]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/log
pkg/ggml/sources/llama.cpp
14 months agollama : fix restoring the number of outputs from state files (#6687)
compilade [Mon, 15 Apr 2024 12:56:55 +0000 (08:56 -0400)]
llama : fix restoring the number of outputs from state files (#6687)

14 months agoserver : revert "minor layout improvements" (#6684)
Pierrick Hymbert [Mon, 15 Apr 2024 12:18:47 +0000 (14:18 +0200)]
server : revert "minor layout improvements" (#6684)

This reverts commit b3a96f27f065a828f08c5d89ff60aab5361188fe.

14 months agoswift : linux support (#6590)
Steven Prichard [Mon, 15 Apr 2024 10:14:46 +0000 (05:14 -0500)]
swift : linux support (#6590)

- Package.swift now supports conditional compilation based on OS
- Allows for package to be used by SPM on Non-Apple platforms

Co-authored-by: Steven Prichard <redacted>
14 months agofix mul_mat_id() for new input, make the ut pass (#6682)
Neo Zhang Jianyu [Mon, 15 Apr 2024 09:12:26 +0000 (17:12 +0800)]
fix mul_mat_id() for new input, make the ut pass (#6682)

14 months agollama : add missing kv clear in llama_beam_search (#6664)
David Renshaw [Sun, 14 Apr 2024 19:24:15 +0000 (15:24 -0400)]
llama : add missing kv clear in llama_beam_search (#6664)

14 months agoAdd Command R chat template (#6650)
Chao Jiang [Sun, 14 Apr 2024 16:16:34 +0000 (00:16 +0800)]
Add Command R chat template (#6650)

* Add chat template for command-r model series

* Fix indentation

* Add chat template test for command-r models and update the implementation to trim whitespaces

* Remove debug print

14 months agoflake.lock: Update (#6669)
Georgi Gerganov [Sun, 14 Apr 2024 13:55:30 +0000 (16:55 +0300)]
flake.lock: Update (#6669)

14 months agoAdded support for GGML_OP_CLAMP in Metal (#6662)
Dave [Sun, 14 Apr 2024 11:14:19 +0000 (07:14 -0400)]
Added support for GGML_OP_CLAMP in Metal (#6662)

* Added support for GGML_OP_CLAMP in Metal

* Corrected size

---------

Co-authored-by: dave-fl <redacted>
14 months agoFix --split-max-size (#6655)
Sigbjørn Skjæret [Sun, 14 Apr 2024 11:12:59 +0000 (13:12 +0200)]
Fix --split-max-size (#6655)

* Fix --split-max-size

Byte size calculation was done on int and overflowed.

* add tests.sh

* add examples test scripts to ci run

Will autodiscover examples/*/tests.sh scripts and run them.

* move WORK_PATH to a subdirectory

* clean up before and after test

* explicitly define which scripts to run

* add --split-max-size to readme

14 months ago[bug fix] convert github repository_owner to lowercase (#6673)
Jaemin Son [Sun, 14 Apr 2024 11:12:36 +0000 (20:12 +0900)]
[bug fix] convert github repository_owner to lowercase (#6673)

14 months agoconvert : enable the `--use-temp-file` cli flag (#6645)
James A Capozzoli [Sun, 14 Apr 2024 08:40:18 +0000 (04:40 -0400)]
convert : enable the `--use-temp-file` cli flag (#6645)

14 months agofix memcpy() crash, add missed cmd in guide, fix softmax (#6622)
Neo Zhang Jianyu [Sun, 14 Apr 2024 02:42:29 +0000 (10:42 +0800)]
fix memcpy() crash, add missed cmd in guide, fix softmax (#6622)

* disable mmap to fix memcpy crash, add missed cmd in guide, fix softmax

* refactor to disable mmap for SYCL backend

* fix compile error in other os

* refactor the solution, use host buf to fix it, instead of disable mmap

* keep to support mmap()

* use host buff to reduce malloc times

* revert to malloc/free solution, for threaad safe

14 months agoCUDA: fix matrix multiplication logic for tests (#6667)
Johannes Gäßler [Sat, 13 Apr 2024 22:21:55 +0000 (00:21 +0200)]
CUDA: fix matrix multiplication logic for tests (#6667)

14 months agomodel: support arch `DbrxForCausalLM` (#6515)
Pierrick Hymbert [Sat, 13 Apr 2024 09:33:52 +0000 (11:33 +0200)]
model: support arch `DbrxForCausalLM` (#6515)

* model: dbrx convert to gguf
#6344

* llama: support dbrx
#6344

* doc: dbrx: add the model as supported

* scripts: get-wikitext-2 add unzip

* llama: increase maximum experts allowed

* llama: factorize moe graph implementation between grok, mixtral and dbrx

---------

Co-authored-by: Megha Agarwal <redacted>
14 months agoJSON schema conversion: ⚡️ faster repetitions, min/maxLength for strings, cap number...
Olivier Chafik [Fri, 12 Apr 2024 18:43:38 +0000 (19:43 +0100)]
JSON schema conversion: ⚡️ faster repetitions, min/maxLength for strings, cap number length (#6555)

* json: rename python schema converter to make import easier

* server: skip null json_schema / grammar fields

* json: deps management for primitive rules (+ allow null values)

* json: optimize repetitions for minItems/maxItems and regexps: `a{,3}` goes from `"a"? "a"? "a"?` (explosive combos) to `(a (a (a)?)?)?`

* grammars: add troubleshooting section to readme

* json: cap length of numbers to 15 digits before/after decimal point

(avoids infinite gen, e.g. "one third" -> `0.333333333333...`)

* json: unify all repetition code (w/ or w/o sep)

* json: support string minLength/maxLength

* server+json: update server/README w/ result_format

* nits

* json: fix type error w/ python 3.8

* json: fix server/README (json_schema in /completion vs. result_format in /v1/chat/completions)

* json: simplify DOT `{"type": "string", "pattern": "^.$"}`

* json: remove recursion in opt_repetitions (avoids Python stack overflow)

* json: rm dead code

* json: rm useless assert & ggml.h import

14 months agometal : unify mul_mv_id kernels (#6556)
slaren [Fri, 12 Apr 2024 16:13:20 +0000 (18:13 +0200)]
metal : unify mul_mv_id kernels (#6556)

14 months agoinfill : add download instructions for model (#6626)
Daniel Bevenius [Fri, 12 Apr 2024 12:11:46 +0000 (14:11 +0200)]
infill : add download instructions for model (#6626)

* infill : add download instructions for model

This commit adds instructions on how to download a CodeLlama model
using the `hf.sh` script. This will download the model and place it
in the `models` directory which is the same model use later by the
infill example.

Signed-off-by: Daniel Bevenius <redacted>
* squash! infill : add download instructions for model

Clarify the reason for using CodeLlama.

Signed-off-by: Daniel Bevenius <redacted>
---------

Signed-off-by: Daniel Bevenius <redacted>
14 months agoserver : coherent log output for KV cache full (#6637)
Pierrick Hymbert [Fri, 12 Apr 2024 11:49:21 +0000 (13:49 +0200)]
server : coherent log output for KV cache full (#6637)

14 months agollama : add gguf_remove_key + remove split meta during quantize (#6591)
jiez [Fri, 12 Apr 2024 10:45:06 +0000 (18:45 +0800)]
llama : add gguf_remove_key + remove split meta during quantize (#6591)

* Remove split metadata when quantize model shards

* Find metadata key by enum

* Correct loop range for gguf_remove_key and code format

* Free kv memory

---------

Co-authored-by: z5269887 <redacted>
14 months agochore: Fix markdown warnings (#6625)
Rene Leonhardt [Fri, 12 Apr 2024 08:52:36 +0000 (10:52 +0200)]
chore: Fix markdown warnings (#6625)

14 months agoimatrix : remove invalid assert (#6632)
Georgi Gerganov [Fri, 12 Apr 2024 08:49:58 +0000 (11:49 +0300)]
imatrix : remove invalid assert (#6632)

14 months agoCorrect free memory and total memory. (#6630)
MasterYi1024 [Fri, 12 Apr 2024 08:28:12 +0000 (16:28 +0800)]
Correct free memory and total memory. (#6630)

Co-authored-by: MasterYi <redacted>
14 months agoeval-callback: use ggml_op_desc to pretty print unary operator name (#6631)
Pierrick Hymbert [Fri, 12 Apr 2024 08:26:47 +0000 (10:26 +0200)]
eval-callback: use ggml_op_desc to pretty print unary operator name (#6631)

14 months agoci : disable Metal for macOS-latest-cmake-x64 (#6628)
Georgi Gerganov [Fri, 12 Apr 2024 08:15:05 +0000 (11:15 +0300)]
ci : disable Metal for macOS-latest-cmake-x64 (#6628)

14 months agoOptimization: eliminate addition of redundant stacks when advancing grammar. (#6616)
Clint Herron [Fri, 12 Apr 2024 01:44:50 +0000 (21:44 -0400)]
Optimization: eliminate addition of redundant stacks when advancing grammar. (#6616)

14 months agoAs suggested by @slaren, disabling Metal for test to fix CI build on OSX from #6576...
Clint Herron [Thu, 11 Apr 2024 21:44:48 +0000 (17:44 -0400)]
As suggested by @slaren, disabling Metal for test to fix CI build on OSX from #6576 (#6619)

14 months agoRefactor Error Handling for CUDA (#6575)
Nikolas [Thu, 11 Apr 2024 19:56:29 +0000 (21:56 +0200)]
Refactor Error Handling for CUDA (#6575)

* Refactor Error Handling for CUDA

Add guidance for setting CUDA_DOCKER_ARCH to match GPU compute capability for CUDA versions < 11.7. Include link to NVIDIA's CUDA GPUs documentation for compute capability reference.

* Update Makefile

Improved wording

Co-authored-by: Johannes Gäßler <redacted>
---------

Co-authored-by: Johannes Gäßler <redacted>
14 months agogrammars: 1.5x faster inference w/ complex grammars (vector reserves / reuses) (...
Olivier Chafik [Thu, 11 Apr 2024 18:47:34 +0000 (19:47 +0100)]
grammars: 1.5x faster inference w/ complex grammars (vector reserves / reuses) (#6609)

* grammars: reserve rejects & next candidates

* grammars: reuse new_stacks

* grammars: fix missing sig change in llama.h

* grammars: fix test (api changed)

* grammars: update gbnf-validator.cpp

* grammars: simpler syntax (no swap)

14 months agoci: download artifacts to release directory (#6612)
Hugo Roussel [Thu, 11 Apr 2024 17:52:21 +0000 (19:52 +0200)]
ci: download artifacts to release directory (#6612)

When action download-artifact was updated to v4, the default download path changed.
This fix binaries not being uploaded to releases.

14 months agoscripts : add --outdir option to hf.sh (#6600)
Daniel Bevenius [Thu, 11 Apr 2024 13:22:47 +0000 (15:22 +0200)]
scripts : add --outdir option to hf.sh (#6600)

* scripts : add --outdir option to hf.sh

This commit adds an option to the hf.sh script that allows the user to
specify an output directory for the downloaded file.

The motivation for this changes is that examples that use the hf.sh
script to download models from huggingface can now specify the output
directory, perhaps to the `models` directory to keep them in one place
and not clutter the root directory.

Signed-off-by: Daniel Bevenius <redacted>
* squash! scripts : add --outdir option to hf.sh

Fix format of the --outdir option in the usage message.

Signed-off-by: Daniel Bevenius <redacted>
---------

Signed-off-by: Daniel Bevenius <redacted>
14 months agoeval-callback: Example how to use eval callback for debugging (#6576)
Pierrick Hymbert [Thu, 11 Apr 2024 12:51:07 +0000 (14:51 +0200)]
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 <redacted>
Co-authored-by: Georgi Gerganov <redacted>
14 months agogguf : add option to not check tensor data (#6582)
Daniel Bevenius [Wed, 10 Apr 2024 18:16:48 +0000 (20:16 +0200)]
gguf : add option to not check tensor data (#6582)

This commit adds an option to the gguf example to not check the tensor
data.

The motivation for this is that it can be nice to use the gguf tool to
read other .gguf files that were not created by the gguf tool.

Signed-off-by: Daniel Bevenius <redacted>
14 months agominor layout improvements (#6572)
Ralph Soika [Wed, 10 Apr 2024 17:18:25 +0000 (19:18 +0200)]
minor layout improvements (#6572)

* minor layout improvements

* added missing file, run deps.sh locally

14 months agollama : add model types for mixtral (#6589)
slaren [Wed, 10 Apr 2024 15:24:14 +0000 (17:24 +0200)]
llama : add model types for mixtral (#6589)

14 months agoconvert.py : add consolidated.safetensors for mixtral 8x22b (#6587)
slaren [Wed, 10 Apr 2024 13:23:12 +0000 (15:23 +0200)]
convert.py : add consolidated.safetensors for mixtral 8x22b (#6587)

14 months agodocs : how to add a model (#6565)
Pierrick Hymbert [Wed, 10 Apr 2024 06:58:48 +0000 (08:58 +0200)]
docs : how to add a model (#6565)

* docs: how to add a model

* docs: model: typo and docs

* docs: model: add prevision on RoPE

* docs: model: rephrasing README.md

* docs: model: rephrasing README.md

* docs: model: README.md fix trailing spaces

* docs : some fixes

* Update README.md

---------

Co-authored-by: Georgi Gerganov <redacted>
14 months agoreadme : fix ROCm link (#6579)
Artem Zinnatullin [Wed, 10 Apr 2024 06:49:12 +0000 (00:49 -0600)]
readme : fix ROCm link (#6579)

14 months agoreadme : update UI list (#6560)
sjxx [Wed, 10 Apr 2024 06:34:00 +0000 (14:34 +0800)]
readme : update UI list (#6560)

14 months agoreadme: fix typo in amdgpu target name (#6573)
Jiří Sejkora [Tue, 9 Apr 2024 22:23:02 +0000 (00:23 +0200)]
readme: fix typo in amdgpu target name (#6573)

14 months agoBERT tokenizer fixes (#6498)
Jared Van Bortel [Tue, 9 Apr 2024 17:44:08 +0000 (13:44 -0400)]
BERT tokenizer fixes (#6498)

Key changes:
* BERT conversion: fix abuse of LlamaHfVocab, do not set BOS or EOS
* Nomic Embed conversion: pad vocab instead of slicing embedding tensor
* llama_tokenize: handle added special tokens like HF does

14 months agosync : ggml
Georgi Gerganov [Tue, 9 Apr 2024 17:29:06 +0000 (20:29 +0300)]
sync : ggml

14 months agoserver : detect search query to start webchat (#6554)
Ed Lee [Tue, 9 Apr 2024 08:31:47 +0000 (01:31 -0700)]
server : detect search query to start webchat (#6554)

14 months agollama : add Command R Plus support (#6491)
Carolinabanana [Tue, 9 Apr 2024 08:16:13 +0000 (09:16 +0100)]
llama : add Command R Plus support (#6491)

* Add Command R Plus GGUF

* Add Command R Plus GGUF

* Loading works up to LayerNorm2D

* Export new tensors in 1D so they are not quantized.

* Fix embedding layer based on Noeda's example

* Whitespace

* Add line

* Fix unexpected tokens on MPS. Re-add F16 fix. ((Noeda)

* dranger003: Fix block index overflow in CUDA dequantizing.

* Reverted blocked multiplication code as it still has issues and could affect other Llama arches

* export norms as f32

* fix overflow issues during quant and other cleanup

* Type convention

Co-authored-by: Georgi Gerganov <redacted>
* dranger003: Fix more int overflow during quant.

---------

Co-authored-by: S <redacted>
Co-authored-by: S <redacted>
Co-authored-by: slaren <redacted>
Co-authored-by: Georgi Gerganov <redacted>
14 months agolicense : update copyright notice + add AUTHORS (#6405)
Georgi Gerganov [Tue, 9 Apr 2024 06:23:19 +0000 (09:23 +0300)]
license : update copyright notice + add AUTHORS (#6405)

* license : add AUTHORS

* authors : update

* scipts : add LICENSE and gen-authors.sh to sync

14 months agollama : fix attention layer count sanity check (#6550)
Georgi Gerganov [Mon, 8 Apr 2024 19:25:49 +0000 (22:25 +0300)]
llama : fix attention layer count sanity check (#6550)

* llama : fix attention layer count sanity check

* llama : fix parentheses in attention layer count sanity check

There was otherwise a warning when compiling.

---------

Co-authored-by: Francis Couture-Harpin <redacted>
14 months agoComment explaining a decision (#6531)
kunnis [Mon, 8 Apr 2024 15:44:19 +0000 (10:44 -0500)]
Comment explaining a decision (#6531)

14 months agoquantize : fix precedence of cli args (#6541)
Georgi Gerganov [Mon, 8 Apr 2024 13:23:01 +0000 (16:23 +0300)]
quantize : fix precedence of cli args (#6541)

14 months agollama : support negative ith in llama_get_ API (#6519)
Rick G [Mon, 8 Apr 2024 13:02:30 +0000 (06:02 -0700)]
llama : support negative ith in llama_get_ API (#6519)

* llama_sampling_sample with default args is more naively usable

* Batches populated by either llama_batch_get_one or llama_batch_add work with default args
  * Previously get_one could use the default argument
  * Previously add should usually have used the last index where logits[idx] == true
* This hopefully encourages the use of llama_batch_add
  * By giving expected results when using default arguments.
* Adds "negative indexing" feature to llama_get_logits_ith and llama_get_embeddings_ith
* Believed to work with any currently well behaved program
  * Default arg now works for both cases (previously would give strange results for add case)
  * Any non-negative number is unaffected and behaves as previously
  * Negative arguments were previously invalid.
* Implemented as a special case of indexing as suggested by @compilade in https://github.com/ggerganov/llama.cpp/pull/6519

* Fixed mismatch type errors

* cited in macOS CI tests
* Missed in original updates based on PR feedback in https://github.com/ggerganov/llama.cpp/pull/6519

14 months agollama : save and restore kv cache for single seq id (#6341)
Jan Boon [Mon, 8 Apr 2024 12:43:30 +0000 (20:43 +0800)]
llama : save and restore kv cache for single seq id (#6341)

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

* remove trailing whitespace

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

* add kv seq save restore to test case

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

* Returning 0 for some cases, instead of asserting.

* cleanup error cases

* rename sequence state functions

* rename state get set functions

* add previous function names back in with DEPRECATED notice

* update doc

* adjust endpoints to preferred style

* fix restoring zero cell count

* handle seq rm return value

* unused param

* keep in the size check

* fix return types

* add server test case for slot save restore

* cleanup

* add cake

* cleanup style

* add special

* removing a whole sequence never fails

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

* catch exceptions on save as well

* error log messages

* check types for stricter restore

* update server doc

* readme : update API changes date

* strict filename validation

* move include, reject bom as well

* also reject empty filename

* reject whitespace and trailing dot

---------

Co-authored-by: Martin Evans <redacted>
Co-authored-by: Georgi Gerganov <redacted>
14 months agoremove row=1 cond (#6532)
Abhilash Majumder [Mon, 8 Apr 2024 08:26:01 +0000 (13:56 +0530)]
remove row=1 cond (#6532)

14 months agoAdding KodiBot to UI list (#6535)
Firat [Mon, 8 Apr 2024 07:48:29 +0000 (00:48 -0700)]
Adding KodiBot to UI list (#6535)

KodiBot is free and open source ai chat app released under the GNU General Public License.

14 months agoChange Windows AMD example to release build to make inference much faster. (#6525)
Mark Fairbairn [Sun, 7 Apr 2024 18:52:19 +0000 (19:52 +0100)]
Change Windows AMD example to release build to make inference much faster. (#6525)

14 months agoflake.lock: Update (#6517)
Georgi Gerganov [Sun, 7 Apr 2024 18:25:30 +0000 (21:25 +0300)]
flake.lock: Update (#6517)

Flake lock file updates:

• Updated input 'flake-parts':
    'github:hercules-ci/flake-parts/f7b3c975cf067e56e7cda6cb098ebe3fb4d74ca2' (2024-03-01)
  → 'github:hercules-ci/flake-parts/9126214d0a59633752a136528f5f3b9aa8565b7d' (2024-04-01)
• Updated input 'flake-parts/nixpkgs-lib':
    'github:NixOS/nixpkgs/1536926ef5621b09bba54035ae2bb6d806d72ac8?dir=lib' (2024-02-29)
  → 'github:NixOS/nixpkgs/d8fe5e6c92d0d190646fb9f1056741a229980089?dir=lib' (2024-03-29)
• Updated input 'nixpkgs':
    'github:NixOS/nixpkgs/d8fe5e6c92d0d190646fb9f1056741a229980089' (2024-03-29)
  → 'github:NixOS/nixpkgs/fd281bd6b7d3e32ddfa399853946f782553163b5' (2024-04-03)

Co-authored-by: github-actions[bot] <redacted>
14 months agoAdd GritLM as supported models. (#6513)
DAN™ [Sun, 7 Apr 2024 17:33:59 +0000 (13:33 -0400)]
Add GritLM as supported models. (#6513)

14 months agosync : ggml
Georgi Gerganov [Sun, 7 Apr 2024 14:05:51 +0000 (17:05 +0300)]
sync : ggml

14 months agoggml: bypass code incompatible with CUDA < 11.1 (whisper/2020)
Slava Primenko [Thu, 4 Apr 2024 12:49:24 +0000 (14:49 +0200)]
ggml: bypass code incompatible with CUDA < 11.1 (whisper/2020)

`cudaHostRegisterReadOnly` parameter was only introduced in CUDA 11.1

See this issue for more details:
https://github.com/ggerganov/examples/whisper/whisper.cpp/issues/2007

14 months agoscripts : sync ggml-cuda folder
Georgi Gerganov [Sun, 7 Apr 2024 13:08:12 +0000 (16:08 +0300)]
scripts : sync ggml-cuda folder

14 months agoRun make to build the project (#6457)
limitedAtonement [Sun, 7 Apr 2024 11:05:40 +0000 (07:05 -0400)]
Run make to build the project (#6457)

14 months agosupport/fix OPs GGML_TYPE_IQ4_NL, GGML_TYPE_IQ4_XS, GGML_TYPE_IQ3_XXS, GGML_TYPE_IQ3_...
Neo Zhang Jianyu [Sun, 7 Apr 2024 02:55:59 +0000 (10:55 +0800)]
support/fix OPs GGML_TYPE_IQ4_NL, GGML_TYPE_IQ4_XS, GGML_TYPE_IQ3_XXS, GGML_TYPE_IQ3_S, GGML_TYPE_IQ2_XXS, GGML_TYPE_IQ2_XS, GGML_TYPE_IQ2_S, GGML_TYPE_IQ1_S, GGML_TYPE_IQ1_M (#6521)

14 months agosync : ggml
Georgi Gerganov [Sat, 6 Apr 2024 14:43:15 +0000 (17:43 +0300)]
sync : ggml

14 months agobackend : fix typo in scheduler documentation (ggml/781)
Daniel Bevenius [Wed, 3 Apr 2024 20:57:20 +0000 (22:57 +0200)]
backend : fix typo in scheduler documentation (ggml/781)

Signed-off-by: Daniel Bevenius <redacted>
14 months agoTests: Added integration tests for GBNF parser (#6472)
Clint Herron [Sat, 6 Apr 2024 14:31:33 +0000 (10:31 -0400)]
Tests: Added integration tests for GBNF parser  (#6472)

* Added integration tests for GBNF parser to validate correctness of parsing, as well as correctness of string matching. Intended for use to pin behavior while working on performance improvements.

* Fixing whitespace errors and cleaning error message alert to be clearer.

* Removing hacky include to llama.cpp from grammar integration test now that needed functions are available via internal API.

* Comment cleanup.

* Reorganizing tests for readability.

* Cleaning up debug message to make a bit more sense.

14 months agoci: bench: support sse and fix prompt processing time / server: add tokens usage...
Pierrick Hymbert [Sat, 6 Apr 2024 03:40:47 +0000 (05:40 +0200)]
ci: bench: support sse and fix prompt processing time / server: add tokens usage in stream OAI response (#6495)

* ci: bench: support sse and fix prompt processing time
server: add tokens usage in stream mode

* ci: bench: README.md EOL

* ci: bench: remove total pp and tg as it is not accurate

* ci: bench: fix case when there is no token generated

* ci: bench: change to the 95 percentile for pp and tg as it is closer to what the server exports in metrics

* ci: bench: fix finish reason rate

14 months agogguf.py : add licence and version to gguf writer (#6504)
Brian [Fri, 5 Apr 2024 18:41:38 +0000 (05:41 +1100)]
gguf.py : add licence and version to gguf writer (#6504)

14 months agoreadme : update UI list (#6503)
Hoang Nguyen [Fri, 5 Apr 2024 18:39:43 +0000 (11:39 -0700)]
readme : update UI list (#6503)

* Add MindMac to UI list

* Update proprietary description

Co-authored-by: slaren <redacted>
---------

Co-authored-by: slaren <redacted>
14 months agobench : make n_batch and n_ubatch configurable in Batched bench (#6500)
Ting Sun [Fri, 5 Apr 2024 18:34:53 +0000 (01:34 +0700)]
bench : make n_batch and n_ubatch configurable in Batched bench (#6500)

* bench: make n_batch and n_ubatch configurable

* bench: update doc for batched bench

14 months ago[SYCL] Fixed minor bug when enabling FP16 for non intel targets (#6464)
Ouadie EL FAROUKI [Fri, 5 Apr 2024 13:35:06 +0000 (14:35 +0100)]
[SYCL] Fixed minor bug when enabling FP16 for non intel targets (#6464)

* moved INTEL_MKL guard from gemm_impl to gemm (wrapper)

* Update ggml-sycl.cpp

Co-authored-by: AidanBeltonS <redacted>
---------

Co-authored-by: AidanBeltonS <redacted>
14 months agoreadme : add Dot to UI list (#6487)
alexpinel [Thu, 4 Apr 2024 17:22:50 +0000 (18:22 +0100)]
readme : add Dot to UI list (#6487)

14 months agoreadme : fix typo (#6481)
Jun Jie [Thu, 4 Apr 2024 17:16:37 +0000 (01:16 +0800)]
readme : fix typo (#6481)

14 months agoserver: add cURL support to server Dockerfiles (#6474)
Ed Lepedus [Thu, 4 Apr 2024 16:31:22 +0000 (17:31 +0100)]
server: add cURL support to server Dockerfiles (#6474)

* server: add cURL support to `full.Dockerfile`

* server: add cURL support to `full-cuda.Dockerfile` and `server-cuda.Dockerfile`

* server: add cURL support to `full-rocm.Dockerfile` and `server-rocm.Dockerfile`

* server: add cURL support to `server-intel.Dockerfile`

* server: add cURL support to `server-vulkan.Dockerfile`

* fix typo in `server-vulkan.Dockerfile`

Co-authored-by: Georgi Gerganov <redacted>
---------

Co-authored-by: Georgi Gerganov <redacted>
14 months agoci: exempt master branch workflows from getting cancelled (#6486)
Minsoo Cheong [Thu, 4 Apr 2024 16:30:53 +0000 (01:30 +0900)]
ci: exempt master branch workflows from getting cancelled (#6486)

* ci: exempt master branch workflows from getting cancelled

* apply to bench.yml

14 months agobuild CI: Name artifacts (#6482)
Ewout ter Hoeven [Thu, 4 Apr 2024 15:08:55 +0000 (17:08 +0200)]
build CI: Name artifacts (#6482)

Name the artifacts in the build CI, so that they get uploaded with separate names, instead of all put into the same `artifact` ZIP.

It might be possible to further simplify the packing step (in future PRs).

14 months agoserver: allow penalizing repetition of newlines on server webpage (#6431)
Shakhar Dasgupta [Thu, 4 Apr 2024 15:03:00 +0000 (11:03 -0400)]
server: allow penalizing repetition of newlines on server webpage (#6431)

14 months agoci: bench fix concurrency for workflow trigger dispatch with sha1 (#6478)
Pierrick Hymbert [Thu, 4 Apr 2024 14:59:04 +0000 (16:59 +0200)]
ci: bench fix concurrency for workflow trigger dispatch with sha1 (#6478)

14 months agoCorrect README link (#6458)
limitedAtonement [Thu, 4 Apr 2024 14:30:02 +0000 (10:30 -0400)]
Correct README link (#6458)

README is called README.md.

14 months agoci: bench: add more ftype, fix triggers and bot comment (#6466)
Pierrick Hymbert [Thu, 4 Apr 2024 09:57:58 +0000 (11:57 +0200)]
ci: bench: add more ftype, fix triggers and bot comment (#6466)

* ci: bench: change trigger path to not spawn on each PR

* ci: bench: add more file type for phi-2: q8_0 and f16.
- do not show the comment by default

* ci: bench: add seed parameter in k6 script

* ci: bench: artefact name perf job

* Add iteration in the commit status, reduce again the autocomment

* ci: bench: add per slot metric in the commit status

* Fix trailing spaces

14 months agocommon: remove duplicate check for curl (#6471)
Daniel Bevenius [Thu, 4 Apr 2024 07:49:21 +0000 (09:49 +0200)]
common: remove duplicate check for curl (#6471)

This commit removes one of the two identical checks for curl being NULL
in llama_load_model_from_url.

Signed-off-by: Daniel Bevenius <redacted>
14 months agoexamples : add GBNF validator program (#5948)
Clint Herron [Thu, 4 Apr 2024 07:44:28 +0000 (03:44 -0400)]
examples : add GBNF validator program (#5948)

* Revising GBNF validator program to be much simpler.

* Changing from streams to using cstdio

* Adding final newline character.

14 months agoserver : remove obsolete --memory-f32 option
Georgi Gerganov [Thu, 4 Apr 2024 06:34:58 +0000 (09:34 +0300)]
server : remove obsolete --memory-f32 option

14 months agoserver : add option to disable KV offload (#6468)
Xiao-Yong Jin [Thu, 4 Apr 2024 06:33:48 +0000 (01:33 -0500)]
server : add option to disable KV offload (#6468)

14 months agoconvert : fix for lint error complaining of bare except (#6470)
Clint Herron [Thu, 4 Apr 2024 06:32:53 +0000 (02:32 -0400)]
convert : fix for lint error complaining of bare except (#6470)

15 months agoA few small fixes to server's README docs (#6428)
Fattire [Wed, 3 Apr 2024 20:22:57 +0000 (13:22 -0700)]
A few small fixes to server's README docs (#6428)

* Typo fix to server's README.md

Fix minor typo ("tonen") in server README.

* server readme grammar/style fixes.

Quickly went through this file to look for inconsistencies in
presentation of defaults, flag options, and looked for typos
and grammar issues.

Not perfect, but hopefully improved.

* Update README.md

Remove an extra space before newline.

15 months agoserver : handle exception on wrong type in request (#6452)
JH23X [Wed, 3 Apr 2024 18:09:52 +0000 (20:09 +0200)]
server : handle exception on wrong type in request (#6452)

Co-authored-by: Jonas Holzner <redacted>
15 months agollama : add SEA-LION support (#6448)
bryanSwk [Wed, 3 Apr 2024 18:05:10 +0000 (02:05 +0800)]
llama : add SEA-LION support (#6448)

* initial commit for sealion support

* add sealion support

* minor fix

* q/k ln and pos_embd only if required

* Apply suggestions from code review

Co-authored-by: Georgi Gerganov <redacted>
* minor : clear whitespaces

---------

Co-authored-by: bryan <redacted>
Co-authored-by: Georgi Gerganov <redacted>
15 months agoci : update checkout, setup-python and upload-artifact to latest (#6456)
Ewout ter Hoeven [Wed, 3 Apr 2024 18:01:13 +0000 (20:01 +0200)]
ci : update checkout, setup-python and upload-artifact to latest (#6456)

* CI: Update actions/checkout to v4

* CI: Update actions/setup-python to v5

* CI: Update actions/upload-artifact to v4

15 months agoserver: add cURL support to `server.Dockerfile` (#6461)
Ed Lepedus [Wed, 3 Apr 2024 17:56:37 +0000 (18:56 +0100)]
server: add cURL support to `server.Dockerfile` (#6461)

15 months agoreadme : add feature-rich rust bindings (#6465)
Francisco Melo [Wed, 3 Apr 2024 17:53:37 +0000 (18:53 +0100)]
readme : add feature-rich rust bindings (#6465)

15 months agosecurity : create policy (#6354)
Joyce [Wed, 3 Apr 2024 17:48:07 +0000 (14:48 -0300)]
security : create policy (#6354)

* Create SECURITY.md

Signed-off-by: Joyce <redacted>
* Fix: link on SECURITY.md

Signed-off-by: Joyce <redacted>
* Fix: link on SECURITY.md

Signed-off-by: Joyce <redacted>
* minor

* fix

* fix

---------

Signed-off-by: Joyce <redacted>
Co-authored-by: Georgi Gerganov <redacted>
15 months agoMissing tokenizer.model error during gguf conversion (#6443)
Abhishek Gopinath K [Wed, 3 Apr 2024 15:42:52 +0000 (21:12 +0530)]
Missing tokenizer.model error during gguf conversion (#6443)

Co-authored-by: Jared Van Bortel <redacted>
15 months agoAdd OpenChat, Alpaca, Vicuna chat templates (#6397)
kaizau [Wed, 3 Apr 2024 15:24:31 +0000 (23:24 +0800)]
Add OpenChat, Alpaca, Vicuna chat templates (#6397)

* Add openchat chat template

* Add chat template test for openchat

* Add chat template for vicuna

* Add chat template for orca-vicuna

* Add EOS for vicuna templates

* Combine vicuna chat templates

* Add tests for openchat and vicuna chat templates

* Add chat template for alpaca

* Add separate template name for vicuna-orca

* Remove alpaca, match deepseek with jinja output

* Regenerate chat template test with add_generation_prompt

* Separate deepseek bos from system message

* Match openchat template with jinja output

* Remove BOS token from templates, unprefix openchat

15 months agoreadme : update hot topics
Georgi Gerganov [Wed, 3 Apr 2024 13:11:15 +0000 (16:11 +0300)]
readme : update hot topics

15 months agoggml : mul_mat_id use the same tensor for all the experts (#6387)
slaren [Wed, 3 Apr 2024 13:07:05 +0000 (15:07 +0200)]
ggml : mul_mat_id use the same tensor for all the experts (#6387)

* ggml : update mul_mat_id to use the same tensor for all the experts

* update cuda

* minor

* update metal

* update test-backend-ops

* fix cuda

* Update ggml-metal.m

Co-authored-by: Georgi Gerganov <redacted>
* update convert.py

* update convert-hf-to-gguf.py

* update convert.py for mixtral hf models

* Update convert-hf-to-gguf.py

Co-authored-by: Georgi Gerganov <redacted>
* cuda : support non-pow-2 number of experts

* allow quantize to work for split and merged experts models in the same way

* cleanup + disable mmap automatically with split tensors models

* update imatrix

* test-backend-ops : test qwen argsort

* update grok model loading

* llama : add merged experts tensors to the grok tensor map

* minor

* gguf : bump version

* fix quantizing of merged experts

* convert-hf-to-gguf.py : update grok (untested)

* make linter happy

* cuda/argsort : use shared memory instead of pool memory

* convert : fix grok tensor names

* metal : add support for non-pow-2 argsort

* llama : more loader cleanup, better error checking

* cuda : fix warning

* llama : still use mmap for loading old models, but copy the data to a host buffer

* add review note

* llama : remove ffn tensor counting + add sanity check

ggml-ci

* convert : fix handling of n_experts == None

ggml-ci

* imatrix : fix ncall counters

* llama : produce error if imatrix size does not match

* quantize : terminate on errors + trace logs

ggml-ci

* metal : pad shared memory to 16 bytes

---------

Co-authored-by: Georgi Gerganov <redacted>
15 months ago[SYCL] Disable iqx on windows as WA (#6435)
Meng, Hengyu [Wed, 3 Apr 2024 02:34:40 +0000 (10:34 +0800)]
[SYCL] Disable iqx on windows as WA (#6435)

* disable iqx on windows as WA

* array instead of global_memory

15 months agoflake.lock: Update (#6402)
Georgi Gerganov [Mon, 1 Apr 2024 16:05:57 +0000 (19:05 +0300)]
flake.lock: Update (#6402)

Flake lock file updates:

• Updated input 'nixpkgs':
    'github:NixOS/nixpkgs/44d0940ea560dee511026a53f0e2e2cde489b4d4' (2024-03-23)
  → 'github:NixOS/nixpkgs/d8fe5e6c92d0d190646fb9f1056741a229980089' (2024-03-29)

Co-authored-by: github-actions[bot] <redacted>
15 months agocompare-llama-bench.py: fix long hexsha args (#6424)
Johannes Gäßler [Mon, 1 Apr 2024 11:30:43 +0000 (13:30 +0200)]
compare-llama-bench.py: fix long hexsha args (#6424)

15 months agoci: server: verify deps are coherent with the commit (#6409)
Pierrick Hymbert [Mon, 1 Apr 2024 10:36:40 +0000 (12:36 +0200)]
ci: server: verify deps are coherent with the commit (#6409)

* ci: server: verify deps are coherent with the commit

* ci: server: change the ref to build as now it's a pull event target

15 months agoreadme : update hot topics
Georgi Gerganov [Sun, 31 Mar 2024 08:56:30 +0000 (11:56 +0300)]
readme : update hot topics

15 months agoci: bench: fix Resource not accessible by integration on PR event (#6393)
Pierrick Hymbert [Sat, 30 Mar 2024 10:36:07 +0000 (11:36 +0100)]
ci: bench: fix Resource not accessible by integration on PR event (#6393)

15 months agoFedora build update (#6388)
Mohammadreza Hendiani [Fri, 29 Mar 2024 21:59:56 +0000 (01:29 +0330)]
Fedora build update (#6388)

* fixed deprecated address

* fixed deprecated address

* fixed deprecated address

* Added 'Apache-2.0' SPDX license identifier due to 'kompute.cc' submodule licensing. Explanation of licensing method: https://docs.fedoraproject.org/en-US/legal/spdx/#_and_expressions

* Added 'Apache-2.0' SPDX license identifier due to 'kompute.cc' submodule licensing. Explanation of licensing method: https://docs.fedoraproject.org/en-US/legal/spdx/#_and_expressions

* Added 'Apache-2.0' SPDX license identifier due to 'kompute.cc' submodule licensing. Explanation of licensing method: https://docs.fedoraproject.org/en-US/legal/spdx/#_and_expressions

* reverted back to only the MIT license

15 months agosplit: allow --split-max-size option (#6343)
Xuan Son Nguyen [Fri, 29 Mar 2024 21:34:44 +0000 (22:34 +0100)]
split: allow --split-max-size option (#6343)

* split by max size

* clean up arg parse

* split: ok

* add dry run option

* error on 0 tensors

* be positive

* remove next_metadata_size