGUF (WIP)

* gguf : start implementing libllama in GGUF (WIP)

* rm binary commited by mistake

* upd .gitignore

* gguf : calculate n_mult

* gguf :  inference with 7B model working (WIP)

* gguf : rm deprecated function

* gguf : start implementing gguf_file_saver (WIP)

* gguf : start implementing gguf_file_saver (WIP)

* gguf : start implementing gguf_file_saver (WIP)

* gguf : add gguf_get_kv_type

* gguf : add gguf_get_kv_type

* gguf : write metadata in gguf_file_saver (WIP)

* gguf : write metadata in gguf_file_saver (WIP)

* gguf : write metadata in gguf_file_saver

* gguf : rm references to old file formats

* gguf : shorter name for member variable

* gguf : rm redundant method

* gguf : get rid of n_mult, read n_ff from file

* Update gguf_tensor_map.py

* Update gptneox-main.cpp

* gguf : rm references to old file magics

* gguf : start implementing quantization (WIP)

* gguf : start implementing quantization (WIP)

* gguf : start implementing quantization (WIP)

* gguf : start implementing quantization (WIP)

* gguf : start implementing quantization (WIP)

* gguf : start implementing quantization (WIP)

* gguf : quantization is working

* gguf : roper closing of file

* gguf.py : no need to convert tensors twice

* convert-gptneox-h5-to-gguf.py : no need to convert tensors twice

* convert-llama-h5-to-gguf.py : no need to convert tensors twice

* convert-gptneox-h5-to-gguf.py : simplify nbytes

* convert-llama-h5-to-gguf.py : simplify nbytes

* gptneox-main.cpp : n_layer --> n_block

* constants.py : n_layer --> n_block

* gguf.py : n_layer --> n_block

* convert-gptneox-h5-to-gguf.py : n_layer --> n_block

* convert-llama-h5-to-gguf.py : n_layer --> n_block

* gptneox-main.cpp : n_layer --> n_block

* Update gguf_tensor_map.py

* convert-gptneox-h5-to-gguf.py : load model in parts to save memory

* convert-llama-h5-to-gguf.py : load model in parts to save memory

* convert : write more metadata for LLaMA

* convert : rm quantization version

* convert-gptneox-h5-to-gguf.py : add file_type key

* gptneox-main.cpp : add file_type key

* fix conflicts

* gguf : add todos and comments

* convert-gptneox-h5-to-gguf.py : tensor name map changes

* Create gguf_namemap.py : tensor name map changes

* Delete gguf_tensor_map.py

* gptneox-main.cpp : tensor name map changes

* convert-llama-h5-to-gguf.py : fixes

* gguf.py : dont add empty strings

* simple : minor style changes

* gguf : use UNIX line ending

* Create convert-llama-7b-pth-to-gguf.py

* llama : sync gguf-llama.cpp with latest llama.cpp (#2608)

* llama : sync gguf-llama.cpp with latest llama.cpp

* minor : indentation + assert

* llama : refactor gguf_buffer and gguf_ctx_buffer

* llama : minor

* gitignore : add gptneox-main

* llama : tokenizer fixes (#2549)

* Merge tokenizer fixes into the gguf branch.

* Add test vocabularies

* convert : update convert-new.py with tokenizer fixes (#2614)

* Merge tokenizer fixes into the gguf branch.

* Add test vocabularies

* Adapt convert-new.py (and fix a clang-cl compiler error on windows)

* llama : sync gguf-llama with llama (#2613)

* llama : sync gguf-llama with llama

* tests : fix build + warnings (test-tokenizer-1 still fails)

* tests : fix wstring_convert

* convert : fix layer names

* llama : sync gguf-llama.cpp

* convert : update HF converter to new tokenizer voodoo magics

* llama : update tokenizer style

* convert-llama-h5-to-gguf.py : add token types

* constants.py : add token types

* gguf.py : add token types

* convert-llama-7b-pth-to-gguf.py : add token types

* gguf-llama.cpp :  fix n_head_kv

* convert-llama-h5-to-gguf.py : add 70b gqa support

* gguf.py : add tensor data layout

* convert-llama-h5-to-gguf.py : add tensor data layout

* convert-llama-7b-pth-to-gguf.py : add tensor data layout

* gptneox-main.cpp : add tensor data layout

* convert-llama-h5-to-gguf.py : clarify the reverse permute

* llama : refactor model loading code (#2620)

* llama : style formatting + remove helper methods

* llama : fix quantization using gguf tool

* llama : simplify gguf_file_saver

* llama : fix method names

* llama : simplify write_header()

* llama : no need to pass full file loader to the file saver

just gguf_ctx

* llama : gguf_file_saver write I32

* llama : refactor tensor names (#2622)

* gguf: update tensor names searched in quantization

* gguf : define tensor names as constants

* gguf : initial write API (not tested yet)

* gguf : write to file API (not tested)

* gguf : initial write API ready + example

* gguf : fix header write

* gguf : fixes + simplify example + add ggml_nbytes_pad()

* gguf : minor

* llama : replace gguf_file_saver with new gguf write API

* gguf : streaming support when writing files

* gguf : remove oboslete write methods

* gguf : remove obosolete gguf_get_arr_xxx API

* llama : simplify gguf_file_loader

* llama : move hparams and vocab from gguf_file_loader to llama_model_loader

* llama : merge gguf-util.h in llama.cpp

* llama : reorder definitions in .cpp to match .h

* llama : minor simplifications

* llama : refactor llama_model_loader (WIP)

wip : remove ggml_ctx from llama_model_loader

wip : merge gguf_file_loader in llama_model_loader

* llama : fix shape prints

* llama : fix Windows build + fix norm_rms_eps key

* llama : throw error on missing KV paris in model meta data

* llama : improve printing + log meta data

* llama : switch print order of meta data

---------

Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com>

* gguf : deduplicate (#2629)

* gguf : better type names

* dedup : CPU + Metal is working

* ggml : fix warnings about unused results

* llama.cpp : fix line feed and compiler warning

* llama : fix strncpy warning + note token_to_str does not write null

* llama : restore the original load/save session implementation

Will migrate this to GGUF in the future

* convert-llama-h5-to-gguf.py : support alt ctx param name

* ggml : assert when using ggml_mul with non-F32 src1

* examples : dedup simple

---------

Co-authored-by: klosax <131523366+klosax@users.noreply.github.com>

* gguf.py : merge all files in gguf.py

* convert-new.py : pick #2427 for HF 70B support

* examples/gguf : no need to keep q option for quantization any more

* llama.cpp : print actual model size

* llama.cpp : use ggml_elements()

* convert-new.py : output gguf (#2635)

* convert-new.py : output gguf (WIP)

* convert-new.py : add gguf key-value pairs

* llama : add hparams.ctx_train + no longer print ftype

* convert-new.py : minor fixes

* convert-new.py : vocab-only option should work now

* llama : fix tokenizer to use llama_char_to_byte

* tests : add new ggml-vocab-llama.gguf

* convert-new.py : tensor name mapping

* convert-new.py : add map for skipping tensor serialization

* convert-new.py : convert script now works

* gguf.py : pick some of the refactoring from #2644

* convert-new.py : minor fixes

* convert.py : update to support GGUF output

* Revert "ci : disable CI temporary to not waste energy"

This reverts commit 7e82d25f40.

* convert.py : n_head_kv optional and .gguf file extension

* convert.py : better always have n_head_kv and default it to n_head

* llama : sync with recent PRs on master

* editorconfig : ignore models folder

ggml-ci

* ci : update ".bin" to ".gguf" extension

ggml-ci

* llama : fix llama_model_loader memory leak

* gptneox : move as a WIP example

* llama : fix lambda capture

ggml-ci

* ggml : fix bug in gguf_set_kv

ggml-ci

* common.h : .bin --> .gguf

* quantize-stats.cpp : .bin --> .gguf

* convert.py : fix HF tensor permuting / unpacking

ggml-ci

* llama.cpp : typo

* llama : throw error if gguf fails to init from file

ggml-ci

* llama : fix tensor name grepping during quantization

ggml-ci

* gguf.py : write tensors in a single pass (#2644)

* gguf : single pass for writing tensors + refactoring writer

* gguf : single pass for writing tensors + refactoring writer

* gguf : single pass for writing tensors + refactoring writer

* gguf : style fixes in simple conversion script

* gguf : refactor gptneox conversion script

* gguf : rename h5 to hf (for HuggingFace)

* gguf : refactor pth to gguf conversion script

* gguf : rm file_type key and method

* gguf.py : fix vertical alignment

* gguf.py : indentation

---------

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

* convert-gptneox-hf-to-gguf.py : fixes

* gguf.py : gptneox mapping

* convert-llama-hf-to-gguf.py : fixes

* convert-llama-7b-pth-to-gguf.py : fixes

* ggml.h : reverse GGUF_MAGIC

* gguf.py : reverse GGUF_MAGIC

* test-tokenizer-0.cpp : fix warning

* llama.cpp : print kv general.name

* llama.cpp : get special token kv and linefeed token id

* llama : print number of tensors per type + print arch + style

* tests : update vocab file with new magic

* editorconfig : fix whitespaces

* llama : re-order functions

* llama : remove C++ API + reorganize common source in /common dir

* llama : minor API updates

* llama : avoid hardcoded special tokens

* llama : fix MPI build

ggml-ci

* llama : introduce enum llama_vocab_type + remove hardcoded string constants

* convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested

* falcon-main.cpp : falcon inference example

* convert-falcon-hf-to-gguf.py : remove extra kv

* convert-gptneox-hf-to-gguf.py : remove extra kv

* convert-llama-7b-pth-to-gguf.py : remove extra kv

* convert-llama-hf-to-gguf.py : remove extra kv

* gguf.py : fix for falcon 40b

* falcon-main.cpp : fix for falcon 40b

* convert-falcon-hf-to-gguf.py : update ref

* convert-falcon-hf-to-gguf.py : add tensor data layout

* cmpnct_gpt2bpe.hpp : fixes

* falcon-main.cpp : fixes

* gptneox-main.cpp : fixes

* cmpnct_gpt2bpe.hpp : remove non-general stuff

* Update examples/server/README.md

Co-authored-by: slaren <slarengh@gmail.com>

* cmpnct_gpt2bpe.hpp : cleanup

* convert-llama-hf-to-gguf.py : special tokens

* convert-llama-7b-pth-to-gguf.py : special tokens

* convert-permute-debug.py : permute debug print

* convert-permute-debug-master.py : permute debug for master

* convert-permute-debug.py : change permute type of attn_q

* convert.py : 70b model working (change attn_q permute)

* Delete convert-permute-debug-master.py

* Delete convert-permute-debug.py

* convert-llama-hf-to-gguf.py : fix attn_q permute

* gguf.py : fix rope scale kv

* convert-llama-hf-to-gguf.py : rope scale and added tokens

* convert-llama-7b-pth-to-gguf.py : rope scale and added tokens

* llama.cpp : use rope scale kv

* convert-llama-7b-pth-to-gguf.py : rope scale fix

* convert-llama-hf-to-gguf.py : rope scale fix

* py : fix whitespace

* gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682)

* First pass at converting GGMLv3 LLaMA models to GGUF

* Cleanups, better output during conversion

* Fix vocab space conversion logic

* More vocab conversion fixes

* Add description to converted GGUF files

* Improve help text, expand warning

* Allow specifying name and description for output GGUF

* Allow overriding vocab and hyperparams from original model metadata

* Use correct params override var name

* Fix wrong type size for Q8_K

Better handling of original style metadata

* Set default value for gguf add_tensor raw_shape KW arg

* llama : improve token type support (#2668)

* Merge tokenizer fixes into the gguf branch.

* Add test vocabularies

* Adapt convert-new.py (and fix a clang-cl compiler error on windows)

* Improved tokenizer test

But does it work on MacOS?

* Improve token type support

- Added @klosax code to convert.py
- Improved token type support in vocabulary

* Exclude platform dependent tests

* More sentencepiece compatibility by eliminating magic numbers

* Restored accidentally removed comment

* llama : add API for token type

ggml-ci

* tests : use new tokenizer type API (#2692)

* Merge tokenizer fixes into the gguf branch.

* Add test vocabularies

* Adapt convert-new.py (and fix a clang-cl compiler error on windows)

* Improved tokenizer test

But does it work on MacOS?

* Improve token type support

- Added @klosax code to convert.py
- Improved token type support in vocabulary

* Exclude platform dependent tests

* More sentencepiece compatibility by eliminating magic numbers

* Restored accidentally removed comment

* Improve commentary

* Use token type API in test-tokenizer-1.cpp

* py : cosmetics

* readme : add notice about new file format

ggml-ci

---------

Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com>
Co-authored-by: klosax <131523366+klosax@users.noreply.github.com>
Co-authored-by: goerch <jhr.walter@t-online.de>
Co-authored-by: slaren <slarengh@gmail.com>
Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 23:07:43 +03:00
slaren
097e121e2f
llama : add benchmark example (#2626)
* llama : add benchmark example

* add to examples CMakeLists.txt

* fix msvc build

* add missing include

* add Bessel's correction to stdev calculation

Co-authored-by: Johannes Gäßler <johannesg@5d6.de>

* improve markdown formatting

* add missing include

* print warning is NDEBUG is not defined

* remove n_prompt and n_gen from the matrix, use each value separately instead

* better checks for non-optimized builds

* llama.cpp : fix MEM_REQ_SCRATCH0 reusing the value of n_ctx of the first call

* fix json formatting

* add sql output

* add basic cpu and gpu info (linx/cuda only)

* markdown: also show values that differ from the default

* markdown: add build id

* cleanup

* improve formatting

* formatting

---------

Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2023-08-18 12:44:58 +02:00
byte-6174
b19edd54d5
Adding support for llama2.c models (#2559) 2023-08-12 01:17:25 +02:00
DannyDaemonic
3498588e0f
Add --simple-io option for subprocesses and break out console.h and cpp (#1558) 2023-08-04 08:20:12 -07:00
Evan Jones
84e09a7d8b
llama : add grammar-based sampling (#1773)
* llama, main : constrain sampling to grammar

* allow loading grammar from file

* fix whitespace errors

* handle & print parser errors

* add comments to grammar syntax and allow newlines where unambiguous

* add missing include

* support alternates in root rule

* fix bugs with empty token and EOS

* adjust JSON grammar

* remove swp file

* rewrite ternary expressions

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

* use struct for grammar elements and add Unicode support

* add unicode escapes

* add inverse char ranges

* only sample full tokens (no peeking or truncation)

* llama : minor style changes

blindly applied in online editor - hopefully I didn't break something

* update help text

* add warning message if EOS is disabled

---------

Co-authored-by: Henri Vasserman <henv@hot.ee>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-23 23:58:10 -04:00
ningshanwutuobang
cfa0750bc9
llama : support input embeddings directly (#1910)
* add interface for float input

* fixed inpL shape and type

* add examples of input floats

* add test example for embd input

* fixed sampling

* add free for context

* fixed add end condition for generating

* add examples for llava.py

* add READMD for llava.py

* add READMD for llava.py

* add example of PandaGPT

* refactor the interface and fixed the styles

* add cmake build for embd-input

* add cmake build for embd-input

* Add MiniGPT-4 example

* change the order of the args of llama_eval_internal

* fix ci error
2023-06-28 18:53:37 +03:00
Georgi Gerganov
051e1b0e6a
llama : fix kv_cache n init (close #1903) 2023-06-17 19:31:20 +03:00
xaedes
e32089b2c2
train : improved training-from-scratch example (#1652)
* add python wrapper

https://gist.github.com/abetlen/2b90e5f153f6efd00931d098de5c73ce

* fix decoding error. adds errors=ignore parameter

* add python bindings for functions to get and set the whole llama state
(rng, logits, embedding and kv_cache)

* update python bindings

* add text generating baby-llama from scratch example

* fix race condition bug in ggml_compute_forward_diag_mask_f32

* implement ggml_soft_max_back for more performant backward pass of soft_max

avoids creating big intermediate matrices of size n_embd x n_embd for llama layers and n_vocab x n_vocab for cross entropy loss

* improve softmax backward pass

go from quadratic runtime to linear runtime by simplifying the formulas

* fix race condition bug in non-inplace ggml_compute_forward_diag_mask_f32

memcpy needs to be synchronized across threads to avoid race conditions.
=> do it in INIT phase

* fix bug in ggml_compute_forward_soft_max_back_f32 on DEBUG build

* improve performance of mul_mat backward pass

avoid transpose by using mul_mat with swapped arguments

* avoid printing too much newlines in baby-llama-text

* activate threading in baby-llama-text

* add ggml_out_prod and use it for mul_mat backward pass for improved performance

performance stats report improvement from 37 seconds to 16 seconds runtime during my training tests

* better weight initialization improves training convergence at start

* better weight initialization improves training convergence at start

* improve ggml_out_prod performance

- change iteration order (>15s -> 10s runtime)
- parallelize over one more dimension: over dst matrix rows (10s -> <5s runtime)

* add llama sampler, shuffle samples and constrain sampling to tokens occurring in train data

* fix get_samples call, add model tensor names, increase model size, start training samples after newline

* save train trained model to checkpoint and load model to be trained from checkpoint

* use inplace functions where possible

* initialize rng with srand

* use different arguments for input and output checkpoint

* ggml fixes to support backward pass on inplace operations

* remove duplicate include

* fix cross entropy loss

- add target probabilities for each sample which is then used in cross entropy loss

* print used memory before and after optimization

* sample with non-greedy sampling parameters at the end of training

* add cmake target for baby-llama-text

* add ggml_add1_inplace to header

* enable gradient propagation for inplace add1 and scale operations

those functions backward passes don't need the original src0, so they also work when forward is inplace

* implement AdamW in ggml_opt_adam by adding weight decay parameter (default 0.001f)

also add a schedule parameter (default 1.0f) that can be used to scale alpha and decay according to learning schedule.
setting the decay parameter to zero disables AdamW resulting in normal Adam optimizer.

since the difference between Adam and AdamW is minimal it is not implemented as another optimizer, but integrated into the existing Adam optimizer.

* use inplace operations in cross_entropy_loss

* fix random weight initialization scale

* add missing default parameters for adam optimizer

* add ggml_opt_context, so that we can properly resume training

otherwise the optimizer states, tracking statistics about the error function and its derivates,
will reset to zero each time ggml_opt is called, hindering convergence on resumed training.

now the optimizer context and all its memory is stored in a separate struct.

* fix bug in llama_sample_token_mirostat_v2

when all candidates are filtered out through mu threshold, the following soft_max operation will fail.
so keep at least one.

* add forward function without using cache, for more performant training

during training on whole samples no cache is required.
removing the cache and simplifying the remaining code results in performance and memory usage improvement.

* print suppressed newline tokens as string "\n"

printing too much actual newlines is suppressed to avoid flooding the console.

* store optimizer state in training checkpoint and add learning schedule

persistent optimizer state allows to resume training without resetting the optimizer
learning schedule consists of linear warmup ramp followed by cosine decay with restarts

* remove unused functions

* fix bug in get_samples which corrupted training targets

* save checkpoint only when it was trained

* simplify code

* remove trailing whitespace

* simplify backward pass for SQRT

* replace inefficient repeat backward pass with dedicated repeat_back operation

* add ggml_cross_entropy_loss with backward pass for faster training

cross entropy loss can also be implemented using softmax and log, but as dedicated operation it is faster and especially avoids unnecessary memory overhead.

* add tests for cross_entropy_loss backward pass

finite differences regularly results in estimated gradient of zero, despite the backward pass giving non zero gradient.
_probably_ the finite differences fails due to numerical issues

* use ggml_cross_entropy_loss in text training example

* remove trailing whitespace

* slightly improve how cross entropy loss is compute

btw: directly implemented cross entropy loss seems to have way lower magnitudes than when implemented with softmax and log.
probably the input to log gets closer to zero due to float numerics.
maybe the multiplication by (1.0-eps)/sum is more accurate..

* add llama_get_vocab to get the vocabulary as output parameters

* set default model.type for unknown models with few layers

* add export of training checkpoint to llama compatible model file

* get vocabulary for exporting training checkpoint to llama compatible model file

* implement backward pass of flash attention

* bugfixes for backward pass of flash attention

* test flash attention backward pass

need to set loose error bounds to pass.
the finitie differences are close to numeric limits and often return quite different values than the backward pass.
reducing eps further lets the gradients vanish completely.
likewise setting eps to big results in wronger values.
the softmax in the middle of the function is probably the most responsible for the numeric issues using finite differences.

* add option to train with flash attention and move options to the top of the main function

training from scratch also works with flash attention
training convergence and generation results after fix number of iterations are worse than when not using flash attention.
maybe there still lingers a bug in the flash attention backward pass?
but training works, just with slower convergence.

flash attention is still worth to use, because it requires way less memory and is faster with high n_ctx

* add train_params and command line option parser

* remove unnecessary comments

* add train params to specify memory size

* remove python bindings

* rename baby-llama-text to train-text-from-scratch

* replace auto parameters in lambda function

* add #include <climits>

* add explicit cast to fix compile error

"error: non-constant-expression cannot be narrowed from type 'int64_t' (aka 'long long') to 'uint32_t' (aka 'unsigned int') in initializer list [-Wc++11-narrowing]"

* remove trailing whitespace

* add ggml_opt_resume_g which accepts forward and backward cgraphs

* fix formulas in comments

* bug fix for ggml_compute_forward_get_rows_back_f32

the result should be set to zero, not to whatever data is in opt0

* improve training memory usage with scratch buffers

instead of relying on the automatic backward pass, we manually create the graph for the backward pass.
it turns out that all backward pass operations need only temporary memory which can be reused after each layer.

will compute backward pass for ALL model parameters

* add option to use scratch buffers in training or not

make it configurable because currently training with scratch buffers implies flash attention and optimization over all parameters.

* ci : disable temporary

* store view offset and permute axes in opt[0] instead of storing it in padding

use memcpy to store offset, because offset is of type size_t.
when storing it as int32_t offset would have to be smaller than 2^31 which is not necessarily true.

* minor : fix compile warnings + minor style changes

* fix bug in threaded indices calculation of ggml_compute_forward_flash_attn_back_f32

* store view offset like in master branch

* bug fix in forward_batch_wo_cache_flash_attn_train

* scratch buffer bug fixes in forward_batch_wo_cache_flash_attn_train

data of permute and reshape is the same as their input.
if we want to preserve the output of permute/reshape, we also need to preserve their inputs.

replace reshape(src0, src1) with reshape_nd calls so that we don't need src1.

replace (temporary) t03 with ggml_repeat(ctx0, layer.attention_norm, t02).
in the future we could also use the new broadcasting ggml_mul to avoid these repeat calls.
for this we need backward pass of broadcasting ggml_mul.

* remove unnecessary scratch buffer 0

buf 0 is persistent memory, so we can just disable scratch for this by using buf -1

* avoid creating unnecessary grad tensors

previously we need to create grads for model parameters, so that expand(..) correctly populates cgraph->leafs & cgraph->grads
this wasted memory, because unnecessary grad for each op were automatically created:
the automatically generated grad was unnecessary because we later manually set the grad (e.g. t35->grad = expand(gb, ...) ).
this discarded the automatically generated grad resulting in wasted memory.

improved this by changing expand(..) to not use ggml_build_forward_expand.
expand set cgraph->nodes but not the leafs.
cgraph->leafs & cgraph->grads are set in another pass after the last expand call.

* print used training seed

* zero initialize gfbuf and gbbuf

* ci : re-enable workflows + add README for training

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-13 22:04:40 +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 98c267fc77.

* 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
Steward Garcia
7e4ea5beff
examples : add server example with REST API (#1443)
* Added httplib support

* Added readme for server example

* fixed some bugs

* Fix the build error on Macbook

* changed json11 to nlohmann-json

* removed some whitespaces

* remove trailing whitespace

* added support custom prompts and more functions

* some corrections and added as cmake option
2023-05-21 20:51:18 +03:00
xaedes
f954edda93
ggml : implement backward pass for llama + small training-llama-from-scratch example (#1360)
* implement 8 of 14 missing backward pass operations used by llama

- GGML_OP_ADD_AT
- GGML_OP_CPY
- GGML_OP_MUL_MAT (src0.grad)
- GGML_OP_PERMUTE
- GGML_OP_RESHAPE
- GGML_OP_SCALE
- GGML_OP_TRANSPOSE
- GGML_OP_VIEW

implement additional ggml operation GGML_OP_ADD_AT, which is necessary for backward pass of GGML_OP_VIEW.

this operation adds src1 to src0 with data offset, i.e. to view(src0, ..., offset).
the values are return in a tensor size of src0. values outside of [data+offset:data+offset+nbytes(src1)] are just the original values from src0.

still missing backward passes for llama:

- GGML_OP_DIAG_MASK_INF
- GGML_OP_GET_ROWS
- GGML_OP_RMS_NORM
- GGML_OP_ROPE
- GGML_OP_SILU
- GGML_OP_SOFT_MAX

* implement 5 of 6 missing backward pass operations used by llama

- GGML_OP_DIAG_MASK_INF
- GGML_OP_GET_ROWS
- GGML_OP_RMS_NORM
- GGML_OP_SILU
- GGML_OP_SOFT_MAX

add necessary ggml operations GGML_OP_ADD1, GGML_OP_SILU_BACK, GGML_OP_RMS_NORM_BACK, GGML_OP_DIAG_MASK_ZERO, and GGML_OP_ROPE_BACK

GGML_OP_ADD1 is necessary to add a scalar value in the backward pass of GGML_OP_SOFT_MAX
GGML_OP_ADD1 could also be replaced by using GGML_OP_ADD and GGML_OP_REPEAT, but the performance would be worse. additionally GGML_OP_REPEAT will return unexpected value when the the input to GGML_OP_SOFT_MAX contains only a single scalar. in this case GGML_OP_REPEAT will not return the value that should be repeated (src1) but the value which shape the result should take (src0). So in this case it can not replace GGML_OP_ADD1.

GGML_OP_SILU_BACK, GGML_OP_RMS_NORM_BACK and GGML_OP_ROPE_BACK are necessary for backward pass of GGML_OP_SILU, GGML_OP_RMS_NORM and GGML_OP_ROPE. The backward pass for these functions cannot be easily composed of existing operations. Since the backward pass builds a computation graph we need operations forward pass implementations of the the required backward passes. Sounds a bit confusing at first, I know...

GGML_OP_DIAG_MASK_ZERO is necessary for backward pass of GGML_OP_DIAG_MASK_INF.

Some operations where previously inplace-only. for backward pass there needs to be non-inplace variants.
staying consistent with other operations that have non-inplace and inplace variants, the operations are changed to non-inplace and
functions with "_inplace" are added which are inplace.
in llama we need to call the inplace variants so that it is implemented as before.
for llama backward pass we need to use the non-inplace variants.

still not completely implemented backward passes for llama:

- GGML_OP_ROPE: needs forward pass for GGML_OP_ROPE_BACK
- GGML_OP_GET_ROWS: only necessary for tokenizer

* norm & rms_norm can not be threaded:

after investigation rms norm for quite some time I come to the conclusion that neither norm, nor rms_norm can be threaded, because we need mean over all items, not just of the slices each thread sees.

* remove already resolved TODO

* implement backward pass of ggml_rope and ggml_rope_back

* implement backward pass for ggml_get_rows and for new operation ggml_get_rows_back

* add test-grad0.c

* use GGML_PRINT_DEBUG for debug messages which will otherwise flood the console

* test both gradients of mul_mat

* disable graph dot export as it floods console

* bug fixes for silu_back

* successfully test silu backward

* bug fix for scale backward pass

use sum instead of mean for gradient of scalar scale parameter

* successfully test scale backward

* improve performance of sum backward pass

use add1(x,y) instead of add(x,repeat(y,x))

* improve performance of sqr backward pass

use scale(x,y) instead of mul(x,repeat(y,x))

* successfully test rope backward

* bug fix for cpy backward pass

* successfully test cpy backward

* bug fix for reshape backward pass

* successfully test reshape backward

* add test-opt.c

this uses ggml_opt to train a,b for minimal e=sum(sqr(c - a*b)) for random initial a,b,c

* correctly implement softmax backward pass using new operation ggml_diag

ggml_diag constructs diagonal matrices with entries.
ggml_diag(shape[a,1,c,d]) -> shape[a,a,c,d]

* successfully test soft_max backward

* align shape annotations

* add shape annotations for llama

* de-duplicate ggml_forward_dup code taking care of contiguous tensors of same type.

with this we can duplicate tensor of any typ as long as they are contiguous.

* fix ggml_compute_forward_dup_same_cont for when nelements < nthreads

when more threads are used than elements exist ie1 was less than ie0, resulting in invalid negative byte count argument in memcpy

* bug fix for add_at forward

required for view backward pass

src0 values must be copied to dst, because during addition we don't touch all dst elements in contrast to the normal add function.

* successfully test view backward

* minor code format improvement

* fix ggml_forward_add functions to work correctly with transposed tensors

uses the same logic as in ggml_compute_forward_add_q_f32, but make it consistent across all ggml_compute_forward_add_... functions.
this also slightly changes the mem access pattern of the different threads to works as in ggml_compute_forward_add_q_f32.

* fix ggml_forward_add1 functions to work correctly with transposed tensors

uses the same logic as in ggml_compute_forward_add1_q_f32, but make it consistent across all ggml_compute_forward_add1_... functions.
this also slightly changes the mem access pattern of the different threads to works as in ggml_compute_forward_add1_q_f32.

* test-grad0.c : add print_elements to help with debugging

* successfully test permute backward

* some minor test-grad0 fixes

* fix sub, mul and div functions to work correctly with transposed tensors

uses the same logic as in add

* implement ggml_cont backward pass

* successfully test transpose backward and permute for all permutations

also test sub, mul and div up to max n_dims

* test-grad0.c add TODO for view_2d and view_3d

add_at (required for view backward pass) is a bit tricky for n_dims > 1.

* fix comments

* successfully test diag_mask_inf and diag_mask_zero backward

* test-grad0 : fix test for div

nargs and ndims was swapped, corrupting the stack

* fix diag_mask to work with non-inplace input

* move dup call into the actual add_at functions

* fix get rows backward pass

* successfully test get_rows backward

* fix view backward pass

add nb parameters to add_at like in view.
together with offset they define how to view dst and src0 during the add_at operation.

* successfully test backward pass of view_1d, view_2d and view_3d

* fix backward pass for rms_norm

I would have used formulas from other frameworks, but they differed so I could not decide which is correct.
Instead it was derived here in comment using manual forward-backward automatic differention of rms_norm and simplification.

* successfully test backward pass of rms_norm

some tests may fail when gradients are large.
could not find a satisfying configuration to check for abs error and relative error that passes all tests while still actually testing the results with tight enough error bounds.
when looking at the values the "failed" tests look actually ok. for example:

rms_norm: ndims=2, i=0, k=2, x0=0.000153, xm=0.000053, xp=0.000253, f0=0.278594, f1=0.086213, g0=961.905457, g1=966.064941, eps=0.000100, error_abs=4.159485, error_rel=0.004324

it is due to the test logic in check_gradients that they fail.

* add todos for llama backward pass

- implementation for ADD1 backward pass should probably use sum instead of mean (but this backward pass is not required)
- repeat is not yet tested and looks like it only works for single element src0 inputs.

* add operation ggml_sum_rows

ggml_sum_rows(shape[a,b,c,d]) -> shape[1,b,c,d]

* add missing GGML_OP_SUM_ROWS

* fix backward pass for repeat

requires ggml_sum_rows

* successfully test backward pass of repeat

* update quantization types in switch-case of add_at and add1

* add baby-llama example training a very small llama model from scratch to output a sinusoidal wave.

had to increase maximum number of optimization parameters to train from scratch.

* fix softmax in baby-llama example

* switching from training with adam to lbfgs produces much better results in the baby-llama example

* train with two examples, creating new tensors each time..

* fix bug when using ggml_opt to optimize params in one context and use a renewable context for eval and opt

when not keeping gradients of model parameters they are overwritten by tensors created by opt, which may be invalid after opt context is renewed.
so we need to keep the original gradients and make dups for opt

* train on multiple examples, generate & print tokens with trained model afterwards

ctx0 for evaluation and optimization is renewed for each sample

* add ggml_reshape_1d, ggml_reshape_4d and ggml_view_4d

* fix soft_max backward pass for input->ne[1] != 1

* add ggml_log operation necessary for cross entropy loss

* add test for ggml_log gradients

* implement backward pass for ggml_sum_rows, necessary for cross entropy loss

* implement ggml_repeat support for rank > 2 tensors

* add test for ggml_sum_rows gradients

* fix training get_example_targets

predict the next token, not the current token!

* add square_error_loss and cross_entropy_loss functions

* optimize loss over multiple samples

this increases computation graph, need parallel batched forward for more efficiency.

* fix backward pass for add_at and change arguments to have same order as in view

* add ggml_set(ctx, a, b) to set b in view of a and return modified a

necessary to set values into kv_self cache and properly propagate the gradients

* fix kv_self gradients for training

use ggml_set instead of ggml_cpy to set kv_self cache with properly propagating gradients

* replace inplace operations for training with copying operations to allow gradient propagation

* add GGML_ASSERT to catch ggml_rope and back value errors

* add trainable lora-only model with all big matrices C split into A,B with A*B=C

this is not a lora-finetune, but the whole model changed to have only low-rank "lora" matrices.

training this instead of the normal model resulted in much worse results though...

* vastly improve training results

instead of logit targets 0 and 1 use -1 and +1.

* shorten code using a variable

* change name of GGML_OP_ADD_AT to GGML_OP_ACC

* smaller default values for baby llama model parameters

* update static assert of GGML_OP_COUNT

* remove shape annotations in llama_eval_internal

* revert disabling of threading for rms_norm and norm

* rename print functions in baby-llama example

* fix call to ggml_set_name

* add missing include for strcmp, etc

* remove trailing whitespace

* reduce number of test-grad0 iterations

avoid exceeding timeout of automated tests

* remove busy loop that was used as sleep for slower sinus wave generation

* disable slow tests grad0 and opt to avoid exceeding timeouts

* c++ in baby-llama example

use c++ includes instead of c includes
use std::min, std::max instead of MIN, MAX macros

* c++ in baby-llama example

use c++ includes instead of c includes
use std::min, std::max instead of MIN, MAX macros

* ggml : fix compiler warnings + cosmetic changes

* ggml : fix nullptr derefs in GGML_OP_CONT and GGML_OP_RESHAPE back

* swap arguments to vDSP_vdiv call

documentation for vDSP_vdiv states: "Note that B comes before A!"

* swap arguments to vDSP_vdiv call

documentation for vDSP_vdiv states: "Note that B comes before A!"

* ggml : swap vDSP_vsub args as per documentation

* add parallel batched forward function for baby-llama training

* cleanup code for batched training

* remove trailing whitespace

* minor : fix compiler warnings + indentation style

* ggml : fix null ptr deref in backward pass

* ggml : remove Q4_2 remnants

* ggml : fix clang-tidy warnings

* baby-llama : couple of clang-tidy warnings

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-05-13 15:56:40 +03:00
Stephan Walter
f0d70f147d
Various fixes to mat_mul benchmark (#1253) 2023-04-30 12:32:37 +00:00
xaedes
0c5692345d
examples : add save_load_state example (#1150)
* add save_load_state example

* use <cstdio> instead of <iostream> and fprintf / printf instead of cout

* renamed save-load-state example files replacing underscores by dashes
2023-04-24 19:23:31 +03:00
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
Georgi Gerganov
a316a425d0
Overhaul the examples structure
- main -> examples
- utils -> examples (renamed to "common")
- quantize -> examples
- separate tools for "perplexity" and "embedding"

Hope I didn't break something !
2023-03-25 20:26:40 +02:00